<%BANNER%>

Evolutionary Significance of Filial Cannibalism in Fishes with Parental Care

Permanent Link: http://ufdc.ufl.edu/UFE0021724/00001

Material Information

Title: Evolutionary Significance of Filial Cannibalism in Fishes with Parental Care
Physical Description: 1 online resource (151 p.)
Language: english
Creator: Klug, Hope M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: cannibalism, care, filial, flagfish, goby, infanticide, microsatellite
Zoology -- Dissertations, Academic -- UF
Genre: Zoology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Parental care typically increases offspring survival, thereby increasing parental fitness. Thus, it is surprising that filial cannibalism, the consumption of one's own offspring, is prevalent in fishes exhibiting parental care. The most widely-accepted hypothesis of filial cannibalism suggests that males gain energy from eggs that they invest into future reproduction (energy-based hypothesis). Recently, an alternative hypothesis suggested that partial-clutch cannibalism increases oxygen availability to remaining eggs, which in turn increases overall egg survival (oxygen-mediated hypothesis). Evidence for both hypotheses is mixed and there are few alternative hypotheses. Thus, the evolutionary significance of filial cannibalism remains unclear. To enhance our understanding of filial cannibalism, I re-examined current theory (i.e., the energy-based and oxygen-mediated hypotheses), developed and evaluated an alternative hypothesis, and developed a mathematical model of filial cannibalism. I experimentally quantified the effect of filial cannibalism on mating success of parental males in the flagfish (Jordanella floridae), and found that filial cannibalism always reduced lifetime reproductive success. In the sand goby (Pomatoschistus minutus), males that were in poorer condition consumed less of their eggs than males that were in better condition. These findings are contrary to predictions of the energy-based hypothesis. In the sand goby, I found that egg survival is density-dependent and filial cannibalism increases when egg density is high. However, this density-dependence is not mediated by oxygen. Therefore, I did not find support for the oxygen-mediated hypothesis. I suggest a more general hypothesis of filial cannibalism mediated by density-dependent egg survival. I hypothesize that the ability to preferentially cannibalize offspring of reduced quality might play a large role in the evolution of filial cannibalism (selective filial cannibalism hypothesis). To begin to understand the importance of selective cannibalism, I evaluated whether males cannibalize selectively in the sand goby and the flagfish. Male sand gobies cannibalized selectively with regard to egg development rate, and male flagfish cannibalized selectively with regard to egg energy and maternal size. Thus, selective filial cannibalism occurs in at least two species and this hypothesis warrants further attention. I then developed a mathematical model of filial cannibalism to isolate factors affecting the evolution of filial cannibalism. The findings of this model highlight the plausibility of a range of alternative hypotheses. Specifically, the evolution of filial cannibalism is enhanced if (1) parents can selectively cannibalize lower quality offspring, (2) filial cannibalism increases egg maturation rate, (3) there are energetic benefits of eggs to cannibalizing males, (4) cannibalism increases a parent's reproductive rate (e.g., through mate attractiveness). Density-dependent egg survivorship alone did not favor the evolution of cannibalism. Additionally, the results of the model suggest that population-level dynamics potentially play a large role in the evolution. Finally, I isolated and characterized six polymorphic microsatellite loci in the flagfish. These microsatellite markers will be useful in paternity assays and estimating heritability of flagfish behaviors, such as filial cannibalism.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Hope M Klug.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Brockmann, H. Jane Jane.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021724:00001

Permanent Link: http://ufdc.ufl.edu/UFE0021724/00001

Material Information

Title: Evolutionary Significance of Filial Cannibalism in Fishes with Parental Care
Physical Description: 1 online resource (151 p.)
Language: english
Creator: Klug, Hope M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: cannibalism, care, filial, flagfish, goby, infanticide, microsatellite
Zoology -- Dissertations, Academic -- UF
Genre: Zoology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Parental care typically increases offspring survival, thereby increasing parental fitness. Thus, it is surprising that filial cannibalism, the consumption of one's own offspring, is prevalent in fishes exhibiting parental care. The most widely-accepted hypothesis of filial cannibalism suggests that males gain energy from eggs that they invest into future reproduction (energy-based hypothesis). Recently, an alternative hypothesis suggested that partial-clutch cannibalism increases oxygen availability to remaining eggs, which in turn increases overall egg survival (oxygen-mediated hypothesis). Evidence for both hypotheses is mixed and there are few alternative hypotheses. Thus, the evolutionary significance of filial cannibalism remains unclear. To enhance our understanding of filial cannibalism, I re-examined current theory (i.e., the energy-based and oxygen-mediated hypotheses), developed and evaluated an alternative hypothesis, and developed a mathematical model of filial cannibalism. I experimentally quantified the effect of filial cannibalism on mating success of parental males in the flagfish (Jordanella floridae), and found that filial cannibalism always reduced lifetime reproductive success. In the sand goby (Pomatoschistus minutus), males that were in poorer condition consumed less of their eggs than males that were in better condition. These findings are contrary to predictions of the energy-based hypothesis. In the sand goby, I found that egg survival is density-dependent and filial cannibalism increases when egg density is high. However, this density-dependence is not mediated by oxygen. Therefore, I did not find support for the oxygen-mediated hypothesis. I suggest a more general hypothesis of filial cannibalism mediated by density-dependent egg survival. I hypothesize that the ability to preferentially cannibalize offspring of reduced quality might play a large role in the evolution of filial cannibalism (selective filial cannibalism hypothesis). To begin to understand the importance of selective cannibalism, I evaluated whether males cannibalize selectively in the sand goby and the flagfish. Male sand gobies cannibalized selectively with regard to egg development rate, and male flagfish cannibalized selectively with regard to egg energy and maternal size. Thus, selective filial cannibalism occurs in at least two species and this hypothesis warrants further attention. I then developed a mathematical model of filial cannibalism to isolate factors affecting the evolution of filial cannibalism. The findings of this model highlight the plausibility of a range of alternative hypotheses. Specifically, the evolution of filial cannibalism is enhanced if (1) parents can selectively cannibalize lower quality offspring, (2) filial cannibalism increases egg maturation rate, (3) there are energetic benefits of eggs to cannibalizing males, (4) cannibalism increases a parent's reproductive rate (e.g., through mate attractiveness). Density-dependent egg survivorship alone did not favor the evolution of cannibalism. Additionally, the results of the model suggest that population-level dynamics potentially play a large role in the evolution. Finally, I isolated and characterized six polymorphic microsatellite loci in the flagfish. These microsatellite markers will be useful in paternity assays and estimating heritability of flagfish behaviors, such as filial cannibalism.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Hope M Klug.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Brockmann, H. Jane Jane.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021724:00001


This item has the following downloads:


Full Text





EVOLUTIONARY SIGNIFICANCE OF FILIAL CANNIBALISM IN FISHES WITH
PARENTAL CARE




















By

HOPE KLUG


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2007


































2007 Hope Klug


































To my family, who has always supported me.









ACKNOWLEDGMENTS

I would like to thank my advisor, H. Jane Brockmann, and my supervisory committee,

Rebecca Kimball, Kai Lindstrom, Steve Phelps, and Clive Wynne, for their guidance and

feedback throughout my graduate career. Their constant support and thoughtfulness has

facilitated this dissertation work and, more generally, my professional development as a scientist.

I am indebted to the lab group of Ben Bolker, Craig Osenberg, and Colette St. Mary for

providing me with logistical support, as well as critical feedback during the first four years of my

graduate career. It was in their lab group that I learned to think and write logically, and I am

thankful for every bit of critical feedback I received from this lab group. I thank Mike Bonsall,

whose support, patience, and passion for science made chapter 6 possible. I am also grateful for

the logistical support provided during two visits to Oxford. Similarly, Kai Lindstrom and his lab

group provided me with space, supplies, and a great deal of assistance at Tvarminne Zoological

Station. My work on the sand goby would have been impossible without their support.

My dissertation research was greatly improved by the funding I received throughout my

graduate career. In particular, I'm grateful for the funding provided by an NSF Graduate

Research Fellowship, a SPICE fellowship (PI: Doug Levey), an NSF Dissertation Improvement

Grant, and the Department of Zoology at the University of Florida. I also thank the Department

of Zoology and the University of Helsinki's Tvarminne Zoological Station for additional

logistical support.

Finally, I thank my family and friends, whose continuous love and support allowed me to

pursue my interests in science. Without their encouragement, none of this would have been

possible.









TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ........................... .................. ............................................................4

LIST OF TABLES ................... ........................ ...............

LIST OF FIGURES ................................... .. .... .... ................. 10

ABSTRAC T ............................... ..................... 12

CHAPTER

1 INTRODUCTION ............... ................. ........... .............................. 14

Background: Theory and Empirical Evidence................... ... ...............14
Re-Evaluation of Current Theory with an Explicit Focus on Fitness Consequences......17
Development and Evaluation of Alternative Hypotheses .............................................17
Development of a Synthetic Model of Filial Cannibalism.................. ............... 18
Summary of Dissertation Objectives ............................ ............................. ............... 18
S tu dy S y stem s ............... .......................... .. .. ................................................. 19
O verview of D issertation C hapters................................................................................ .... 20
Reproductive Fitness Consequences of Filial Cannibalism in the Flagfish ..................20
Parents Benefit from Eating Offspring: Density-Dependent Egg Survivorship
C om pensates for Filial C annibalism .................................................. .....................21
Selective Filial Cannibalism in the Sand Goby ........... .......................................21
Selective Filial Cannibalism in the Flagfish................. ................... ............... ... 22
A Model of the Evolution of Parental Care and Filial Cannibalism. .............................23
C o n c lu sio n s ........................................................................... 2 4

2 REPRODUCTIVE FITNESS CONSEQUENCES OF FILIAL CANNIBALISM IN
THE FLAGFISH, JORDANELLA FLORIDAE....................... ...............25

Introduction.........................................25
M e th o d s ...........................................................................2 8
S tu dy S p ecie s.................................................................2 8
Experimental Design and Data Collection .......................................... 29
S tatistic a l A n aly se s .................................................................................................... 3 1
Results ......... ...... ......... ....... .................. .32
Occurrence of Filial Cannibalism.................................. .......... 32
Effect of Diet on Filial Cannibalism .............. .. .... .....................32
Costs and Benefits of Filial Cannibalism for Reproduction ............. ......... .......32
Effects of Food and Access to Eggs on Components of Fitness ............................... 33
R ep ro du action ...............................................................3 3
Male weight and length ..... ........... ........ ................33
D iscu ssion .......... .. ........................................................34









3 PARENTS BENEFIT FROM EATING OFFSPRING: DENSITY-DEPENDENT EGG
SURVIVORSHIP COMPENSATES FOR FILIAL CANNIBALISM ..................................42

In tro d u c tio n ............................................................................................................................. 4 2
M e th o d s ....................................................................................4 6
Study Species and Experim ental Site .................................... .......................... .......... 46
Experim ental D esign .................. ..... ........... ...................................... .. ....... ....... ........ 46
Experiment 1: Effect of oxygen and egg density on filial cannibalism ...................46
Experiment 2: Effect of simulated filial cannibalism on egg survivorship..............49
D ata A n aly sis.............................................. ..... ............ ......... ............. 50
Experiment 1: Effect of oxygen and egg density on filial cannibalism ...................50
Experiment 2: Effect of simulated filial cannibalism on egg survivorship .............52
R results ............................ ...... ........ .... .................... ..... ........... ............ 52
Experiment 1: Effect of Oxygen, Egg Density, and Male Condition on Filial
C annibalism ............................................ ....................... ........... 52
Occurrence of whole clutch cannibalism ...................................... ............... 52
E gg su rviv worship ...............................................................53
Male condition........................................ ....... ...... ........ 54
Experiment 2: Effect of Simulated Filial Cannibalism on Egg Survivorship ..............55
Effect of oxygen and egg removal on remaining egg survivorship .......................55
Effect of oxygen and egg removal on total number of eggs surviving ..................55
D discussion .................................... ..................................... ................. 56

4 SELECTIVE FILIAL CANNIBALISM IN THE SAND GOBY ............... ...............67

In tro d u c tio n ............................................................................................................................. 6 7
M materials and M methods ........................ .. ........................ .. .... ........ ........ 68
E xperim mental D design ........................ .................... .. .. .... ........ .... .. ... 68
Single-fem ale set-up .................. .............................. .. .. ........ .............. .. 70
Im ag e A n a ly sis .................................................................. ...............................7 0
P reference C alculation ......... ................................................................ ......... ....... 7 1
Statistical A naly ses............ .............................................................................. ...... 72
R e su lts ......................................7...................3..........
Differences in Egg Size, Egg Density, and Cannibalism Rates between Years..............73
Egg Size, Survivorship, and Development Time in Eggs Reared in the Absence of
M ale s .................................................................7 4
Cannibalistic Preferences by M ales........................................... .......................... 74
D iscu ssion .......... ..........................................................74

5 SELECTIVE FILIAL CANNIBALISM IN THE FLAGFISH .............................................80

In tro d u c tio n ............................................................................................................................. 8 0
M e th o d s ...........................................................................8 2
S tu d y S p e c ie s ............................................................................................................. 8 2
E xperim mental D design ........................ .................... .. .. .... ........ .... .. ... 82
E energy A says ..............8............................84
S ta tistic s ................................ .......................................................8 4


6









R e su lts ......................... .. .... .... ...... .......... ......... ......... ....... ......... ............... 8 5
Parental Condition and Size, Egg Energetic Content, and Egg Number......................... 86
W hole C lutch C annibalism ...................................................................... ..................87
P artial C lutch C annibalism ...................................................................... .................. 87
D discussion .................................... ..................................... ................. 88

6 WHEN TO CARE FOR, ABANDON, OR EAT YOUR OFFSPRING: A MODEL OF
THE EVOLUTION OF PARENTAL CARE AND FILIAL CANNIBALISM .....................96

Introduction .......... ............................... ................................................96
M eth o d s ............................................................................................. 9 9
M odel D ynam ics .................. ............................................ ................ .......... ...... 100
Resident and M utant Trade-Offs ........................................................ ............. 101
Invasion D ynam ics and Fitness ......................................................... ............... 103
Biologically R relevant Com parisons.................................... ......................... .. ......... 105
R results ................... ...............................................................................108
Invasion of Parental Care ........................... ..... ...................................................108
Effects of egg maturation rate, egg death rate, adult reproductive rate, and
carrying capacity.................................... ...... ... .... .............. .. 108
Effect of cannibalism on the evolution of care .............................. ............... 109
Invasion of Filial Cannibalism (With and Without Parental Care) ............................109
Effects of Egg Maturation Rate, Reproductive Rate, and Selective Cannibalism .109
Effects of Density-Dependent Egg Survivorship ......................................................... 110
Effects of Energetic Benefits of Consuming Offspring ............................. ................111
Effects of Carrying Capacity ....................................... .................................... 1
D iscu ssion ......... ................................ ................................................112

7 GENERAL CONCLUSIONS AND SYNTHESIS ................................... ............... 127

In tro d u c tio n ....................................... ............................. ................. ................. 12 7
Are the Current Energy-Based and Oxygen-Mediated Hypotheses Sufficient? ................129
An Alternative Hypothesis: Selective Filial Cannibalism............................... ...............131
The Plausibility of M multiple Hypotheses ................................................... .. ... .......... 134
Future D directions ............................................................... ................. 135
Determining the Relative Importance of Varying Factors ................. ... ..................136
Role of Environmental Variation .................................. .......................... 136
The N on-Cannibalistic Parent ...................................... ..........................................137
Identification of Additional Species Practicing Filial Cannibalism............. ..............137
A Comparative Framework of Filial Cannibalism ................................ .................. 137
Why Don't All Parents Exhibit Filial Cannibalism? ...............................................138

APPENDIX

ISOLATION AND CHARACTERIZATION OF MICROSATELLITE DNA
MARKERS FOR THE FLAGFISH 1...................... ............. 39









L IST O F R E F E R E N C E S ............................................................................. ..........................143

B IO G R A PH IC A L SK E T C H ......................................................................... .. ...................... 151





















































8









LIST OF TABLES


Table page

6-1 Trade-off functions associated with parental care and filial cannibalism......................1..19

6-2 Alternative hypotheses regarding the evolutionary significance of filial cannibalism
(FC)................... ......................................... ......... 120

A-i Characteristics of flagfish microsatellite loci............................................ .................. 142









LIST OF FIGURES


Figure pe

2-1 Expected and observed benefit of filial cannibalism in eggs received by males...................40

2-2 Effect of filial cannibalism on components of fitness.. .........................................................41

3-1 Effect of oxygen and egg density on the prevalence of whole clutch cannibalism by
p rental m ales.. .................................................................................62

3-2 Effect of oxygen and egg density on the mean (+/- SE) egg survivorship (i.e., proportion
of the clutch that survived until hatching) when males were present with eggs,
including cases of both whole and partial clutch cannibalism.............................63

3-3 Relationship between A) male condition (i.e., K=100*g/cm3) and the proportion of the
clutch consumed by parental males, and B) partial clutch cannibalism and change in
m ale condition................................... .................................. ........... 64

3-4 Effects of simulated filial cannibalism ............................................................................65

3-5 Simulated filial cannibalism: The effect of oxygen and egg removal on the total number
of eggs surviving, including cases of both whole and partial clutch death......................66

4-1 Relationship between initial egg size and development time (i.e. the number of days
from spawning until hatching) in eggs reared in the absence of males. ............................77

4-2 Preferences in egg consumption by parental males.................................... ......................78

4-3 Distribution of egg size initially and in the male diet. proportion of eggs in A) 2004 and
B ) 2006 .............. ..................... ..................................... ......... ...... 79

5-1 Relationship between the mean energy per egg (J egg-1) within a clutch and A) female
w eight and B ) m ale w eight. ...................................................................... ...................91

5-2 Relationship between the frequency of whole clutch cannibalism and A) the mean
energy per egg (J egg-1) within a clutch and B) female weight..................................92

5-3 Relationship between male weight and A) the proportion and B) the number of eggs
consumed for cases of partial clutch cannibalism.............. ...............................................93

5-4 Relationship between female weight and A) the proportion and B) the number of eggs
consumed for cases of partial clutch cannibalism.............. .............................................94

5-5 Relationship between the mean energy per egg (J egg-') within a clutch and the number
of eggs consumed for cases of partial clutch cannibalism............. ................................ 95

6-1 D iagram of the m odel.. ........................ ........................... .... ................. 121









6-2 Invasion dynam ics of parental care. ............................................ ............................. 122

6-3 Invasion dynamics of filial cannibalism...... ............................................ ...............123

6-4 Effect of density-dependent egg survivorship on the evolution of parental care and filial
can n ib alism ............................................................................................................... 12 4

6-5 Effect of energetic benefits on the evolution of filial cannibalism............... ...............125

6-6 Effect of carrying capacity on the evolution of filial cannibalism. ................................. 126









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

EVOLUTIONARY SIGNIFICANCE OF FILIAL CANNIBALISM IN FISHES WITH
PARENTAL CARE

By

Hope Klug

December 2007

Chair: H. J. Brockmann
Major: Zoology

Parental care typically increases offspring survival, thereby increasing parental fitness.

Thus, it is surprising that filial cannibalism, the consumption of one's own offspring, is prevalent

in fishes exhibiting parental care. The most widely-accepted hypothesis of filial cannibalism

suggests that males gain energy from eggs that they invest into future reproduction (energy-

based hypothesis). Recently, an alternative hypothesis suggested that partial-clutch cannibalism

increases oxygen availability to remaining eggs, which in turn increases overall egg survival

(oxygen-mediated hypothesis). Evidence for both hypotheses is mixed and there are few

alternative hypotheses. Thus, the evolutionary significance of filial cannibalism remains unclear.

To enhance our understanding of filial cannibalism, I re-examined current theory (i.e., the

energy-based and oxygen-mediated hypotheses), developed and evaluated an alternative

hypothesis, and developed a mathematical model of filial cannibalism.

I experimentally quantified the effect of filial cannibalism on mating success of parental

males in the flagfish (Jordanellafloridae), and found that filial cannibalism always reduced

lifetime reproductive success. In the sand goby (Pomatoschistus minutus), males that were in

poorer condition consumed less of their eggs than males that were in better condition. These

findings are contrary to predictions of the energy-based hypothesis.









In the sand goby, I found that egg survival is density-dependent and filial cannibalism

increases when egg density is high. However, this density-dependence is not mediated by

oxygen. Therefore, I did not find support for the oxygen-mediated hypothesis. I suggest a more

general hypothesis of filial cannibalism mediated by density-dependent egg survival.

I hypothesize that the ability to preferentially cannibalize offspring of reduced quality

might play a large role in the evolution of filial cannibalism (selective filial cannibalism

hypothesis). To begin to understand the importance of selective cannibalism, I evaluated whether

males cannibalize selectively in the sand goby and the flagfish. Male sand gobies cannibalized

selectively with regard to egg development rate, and male flagfish cannibalized selectively with

regard to egg energy and maternal size. Thus, selective filial cannibalism occurs in at least two

species and this hypothesis warrants further attention.

I then developed a mathematical model of filial cannibalism to isolate factors affecting

the evolution of filial cannibalism. The findings of this model highlight the plausibility of a range

of alternative hypotheses. Specifically, the evolution of filial cannibalism is enhanced if (1)

parents can selectively cannibalize lower quality offspring, (2) filial cannibalism increases egg

maturation rate, (3) there are energetic benefits of eggs to cannibalizing males, (4) cannibalism

increases a parent's reproductive rate (e.g., through mate attractiveness). Density-dependent egg

survivorship alone did not favor the evolution of cannibalism. Additionally, the results of the

model suggest that population-level dynamics potentially play a large role in the evolution.

Finally, I isolated and characterized six polymorphic microsatellite loci in the flagfish.

These microsatellite markers will be useful in paternity assays and estimating heritability of

flagfish behaviors, such as filial cannibalism.









CHAPTER 1
INTRODUCTION

Background: Theory and Empirical Evidence

Adaptive theories of evolution suggest that parents should exhibit strategies that increase

offspring survival. Parental care, which is common across taxa, is one way in which parents are

thought to do this (reviewed in Clutton-Brock 1991). Because parental care is assumed to

increase offspring survival, it is surprising that filial cannibalism, the consumption of one's own

offspring, may co-occur with parental care. Indeed, filial cannibalism is prevalent in a range of

taxa (Polis 1981) and is particularly common in teleost fishes exhibiting paternal care (reviewed

in Manica 2002). Because parental males often consume more eggs than would die naturally,

filial cannibalism does not solely serve to clean the nest of dead or diseased eggs (e.g. Klug et al.

2005). Likewise, egg cannibalism cannot be attributed to the removal of eggs that were fertilized

by other males (DeWoody et al. 2001; but see Neff and Sherman 2003), and filial cannibalism

remains common in the absence of possible cuckolding events (Kvamemo et al. 1998; Kume et

al. 2000; Lissiker et al. 2002; Klug et al. 2005; Svensson ad Kvamemo 2007). Despite some

theoretical development, the evolutionary significance of filial cannibalism remains unclear.

Early ethologists considered filial cannibalism a social pathology of little or no adaptive

significance. However, more recent studies suggest that filial cannibalism reflects an adaptive

trade-off between current and future reproductive success, in which males gain energy and

nutrients from consumed eggs that are then reinvested into future reproduction (Rohwer 1978;

Sargent 1992). According to this energy-based hypothesis, whole clutch cannibalism (i.e. the

consumption of all eggs present) is expected to be more frequent within a species when clutch

size is relatively small because the energy requirement of the males can be satisfied only by

clutches larger than a certain size (i.e., when clutch size is small, males must consume all eggs to









satisfy their energy requirements). Empirical studies support this prediction (reviewed by Manica

2002). Theory also predicts that cannibalism should increase as food availability decreases

and/or when male condition is poor (Rohwer 1978; Sargent 1992). However, evidence with

regard to these predictions is equivocal. Supplementary feeding of parental males reduces partial

clutch cannibalism (i.e., the consumption of only some eggs in a nest) in the common goby

(Pomatoschitus microps, Kvarnemo et al. 1998), the scissortail sergeant (Abudefdufsexfasciatus,

Manica 2004), and the Cortez damselfish (Stegastes rectifraenum, Hoelzer 1992), but food

availability and/or male condition did not affect filial cannibalism in fantail darters (Etheostoma

flabellare, Lindstrom and Sargent 1997) and the three-spined stickleback (Gasterosteus

aculeatus, Belles-Isles and Fitzgerald 1991). Still other studies have quantified the energetic

value of eggs to determine if eggs can provide a caring male with sufficient energy to offset the

direct fitness costs of care. In one species, energy attained from partial brood cannibalism was

found to be sufficient to offset costs related to care (Apogon lineatus, Kume et al. 2000), while in

another, energy from eggs was found to be insufficient (Gasterosteus aculeatus, Smith 1992).

Thus, there is no consistent support for the energy-based explanation of filial cannibalism.

Payne et al. (2002) recently suggested that by consuming some eggs in their nest, parental

males can improve oxygen availability of the remaining eggs, thereby improving overall egg

survivorship (oxygen-mediated hypothesis; Payne et al. 2002, 2004). However, there has been

only limited empirical examination of this hypothesis, which does not account for the common

occurrence of filial cannibalism in high oxygen environments. For example, filial cannibalism

occurs when oxygen availability is relatively high in the sand goby (Pomatoschistus minutus,

Klug et al. 2006), the flagfish (Jordanellafloridae, Klug et al. 2005), and fantail darters

(17lhe',,it ,ulflabellare, Lindstrom and Sargent 1997). Additionally, Lissaker et al. (2002) found









no differences in partial or whole clutch cannibalism across oxygen levels. Thus, neither the

energy-based hypothesis nor the oxygen-mediated hypothesis can explain the occurrence of filial

cannibalism. Furthermore, there has been a general lack of alternative explanations for the

widespread occurrence of filial cannibalism in fishes.

As an alternative to the energy-based and oxygen-mediated hypotheses, I propose the

hypothesis of selective filial cannibalism (i.e., non-random consumption of offspring with regard

to some aspect of offspring phenotype). Specifically, I hypothesize that the ability to cannibalize

offspring selectively in relation to aspects of offspring quality can directly favor the evolution of

filial cannibalism (see also Klug and Bonsall 2007 and Chapter 6). The idea of weeding out

inferior offspring has been documented in other contexts (brood reduction: Forbes & Mock

1998; selective abortion in humans: Steams, 1987; Forbes 1997; Diamond 1987; Hesketh &

Xing 2006; selective abortion in plants: Burd 1998; Karkkainen et al. 1999; Melser &

Klinkhamer 2001), but it has not yet been explicitly evaluated in relation to filial cannibalism

(but see Mrowka 1987 and Kraak 1996 for work on consumption of unfertilized or diseased

eggs). Indeed, selective elimination of lower quality offspring is thought to play a large role in

the evolution of selective embryo abortion, brood reduction, and offspring abandonment (Stearns

1987; Forbes and Mock 1998; Stearns 1992). However, the relationship between offspring

phenotype and filial cannibalism remains unknown (but see Chapters 4, 5, and 6), and thus, the

relative importance of selective filial cannibalism is unclear.

Despite mixed empirical evidence and a lack of alternative hypotheses, the energetic

hypothesis of filial cannibalism is typically accepted as a valid explanation for the prevalence of

filial cannibalism in fishes (e.g. Vinyoles et al. 1999; Kume at al. 2000; Manica 2002). Overall,

the support for this hypothesis is unsatisfying and it remains unclear in many systems whether or









not (and potentially how) filial cannibalism is an adaptive strategy. Despite nearly three decades

of empirical and theoretical work on the subject, filial cannibalism in fishes still remains an

evolutionary conundrum. Understanding the evolutionary significance of filial cannibalism

requires both additional theoretical development and empirical research.

An enhanced understanding of the evolution of filial cannibalism necessitates three

research strategies.

Re-Evaluation of Current Theory with an Explicit Focus on Fitness Consequences

Previous studies have failed to quantify fitness consequences of filial cannibalism (i.e., the

parent and offspring survivorship and/or reproductive benefits of egg eating). Most studies of

filial cannibalism have focused on the relationship between male condition and filial

cannibalism, or the effect of supplemental feeding on filial cannibalism. Simply showing that

males with less food eat more eggs does not demonstrate that an adaptive trade-off is occurring.

Thus, there is limited support for the energy-based hypothesis, and as mentioned previously, the

oxygen-mediated hypothesis has received little empirical evaluation. The first step in

understanding the evolutionary significance of filial cannibalism is to explicitly evaluate the

energy-based and oxygen-mediated hypotheses by focusing on potential fitness consequences of

filial cannibalism (i.e., the parent and offspring survivorship and reproductive costs and

benefits).

Development and Evaluation of Alternative Hypotheses

The energy-based hypothesis continues to be accepted despite mixed evidence. Thus the

development and empirical examination of new hypotheses is critical. As mentioned above, I

propose the alternative hypothesis of selective filial cannibalism (discussed further in Chapters 4,

5, and 6). Specifically, I hypothesize that 1) parental males might preferentially cannibalize

lower quality offspring and 2) the ability to cannibalize selectively potentially plays a large role













Development of a Synthetic Model of Filial Cannibalism

The two current hypotheses (i.e., the energy-based and oxygen-mediated) are currently

presented as mutually exclusive ideas, leading to a debate that has resulted in very little

resolution. Instead, I argue that the prevalence of filial cannibalism cannot be explained by single

factors (i.e., energy or oxygen). A model incorporating a variety of trade-offs associated with

current and future reproductive success (i.e., a multivariate approach) is essential to identify and

evaluate the plausibility of a range of alternative hypotheses.

Summary of Dissertation Objectives

The goals of my dissertation are three-fold, and consist of four experiments and one

mathematical model.

First, I re-evaluate current theory by focusing on fitness consequences of cannibalism. In

my first experiment, I evaluate the reproductive fitness consequences of filial cannibalism. To

evaluate the energy-based hypothesis, I quantify the effect of filial cannibalism on long-term

mating success of parental males (i.e., current and future reproductive success) and determine

whether energy consumed from eggs is directly translated into future reproduction in the flagfish.

In my second experiment, I examine the relationship between partial clutch cannibalism and egg

survivorship in relation to oxygen availability. I evaluate the oxygen-mediated hypothesis by

quantifying the effect of oxygen and egg density on filial cannibalism and net egg survival.

Second, I develop and evaluate an alternative hypothesis. In my third and fourth

experiments, I assess the relative importance of aspects of egg phenotype on filial cannibalism in

the sand goby and flagfish. To begin to evaluate the hypothesis of selective filial cannibalism

proposed above, I assess whether parental males preferentially consume eggs with regard to









some aspect of offspring phenotype. Specifically, I focus on the relationship between filial

cannibalism and egg size or energetic content, because egg size and energetic content have been

correlated with post-hatching survival and growth in fishes (Kamler 2005).

Finally, I fully develop a mathematical model incorporating a range of costs and benefits

associated with filial cannibalism (i.e., the multivariate approach). I use this model to isolate the

pivotal factors affecting the evolutionary dynamics of filial cannibalism, and I evaluate the

plausibility of a range of non-mutually exclusive alternative hypotheses.

Study Systems

Filial cannibalism likely functions differently among systems, and because there are

potentially multiple explanations for the occurrence of filial cannibalism, I focused on two

distinct systems: the flagfish (Jordanellafloridae) and the sand goby (Pomatoschistus minutus).

Both systems are exemplary of care in fishes: (i) males provide care (i.e., nest guarding, fanning,

and cleaning), (ii) filial cannibalism is prevalent, and (iii) they live approximately one year,

allowing fitness measurements quantified over one breeding season to be representative of

lifetime fitness. However, the species have different life histories (egg laying patterns, nest

structure, rates of mating), thus making them suitable for testing particular hypotheses related to

filial cannibalism.

Flagfish males defend open nests consisting of filamentous algae or bare substratum. Male

flagfish fan eggs but fanning is not necessary for egg survival, and flagfish eggs are not clumped

together (Mertz & Barlow 1966). Thus, limited oxygen at the nest is not a main source of egg

mortality, and there is no a priori expectation in flagfish that filial cannibalism leads to increased

survivorship of remaining eggs through increased oxygen availability. Unlike flagfish, sand goby

males build nests under shells and cover the nest with sand, leaving only a small (-lcm)

opening. Sand goby eggs are clumped and oxygenation of the eggs by male fanning is required









for egg survival. Thus, the sand goby is an ideal system to test the hypothesis of oxygen-

mediated filial cannibalism. Males of both species sequentially receive eggs from multiple

females, increasing the possibility that eggs vary in quality.

Overview of Dissertation Chapters

Reproductive Fitness Consequences of Filial Cannibalism in the Flagfish

In Chapter 2, I investigate the effects of filial cannibalism on components of fitness in the

flagfish. Specifically, I expect filial cannibalism to increase future reproductive success. To

quantify the reproductive fitness consequences of filial cannibalism, I experimentally manipulate

(1) the diet of parental males (i.e., males either receive a high quality or low quality diet) and (2)

the ability to practice filial cannibalism (i.e., males either have full access to eggs or access to

eggs is prevented by a nest cover). Each male experiences a single experimental treatment (i.e.,

high food, access to eggs; high food, no access to eggs; low food, access to eggs; low food, no

access to eggs) for the duration of the experiment. Specifically, I follow males over 90 days,

approximately one breeding season (i.e., the expected reproductive lifetime of flagfish), and

consider three components of fitness: 1) the total number of eggs received, 2) the number of

spawnings, and 3) the frequency of spawning. Contrary to predictions, filial cannibalism reduces

male reproductive success. While an enhanced diet increases the number of eggs received, the

number of spawnings, the frequency of spawning, and male weight gain, there is no effect of

filial cannibalism on any component of reproductive success or male weight. Thus, in the

flagfish there is no evidence that energy or nutrients gained from filial cannibalism are directly

translated into increased future reproductive success. See also Klug and St. Mary (2005).









Parents Benefit from Eating Offspring: Density-Dependent Egg Survivorship Compensates
for Filial Cannibalism

In Chapter 3, I evaluate the hypothesis of oxygen-mediated filial cannibalism in the sand

goby by 1) examining the effect of oxygen and egg density on the occurrence of filial

cannibalism, 2) evaluating the effects of partial clutch cannibalism on the survivorship of the

remaining eggs, and (3) comparing potential costs and benefits of filial cannibalism related to the

net number of eggs surviving. I find that oxygen level and egg density affect the occurrence of

cannibalism and that simulated partial clutch cannibalism improves survivorship of the

remaining eggs. Additionally, because increased egg survivorship, stemming from partial egg

removal, compensates for the cost of cannibalism (i.e., number of eggs removed) at a range of

cannibalism levels, filial cannibalism potentially results in no net losses in reproductive success.

However, oxygen does not affect egg survivorship. Thus, in this chapter, I suggest a more

general hypothesis of filial cannibalism mediated by density-dependent egg survivorship. See

also Klug et al. 2006.

Selective Filial Cannibalism in the Sand Goby

In Chapter 4, I evaluate the novel hypothesis of selective filial cannibalism. Specifically, I

examine the relationship between aspects of egg phenotype and partial clutch filial cannibalism

in the sand goby. Males are either allowed to spawn with one or two females, and I then evaluate

cannibalistic preferences in relation to the order in which eggs were spawned and egg size. I

focus on egg size because egg size has been correlated with post-hatching survival and growth in

a range of fishes (Kamler 2005). In this experiment, I find that males selectively cannibalize eggs

with respect to egg phenotype, but only in some cases. When males mated with two females

sequentially, they preferentially consume the larger eggs of the second female. Because egg size

is correlated with development time (and because female 2's eggs were slightly younger), the









patterns of cannibalism appear to represent a strategy that reduces the duration of care necessary

for the current brood. By reducing the duration of care required for each brood, it is possible that

males can attain additional brood cycles during their breeding season, thereby increasing their

reproductive success. This work highlights the potential role of selectivity in understanding the

adaptive significance of filial cannibalism.

Selective Filial Cannibalism in the Flagfish

In Chapter 5, I examine the relationship between maternal size and condition, egg

energetic content, and filial cannibalism. Both egg energetic content and maternal size have been

correlated with post-hatching survival in a range of fishes (Kamler 2005). In this experiment, I

find that males exhibit preferential cannibalism with regard to mean energetic content per egg

and maternal size. However, the patterns of selective cannibalism differ between whole and

partial clutch cannibalism. Specifically, males cannibalize the whole clutch more often when the

mean per-egg energetic content is relatively high. Because I find no relationship between mean

egg energy and egg number, the increase in whole clutch cannibalism when mean egg energetic

content is relatively great is not explained by any differences in egg number. In contrast, there is

a negative relationship between mean egg energy content and the number of eggs consumed for

the case of partial clutch cannibalism. Similarly, maternal size is negatively correlated with the

proportion of and the number of eggs consumed. Because female size has been positively related

to post-hatching survival in other fishes, it seems that for the case of partial clutch cannibalism,

males in this experiment preferentially cannibalize offspring that have lower survival post-

hatching. In contrast, when they practice whole clutch cannibalism, they appear to be

maximizing their energetic gain. The findings in this chapter (in combination with those of

Chapter 4) suggest that males practice selective filial cannibalism with regard to aspects of

offspring phenotype.









A Model of the Evolution of Parental Care and Filial Cannibalism

In Chapter 6, I evaluate the importance of a range of factors on the evolution of parental

care and filial cannibalism using an evolutionary ecology approach (see also Klug and Bonsall

2007). Parental care, no care/total abandonment, and filial cannibalism evolve and often co-exist

over a range of life-history strategies (i.e., strategies associated with a wide range of adult death

rates, egg death rates, egg maturation rates, population carrying capacities, juvenile survival

rates). While no single benefit is essential for the evolution of filial cannibalism, benefits

associated with adult or offspring survival and/or reproduction facilitate the evolution of

cannibalism. This model highlights the plausibility of a range of alternative hypotheses.

Specifically, the evolution of filial cannibalism is enhanced if 1) parents can selectively

cannibalize lower quality offspring, 2) filial cannibalism increases egg maturation rate, 3) there

are energetic benefits of eggs to cannibalizing males, 4) cannibalism increases a parent's

reproductive rate (e.g., through mate attractiveness). Density-dependent egg survivorship alone

does not favor the evolution of cannibalism. When egg survivorship is density-dependent,

parents are expected to simply lay eggs at densities that maximize offspring survival. While

density-dependent egg survival does not directly favor the evolution of filial cannibalism in the

model, density-dependent egg survival does not preclude the evolution of filial cannibalism

entirely. For the case of density-dependent egg survival, filial cannibalism evolves more often

when the density-dependence is relatively more intense (i.e., when there is a relatively large

increase in offspring mortality with increasing density). Additionally, the evolution of filial

cannibalism and/or parental care is highly sensitive to population carrying capacity in the model,

and care and/or cannibalism are more likely to evolve if they allow an organism to use resources

more effectively. These results suggest that population-level resource competition potentially

plays an important role in the evolution of both parental care and filial cannibalism.









Conclusions

In this final chapter, I summarize the major conclusions of my dissertation. I discuss my

findings in relation to previous theory, and highlight future directions of research on filial

cannibalism.

Isolation and Characterization of Microsatellite Markers in the Flagfish

In the Appendix (Klug, St. Mary and Clark, in preparation), I describe six polymorphic

microsatellite loci in the flagfish. These microsatellite markers will be useful in future behavioral

research on the flagfish for estimating paternity, reproductive success, and heritability.









CHAPTER 2
REPRODUCTIVE FITNESS CONSEQUENCES OF FILIAL CANNIBALISM IN THE
FLAGFISH, JORDANELLA FLORIDAE

Introduction

Parental care, which is assumed to increase the fitness of a parent by increasing survival or

quality of the offspring (Clutton-Brock 1991), is common across animal taxa (reviewed in

Clutton-Brock 1991; Rosenblatt & Snowdon 1996). In fishes care is typically paternal, and

includes territory guarding and nest care (Gross and Sargent 1985). Because parental care is

assumed to increase the survival of offspring, it is surprising that filial cannibalism, the

consumption of one's own offspring, is prevalent in fishes exhibiting parental care (reviewed in

Manica 2002). Since parental males often consume more eggs than would die naturally, filial

cannibalism does not solely serve to clean the nest of dead or diseased eggs (e.g., Klug et al.

2005). While early ethologists considered filial cannibalism a social pathology with little or no

adaptive significance, filial cannibalism is currently thought to reflect an adaptive trade-off

between current and future reproductive success (Rohwer 1978).

Whole clutch cannibalism (i.e., the consumption of all eggs present) and partial clutch

cannibalism (i.e., the consumption of some eggs present) are thought to occur for different

reasons. Whole clutch cannibalism theoretically represents a termination of parental care

(Rohwer 1978; reviewed in Manica 2002). Because the net reproductive gain associated with

caring for a small clutch is expected to be less than that of a large clutch, whole clutch

cannibalism is expected to be more frequent when clutch size is relatively small. Indeed, in many

systems with paternal care, smaller clutches are subject to whole clutch cannibalism more

frequently than larger clutches (reviewed in Manica 2002).

The most widely accepted explanation for partial-clutch filial cannibalism is that it

provides a male with energy or limited nutrients allowing him to care for the remaining brood or









to increase future reproduction (Rohwer 1978; Sargent 1992). According to this energy-based

hypothesis of filial cannibalism, food availability and male condition should affect the

occurrence of filial cannibalism (Rohwer 1978; Sargent 1992; Sargent 1997). Evidence regarding

these predictions is mixed. Contrary to the theory, neither initial male condition nor food

availability predicted the number of eggs consumed in fantail darters, Etheostomaflabellare

(Lindstrom and Sargent 1997). Similarly, food availability was not related to the occurrence of

filial cannibalism in the three-spined stickleback Gasterosteus aculeatus (Belles-Isles and

Fitzgerald 1991). However, evidence in other species suggests food availability does affect filial

cannibalism. Kvarnemo et al. (1998) found that starved male common gobies consumed more of

their eggs than males whose diet was supplemented with either mussel meat or both mussel meat

and the eggs of conspecifics. Similarly, Manica (2004) found that supplementary feeding

significantly reduced partial clutch filial cannibalism in the scissortail sergeant (Abudefduf

sexfasciatus). In the Cortez damselfish (Stegastes rectifraenum), supplementation of a male's

diet with conspecific eggs reduced cannibalism but did not fully inhibit it (Hoelzer 1992). Other

studies have quantified the energetic value of eggs to determine if eggs can provide a caring male

with sufficient energy to offset the direct fitness costs of cannibalism. In one species energy

attained from partial brood cannibalism was found to be sufficient to offset costs related to care

(Apogon lineatus, Kume et al. 2000), while in another, energy from eggs was found to be

insufficient (Gasterosteus aculeatus, Smith 1992). Thus, there is a lack of consistent support

across fish species for the energy-based hypothesis of filial cannibalism. Despite the mixed

empirical support, the energy-based hypothesis of filial cannibalism is generally accepted as

valid (e.g. Manica 2002), and it remains unclear in many systems whether (and potentially how)

filial cannibalism is an adaptive strategy. While previous studies have focused on testing some of









the assumptions of energy-based models, simply showing that males with less food eat more

eggs does not demonstrate that an adaptive trade-off is occurring. A thorough examination of

current theory instead necessitates an explicit examination of the fitness consequences of filial

cannibalism (i.e., the survivorship and/or reproductive consequences of egg eating).

According to the energy-based explanation of filial cannibalism (Rohwer 1978), we would

expect cannibalism to increase lifetime reproductive success, and in a short-lived species we

might expect filial cannibalism to increase reproductive success over the course of one breeding

season. The present study quantified reproductive fitness consequences of filial cannibalism over

approximately one breeding season in a short-lived species of fish exhibiting filial cannibalism.

Specifically, I was interested in whether males that eat eggs experience an increase in

reproduction. The flagfish, Jordanellafloridae, is an ideal subject for such a study because 1)

males show care behavior, i.e., nest guarding, fanning, and cleaning (Mertz and Barlow 1966), 2)

filial cannibalism is prevalent (Mertz and Barlow 1966; Foster et al. 1969; Klug et al. 2005), and

3) they live approximately one year (H. Klug, unpublished data), allowing fitness measurements

quantified over one breeding season to be representative of lifetime fitness.

In order to evaluate the effect of filial cannibalism on reproductive output, it is necessary to

compare the reproductive success of males who are able to consume eggs and males that are

prevented from consuming eggs. To do so effectively, males should be the only egg consumers

and thus the experimental design must limit the access of potential egg predators, even females,

to the eggs. Males not allowed to cannibalize must be excluded from their eggs yet still

encouraged to provide parental care. Thus, these males must have access to their nest (with eggs

covered) and yet not be allowed to continue spawning (as they would then have the opportunity

to eat some eggs). Additionally, such a study must be designed so that males are energy-limited









and consume a substantial number of eggs in order to adequately evaluate benefits of egg

consumption. The evaluation of the effect of filial cannibalism on male survivorship requires

additional environmental controls; such a study must be done as above (to control egg

consumption) but also in the presence of predators of the male and with other natural stressors.

Because of the difficulty of evaluating the survivorship effects of filial cannibalism, I focus here

on the reproductive effects of filial cannibalism.

I explore the reproductive fitness consequences of filial cannibalism in the flagfish by

comparing two cannibalism treatments, one in which males have the opportunity to consume

their eggs and another in which filial cannibalism is prevented, over 90 days, the approximate

duration of a flagfish breeding season in north central Florida (Hale, unpublished data). Three

components of reproductive fitness were considered: 1) total number of eggs received, 2) total

number of spawnings, and 3) the frequency of spawning. Specifically, I evaluated whether

energy gained from the consumption of eggs was directly translated into an increase in the total

number of eggs received by a male. In addition, I examined the effect of filial cannibalism on

components of male condition. While I only focus on reproduction here, systematically

quantifying specific effects of filial cannibalism on components of fitness will provide key

insight into the validity of current models of filial cannibalism.

Methods

Study Species

Flagfish live approximately one year, and flagfish males care for their nest by guarding,

cleaning, and fanning. The incubation period of eggs is typically less than one week, and males

are able to feed on vegetation and invertebrates, which are common near nesting sites. Flagfish

females spawn with multiple males, and brood cycling (i.e., the alteration of courtship and

mating periods with periods of full brood care) is absent in this species. Nesting males









potentially receive eggs continuously but may also care for only a single brood. In north central

Florida, I have observed flagfish breeding approximately May through September. Flagfish have

been successfully used in laboratory studies before (e.g., Mertz and Barlow 1966; Hale et al.

2003; St. Mary et al. 2001; Klug et al. 2005) and readily adapt to being housed in aquaria.

Experimental Design and Data Collection

The experiment was conducted in Gainesville, Florida beginning in May and ending in

November 2002. Due to space restrictions, two blocks of the experiment were completed, the

first beginning in May and the second beginning in September. To replicate peak breeding

season conditions in both blocks, the experiment was conducted in an environmentally-

controlled room where temperature and lighting could be held constant. The two blocks did not

differ significantly (i.e., there were no significant block effects in any of the analyses) so the data

were pooled for all analyses. Males were used only once in the experiment and were euthanized

after the experiment. Flagfish were collected from the Otter Creek/ Waccasassa River drainage

just prior to use and both sexes were housed in separate 150 1 freshwater holding tank maintained

at approximately 280C prior to use. During this time the fish were fed ad libitum a diet consisting

of algae tablets and frozen brine shrimp.

In all cases the experiment began by placing one male and three females in a 36 1

freshwater aquarium (measuring 26 cm x 30 cm x 30 cm) equipped with air-driven, activated

carbon and Dacron floss filtration, a spawning mat, and three artificial Ludwigia plants. The mat

consisted of a 100 cm2 tile covered with heavy, green acrylic felt carpet. The fish experienced a

regular 14-hour daylight period and temperature was maintained at a constant 290C. All males

were randomly assigned to a treatment. Each male was allowed to spawn 90 days from his initial

spawning date (i.e., date of first spawning for each male). I crossed two filial cannibalism

treatments (access and no access to eggs) with two feeding regimes (low and high food). Thus,









there were four treatments: 1) high food and no filial cannibalism (HF NFC), 2) high food and

filial cannibalism (HF FC), 3) low food and no filial cannibalism (LF NFC), and 4) low food and

filial cannibalism (LF FC). Fish in a low food treatment were fed one algae tab weighing

approximately 0.29 g every second day. Fish in a high food treatment received one algae tab

weighing approximately 0.52 g and 1.2 g (wet weight) of frozen brine shrimp daily. During the

experiment, fish were fed at approximately 1500 hrs daily. As mentioned above, males either had

access to their eggs or access was denied by covering the eggs with a screen nest cover. The nest

covers used in this experiment were 182.5 cm2 and consisted of plastic netting with 1mm mesh.

Males continue to actively care for the nest despite the addition of the cover.

I checked the nests for eggs four times each day at approximately 8am, 1 am, 3pm, and

7pm. They were frequently checked in excess of four times daily, and if it was not possible to

check them at least three times on a given day, a clear acrylic partition was placed in each tank

separating males and females to prevent any undetected spawning. There were six days during

the first block of the experiment and five days during the second block during which I was

unable to check the nests at least three times per day.

When eggs were discovered (day 0), I removed the nest from the tank, counted the eggs,

and recorded the developmental stage to ensure eggs were discovered soon after spawning. I am

confident that in all cases eggs were discovered immediately after spawning. In all treatments, I

then placed a clear acrylic divider containing nine holes in the rear half of the tank immediately

after eggs were discovered, physically separating males and females while still allowing them to

remain in visual and chemical contact, thus allowing males to continuously court females.

Physically separating males and females with the partition (which was done in all treatments)

was necessary to prevent females from consuming eggs. I then returned the eggs to the male, and









if the male was in a no filial cannibalism treatment (i.e., NFC, no access to eggs), I immediately

placed the mesh nest cover over the eggs. In all treatments, nests with eggs were briefly removed

from the tank each day in order to count the eggs. On day five (all eggs either hatched or were

consumed by day 5), I removed the nest cover and/or the partition, allowing males the

opportunity to spawn again. This procedure was repeated each time a male spawned for the

duration of the 90 days.

Males remained in the same aquaria during the entire experiment. Females were used

repeatedly throughout the experiment, but were frequently and randomly moved amongst the

tanks to ensure that all females 1) experienced similar feeding regimes and 2) spent

approximately equal amounts of time with each male (to limit any female-specific effects).

Males were weighed at several intervals during the experiment.

Statistical Analyses

I used a 2-way ANOVA to examine the effect of food and filial cannibalism on correlates

of fitness and male weight and length. These analyses were performed in SYSTAT 9.0 (SPSS,

Inc.). I used logistic regression to evaluate the effect of food on proportion of eggs consumed;

this analysis was performed using SAS 8.2 (SAS institute). I performed all analyses including

and excluding data from males that died during the 90 days. Since the findings did not differ for

any of the comparisons, I only present results excluding dead males. Because filial cannibalism

treatment could not have affected the first clutch received, I evaluated the effect of filial

cannibalism on total number of eggs received both including and excluding the first clutch

received. The findings did not differ qualitatively, so results including all clutches are presented.

In order to explicitly evaluate whether cannibalistic males experienced a net gain in

reproductive success, I compared the observed number of eggs cannibalistic males received with

the number of eggs they were expected to receive to achieve, as a minimum, no net loss in









offspring produced. Specifically, I expected males that consumed eggs to receive, on average,

the same number of eggs as non-cannibalistic males plus at least the number of eggs that were

consumed. Thus, the expected egg benefit necessary for males to receive no reduction in

reproductive output is defined as the mean number of eggs consumed by FC males, and the

observed egg benefit is defined as the difference between the mean number of eggs received by

FC males and the mean number of eggs received by NFC males. I used two-sample, one-tailed t-

tests to compare the expected and observed egg benefit for high food and for low food males.

Results

Occurrence of Filial Cannibalism

Filial cannibalism was prevalent when males had access to eggs. Indeed, when food was

high males consumed approximately 95% +/- 3.9% (X +/- SE) of their eggs and when food was

low males consumed approximately 84% +/- 3.6% (X +/- SE) of their eggs. Such high rates of

cannibalism gave me confidence that in all cases where access to eggs was allowed, I would

expect to see benefits of filial cannibalism if they exist.

Effect of Diet on Filial Cannibalism

Diet significantly affected the proportion of eggs consumed (logistic regression food


effect: X1 = 7.11, P = 0.008). Specifically, males in the low food treatment consumed a smaller

proportion of eggs than males in the high food treatment.

Costs and Benefits of Filial Cannibalism for Reproduction

I compared the difference between the mean number of eggs received for FC males and

NFC males (i.e. the observed egg benefit) and the mean number of eggs consumed by FC males

(i.e., the minimum expected egg benefit) (Figure 2-1). For high food males, the egg benefit of

filial cannibalism necessary for filial cannibalism to result in no reduction in reproductive output









was 205 eggs; surprisingly, high food FC males received approximately 40 fewer eggs than high

food NFC males (Figure2-1). Consequently, high food males received significantly fewer eggs

than what was necessary to achieve no reduction in net eggs received (2-sample t test: tio =

7.889, P < 0.0001). Similarly, the egg benefit necessary for low food males to break even was

107 eggs, but they received only 38 more eggs than low food NFC males. Again, low food males

received significantly fewer eggs than was necessary for no reduction in fitness (2-sample t test:

tio = 2.504, P = 0.016).

Effects of Food and Access to Eggs on Components of Fitness

Reproduction: There was a significant effect of food on all components of fitness I

measured. In comparison to males receiving the low food diet, high food males received

significantly more eggs (2-way ANOVA, food: F1,23 = 10.047, P = 0.005), spawned more times

over the 90 day period (F1,23= 11.861, P = 0.002), and had a greater frequency of spawning (F1,23

= 11.250, P = 0.003). In contrast, there was no effect of filial cannibalism on any of these

variables (2-way ANOVA, FC: F1,23 = 0.001, P = 0.975; F1,23 = 0.533, P = 0.474; F1,23 = 1.103, P

= 0.306, respectively) and no interaction (F1,23 = 0.911, P = 0.351; F1,23 = 0.035, P = 0.853; F1,23

= 0.559, P = 0.463, respectively) (Figure 2-2).

Male weight and length: There was no significant difference in initial male weight

between the treatment groups (2-way ANOVA: food, F1,23 = 0.200, P = 0.659; FC, F1,23 = 0.743,

P = 0.398; food x FC, F1,23 = 0.416, P = 0.526). During the experiment, high food cannibalistic

males gained 1.73 +/- 0.73 g (X +/- SD), high food non-cannibalistic males gained 1.75 +/- 0.47

g (X +/- SD), low food cannibalistic males gained 0.79 +/- 0.60 g (X +/- SD), and low food non-

cannibalistic males gained 1.15 +/- 0.77 g (X +/- SD). Although males in the high food treatment

gained significantly more weight than low food males (F1,23 = 8.538, P = 0.008), there was no









effect of filial cannibalism (F1,23 = 0.532, P = 0.474) and no interaction between diet and

cannibalism (F1,23 = 0.429, P = 0.520) on weight gained.

Additionally, there was no significant difference in initial male standard length between

the treatment groups (2-way ANOVA: food, F1,25 = 0.075, P = 0.786; FC, F1,25 = 0.402, P =

0.533, food x FC, F1,25 = 0.288, P = 0.597), and there was no significant effect of food or filial

cannibalism on change in male standard length (food, F1,23 = 1.881, P = 0.185; FC, F1,23 = 0.105,

P = 0.749; food x FC, F1,23 = 0.105, P = 0.749).

Discussion

Parental investment, and parental care in particular, is expected to increase offspring

survival (Trivers 1972; Sargent 1988; Clutton-Brock 1991; Balshine et al. 2002). Indeed, care

has been shown to increase parental and offspring fitness in various animal taxa (Forester 1979;

Dominey 1981; Simon 1983; Fairbanks and McGuire 1986; discussed in Balshine et al. 2002).

Thus, the ambiguity regarding the adaptive significance of filial cannibalism coupled with the

prevalence of filial cannibalism in fishes exhibiting parental care is surprising. Indeed, previous

research in the flagfish suggests that the offspring survivorship benefits of parental care may be

offset by the occurrence of filial cannibalism (Klug et al. 2005). If filial cannibalism is an

adaptive trade-off between current and future reproductive success (Rohwer 1978), filial

cannibalism should lead to an increase in net reproductive success. One way that an increase in

net reproductive output could occur is through an increase in number of eggs fertilized (e.g., if

males reinvest energy from cannibalism into increased courtship behavior).

Contrary to the expectation that filial cannibalism will increase male mating success, I

found that the number of additional eggs received by males who cannibalized never compensated

for the loss resulting from cannibalism. Among males on the high food diet, those that were

allowed to cannibalize eggs actually received fewer eggs from females than those that were not









allowed to cannibalize eggs. For males on the low food diet, those that were allowed to

cannibalize eggs received more eggs from females than those that were not allowed to

cannibalize eggs, but the number of extra eggs received never compensated for the number lost

to cannibalism. Since the flagfish life span is approximately one year in the wild, such a

reduction in the number of eggs over the course of one breeding season likely has a large impact

on lifetime reproductive success. In order to explain these findings in the context of the current

energy-based explanation of filial cannibalism (Rohwer 1978; Sargent 1992), egg cannibalism

would need to have a substantial positive effect on male survivorship, the survivorship of the

eggs that remain in the nest, or on the survivorship of any future eggs that a male receives.

In contrast to cannibalism, food availability greatly affected correlates of fitness. Indeed,

an enhanced diet was related to an increase in eggs received, spawning frequency, number of

clutches received, and increased weight gain. Thus, I have evidence that males in my study were

energy-limited and that food and/or nutritional level affects reproductive success. The finding

that food availability affects reproductive success coupled with the lack of benefits of filial

cannibalism suggests that eggs do not have substantial energetic or nutrient content relative to

the costs of reproduction. This finding is also inconsistent with the energy-based hypothesis of

filial cannibalism (Rohwer 1978; Sargent 1992). Furthermore, I found that males in the low food

treatment consumed significantly fewer eggs than males in the high food treatment, which

further contradicts the predictions of energy-based hypothesis (Rohwer 1978; Sargent 1992).

However, reduced consumption of eggs by males in the low food treatment is consistent with life

history theory if low food males have a reduced expectation of future reproduction and thus

invest more in current reproduction. A similar trend (i.e., increased investment in current clutch

as expected future reproduction declines) has been associated with seasonal patterns of









cannibalism in the cardinal fish (Apogon doederleini). Takeyama et al. (2002) found that 1-year

old males cannibalised less at the end of the breeding season.

There were several limitations stemming from the controlled nature of the present study. In

order to experimentally manipulate cannibalism while still allowing males to care for eggs I used

a nest cover. After eggs were discovered, males and females were separated for 4 days in all

treatments, thus preventing continuous spawning. Such separation was necessary to prevent NFC

males and females from consuming eggs. Not allowing males to receive eggs continuously

possibly led to clutch sizes that were small in comparison to those found in nature, and relatively

small clutches, which have been associated with increased whole clutch cannibalism in other

species (reviewed in Manica 2002), likely contributed to the high rates of cannibalism observed.

Indeed, I do not believe that such rates of cannibalism are necessarily representative of natural

filial cannibalism rates. Due to the nature of the spawning substrate in the field, it's currently

impossible to accurately measure clutch size in the wild, and thus, I do not have reliable

estimates of natural rates of filial cannibalism. Regardless, such high rates of cannibalism should

result in even greater fitness effects of cannibalism and thus gave me confidence that I would be

able to detect any benefits of filial cannibalism. Yet, I found no net benefits related to

reproduction and I have no reason to believe that there would be any benefit at lower levels of

filial cannibalism. Additionally, I only evaluated one major component of reproductive success

and therefore, cannot evaluate alternative ways in which filial cannibalism could affect lifetime

fitness. For example, separating males and females after spawning prevented me from assessing

the effect of filial cannibalism on the number of simultaneous broods a male would subsequently

receive, which may be an important component of reproductive success. Similarly, I did not

measure hatching success, predation, and other potentially important components of reproductive









success. Further work is clearly needed to evaluate the adaptive significance of filial cannibalism

in the flagfish. Nonetheless, experimentally manipulating cannibalism while still allowing males

to care for eggs allowed me to evaluate key predictions of energy-based explanations of filial

cannibalism. Indeed, this study is the first to experimentally manipulate filial cannibalism while

still allowing males to care for eggs, thus allowing for the specific examination of some longer-

term fitness consequences of filial cannibalism in relation to the energetic hypothesis (Rohwer

1978).

The hypothesis that filial cannibalism in fishes reflects an adaptive trade-off in which

energy or nutrients gained from eggs is invested into future reproduction is widely accepted as

valid (e.g. Manica 2002) and has rarely been questioned since it was first proposed 25 years ago

(but see Smith 1992). While whole clutch cannibalism may be explained as the termination of

care, it does not appear that we have an adequate explanation for the widespread occurrence of

partial clutch cannibalism. In the present study, flagfish males consumed a large number of their

eggs and no benefits related to increased reproduction or physical condition were observed.

Thus, with regard to the flagfish, there is no evidence that Rohwer's (1978) hypothesis provides

adequate explanation for partial clutch filial cannibalism. More generally, there is mixed support

for Rohwer's theory of partial clutch cannibalism, as is evident from the inconsistent results of

studies examining the effect of food availability and parental condition on filial cannibalism (as

discussed above). In the case of flagfish, food sources other than eggs are available and the

incubation period is relatively short (less than 4 days at 290C), making it even more difficult to

understand why males would consume eggs purely for caloric or nutritional purposes. Thus,

future experiments should evaluate further specific ways in which energy gained from eggs

could be translated into net fitness benefits (i.e., male survivorship and increased quality of









parental care). Furthermore, I suggest alternatives to Rohwer's (1978) energy-based hypothesis

should be considered.

Currently, I can envision several adaptive explanations for why parental flagfish males

would consume eggs; 1) energy or nutrients from egg may increase male survival (consistent

with Rohwer 1978), 2) partial clutch cannibalism may improve survival of the remaining clutch

(e.g. Payne et al. 2002), and 3) males may selectively cannibalize eggs with reduced survivorship

or quality. These alternatives have received relatively little attention. Currently, the effect of

filial cannibalism on male survival remains untested and further research is needed to evaluate

this hypothesis. The effect of cannibalism on survival should be evaluated in the presence of

predators and other natural stressors. In the present study, I did not measure the effect of filial

cannibalism on offspring survival, and this idea should also be explicitly examined in separate

experiments. Recently, Payne et al (2002) suggested that partial clutch cannibalism increases

survivorship of remaining eggs through increased oxygen availability. This idea has received

support in one system (Stegastes leucostictus, Payne et al. 2002) but not in another

(Pomatoschistus minutus, Lissaker et al. 2002). However, it is possible that partial clutch

cannibalism improves survivorship of remaining offspring through other mechanisms. For

instance, if density-dependent egg predation exists, partial clutch cannibalism might reduce the

male's risk of losing some or all of his eggs to egg predators. It is also possible that parental

males use energy attained from eggs to increase the quality and/or quantity of parental care,

thereby increasing remaining offspring survivorship or quality. This idea is still consistent with

Rohwer's hypothesis (1978) but has not yet been evaluated. The idea that males may selectively

cannibalize eggs has rarely been considered. Indeed, males of many species eat more eggs than

would die naturally (e.g. Klug et al. 2005), but survival is typically measured only to hatching. It









is possible that males consume eggs with decreased post-hatching survivorship. Finally, it is

possible that filial cannibalism is maladaptive. This idea has been dismissed in recent literature,

but the validity of this dismissal remains unclear. In general, future work should focus on

developing and examining alternatives to Rohwer's energy-based explanation of filial

cannibalism.













MM w EXPECTED
OBSERVED


HIGM FOOD LOW FOOO
FOOD LEVEL
Figure 2-1. Expected and observed benefit of filial cannibalism in eggs received by males. The
minimum expected benefit (hatched bars) of filial cannibalism necessary to overcome
the loss resulting from the consumption of eggs is defined as the number of eggs
consumed by FC males; the observed benefit (solid bars) of filial cannibalism is the
difference in the mean number of eggs received by FC males and the mean number of
eggs received by NFC males. Bars represent mean number of eggs and error bars are
standard error.













A
NO FC








0I
-,



W" C*D UEEI..
FOOD LEVEL


B *
e'


34



z 2




C













FOOD LEVEL
U
Z a-



ui




niD FOOD O low oo
FOOD LEVEL

Figure 2-2. Effect of filial cannibalism on components of fitness. A) The mean total number of
eggs received over 90 days for males in each of the four treatments. The treatments
were high food and FC (i.e. access to eggs), high food and NO FC (i.e. no access to
eggs), low food and FC, & low food and NO FC. B) The total number of clutches
received over the 90 days across treatments. C) The frequency of spawning during the
90 days across treatments. Bars represent means and error bars are standard error.









CHAPTER 3
PARENTS BENEFIT FROM EATING OFFSPRING: DENSITY-DEPENDENT EGG
SURVIVORSHIP COMPENSATES FOR FILIAL CANNIBALISM

Introduction

Filial cannibalism is an evolutionary conundrum. How is eating one's own offspring ever

an adaptive strategy? Indeed, it is hard to imagine many circumstances in which regularly

consuming one's own offspring leads to increased net reproductive success. Thus, it is surprising

that filial cannibalism, which occurs in a range of taxa (Polis 1981), is particularly common in

fishes exhibiting paternal care (reviewed in Manica 2002), a behavior assumed to increase an

individual's fitness through increased offspring survivorship or quality (Clutton-Brock 1991).

While early ethologists considered filial cannibalism to be a rare behavior with little or no

evolutionary significance, filial cannibalism in fishes has now been well documented in both the

laboratory and the field (reviewed in Manica 2002), and currently, filial cannibalism in fishes is

thought to represent an adaptive strategy in which males maximize lifetime reproductive success.

Specifically, filial cannibalism is thought to reflect a trade-off between current and future

reproductive success, in which males gain energy and nutrients from eggs that are reinvested into

current and future reproduction (the energy-based hypothesis; as articulated by Rohwer 1978 and

Sargent 1992). According to this hypothesis, whole clutch cannibalism (i.e., the consumption of

all eggs present) is expected to be more frequent when clutch size is relatively small because the

energy requirements of caring males can be satisfied only by clutches larger than a certain size

(Rohwer 1978). Specifically, this hypothesis suggests that males should consume some specific

number of eggs to satisfy their energetic needs, and when initial clutch size is smaller than this

critical number, males should consume the whole clutch. Consistent with this prediction, several

studies have found that whole clutch cannibalism is more frequent when clutch size is relatively

small (reviewed in Manica 2002; but see Payne et al. 2003, who found that smaller clutches were









not preferentially eaten). The energy-based hypothesis also predicts that cannibalism should

increase as food availability decreases and/or when male condition is poor (Rohwer 1978;

Sargent 1992). Evidence regarding these predictions is equivocal. Consistent with the energy-

based hypothesis, supplementary feeding parental males reduced filial cannibalism in the

common goby (Pomatoschitus microps, Kvarnemo et al. 1998), the scissortail sergeant

(Abudefdufsexfasciatus, Manica 2004), and the Cortez damselfish (Stegastes rectifraenum,

Hoelzer 1992), and in some cases (e.g., Manica 2004) males do appear to simply be cleaning the

nest of dead eggs (i.e., mortality resulting from filial cannibalism is similar to background

mortality) when food is abundant. Contrary to predictions of the energy-based hypothesis, there

was no relationship found between cannibalism and food availability and/or male condition in

fantail darters (irhi,,iitaiflabellare, Lindstrom and Sargent 1997) and the three-spined

stickleback (Gasterosteus aculeatus, Belles-Isles & Fitzgerald 1991). Furthermore, male flagfish

(Jordanellafloridae) with reduced food availability actually consumedfewer eggs than males

with high food availability (Klug and St. Mary 2005 and Chapter 2).

Other studies have taken a different tack and examined the energetic content of eggs-- one

study suggested that energy from partial clutch cannibalism could potentially offset costs related

to care (Apogon lineatus, Kume et al. 2000), while another claimed that energy from eggs would

be insufficient (Gasterosteus aculeatus, Smith 1992). Also inconsistent with the energy-based

hypothesis, Payne et al. (2002) found that filial cannibalism increased in the later stages of egg

development, when egg energetic value is much lower. Thus, there is a lack of general support

for the energy-based explanation of filial cannibalism (Rohwer 1978; Sargent 1992) and at best

current theory can only explain cannibalism in some cases. Despite such mixed evidence and a

lack of many alternative hypotheses, the energy-based hypothesis of filial cannibalism is often









accepted as valid, and filial cannibalism is commonly considered an adaptive strategy (e.g.,

Vinyoles et al. 1999; Kume at al. 2000; Manica 2002). Overall, the support for this hypothesis is

unsatisfying and it remains unclear in many systems if (and potentially how) filial cannibalism is

an adaptive strategy.

Recently, Payne et al. (2002) proposed an alternative hypothesis suggesting that filial

cannibalism is an adaptive strategy in which partial clutch cannibalism improves survivorship of

remaining eggs by increasing oxygen availability to remaining eggs. In several systems low

dissolved oxygen levels have been related to increased egg mortality (Kamler 1992), and

according to the oxygen-mediated hypothesis of filial cannibalism (as articulated by Payne et al.

2002, 2004), males potentially improve overall clutch survivorship by removing some of their

eggs. Through a reduction in egg density, cannibalism can increase the surface area of the

developing embryos exposed to the water, thereby improving oxygen exchange and overall

survivorship of the remaining eggs. In other words, when egg density is relatively low, each

individual egg is expected to have greater oxygen availability than when egg density is high.

This hypothesis has received relatively little empirical examination. In their initial paper

proposing the idea of oxygen-mediated filial cannibalism, Payne et al. (2002) found that

reducing egg density in the beaugregory damselfish (Stegastes leucostictus) increased

developmental rate of embryos and that partial clutch cannibalism was significantly reduced

when oxygen levels were high. However, the reefs inhabited by beaugregory damselfish have

undergone significant environmental changes, particularly in relation to oxygen levels, over the

past twenty years, and beugregory damselfish do not oxygenate their eggs by fanning; thus, with

respect to other systems, it is unclear how general we would expect oxygen-mediated

cannibalism to be, particularly in species that are thought to oxygenate their eggs by fanning. In









the sand goby (Pomatoschistus minutus), a system in which males fan eggs, Lissdker et al.

(2002) found no differences in whole clutch or partial clutch cannibalism across oxygen levels,

although the aim of that study was not to explicitly evaluate oxygen-mediated filial cannibalism.

Thus, the importance of oxygen-mediated cannibalism remains unclear.

Effectively evaluating the importance of oxygen-mediated filial cannibalism necessitates

both an evaluation of predictions of the oxygen-mediated hypothesis of filial cannibalism (i.e.,

does cannibalism vary across oxygen/egg density levels) and an examination of potential fitness

consequences of filial cannibalism (i.e., does the potential benefit of cannibalism in terms of

additional number of eggs surviving outweigh the cost in number of eggs consumed by the

male). Thus, the goals of my study were two-fold. First, I evaluated the effect of oxygen, egg

density, and male condition on the occurrence of filial cannibalism and compared my findings to

predictions of both the energy-based and oxygen-mediated hypotheses of filial cannibalism.

Specifically, the oxygen-mediated hypothesis predicts that: 1) cannibalism will increase as

oxygen decreases (because of reduced egg survivorship when oxygen is relatively low); 2)

cannibalism will increase as egg density increases (because of decreased oxygen availability and

survivorship when egg density is high); and 3) there is no apriori expectation that male

condition will affect the occurrence of filial cannibalism. Likewise, the energy-based hypothesis

predicts that 1) cannibalism will increase as oxygen decreases (because energetic costs increase

when oxygen is low due to increased male fanning and increased metabolic costs to the males,

Jones and Reynolds 1999); 2) there is no apriori expectation that egg density will affect

cannibalism; and 3) cannibalism is expected to increase as initial male condition decreases.

Secondly, by experimentally manipulating cannibalism, I evaluated the effect of simulated

partial clutch cannibalism on egg survivorship and quantified the net benefits of partial clutch









cannibalism. If partial clutch cannibalism is an adaptive mechanism in which removing some

eggs improves overall egg survivorship, we would expect partial clutch cannibalism to be related

to no net reduction in offspring produced. The sand goby, Pomatoschistus minutus, is an ideal

system for evaluating the oxygen-mediated hypothesis of filial cannibalism because eggs are laid

in a dense layer, nests often have relatively low oxygen availability, and egg development is

dependent on male fanning (and therefore thought to be highly oxygen dependent); thus, if

oxygen-mediated filial cannibalism occurs generally, I would expect to find evidence of it in a

system such as the sand goby.

Methods

Study Species and Experimental Site

Sand goby males care for their eggs through guarding, cleaning, and fanning. Males build

nests under suitable substrates and cover the nest with sand, leaving only a small (-lcm)

opening. Sand goby eggs are clumped and male fanning is required for egg development and

survival. I conducted this study at Tvarminne Zoological Station, University of Helsinki, in

southern Finland. Sand gobies were collected in shallow brackish water using a seine, and males

and females were housed in separate holding tanks (100 1) with continuous seawater flow prior to

use. During this time the fish were fed ad libitum live Mysid shrimp and frozen Chironomidae

larvae. The experiment was conducted during the sand goby breeding season (June and July) of

2004.

Experimental Design

Experiment 1: Effect of oxygen and egg density on filial cannibalism: I crossed two

oxygen treatments (high and low oxygen concentration) with two egg density levels (high and

low egg density) to evaluate the effect of oxygen and egg density on the occurrence of filial

cannibalism. Thus, there were four treatments: 1) high oxygen, high egg density, 2) high oxygen,









low egg density, 3) low oxygen, high egg density, 4) low oxygen, low egg density. I used natural

variation in initial male condition (measured as K=100*weight/length3;Williams 2000) to

evaluate the relationship between parental condition and the occurrence of filial cannibalism.

Each experimental tank was 60 1 and equipped with continuous seawater flow through. The

tanks contained either a large (8 cm diameter) or a small (4 cm diameter) half-flowerpot, which

served as an artificial nest site. The two nest size treatments corresponded to my two

experimental egg density levels: high and low. Because females spawn their eggs in a monolayer

on the ceiling of the half-flowerpot and because egg number is approximately equal amongst

females, egg density in small nests can be much greater than egg density in large nests (K.

Lindstrom, personal observation). I fitted the inside of each nest with a transparent piece of

plastic onto which females spawn their eggs; the transparent plastic allowed me to remove and

photograph the clutch, when necessary, without disturbing the eggs. Specifically, there were 1.8

+/- 0.05 eggs per mm2 (X +/- SE) in small, high density nests and 1.4 +/- 0.12 eggs per mm2 in

the large, low density nests, and this difference was significant (2-tailed t-test, t=-3.09, df=38,

p=0.004). These egg density measurements are comparable to those observed in the wild (egg

density of nests in nature: range = 0.79 2.8 eggs per mm2, X +/- SE = 1.77 +/- 0.26 eggs per

mm2, N = 9).

I began each experimental by placing one male and one female in a tank with the randomly

assigned egg density treatment (small nest or large nest). After spawning occurred, I removed the

nest from the tank and photographed the eggs using a digital camera such that individual eggs

could be counted. I subsequently quantified initial egg numbers using these digital images. I then

transferred the clutch, on its plastic sheet, to a nest of intermediate size (6 cm diameter) and

returned the nest with eggs to the male. Transferring the eggs to a nest of intermediate size









ensured that males in both egg density treatments fanned nests of similar size and thus had

similar costs of care, which allowed me to isolate effects of egg density from effects of nest size.

Immediately after returning the eggs to the male, I randomly assigned an oxygen treatment (high

or low). In the low oxygen treatment, low oxygen water constantly flowed directly into the

male's nest, reducing the oxygen concentration inside the nest to approximately 32.2 +/- 10.1 %

(i.e., 3.2 +/- 1.0 mg/L at 14.30C, X +/- SE) of fully air-saturated water. This was done by

continuously bubbling nitrogen gas into a covered holding tank; using airline tubing, the

reduced-oxygen seawater was then allowed to flow continuously in through the rear of the

male's nest. In the high oxygen treatment, high oxygen water (from a flow through seawater

system) was continuously allowed to flow into the rear of the male's nest; in this case oxygen

concentration in the males nest was maintained at approximately 96.1 +/- 1.4% (i.e., 9.3 +/- 0.18

mg/L at 14.70C, X +/- SE ) of fully air-saturated water. Because oxygen concentration was only

manipulated in the nest and because all tanks had continuous flow through of sea water, oxygen

concentration outside of the nest was approximately the same for both high and low oxygen

treatments. I quantified the oxygen levels in the high and low oxygen nests for a sample of the

males in the experiment using an IS02 oxygen meter equipped with an OXELP oxygen electrode

(World Precision Instruments); oxygen levels inside the high oxygen treatment nests were

significantly greater than those in the low oxygen treatments (two-tailed t-test, t = -4.7, df = 9,

p=0.001).

I followed eggs until hatching and visually inspected nests daily by shining a light into the

nest. When eye shine was visible in the nest (approximately 1-3 days prior to hatching), I

removed the nest, photographed the plastic transparency with eggs, and later counted the number

of eggs using the digital photograph. Male and female weight and standard length were measured









at the beginning and end of each replicate, and the condition measure K (Williams 2000) was

used as an indicator of male condition.

Experiment 2: Effect of simulated filial cannibalism on egg survivorship: I examined

the effect of two levels of simulated filial cannibalism (egg removal and no egg removal) in high

and low oxygen environments. In this experiment, one male and one or two females were placed

in an aquarium (described above) with an intermediate-size artificial nest (described above)

containing a transparent plastic sheet for spawning; after spawning occurred, I removed the nest

from the tank. I then cut the plastic sheet into four approximately equal pieces and each quarter

was randomly assigned to one of the four treatments: 1) high oxygen, egg removal, 2) high

oxygen, no egg removal, 3) low oxygen, egg removal, 4) low oxygen, no egg removal.

In the egg removal treatments, I simulated filial cannibalism by using a pair of fine forceps

to haphazardly remove some proportion of the eggs corresponding to actual levels of filial

cannibalism observed in experiment 1 (simulated cannibalism: 10 66% of eggs removed, 39.6 +

2.5 % = mean SE; observed partial cannibalism: 6 93%, 37.5 7.9%). In the no egg removal

treatments, I removed a trivial number of eggs (approximately 5-10 eggs) from the nest and

gently touched a proportion of the remaining eggs (corresponding to the approximate number of

eggs that were removed in the egg removal treatment) with the forceps. Digital images were then

taken of the transparencies and I later counted the eggs. Each transparency with eggs was then

placed in either a high oxygen water or low oxygen water container, each approximately 18 cm x

13 cm x 3.5 cm. Specifically, using a pin, I attached the transparencies with eggs to a styrofoam

lid designed to snuggly fit on each of the containers. I then placed the lid on the container so that

the eggs were always in water. Because eggs are dependent on fanning, an air stone continuously

bubbled air approximately 5 cm directly below the eggs at all times in both treatments. This









method allows eggs to develop at rates comparable to those observed when males fan the eggs

(Maria Jarvi-Laturi, personal communication). The high oxygen water contained a second air

stone that continuously bubbled air into the water. This air stone was placed in the corner of the

container so that the bubbles did not directly hit the eggs, and the oxygen concentration was

maintained at 93.0 +/- 3.1% (i.e., 8.9 +/- 0.28 mg/L, X +/- SE) of fully saturated water. Likewise,

in the low oxygen treatment, I placed a second air stone in the corner of the tank and nitrogen

was continuously bubbled into this container through this air stone. Again, the bubbles from the

second air stone did not directly flow onto the eggs, and thus, I eliminated any potential effects

of direct contact of nitrogen gas with the eggs. Oxygen concentration in this tank was maintained

at approximately 24.9 +/- 6.4% (i.e., 2.4 +/- 0.61 mg/L, X +/- SE ) of fully saturated water when

the probe was placed 5cm above the air stone bubbling air (i.e., where the eggs were placed and

in direct flow of the air bubbles). I checked eggs and oxygen concentration daily by briefly

removing the lids of the containers. When measuring oxygen level, the Styrofoam lid was

carefully raised and the probe was placed against the Styrofoam near the eggs, directly in the

flow of the air bubbles. Despite bubbling air underneath the eggs in all treatments, oxygen

concentration was significantly greater in the high oxygen containers than in the low oxygen

containers (two-tailed t-test, t = -8.5, df = 5, p<0.001), and I am thus confident that eggs in the

two oxygen treatments experienced very different oxygen levels. I recorded the proportion of the

clutch surviving until eye shine was present, the day on which eye shine was visible, and the

proportion of the clutch covered in fungus. Three to six clutches were placed in a container at

one time and four replicates were completed.

Data Analysis

Experiment 1: Effect of oxygen and egg density on filial cannibalism: I analyzed the

effect of oxygen and egg density on the occurrence of whole clutch cannibalism using a stepwise









logistic regression (remove ifp > 0.15); oxygen and egg density were treated as categorical

variables, and date of spawning, initial male condition, change in male condition, and initial egg

number were used as covariates in the analysis. I analyzed the effect of oxygen and egg density

on the day of whole clutch cannibalism, egg survivorship, and male condition using stepwise

ANCOVA (remove if p > 0.15), and thus, for simplicity's sake, I only present significant effects

of the final model, unless otherwise stated. In all analyses, I initially evaluated all two-way

interactions between factors and covariates and found them to be non-significant. I then removed

these factor-covariate interactions and proceeded with the stepwise ANCOVA. For day of whole

clutch cannibalism and egg survivorship analyses, oxygen and egg density were treated as

categorical fixed factors, and initial egg number, date of spawning, initial male condition, and

change in male condition were used as covariates. The interaction between oxygen and egg

density was also included in the model. For male condition analyses (i.e., change in condition

and final condition), oxygen and egg density were again treated as categorical fixed factors, and

date of spawning, initial male condition, and number of eggs consumed were used as covariates,

and again, the interaction between oxygen and egg density was included. For analyses of

survivorship, proportion surviving was first arcsin square root transformed.

Additionally, whole clutch cannibalism and partial clutch cannibalism likely have distinct

biological significance; whole clutch cannibalism is a termination of care and benefits of whole

clutch cannibalism can only be seen in future reproductive success, whereas benefits of partial

clutch cannibalism are potentially related to current and/or future reproductive success. Thus, I

performed analyses both including and excluding cases of whole clutch cannibalism, when

applicable.









Experiment 2: Effect of simulated filial cannibalism on egg survivorship: I evaluated

the effect of egg removal and oxygen on survivorship of remaining eggs using a completely

randomized block design. For the analysis of survivorship, the proportion of the clutch surviving

was first arcsin square root transformed. I treated oxygen and egg removal as fixed factors, and

clutch (i.e., the original clutch eggs were from) was treated as a random factor. Blocking by

clutch allowed me to account for inherent differences among clutches (e.g., effects of parents)

and also implicitly accounts for any effect of time (i.e., replicate); thus, blocking by replicate

(i.e., time) is unnecessary (G. Wallace, University of Florida Statistics Department, pers.

comm.). I compared the number of eggs surviving with and without simulated cannibalism in

order to explicitly evaluate whether the potential benefit of cannibalism (i.e., increased

survivorship of remaining eggs stemming from egg removal) outweighed the associated cost

(i.e., number of eggs removed). For cannibalism to result in no net reduction in current

reproductive success, the total number of eggs surviving with cannibalism must, on average, be

greater than or equal to the total number of eggs surviving without cannibalism. As above, I

analyzed this data as a completely randomized block design, and treated oxygen and egg removal

as fixed factors and clutch as a random factor. All analyses were performed both including and

excluding cases of whole clutch death.

Results

Experiment 1: Effect of Oxygen, Egg Density, and Male Condition on Filial Cannibalism

Occurrence of whole clutch cannibalism: Whole clutch cannibalism was more

prevalent when oxygen was low and/or egg density was high (stepwise binary logistic regression,

oxygen effect, X = 3.81, p = 0.05; egg density effect, X2 = 3.78, p = 0.05; Figure 3-1).

Similarly, smaller clutches were subject to whole clutch cannibalism more frequently than larger










clutches (initial egg number effect, X = 4.43, p = 0.035). It is important to note that there was

no significant difference in initial egg number between the oxygen or egg density treatments (2-

way ANOVA, oxygen effect, F1,39 = 0.1, p = 0.75, egg density effect, F1,39 = 0.001, p = 0.98).

While there was a trend for whole clutch cannibalism to decrease as the breeding season

progressed, there was no significant effect of the date of spawning on whole clutch cannibalism

(X = 3.20, p = 0.07). Also noteworthy in relation to the energy-based hypothesis, there was no

effect of initial male condition or change in male condition on the occurrence of whole clutch

cannibalism (p > 0.15). Specifically, the initial condition (K) of males that cannibalized their

whole clutch was 0.64 +/- 0.014 g/cm3 (X +/- SE), and the initial condition of males that did not

practice whole clutch cannibalism was 0.63 +/- 0.024 g/cm3 (X +/- SE).

To further analyze patterns of whole clutch cannibalism, I examined the day on which

whole clutch cannibalism occurred. The timing of whole clutch cannibalism occurred earlier as

the breeding season progressed (F1,24 = 9.85, p= 0.005) and when initial egg number was

relatively small (F1,24 = 4.559, p = 0.05). Additionally, males in poorer condition cared for their

clutches longer before cannibalizing them entirely (initial condition effect, F1,24 = 7.27, p =

0.01). There was significant interaction between the date spawning occurred and the initial

number of eggs received (F1,24 = 17.6, p < 0.001).

Egg survivorship: High oxygen tended to increase egg survivorship (oxygen effect, F1,37

= 3.62, p = 0.06; Figure 3-2), and egg survivorship also increased as initial egg number increased

(initial egg number effect, F1,37 = 7.96, 0.008); however, there was no significant effect of initial

condition, egg density, or spawning date on egg survivorship (p > 0.15 for all). Because the trend

for oxygen and initial egg number to affect egg survivorship was almost certainly the result of

whole clutch cannibalism (which was affected by oxygen and initial egg number), I then









repeated the analysis excluding cases of whole clutch cannibalism. In this case, I found a

significant effect of oxygen (F1,12 = 9.9, p = 0.02) and spawning date (F1,12 = 6.1, p = 0.05).

Additionally, egg survivorship was higher for males that were initially in poorer condition (initial

condition effect, F1,12 = 15.0, p = 0.008; Figure 3-3 A), and there was also a relationship between

change in male condition and egg survivorship (F1,12 = 15.5, p = 0.008). Specifically, males that

were in poorer condition to begin with ate a smaller proportion of (and fewer) eggs than males in

better condition (Figure 3-3 A), and males that did eat eggs had less of a decline in condition

(Figure 3-3 B). Again, there was no effect of egg density. However, it's important to note this

analysis is, in a sense, not performed on a random sample of eggs; males have already decided

whether or not to cannibalize the whole clutch, and potentially, they have decided to continue to

care for eggs that they expect to have the greatest survivorship. For example, only one male in

the low oxygen, high egg density did not cannibalize his entire clutch. We have no idea what

survivorship of the other low oxygen, high egg density clutches would have been. Effectively

analyzing the direct effects of oxygen and egg density on egg survivorship necessitates that this

is done in the absence of whole clutch cannibalism (as in Experiment 2).

Male condition: When males consumed their entire clutch (i.e., including only cases of

whole clutch cannibalism), there was a significant effect of initial condition on final condition

(F1,24 = 49.5, p < 0.001) but not on change in male condition (p> 0.15). Additionally, final male

condition worsened as the breeding season progressed (spawning date effect, F1,24 = 4.5, p =

0.05).

When I considered only partial clutch cannibalism, providing care in low oxygen nests

resulted in a poorer final condition and a greater decrease in condition, in comparison to

providing care in high oxygen nests (final condition, F1,12 = 19.4, p = 0.003; change in condition,









F1,12 = 16.4, p = 0.007). Additionally, condition worsened as the breeding season progressed

(final condition, F1,12 = 10.6, p = 0.01; change in condition, F1,12 = 9.3, p = 0.02). The number of

eggs a male consumed was positively related to condition, suggesting that eggs provide energetic

benefits to caring males (final condition, F1,12 = 17.2, p = 0.004; change in condition, F1,12

11.0, p = 0.02).

Experiment 2: Effect of Simulated Filial Cannibalism on Egg Survivorship

Effect of oxygen and egg removal on remaining egg survivorship: Egg removal

significantly improved egg survivorship (F1,33 = 4.64, p=0.04; Figure 3-4 A). Additionally, there

were significant differences among clutches in egg survivorship (i.e., eggs from particular

clutches had higher egg survivorship than eggs from other clutches, regardless of treatment; F1,16

= 4.02, p = 0.004) and significant interaction between the clutch that eggs were from and oxygen

(i.e., some clutches did better in the low oxygen environment, while others did better in the high

oxygen environment; F16,33 = 5.06, p < 0.001). Surprisingly, there was no effect of oxygen on

egg survivorship. This analysis was performed including cases in which the whole clutch died. In

all cases in which the whole clutch died, fungus had taken over the nest. However, in the

presence of the male, fungus is never observed to attack the entire clutch. Thus, I repeated this

analysis excluding whole clutch death, which might be more representative of natural conditions.

The results were unchanged; egg removal increased egg survivorship (F1,1o.1 = 13.83, p =0.004;

Figure 3-4 B) and there was a significant interaction between clutch and oxygen (F8,6 = 9.36, p =

0.007), but again, there was no effect of oxygen on egg survivorship.

Effect of oxygen and egg removal on total number of eggs surviving: I compared the

number of eggs surviving with and without cannibalism to evaluate potential fitness

consequences of cannibalism. Regardless of whether I included or excluded cases of whole

clutch death, I found no effect of oxygen or egg removal on the number of eggs surviving









(Figure 3-5). In both cases, there is a significant interaction between oxygen and the clutch that

the eggs were from (including whole clutch death, F17,33 = 4.60, p <0.001; excluding whole

clutch death, F18,16 = 3.89, p =0.004).

Discussion

Here, I present evidence that filial cannibalism can potentially be an adaptive mechanism

associated with density-dependent egg survivorship. When I simulated partial clutch cannibalism

by removing eggs, the survivorship of the remaining eggs increased, and more importantly, egg

removal at a range of cannibalism levels (10 66 % of clutch) did not reduce the total number of

offspring produced. Indeed, my results show that males can consume, on average, 40% of their

eggs with no reduction in current reproductive success! Thus, partial clutch filial cannibalism is

potentially a mechanism by which males improve survivorship of remaining eggs. This finding is

consistent with the theoretical predictions of Payne et al. (2004), whose modeling results suggest

that under some conditions, males can consume up to 80% of their clutch without a reduction in

reproductive output. Under this scenario, males potentially also gain energy from eggs with no

net loss in reproductive success. My results and previous work in the sand goby (Lindstrom

1998) suggest that consuming eggs is energetically beneficial. Additionally, actual partial clutch

cannibalism is likely to be much more precise and selective than my simulated cannibalism and

males can potentially track the conditions in their nest, suggesting that partial clutch cannibalism

may be even more efficient at enhancing the survival of the remaining brood.

Consistent with previous theory (Rohwer 1978; Sargent 1992), I found that whole clutch

cannibalism in the sand goby is more prevalent when initial clutch size is relatively small.

However, male condition did not affect the occurrence of whole clutch cannibalism; this finding

is inconsistent with the current energy-based hypothesis, which suggests that whole clutch

cannibalism should be more frequent when clutch size is relatively small because the energy









requirement of males can be satisfied only by clutches larger than a certain size (Rohwer 1978).

My results suggest that whole clutch cannibalism depends on costs and expected benefits of care,

as suggested by Manica (2004). Parental care in fishes is often assumed to be shareable among

offspring in a nest (i.e., a unit of parental care may be given to one or several offspring;

Wittenberger 1981, Williams 1975). Thus, the costs of care for small and large clutches is

assumed to be comparable, whereas the benefit (i.e., offspring produced) of caring for a large

clutch is assumed to be much greater than that of a small clutch. In my study, whole clutch

cannibalism increased as the expected benefit of care decreased (i.e., as initial egg number

decreased) and as the cost of care increased (i.e., with decreasing oxygen). This finding that low

oxygen led to more whole clutch cannibalism is consistent with both the previous energy-based

hypothesis (Rohwer 1978; Sargent 1992) (because cost of care to the male increases as oxygen

decreases) and with the oxygen-mediated hypothesis (Paynet et al 2002; 2004) (because the

expected benefit would be less when oxygen is low, according to this hypothesis). However, the

finding that males altered their whole clutch cannibalism according to egg density is only

consistent with the oxygen-mediated hypothesis (since the benefit from a high density clutch

would be less than that of a low density clutch, assuming egg numbers are equal, Figure3-4) and

is not predicted by the energy-based hypothesis.

The observed patterns of filial cannibalism, and in particular partial clutch cannibalism, are

thus consistent with cannibalism mediated by density-dependent egg survivorship (e.g., the

oxygen-mediated hypothesis, Payne et al 2002) and inconsistent with the energy-based

hypothesis (Rohwer 1978; Sargent 1992). For example, the energy-based hypothesis assumes

that there is an adaptive trade-off occurring between current and future reproductive success, in

which males gain energy for future reproduction at a cost to current reproductive success. In the









sand goby, I found that there is no cost to current reproduction (i.e., no trade-off between current

and future reproductive success) because density-dependent egg survivorship compensates for

filial cannibalism. Furthermore, the energy-based hypothesis predicts a negative relationship

between male condition and filial cannibalism. Contrary to this prediction, I found that males in

poorer condition actually consumedfewer of their eggs than males in better condition. A similar

finding in the flagfish was reported by Klug and St. Mary (2005; Chapter 2), who suggested that

males in poor condition possibly have reduced expected future reproduction and thus invest more

into their current clutch. While inconsistent with the current energy-based hypothesis, my

findings are similar to those in some other systems (e.g., Belles-Isles and Fitzgerald 1991;

Lindstrom and Sargent 1997; Klug and St. Mary 2005), but contrary to results in others (e.g.,

Hoelzer 1992; Kvarnemo et al. 1998; Manica 2004). However, my results are also, in part,

contrary to oxygen-mediated filial cannibalism. Although whole clutch cannibalism was affected

by oxygen and egg density, and egg removal improved egg survivorship, oxygen did not affect

egg survivorship in the simulated cannibalism experiment. Thus, it does not appear that partial

clutch cannibalism in the sand goby improves egg survivorship by increasing oxygen availability

of remaining eggs. Indeed, partial clutch cannibalism does appear to be a mechanism in which

the consumption of some eggs improves survivorship of remaining eggs; however, the cause of

density dependent egg survivorship is unknown in this system. Additionally, the lack of an effect

of oxygen on egg survivorship also indicates that male fanning might not primarily serve to

oxygenate eggs, as is typically assumed.

I therefore suggest a general explanation of filial cannibalism that is mediated by density-

dependent egg survivorship, but that is not solely related to oxygen, as in current theory (Payne

et al. 2002, 2004). There are many ways in which density-dependent egg survivorship can occur.









For example, in systems such as the sand goby in which eggs are clumped, it is plausible that

waste from developing embryos negatively affects the development and/ or survival of other

embryos, and that decreased egg density reduces the negative effects of such waste. In species

whose eggs are not highly clumped (e.g., flagfish) other density-dependent factors, such as

density-dependent egg predation might affect egg survivorship. For instance, if male nest size

has some limit and predators are attracted to nests with more eggs, a male might reduce the

probability of losing some or all of his eggs to predators by consuming some proportion of them.

In systems such as beaugregory damselfish, the system on which Payne et al. (2002) based their

hypothesis of oxygen-mediated cannibalism, which lack male fanning of eggs, it does seem

likely that density-dependent egg survivorship is due to limited oxygen availability. Specifically,

the reefs inhabited by beaugregory damselfish recently underwent changes in oxygen levels, and

thus it seems likely that males cannibalize their eggs to increase oxygen availability of remaining

eggs. More work is needed to assess the relative importance of the evolution of filial cannibalism

and fanning, and future work should focus on systems with different fanning and parental care

strategies. Indeed, the actual costs and benefits of fanning and filial cannibalism are unknown in

many systems. More generally, other hypotheses associated with filial cannibalism mediated by

density-dependent egg survivorship should be explored further.

Regardless of the mechanism, filial cannibalism mediated by density-dependent egg

survivorship raises an interesting question: why don't females just lay fewer eggs? If egg

survivorship is density dependent and laying tightly packed eggs reduces overall egg

survivorship, why haven't females been selected to lay less dense clutches? In the sand goby, it

seems unlikely that laying dispersed egg batches would be an evolutionarily stable strategy. If a

female were to lay fewer, more dispersed eggs, it is likely that the remaining space would be









taken up by another female's eggs, resulting in the same amount of cannibalism. If all nests (or

nests of high quality males) were full of sparsely laid eggs, a female would likely increase her

fitness by laying a denser clutch in a nest already containing eggs, as opposed to not laying at all

or laying eggs with a low quality male. Additionally, it is important to note that females do, to

some extent, mediate the egg density of their clutch; in the present study, egg density in larger

nests was significantly less than egg density in smaller nests. Thus, females do reduce the egg

density of their clutch when it is possible to do so. Why females lay many, densely packed eggs

is a complex problem, and necessitates consideration of many factors affecting egg survivorship,

including the costs associated with density-dependent egg survivorship and the costs and benefits

associated with mate quality. However, it does not appear that selection will necessarily favor

females laying clutches with fewer, less tightly packed eggs when egg survivorship is density-

dependent.

In conclusion, alternative explanations of filial cannibalism need to be explored further,

and new hypotheses should be developed and evaluated. For example, selective filial

cannibalism seems plausible. Numerous studies show that males consume healthy and viable

eggs (discussed in Manica 2002). However, if variation in offspring quality exists and if care is

not entirely shareable among offspring, or if egg survivorship is density-dependent, males could

potentially benefit by consuming viable eggs of reduced quality (e.g., eggs with reduced post-

hatching survival; see also literature on selective embryo abortion in plants, e.g., Burd 1998). To

more explicitly evaluate the potential fitness consequences of cannibalism, it would be useful to

estimate the average offspring fitness in nests in which males have removed eggs and nests in

which experimenters have removed eggs, to determine if males are eating poor-quality eggs.

Furthermore, the future benefit of eating eggs in this system remains unclear and should be









evaluated in an experiment specifically designed to quantify future reproductive and survivorship

benefits of cannibalism. Eating eggs clearly provides a male with some energetic gain, but the

relative importance of this benefit for future reproduction is unknown in this system. Finally, the

evolutionary significance of filial cannibalism is likely not due solely to one factor (e.g.,

energetic benefit of eggs, density-dependent egg survivorship), and once the current and future

costs of filial cannibalism are better understood, a more synthetic model of filial cannibalism will

be necessary.
















1.0 LOW DENSITY
I I HIGH DENSITY
E

3 0.8-



o 0.6 -
O
rc
-c




O
0.4
.-5


0 0.2




0.0
LOW OXYGEN HIGH OXYGEN

OXYGEN LEVEL

Figure 3-1. Effect of oxygen and egg density on the prevalence of whole clutch cannibalism by
parental males. Bars represent the proportion of clutches in which the male consumed
the entire clutch, and error bars are standard error based on a binomial distribution.
















17/ LOW DENSITY
I-I HIGH DENSITY


4 33


LOW OXYGEN


HIGH OXYGEN


OXYGEN LEVEL

Figure 3-2. Effect of oxygen and egg density on the mean (+/- SE) egg survivorship (i.e.,
proportion of the clutch that survived until hatching) when males were present with
eggs, including cases of both whole and partial clutch cannibalism. It is important to
note that in the low oxygen, high egg density treatment, only 1 male cared for his
clutch until hatching; the other 8 males in this treatment cannibalized their whole
clutch.


0.5




m 0.4
,t


3;
()
c 0.3-

. U
4-_
0
O
c 0.2-
o
0
o

a 0.1 -


p7


777


















































050 060 070 080
Initial Condition (K)





0

o












o





0


I I I I I
000 020 040 060 080
Proportion of Clutch Consumed


Figure 3-3. Relationship between A) male condition (i.e., K=100*g/cm ) and the proportion of

the clutch consumed by parental males, and B) partial clutch cannibalism and change

in male condition.



3


1 00-
IO0-





0 080-
E


O
C
0
060-





S040-
.2
r
0
0.
0.. 020-




000-


010-
C







0- -
0
000-
0
C







C -010-
S00o5-
c

-0

-
0)




-015-
















a NO EGG REMOVAL
S EGG REMOVAL


LOW OXYGEN HIGH OXYGEN

OXYGEN LEVEL


2I22 NO EGG REMOVAL
= EGG REMOVAL


Figure 3-4. Effects of simulated filial cannibalism. A) The effect of oxygen and egg removal on

egg survivorship (i.e., the mean +/- SE proportion of the clutch surviving), including

cases of both whole and partial clutch death. B) The effect of oxygen and egg

removal on egg survivorship (i.e., the mean +/- SE proportion of the clutch

surviving), excluding cases of whole clutch death.


0.5


S0.4
Uo


0.3


0
e 0.2

a-
0.1


0.0 I


1.0



0.8
.>


0.6


as
c 0.4


0
nQ 0.2


00o


LOW OXYGEN HIGH OXYGEN
OXYGEN LEVEL














2 NO EGG REMOVAL
I EGG REMOVAL


LOW OXYGEN HIGH OXYGEN


OXYGEN LEVEL
Figure 3-5. Simulated filial cannibalism: The effect of oxygen and egg removal on the total
number of eggs surviving, including cases of both whole and partial clutch death. Egg
removal did not significantly reduce the total number of eggs surviving. Bars
represent means and error bars are standard error.









CHAPTER 4
SELECTIVE FILIAL CANNIBALISM IN THE SAND GOBY

Introduction

Filial cannibalism is an evolutionary mystery. It is difficult to imagine how regularly

consuming one's own viable young represents an adaptive strategy, yet filial cannibalism is

prevalent in a range of animals, particularly fishes exhibiting paternal care of eggs (discussed in

Manica 2002; Klug and Bonsall 2007). Typically, filial cannibalism is viewed as an adaptive

trade-off in which energy gained from eggs is used to better care for remaining offspring, or for

increasing future reproduction (Rohwer 1978; Sargent 1992; Manica 2002). Because energy is

such an obvious and direct benefit of filial cannibalism, much of the work aimed at

understanding the adaptive significance of filial cannibalism has focused on energetic benefits

(reviewed in Manica 2002). However, some have suggested that energetic benefits alone are

unlikely to explain the prevalence of filial cannibalism in natural systems (e.g. Smith 1992;

Payne et al. 2002; Klug et al. 2006 and Chapter 3).

Recent theoretical work suggests that the ability to cannibalize offspring selectively in

relation to aspects of offspring phenotype (e.g., quality or egg maturation rate) can directly favor

the evolution of filial cannibalism (Klug and Bonsall 2007 and Chapter 6). While the idea of

weeding out inferior offspring has been documented in other contexts (Forbes & Mock 1998; e.g.

selective abortion in humans: Stears, 1987; Forbes 1997; Diamond 1987; Hesketh & Xing

2006; selective abortion in plants: Burd 1998; Karkkainen et al. 1999; Melser & Klinkhamer

2001), it has not yet been explicitly evaluated in relation to filial cannibalism (but see Mrowka

1987 and Kraak 1996 for work on the consumption of unfertilized or diseased eggs, and Neff and

Sherman 2003 regarding egg cannibalism of non-kin). Indeed, the relationship between offspring

phenotype and filial cannibalism remains unknown. One study found preferential cannibalism of









younger eggs when within-brood variation in egg age existed (Salfert and Moodie 1985), but in

general, little is known about the specifics of which eggs are consumed when a parent does

decide to cannibalize. Hence, the importance of selective filial cannibalism remains unknown.

To begin to understand the potential importance of selectivity in filial cannibalism, I

evaluated patterns of within-brood cannibalism in the sand goby, Pomatoschistus minutus, a fish

in which males alone provide parental care and practice filial cannibalism during the egg stage.

My primary goal was to determine if males practice selective cannibalism (i.e., non-random

consumption of eggs with regard to some aspect of egg phenotype). Specifically, I focused on

the relationship between egg size and partial clutch filial cannibalism (i.e., the consumption of

some eggs present during a given reproductive bout) by males. Egg size is correlated to larva

size in this species (H.K., unpublished data) and has been correlated with post-hatching

survivorship in a range of fishes (discussed in Kamler 2005). I considered cases in which males

received eggs from one or two females. For the case in which males had eggs from only one

female, I asked whether males exhibited a preference for eggs of a particular size range. When

males had eggs from multiple females, I was interested in whether males 1) preferentially

consumed eggs of the first or the second female that they spawned with, and 2) showed any

preference with regard to egg size for each of the clutches in a nest (i.e., female l's and female

2's clutches). I also compared the size of female l's and female 2's eggs to determine if any

differences in egg size existed between the females, and if so, whether these differences might

explain the patterns of cannibalism observed.

Materials and Methods

Experimental Design

I evaluated selective filial cannibalism in the sand goby using data from two years. In both

years adult sand gobies were collected in shallow brackish water using a seine near Tvarminne









Zoological Station (University of Helsinki) in southern Finland. All fish were fed ad libitum live

Mysid shrimp and frozen Chironomidae larvae throughout the studies.

Multiple-female set-up: In 2006, I initiated each replicate by placing a single male and a

single female (female 1) in an aquarium equipped with continuous flow-through seawater

system. Each nest contained a half-flowerpot (8 cm diameter) which served as the nesting site.

The inside of each nest was fitted with a transparent piece of plastic onto which females spawn

their eggs. The transparent plastic allowed me to remove and photograph the eggs, when

necessary. The male-female pair was allowed to spawn, and immediately after spawning, I

removed the nest with eggs, digitally photographed the eggs, and cut out and removed a small

subset of eggs (20-30 eggs) from the plastic transparency. I reared the subset of eggs in the

absence of the male (described below) to determine if any size-specific patterns of egg mortality

existed. I then returned the nest with eggs to the male, and a second female (female 2) was

placed in the tank. The male and second female were then allowed 24 h to spawn (only clutches

in which the second female spawned within 24 h of the first spawning were used in this

experiment). After spawning, I again removed and photographed the eggs, and I removed a small

subset of eggs. The nest with eggs was then returned to the male. Only cases of partial-clutch

cannibalism were considered in the present study, as I was interested in within-brood patterns of

cannibalism. I followed all eggs until hatching and visually inspected the nests daily by shining a

light into the nest. Just prior to hatching (i.e., 1-3 days before hatching), eye shine (i.e., reflection

from a flashlight) becomes visible in the developing embryos, and this is an indication that the

eggs are about to begin hatching. When eye shine was visible in the eggs, I removed the nest and

photographed the plastic transparency with eggs.









The subsets of removed eggs were reared in individual plastic containers, each equipped

with an airstone. Water in the plastic containers was changed daily using water from the flow-

through system, and the water temperature in these containers was maintained at approximately

160C.

Single-female set-up: For the single male-single female trials, I retrospectively analyzed

digital images collected from a previous experiment conducted in 2004 (Klug et al. 2006 and

Chapter 3) in the same location using the same population of fish. In this case, I placed a male

and female in an aquarium equipped with continuous sea water flow-through and a half

flowerpot nest (8 cm diameter) equipped with a plastic transparency. Immediately after

spawning, the nest and eggs were removed and photographed. In this case, I then transferred the

clutch, on its plastic sheet, to a nest of intermediate size (6 cm diameter) and returned the nest

with eggs to the male (see Chapter 3 for additional details). I only used clutches from the low

density, high oxygen treatments (Klug et al. 2006 and Chapter 3) in the present study, as this set-

up was most comparable to the design used in 2006 (described above). Again, only cases of

partial clutch cannibalism were considered. I followed all eggs until hatching, and I visually

inspected the eggs daily by shining a light into the nest. When eye shine was visible in the nest, I

removed the nest and photographed the plastic transparency with eggs.

Image Analysis

For each male's eggs in 2006 (the multiple-female scenario), I superimposed the image

immediately following female l's spawning and the image immediately following female's 2. I

then identified all eggs and labeled them as belonging to either female 1 or female 2. The image

following female 2's spawning (in which all eggs have now been identified) was then

superimposed with the final image taken just before hatching. I then determined the specific eggs

that had been consumed. Using Sigma Scan Pro 5.0 (SPSS, Inc.) and the image containing the









spawn of female 1 and 2, I quantified the initial diameter of 1) a random subset of female l's

eggs and female 2's eggs (range: 25-75 eggs of each female) and 2) a subset of the specific eggs

consumed (range: 5-45 eggs of each female). For each male's eggs in 2004 (the single-female

scenario), I superimposed the initial and final image to determine which eggs had been

consumed. I then used Sigma Scan Pro 5.0 (SPSS, Inc.) to quantify the initial diameter of 1) a

random subset of the eggs (range: 35-80 eggs) and 2) a subset of the specific eggs consumed

(range: 10-38 eggs). These data allowed me to quantify the initial size distributions of 1) all eggs

in a nest and 2) the eggs that were consumed. This, in turn, allowed me to estimate cannibalistic

preference for eggs of varying size classes (described below), while taking into account the

initial abundance of eggs of varying sizes.

Preference Calculation

I was interested in whether males preferentially consumed 1) eggs of female 1 or female 2

(for the multiple-female data), and/or 2) eggs of a particular size class (for both the multiple- and

single-female data). To assess whether males consume eggs in some non-random way, I used the


preference measure a (with the ith component a,; Manly et al. 1972; Chesson 1983). This

preference measure has been used widely in studies of foraging (discussed in Chesson 1983). It

is an ideal measure of preference for my purposes because it allowed me to account for 1) the

initial abundance of eggs of varying sizes and 2) the depletion of eggs of varying size due to

cannibalism. I calculated measures of egg preference in relation to preference for female l's eggs

and for female 2's eggs, and for eggs of given size classes. Preference for each egg type (i) was

calculated as follows:


where i= 1,...,m (4-1)
ln((n, -~ )/(no)
J-1












where nio is the number of eggs of type i present initially, ri is the number of eggs of type i

consumed by the male, and m is the total number of different egg types present (modified from

Manly et al. 1972 and Chesson 1983). For this measure of preference, 0 indicates no preference

(i.e. consumption is equivalent to what is expected if males randomly consume eggs), a positive

value reflects consumption that is greater than would be expected from random consumption (i.e.

a preference for that egg type exists), and a negative value suggests that consumption is less than

what would be expected from random consumption.

Based on my data (i.e., the range in observed egg size and patterns of consumption), I had

sufficient resolution to identify 4 size classes for each brood of eggs. I calculated four equal size

classes (i.e., small, small-medium, medium-large, and large) for each brood by dividing the

range in egg diameter for a given brood by four. For the single-female data, I calculated


preference for each of the 4 size classes ( sma, sma" -medum medzum-large, and argue) In this

case, if there was no preference for any particular size class of eggs, I would expect

amaii = smaii-meum amedum-large large = For the 2006 data, I estimated preference for 4 size

classes for each of the two females, and thus, there were a total of 8 egg types. Again, if there

was no preference for eggs of a particular female or egg size, I would expect

fem small fem 1small-medium fem 1medium-large fem 11arge fem small fem 2small-medium
a =a =
fem 2medium-large fem 21arge 0

Statistical Analyses

I analyzed all preference data using non-parametric Friedman ANOVA. I used t-tests to

examine differences in mean egg diameter between female 1 and female 2 in 2006, and to

compare mean egg diameter in 2004 versus 2006. Given the differences in experimental set-up









between years, I compared mean egg density between 2004 and 2006 using a t-test. Linear

regressions were used to examine the relationship between mean egg diameter and egg survival,

and between mean egg diameter and egg development rate (i.e. time from spawning until

hatching) for the subsets of eggs reared in the absence of males. For the regression between egg

diameter and development time, one clutch contained visible fungus and was excluded from this

analysis, as fungus increases the rate of egg development in this species (H. Klug, personal

observation). In both cases, means were taken for the subsets from female 1 and female 2 for a

given male to avoid pseudoreplication.

Results

Differences in Egg Size, Egg Density, and Cannibalism Rates between Years

There was no significant difference between the mean egg diameter of female 1 and of

female 2 in 2006 (paired t-test, t = -0.51, df = 6, p = 0.62; female 1 mean +/- SE egg diameter:

0.61 +/- 0.013 mm; female 2 mean +/- SE egg diameter: 0.62 +/- 0.011 mm). Eggs tended to be

slightly larger in 2006 in comparison to 2004 (2004 mean +/- SE egg diameter: 0.58 +/- 0.018

mm; 2006 mean +/- SE egg diameter: 0.62 +/- 0.011 mm), but this difference was not significant

(independent samples t-test, t = -1.79, df = 10, p = 0.10). Additionally, egg density did not differ

significantly between years (t = -0.12, df = 10, p = 0.91; 2004 mean +/- SE: 1.42 +/- 0.27

eggs/mm2; 2006 mean +/- SE: 1.45 +/- 0.12 eggs/mm2). In 2004 males consumed 32.1 +/- 0.12

% (mean +/- SE) of their eggs, and in 2006 males consumed 36.2 +/- 0.079 % (mean +/- SE) of

their eggs. There was no significant difference in the proportion of eggs cannibalized between

the years (t-test, t = -0.30, df = 10, p = 0.77).









Egg Size, Survivorship, and Development Time in Eggs Reared in the Absence of Males

There was no relationship between egg size and egg survivorship (linear regression, F1, =

0.14, p =0.72). However, egg size was positively correlated with development time, suggesting

that larger eggs take longer to develop (linear regression, F 1,5 = 13.85, p = 0.02; Figure 4-1).

Cannibalistic Preferences by Males

In the single-female scenario, males exhibited no significant size preferences (x2 = 2.25, df

= 3, p = 0.522; Figure 4-2 A). In other words, for each given size class (i.e. small, small-medium,

medium-large, and large), the relative abundance of eggs consumed was comparable to the initial

relative abundance of eggs of that size class (Figure 4-3 A), and this pattern of cannibalism is

consistent with random consumption of eggs with regard to size. However, males exhibited

significant preferences in the multiple-female scenario (x2 = 15.13, df = 7, p = 0.034; Figure 4-2

B), and specifically, males exhibited a significant preference for the larger eggs of female 2

(Figure 4-2 B). In this case, the relative abundance of female 2's medium-large and large eggs

that were consumed was greater than the relative abundance of those eggs that were present

initially (Figure 4-3 B).

Discussion

Male sand gobies cannibalized eggs selectively, but only in some cases. When males

received eggs sequentially from two females, they preferentially consumed the larger eggs from

the second female only. Thus, my results suggest that sand goby males exhibit non-random

preferences for eggs when they spawn sequentially with two females. These patterns raise

several questions. First, why do males prefer larger eggs in some cases, but not others, and in

particular, what distinguishes female 2's larger eggs from those of female 1?

It is possible that energetic benefits play a role in cannibalistic preferences-- larger eggs

likely provide a male with more energy (Kamler 2005), and filial cannibalism is thought to be a









way in which caring parents attain energy to offset costs of care (Rohwer 1978; Manica 2002).

Thus, it wouldn't be surprising if males maximized their per-offspring energetic gain. However,

males didn't always consume larger eggs-- they exhibited no size preference for female l's eggs

in either 2004 or 2006. If males were attempting to maximize their per-offspring energetic gain,

we would have expected them to consume larger eggs in all cases. This was not the case, and

thus, it does not appear that energetic gain alone can explain the patterns of cannibalism

observed in the sand goby.

Alternatively, it is possible that the preference for the larger eggs of female 2 is associated

with decreased duration of care and the ability to re-enter the mating pool sooner. Larger eggs

took longer to develop, and the eggs of female 2 were already several hours (i.e., up to 24 h)

behind those of female 1. Thus, the larger eggs of female 2 would likely hatch later and require a

longer duration of care than female 1's eggs and the smaller eggs of female 2. Perhaps

consuming the larger eggs of female 2 allows a male to decrease time spent caring for the current

brood, thereby allowing him to re-enter the mating pool sooner. This hypothesis seems

particularly relevant for sand gobies, which live only one year and have multiple brood cycles

over a six-week period. During a given brood cycle, males receive eggs for just a few days and

then enter a 'care-only' phase during which they do not receive additional eggs until their current

brood hatches (typically 7 15 days from spawning). Indeed, my results suggest that

preferentially consuming larger eggs potentially reduces the duration of care for a given clutch

by several days (Figure 4-1). It is easy to imagine how even a small reduction in the duration of

care over multiple brood cycles might allow a male an additional brood cycle, which in turn

might increase the total number of eggs he receives over the breeding season. This hypothesis is

consistent with some theoretical work (Chapter 6 and Klug and Bonsall 2007), which suggests









that parental fitness is highly sensitive to the maturation rate of eggs. Specifically, I suggest

(Chapter 6 and Klug and Bonsall 2007) that males potentially benefit by consuming eggs that

take longer to develop, and that the consumption of slower developing eggs can directly facilitate

the evolution of filial cannibalism. However, consuming larger eggs likely comes at a cost, as

egg size is correlated with larva size in the sand goby and larger eggs have been shown to have

higher post-hatching survival in a range of fishes (Kamler 2005). These ideas warrant further

attention, and in particular, more work is needed to understand factors affecting the optimal

duration of care in this species and others.

In summary, I have demonstrated that male sand gobies can cannibalize eggs selectively

with regard to the order in which those eggs are laid and the size of eggs in some contexts.

However, in this case, size per se does not appear to be the factor influencing cannibalism.

Rather, males seem to be sensitive to the expected development rate of eggs. More work is

needed to further understand specific costs and benefits of consuming eggs of a particular size

and expected developmental rate. Additionally, it will be important to assess the relationship

between other aspects of egg phenotype and filial cannibalism in this and other species.















16.00-





r 14.00-
0
E e
-


t a
S 12.00-


E


I 10.00-
E
a
0


8.00-

0.56 0.58 0.60 0.62 0.64 0.66
Mean Egg Diameter (mm)
Figure 4-1. Relationship between initial egg size and development time (i.e. the number of days
from spawning until hatching) in eggs reared in the absence of males.














0.20

0.15

0.10

0.05
0 00

-0 05

-0 10

-0.15

-0.20


S Small
BZ Small-Medium
SMedium-Large
S Large


2004 Female


-0.1


2006 Females
Figure 4-2. Preferences (a) in egg consumption by parental males. Male preferences for A) 4
size classes of eggs (labeled here as small, small-medium, medium-large, and large)
when each male mated with a single female in 2004 and B) 4 size classes of eggs
from two females when males mated sequentially with 2 females in 2006. Bars
represent means and error bars are standard error.


I!


1 N~








Initially
Consumed by male


-





Medium-
Large


0.5
U)
S0.4
4-_
0
O
c 0.3
0
0
0 0.2

0.1


Egg Size


Female 2


Small Small- Medium- Large Small Small- Medium- Large
Medium Large Medium Large


Egg Size

Figure 4-3. Distribution of egg size initially and in the male diet. The mean (+/- S.E.) proportion
of eggs that were either small, small-medium, medium-large, or large initially, i.e. on
Day 1 of each replicate (filled bars), and the mean (+/- S.E.) proportion of eggs
consumed by males (open bars) that were either small, small-medium, medium-large,
or large in A) 2004 and B) 2006.


Large


0.0 -


Small


Small-
Medium


Female 1









CHAPTER 5
SELECTIVE FILIAL CANNIBALISM IN THE FLAGFISH

Introduction

Filial cannibalism commonly co-occurs with parental care in many animals and has been

particularly well-documented in fishes exhibiting paternal care during the egg stage (Polis 1981;

Manica 2002). While parental care typically increases offspring survival (discussed in Clutton-

Brock 1991), filial cannibalism involves the killing of one's own young. It is difficult to imagine

how such a behavior could represent an adaptive strategy. Indeed, prior to the 1970s filial

cannibalism was dismissed as a rare behavior with little or no evolutionary significance

(discussed in Manica 2002). However, filial cannibalism has now been well-documented in

nature, and in many species caring parents consume more offspring than would die naturally. In

recent years, much empirical and theoretical work has focused on understanding the adaptive

significance of filial cannibalism (reviewed in Manica 2002). Currently, filial cannibalism is

thought to be an adaptive strategy, and specifically, some have suggested that filial cannibalism

involves an adaptive trade-off in which parents gain energy or nutrients from eggs, which they

then use to better care for their remaining offspring or re-invest in future reproduction (energy-

based hypothesis: Rohwer 1978; Sargent 1992; reviewed in Manica 2002). According to this

hypothesis, whole clutch cannibalism (i.e., the consumption of all offspring present during a

given reproductive bout) represents an investment in future reproduction, whereas partial clutch

cannibalism (i.e., the consumption of only some offspring present) can either be an investment in

future or current reproduction. Alternatively, others have suggested that by reducing egg density

in the nest, partial clutch filial cannibalism can improve egg survivorship of remaining offspring,

thereby increasing overall egg survival (oxygen-mediated hypothesis: Payne et al. 2002, 2004;

density-dependent egg survival hypothesis: Klug et al. 2006). However, neither energetic need









nor density-dependent egg survival can explain the prevalence of filial cannibalism in natural

systems. Indeed, filial cannibalism is not affected by energetic need in some species (Belles-Isles

and Fitzgerald 1991; Lindstrom and Sargent 1997) and continues to occur when egg density is

relatively low in others (Klug et al. 2006 and Chapter 3). Thus, the evolutionary significance of

filial cannibalism remains unclear in many systems.

Alternatively, recent theoretical work suggests that the ability to selectively cannibalize

offspring that have reduced phenotypic quality can independently facilitate and play a key role in

the evolution of filial cannibalism (Klug and Bonsall 2007 and Chapter 6). The elimination of

lower quality offspring has been demonstrated in relation to selective embryo abortion in humans

and plants (Forbes 1997; Diamond 1987; Burd 1998; Karkkainen et al. 1999), brood reduction

(Mock and Forbes 1995; Mock and Parker 1997; Forbes and Mock 1998), and parents allowing

or encouraging siblicide of low quality offspring (Mock and Parker 1997; Stearns 1987).

Because the elimination of low quality offspring is thought to play a central role in explaining

the evolutionary significance of offspring abandonment and brood reduction (e.g., Stearns 1987,

1992; Forbes and Mock 1998), it is surprising that the relationship between offspring quality and

filial cannibalism has received little attention. While some studies have found a relationship

between filial cannibalism and uncertainty of paternity (Neff 2003; Gray et al. 2007; Frommen et

al. 2007) or egg age (Salfert and Moodie 1985; Sikkel 1994), little is known about the

relationship between offspring quality and filial cannibalism of viable young (see also Kraak

1996, for discussion of cannibalism of diseased eggs). Thus, the general importance of selective

filial cannibalism remains unclear.

The first step in understanding the potential importance of selective filial cannibalism is to

determine whether parents that provide care preferentially consume eggs with regard to some









aspect of offspring quality. Here, I examine the relationship between one aspect of offspring

quality, egg energy content, and filial cannibalism in the flagfish (Jordanellafloridae). Egg

energy content was used as a proxy for quality because energy content and size have been

strongly and positively correlated with post-hatching survival and growth (and hence fitness) in a

range of fish species (reviewed in Kamler 1992, 2005; Keckeis et al. 2000; Brownman et al.

2003). In addition, I examined the relationship between filial cannibalism and maternal condition

and size, because maternal effects on egg quality have been well-documented (Kamler 2005),

and specifically, because a positive relationship between female size and offspring quality has

been found previously in fishes (reviewed in Kamler 2005).

Methods

Study Species

Flagfish males alone provide parental care of eggs (including nest guarding, cleaning, and

fanning), filial cannibalism is prevalent (Klug & St. Mary 2005), and parental males are known

to consume more eggs than die naturally (i.e., the rate of filial cannibalism is greater than the

mortality rate; Klug et al. 2005). Thus, filial cannibalism does not function soley to clean the nest

of dead or diseased eggs. Indeed, egg survival in the absence of parental males and predators is

very high (> 90% egg survival), and thus, most egg mortality can be attributed to parental males.

Flagfish typically live only one year in the wild, and both males and females mate multiply

during a several month breeding season. Additionally, eggs are spawned and fertilized

individually, and sneaking is not thought to occur in this species.

Experimental Design

The study was conducted April July 2005 in Gainesville, Florida. All fish were collected

from the Otter Creek/Waccasassa River drainage in Levy County, Florida within 20 days of the

experiment. Fish of both sexes were housed in separate freshwater holding tanks. All









experimental aquaria were 36 L and equipped with air-driven filtration, a spawning mat (i.e. a

100 cm2 ceramic tile with heavy, green acrylic felt carpet glued to the top of the tile), and three

artificial plants. Throughout the experiment, all fish were fed ad libitum a diet consisting of algae

tablets and frozen brine shrimp. During the experiment, all fish experienced a 14h: 10h light:dark

cycle and temperature was maintained at 260C.

I initiated each replicate by randomly selecting and placing one male and one female in an

aquarium. The male and the female were allowed to spawn, and immediately following

spawning, I briefly removed the nest with eggs from the tank. I counted the number of eggs, and

for a subset of the clutches (N = 18), I removed three eggs from the nest and used them in

subsequent energy assays. No eggs were removed from the nests of six males, which allowed me

to evaluate whether there was an effect of egg removal on filial cannibalism. In all cases, a clear

acrylic divider containing multiple holes was used to physically separate the male and female

following spawning, which ensured that all cannibalism was done by the male. After counting

the eggs, I returned the nest with eggs to the male, who was allowed to care for the eggs until

hatching. I counted the eggs daily by visually inspecting the nest. Eggs usually began to hatch on

day four, and thus I measured egg survivorship through day three of each replicate. Eggs never

became diseased or infected with fungus during the course of the study. I weighed and measured

all experimental fish just prior to the start of each replicate. For five replicates, I did not obtain

reliable weight measurements, and thus these fish were excluded from analyses involving

parental condition or size. I used the condition measure K (where K=100*weight/(length)3;

Williams 2000), which provides a size-independent estimate of physical condition, to evaluate

the relationship between parental condition and the occurrence of filial cannibalism. I also









estimated the relationship between parental size and filial cannibalism, and weight and standard

length (which were highly correlated) were used as estimates of size.

Energy Assays

I used dichromate oxidation technique (modified from McEdward and Carsons 1987) to

quantify total energy content (i.e. J egg-1) of each sampled egg. Egg energy content was

compared to a glucose standard (1 4 J mL-1). Specifically, I incubated each egg in 0.5 mL 70%

phosphoric acid at 1050C for 15 min. I allowed the solution to cool to room temperature, and

then oxidized the sample with 1 mL of 0.3% potassium dichromate in concentrated sulphuric

acid at 1050C for 15 min. Samples were then diluted with 3.5 mL distilled water and I measured

absorbance using a spectrophotometer (X 440 nm). I calculated total energy by comparing the

absorbance of each sample with that of the glucose standards. I performed energy assays on eggs

from 18 clutches. Only intact eggs were used for the energy assays, and I was able to quantify

the total energy of one egg in 12 clutches, two eggs in three clutches, and three eggs in two

clutches. While it would have been ideal to measure a larger sample from each clutch, this was

impossible because flagfish spawn relatively few eggs (typically < 100) and I wanted to

minimize any effects of egg removal. Additionally, because there was greater variation in the

per-egg energy content between clutches than within clutches (discussed in Results), a small

within clutch sample of egg energetic content provided an estimate of the mean energy content

of eggs within a given clutch.

Statistics

I used linear regression to evaluate the relationship between male and female condition

(i.e. K) and size (i.e. weight and length); the relationship between the mean energy content per

egg within a clutch and the number of eggs spawned; the relationship between the number of









eggs spawned or received and female and male condition and size; and the relationship between

mean egg energy content and male and female condition and size.

Whole clutch cannibalism represents a termination of current reproduction, and therefore

any benefit of whole clutch cannibalism is associated with future reproductive success. In

contrast, benefits of partial clutch cannibalism can be associated with either increased current or

future reproductive success. Because whole and partial clutch cannibalism likely represent

different biological phenomena, I analyzed these data separately. First, I used stepwise logistic

regression (remove if p>0.15) to evaluate the relationship between whole clutch cannibalism and

1) male condition (i.e. K), 2) male size (i.e. weight), 3) female condition (i.e. K), 4) female size

(i.e. weight), 5) the number of initial eggs present, and 6) mean energetic content per egg within

a clutch. I then considered only cases of partial clutch cannibalism and used stepwise linear

regression (remove if p>0.15) to evaluate the relationship between the proportion of eggs

consumed and 1) male condition (i.e. K), 2) male size (i.e. weight), 3) female condition (i.e. K),

4) female size (i.e. weight), 5) mean egg energy, and 6) the initial number of eggs present. To

meet assumptions of normality, the proportion of eggs consumed was arcsin square root

transformed. Data associated with the number of eggs consumed could not be transformed to

meet assumptions of normality. Thus, I used spearman rank correlation tests to evaluate the

relationship between the number of eggs consumed and male and female size (i.e. weight) and

condition (i.e. K), mean egg energy, and the initial number of eggs present.

Results

Of the 24 males, 10 exhibited whole clutch cannibalism, 12 exhibited partial clutch

cannibalism, and 2 males consumed no eggs. Excluding cases of whole clutch cannibalism, the

mean (+/- SE) percentage of eggs cannibalized was 38.7 +/- 0.097 % (or 45.2 +/- 0.10 %

excluding males who didn't exhibit any cannibalism). As mentioned previously three eggs were









removed from the nests of 18 males and no eggs were removed from the nests of 6 males. There

was no effect of egg removal (i.e., removal of 3 eggs for subsequent energy assays) on whole

clutch cannibalism (logistic regression, x2 = 0.523, df = 1, p = 0.465) or partial clutch

cannibalism (ANCOVA, F1,12 = 2.20, p = 0.164). On average, eggs contained 1.94 +/- 0.57 J egg

1 (mean +/- SD). As mentioned previously, energy measurements were taken for multiple eggs

for 5 clutches. In these cases, the mean energy content (+/- SD) was 2.01 +/- 1.07 J egg-1. The

standard deviation within a clutch ranged between 0.0094 to 0.089 J egg-1, and was on average

0.042 J egg-1. In all cases, the within clutch variation in energy content was much less than the

between clutch variation in energy content.

Parental Condition and Size, Egg Energetic Content, and Egg Number

There was no significant relationship between male and female condition (F1,19 = 1.84, p =

0.19) or male and female size (weight: F1,19 = 0.91, p = 0.35; length: F1,19 = 0.97, p = 0.34).

Additionally, there was no significant relationship between mean energy content of eggs and the

number of eggs spawned (F1,16 = 0.84, p = 0.37), suggesting that there was not a clear trade-off

between the number of offspring produced and the mean energy invested into those offspring

within a given reproductive episode in this experiment.

There was no significant relationship between female condition and mean egg energy

content (linear regression, F1,25 = 0.37, p = 0.55) or the number of eggs spawned (F1,19 = 1.92, p

= 0.18). Likewise, there was no significant relationship between female size and the number of

eggs spawned (weight: F1,19 = 1.69, p = 0.21; length: F1,19 = 0.88, p = 0.36). However, there was

a significant relationship between female size and the mean egg energy content (weight: r2

0.41, F1,11 = 7.63, p = 0.02; length: r2 = 0.42, F1,11 = 7.92, p = 0.02; Figure 5-1 A).

There was no significant relationship between male condition and mean energetic content

of eggs received (linear regression, F1,11 = 0.08, p = 0.78) or the number of eggs received (F1,19 =









0.53, p = 0.53). Likewise, male size was unrelated to the number of eggs received (weight: F1,19

= 0.70, p = 0.41; length: F1,19 = 1.35, p = 0.26). However, larger males received eggs that were

on average more energetic (weight: r2 = 0.27, F1,11 = 5.33, p = 0.041; length: r2 = 0.33, F1,11=

7.29, p = 0.021; Figure 5-1 B). As mentioned above, there was no relationship between male and

female size, and thus, assortative mating does not explain these patterns.

Whole Clutch Cannibalism

There was no significant effect of initial egg number, male condition, male size, or female

condition on the occurrence of whole clutch cannibalism, i.e., the proportion of clutches that

were entirely eaten (stepwise logistic regression, p > 0.15 in all cases). However, whole clutch

cannibalism was more frequent when the mean energetic content of eggs was relatively great

(logistic regression, x2 = 6.71, df = 1, p = 0.01; Figure 5-2 A) and when female size was greater

(x2 = 4.73, df= 1, p = 0.03; Figure 5-2 B).

Partial Clutch Cannibalism

When only cases of partial clutch cannibalism were considered, there was no relationship

between male condition and the proportion of eggs surviving (p > 0.15), or the number of eggs

consumed (df = 10, t = -0.41, p >0.05). However, larger males tended to consume a smaller

proportion of eggs (F1,s = 3.90, p = 0.10; Figure 5-3 A), and they consumed significantly fewer

eggs than smaller males (weight: df = 10, t = -13.54, p < 0.01; length: df = 10, t = -2.45, p <

0.05; Figure 5-3 B). Female condition was unrelated to the proportion of(p > 0.15) or the

number of eggs consumed by males (df = 10, t = -0.72, p >0.05). However, female size was

negatively correlated with the proportion of eggs consumed (weight: F1,s = 11.77, p = 0.009;

Figure 5-4 A) and the number of eggs consumed (weight: df = 10, t = -8.36, p <0.01; length: df=

10, t = -5.21, p <0.01; Figure 5-4 B).









For cases of partial clutch cannibalism, egg energy was unrelated to the proportion of eggs

consumed (p > 0.15), but there was a negative relationship between mean egg energy and the

number of eggs consumed (df = 8, t = -6.49, p <0.01; Figure 5-5). There was no significant

relationship between the number of eggs initially present and the proportion of eggs (p > 0.15) or

the number of eggs (df = 12, t = 0.0266, p > 0.05) consumed.

Discussion

Male flagfish preferentially cannibalized eggs laid by females of larger body size and when

the mean energy content of eggs was high for cases of whole clutch cannibalism. Because egg

size, egg energy, and maternal size are correlated with post-hatching survival and growth in

fishes (reviewed in Kamler 2005), it appears that males are sacrificing high quality offspring for

a relatively large energetic gain when they practice whole clutch cannibalism. With regard to

partial clutch cannibalism, the number of eggs consumed increased as the mean energy content

of eggs decreased. The energy-based hypothesis of filial cannibalism (Rohwer 1978; Sargent

1992) suggests that partial clutch cannibalism is a way in which males attain energy to offset

costs of care (Manica 2002). If the function of filial cannibalism is to attain energy that can be

reinvested in increased current or future reproduction (which I did not evaluate here), it seems

likely that a male's energetic need can be satisfied by consuming a smaller number of eggs when

those eggs have a relatively high energetic content. However, males in this experiment received

food (i.e., algae and brine shrimp) ad libitum, and thus energetic need alone cannot explain the

filial cannibalism observed in the present study. Alternatively, because egg energy content is

likely correlated with subsequent offspring survival (Kamler 2005), it is possible that the

negative relationship between the number of eggs consumed and mean energetic content of eggs

suggests that males are investing more into offspring of relatively high quality. Similarly, males

consumed a greater proportion of eggs spawned by relatively small females. While female size









was correlated with mean egg energy content, egg energy alone did not explain the proportion of

eggs cannibalized. Because female body size is positively correlated with egg size and resistance

to starvation and predation in a range of fishes (reviewed in Kamler 2005), it appears that males

consumed a greater proportion of relatively low quality eggs when they practiced partial clutch

cannibalism.

In summary, this experiment suggests that at least some aspects of offspring quality (i.e.

egg energy, maternal size) affect both whole and partial clutch filial cannibalism, albeit in

different ways. With regard to partial clutch cannibalism, males consume more of the lower

quality offspring. This finding is consistent with previous work on the elimination of offspring

via abandonment, siblicide, or infanticide, which focuses on the removal of low quality offspring

(e.g., Mock and Parker 1997; Forbes and Mock 1998). Additionally, this finding is consistent

with theoretical work suggesting that selective filial cannibalism of low quality offspring is

beneficial to caring parents (Klug and Bonsall 2007 and Chapter 6).

However, with regard to whole clutch cannibalism, males are more likely to cannibalize

higher quality offspring. This finding suggests that males are sensitive to the nutritional benefits

of cannibalism, which is consistent with the idea that cannibalistic parents use eggs as an

alternative food source (Rohwer 1978; Manica 2002). However, the specific finding that males

sacrificed their higher quality offspring for increased energetic gain is not explicitly predicted by

current theory. Indeed, previous work in the flagfish (Klug and St. Mary 2006 and Chapter 2)

suggests that the energetic benefits of consuming eggs to male size or reproduction are relatively

small in comparison to those of food. Thus, additional theoretical and empirical work is needed

to better understand the expected trade-offs associated with filial cannibalism. In future studies,









it will be important to consider within-clutch patterns of cannibalism to better understand

patterns of parental investment and filial cannibalism.

Finally, there was no relationship between initial egg number and whole clutch

cannibalism in the present study. This finding is in contrast to some previous theoretical

predictions (Rohwer 1978) and empirical findings (reviewed in Manica 2002) suggesting that

whole clutch cannibalism is more common when clutch size is relatively small (but see Payne et

al. 2003, who found that smaller clutches were not preferentially consumed). Similarly, there

was no clear trade-off between the number of eggs spawned and the mean energetic content of

those eggs. This trade-off is a key assumption of life-history evolution (Stearns 1987). However,

it is likely that the scale of the present experiment, as well as other sources of variation, made it

difficult to detect such a trade-off if one does indeed exist in the flagfish. Indeed, better

understanding of female investment will necessitate an experiment specifically designed to

assess such trade-offs over a longer time frame.











A 3.50-





-
,, 3.00-
t-


2.50-
-
LO

52.00-
C

S1.50-



1.00-






B 3.50-



3.00-
-

2.50-
w
0)
u 2.00-
C

S1.50-



1.00-


o o


I I I I I


0.80 1.00 1.20 1.40 1.60
Female Weight (g)


0


I I I I I I
1.00 2.00 3.00 4.00 5.00 6.00
Male Weight (g)

Figure 5-1. Relationship between the mean energy per egg (J egg-') within a clutch and A)
female weight and B) male weight.


10
1.80 2.00














0 O


0.9 1A 1.9 2.4 2.9 3A 3.9


Mean Energy Content of Eggs (J)


0 O0 00 0


1 2 3


Female Weight (g)


Figure 5-2. Relationship between the frequency of whole clutch cannibalism and A) the mean
energy per egg (J egg') within a clutch and B) female weight. Lines represent the
predicted probability of whole clutch cannibalism as a function of A) mean egg
energy or B) female weight, as determined by a logistic regression; for A)
1 1
y = 1- -15 3e+797.x and (B) y = + e 5871 86x where y is equal to the

probability of whole clutch cannibalism and x is equal to either mean egg energy A)
or female weight B).


O
o




-cE
o
a -




0)

u-


B_.
o



o4-
0
E

.





Uc
S0


LL
*o











A 1oo0-

'o
-

E
S080-

0

*, 060-


U
4-
0 040-
C
0
,i
O 20-
t0
0
L

000-


E
30
"U
C
0
U
U)
020
LU
0
4-

10
1-
E
3
z


I I I I
1.00 200 300 400
Male Weight (g)


1.00 2.00 3.00 4.00
Male Weight (g)


Figure 5-3. Relationship between male weight and A) the proportion and B) the number of eggs
consumed for cases of partial clutch cannibalism.


10
5 00 600


5.00 6.00











1.00-

E

M 080-
O8-
C
0
U

S0.60-


U
4- 0.40-
C
0

t0 20-
0
O

0.00-


I I I I I
0.80 1.00 1.20 1.40 1.60
Female Weight (g)


0
o o


I I I I I
0.80 1.00 1.20 1.40 1.60
Female Weight (g)


0 0


Figure 5-4. Relationship between female weight and A) the proportion and B) the number of
eggs consumed for cases of partial clutch cannibalism.


0 0


E
30-
-,
C
O
0




O
C-
gj20-






3
4-
0



I-
0-































I I I I


1.00 1.20 1.40
Mean Egg


I I
1.60 1.80
Energy (J)


Figure 5-5. Relationship between the mean energy per egg (J egg-') within a clutch and the
number of eggs consumed for cases of partial clutch cannibalism.


40-
'a
E
30-
0

-)
020-
LU
4-

Q 10-
E
3
z
0-









CHAPTER 6
WHEN TO CARE FOR, ABANDON, OR EAT YOUR OFFSPRING: A MODEL OF THE
EVOLUTION OF PARENTAL CARE AND FILIAL CANNIBALISM

Introduction

Adaptive theories of evolution typically suggest that parents should exhibit strategies that

increase offspring survival, and parental care is one way in which parents are thought to achieve

this (reviewed by Clutton-Brock 1991). Although parental care is assumed to increase offspring

survival, filial cannibalism, the consumption of one's own viable offspring, commonly co-occurs

with parental care. Indeed, filial cannibalism is prevalent in a range of taxa exhibiting parental

care (Polis 1981; Elgar and Crespi 1992). For example, caring females consume some of their

young in the bank vole (Cleith ii,,iny% glareolus, Klemme et al. 2006), the house finch

(Carpodacus mexicanus, Gilbert et al. 2005), and the wolf spider (Pardosa milvina, Anthony

2003), and both parents of the burying beetle (Nicrophorus orbicollis) are known to consume

their offspring (Bartlett 1987). Filial cannibalism has been particularly well-documented in fish

species with paternal care during the egg stage (reviewed in Manica 2002). Indeed, because of its

prevalence in fish systems, most theoretical and empirical work on filial cannibalism has focused

on fish (but see Bartlett 1987, Thomas and Manica 2003, Creighton 2005). While early

ethologists considered filial cannibalism a social pathology with little or no evolutionary

significance, filial cannibalism is now typically thought to reflect an adaptive trade-off between

current and future reproductive success (e.g., Manica 2002, 2004). However, despite much

theoretical development and empirical work over the last few decades, the evolutionary

significance of filial cannibalism remains unclear in many systems.

The most widely accepted hypothesis of filial cannibalism as an adaptive strategy suggests

that energetic need is the primary factor leading to filial cannibalism, and that a caring parent

gains energy and nutrients from consuming their offspring that are then reinvested into future









reproduction, thereby increasing net reproductive success (Rohwer 1978; Sargent 1992).

Specifically, whole-clutch cannibalism (i.e., the consumption of all offspring during a given

reproductive bout) is assumed to be an investment in future reproduction, whereas partial-clutch

cannibalism (i.e., the consumption of only some offspring present) can represent an investment

in either current or future reproduction. This energy-based hypothesis predicts that cannibalism

will increase as food availability decreases and when parental condition is poor (Rohwer 1978;

Sargent 1992). While food availability and/or parental condition affect the amount of

cannibalism in some species (e.g., Stegastes rectifraenum, Hoelzer 1992; Pomatoschitus

microps, Kvarnemo et al. 1998; Abudefdufsexfasciatus, Manica 2004), it has no effect in others

(e.g., Gasterosteus aculeatus, Belles-Isles and Fitzgerald 1991; Etheostomaflabellare,

Lindstrom and Sargent 1997), and in two systems cannibalism declines as male condition or food

availability decreases (Pomatoschistus minutus, Klug et al. 2006 and Chapter 3; Jordanella

floridae, Klug and St. Mary 2005 and Chapter 2). Other studies have examined whether eggs can

provide a caring parent with sufficient energy to offset the costs of care. Again, the evidence is

mixed-- two studies concluded that energy attained from filial cannibalism is sufficient to offset

costs related to care (Kume et al. 2000; Thomas and Manica 2003), while in another, energy

from eggs was found to be insufficient (Smith 1992). Thus, parental energetic need alone cannot

explain the prevalence of filial cannibalism.

Alternatively, Payne et al. (2002) and Klug et al. (2006 and Chapter 3) suggested that filial

cannibalism is mediated by density-dependent egg survivorship, and that by consuming some

eggs in their nests, caring parents can improve the survivorship of the remaining eggs and

increase their net reproductive success. Such density-dependent egg survivorship is potentially

related to the physical environment (e.g., oxygen availability, Payne et al. 2002) or increased









benefits of parental care to the remaining offspring. The hypothesis of filial cannibalism

mediated by density-dependent egg survivorship has received support in two marine fish species

(Stegastes leucostictus, Payne et al. 2002; Pomatoschistus minutus, Klug et al. 2006 and Chapter

3), but has in general received little further empirical or theoretical examination (but see Payne et

al. 2004). Likewise, some have suggested that filial cannibalism is a mechanism by which

parents reduce brood size in response to anticipated resource competition amongst their adult

offspring (Bartlett 1987; Creighton 2005) or kill offspring of reduced quality (Forbes and Mock

1998; see also Kozlowski and Stearns 1989). While the former hypothesis has received some

attention in the burying beetle (Creighton 2005), neither of these hypotheses of filial cannibalism

has been explicitly evaluated.

Because of the mixed empirical support for the energy-based hypothesis and the lack of

empirical evidence regarding alternative hypotheses, filial cannibalism remains an evolutionary

conundrum. Indeed, previous work suggests that a parent's energetic need (Rohwer 1978;

Sargentl992; Manica 2002), expectations regarding offspring survival or reproductive value

(Payne et al. 2002; Neff 2003; Klug et al. 2006 and Chapter 3), competition for mates (Sikkel

1994; Kondoh and Okuda 2002), and anticipated offspring resource competition (Creighton

2005) are potentially important factors for explaining the adaptive significance of filial

cannibalism. However, previous theory has tended to focus on each of these factors in separate

theoretical contexts (e.g., energetic benefits of consuming offspring: Rohwer 1978; Sargent

1992; expectations regarding offspring survivorship: Payne et al. 2004; variation in offspring

quality: Forbes and Mock 1998; mate availability: Kondoh and Okuda 2002), despite empirical

evidence suggesting that it is unlikely that any single factor alone can explain the prevalence of

filial cannibalism (e.g., Manica 2004; Klug et al. 2006 and Chapter 3).









Here, I develop a model of parental care, total offspring abandonment (i.e., no care), and

filial cannibalism to begin to isolate the pivotal factors affecting the evolution of care and filial

cannibalism. First, I determine the general conditions under which we would expect these

strategies (i.e., care, no care/total abandonment, filial cannibalism) to evolve alone or in

combination. I then evaluate the plausibility of multiple alternative hypotheses within a single

theoretical context by assessing the importance of a range of potential costs and benefits of care

and cannibalism. Specifically I focus on costs and benefits related to energetic, offspring

survival and quality, mate competition, and general resource competition.

Methods

The model is set up as an ecological problem in which a rare mutant with a unique life-

history strategy is allowed to invade a resident population (e.g., Vincent and Brown 2005).

Specifically, the resident strategy represents the strategy that is currently exhibited by individuals

in a population, and the mutant strategy is some alternative strategy not currently exhibited by

individuals in the population. I assume that the resident strategy is in equilibrium (i.e., it is the

strategy that currently prevails in the population) and that an alternative mutant strategy invades

from rare into the population.

Because the general characteristics of an organism in one life-history stage (e.g.,

maturation rate, juvenile survival, adult mortality during the egg, juvenile, and adult stages)

potentially affect the costs and benefits of strategies occurring in another life-history stage, I

assume a stage-structured system in which individuals develop through an egg stage and a

juvenile stage, and then mature and reproduce as adults. While in the egg stage, individuals can

either be abandoned by parents, receive parental care, suffer filial cannibalism, or receive

parental care and suffer filial cannibalism (Figure 6-1). Below, I outline the dynamics of a

system in which a mutant with care and/or cannibalism invades a resident population that either









lacks or provides parental care. I then use mutual invasion analysis to explore the effects of costs

and benefits of varying strategies on lifetime fitness and the evolution (i.e., invasion from rare

and subsequent fixation) of parental care and/ or filial cannibalism.

Model Dynamics

I consider a stage-structured system (which is appropriate for many fish, bird, and insect

systems) in which individuals pass through an egg (E), juvenile, and adult stage (A). The number

of eggs increase as adults reproduce and decrease as eggs mature and as eggs die, such that:

S= r.A(t).[ A-( -dE E(t) mE E(t), (6-1)
dt K

where r represents the rate of egg fertilization (i.e., a proxy for mean reproductive rate of adults),

dE represents death rate of eggs, and mE is the rate at which eggs mature. I assume logistic

population growth, where K represents population carrying capacity, and density-dependence

associated with resource competition affects adult reproduction (i.e., the rate of fertilization).

Adults in the population increase as eggs mature and survive the juvenile stage, and decrease as

adults die, such that:

dA
dt m E(t- r)-.o dA, A(t), (6-2)


where r is a time delay representing the juvenile stage, c, is survival rate through the juvenile

stage, and dA is the death rate of adults (Figure 6-1).

As mentioned above, I assume that the population exhibiting the resident strategy is in

equilibrium. Specifically, the resident strategy is assumed to be fixed in the population (and on

average, all individuals therefore have relative fitness equal to 1). Because 1) the resident

strategy is fixed and 2) the resident population (which exhibits the strategy of interest) is

regulated by density-dependence, the resident population is assumed to be in ecological









equilibrium (i.e., the population density is not increasing or decreasing). Because we know that a

population in equilibrium is not increasing or decreasing (i.e. A(t) and E(t) both equal zero), I can

analytically solve for the equilibrial densities by setting A(t) equal to E(t), and then algebraically

solving for the adult and egg densities at equilibrium (A and E*) The equilibrial densities in

this model are thus

d .A*
E* (6-3)
ME Oj
(dE -d dA

A*= K 1- ME J J J(6-4)
r

Resident and Mutant Trade-Offs

To explore the fixation of different strategies, I allowed rare mutants with different life

histories to invade a resident population. I considered the following cases: 1) a rare mutant who

provides parental care invades a resident population with no care (and no cannibalism), 2) a rare

mutant who practices filial cannibalism invades a resident population with no care (and no

cannibalism), 3) a rare mutant who provides parental care and practices filial cannibalism

invades a resident population with no care and no cannibalism, and 4) a rare mutant who

provides parental care and practices filial cannibalism invades a resident population that provides

parental care (but does not cannibalize). The different life-history strategies are represented

through the incorporation of appropriate trade-offs into the model (described below and in Table

6-1), and the model was analyzed using linear additive trade-offs and non-linear trade-offs

(Table 6-1).

In cases in which parental care was provided (either by individuals in the resident

population and/or by the rare mutant), I assumed that parental care increases the survivorship of

eggs (i.e., as dE decreases, parental care increases) and that receiving parental care during the









egg stage increases an individual's likelihood of surviving through the juvenile stage (i.e., the

level of care received as an egg affects quality such that U, increases as dE decreases). Providing

parental care is assumed to be costly to the parent providing it, and thus, I assumed that the

reproductive rate of adults (i.e., their rate of producing fertilized eggs) decreases and that the

death rate of adults increases as care increases (i.e., r decreases and dA increases as

dE decreases). Furthermore, in all cases in which care is provided, a decrease in the maturation

rate of the eggs was associated with an increase in the reproductive rates of adults: the less time

that an individual has to spend caring for a clutch of eggs, the greater that individual's

reproductive rate will be (i.e., as mE decreased, r increased).

When I considered the case of a rare mutant practicing filial cannibalism, I assumed some

energetic benefit of cannibalism, such that the death rate of adults decreased and the reproductive

rate of adults increased as cannibalism increased (i.e., dA decreases and r increases as the rate of

cannibalism, fl, increases). A further goal of these analyses was to determine if cannibalism

could evolve in the absence of a substantial benefit of cannibalism. In these analyses, I assume

no direct benefit of cannibalism (i.e., there is no effect of f on dA or r).

I analyzed the model both assuming that egg survivorship was density-independent and

for the case in which egg survivorship was density-dependent. For the cases in which egg

survivorship was assumed to be density-dependent, the death rate of eggs follows an increasing

function in E, and I considered two functions:

dE *E2, and (6-5)
dE (1+ E)-', (6-6)
where o) is the strength of density-dependence (following Bellows 1981). I considered these two

particular density-dependent functions because they are very general forms of density









dependence that are common in nature (Bellows 1981). The first function (eqn. 6-5) assumes that

egg mortality increases exponentially as egg density increases; the second equation (eqn. 6-6)

assumes that egg mortality increases with increasing egg density, but the increase is not

exponential and the precise nature of this relationship is determines by co. Specifically, a

relatively large co represents more intense density-dependence (i.e., a relatively great increase in

mortality as density increases) and a relatively small co represents relatively weak density-

dependence (i.e., a relatively small increase in mortality as density increases).

Invasion Dynamics and Fitness

Under the density-independent egg survivorship scenario and the assumptions given

above, the dynamics of the rare mutant are given by the following equations and by

incorporating relevant trade-offs (Table 6-1):

dE A *
S= rm Am(t) 1-- d- E(t -dmE,,-.Em(t) Em(t). A (t) (6-7)
dt Km
dam
d mEL Em (t r)fQ ,, dAm Am (t), (6-8)
dt
where /is the mutant's rate of cannibalism (/ equals an average rate of cannibalism, which

could represent some combination of whole- and partial-clutch cannibalism; = 0 if the mutant

does not cannibalize). Specifically, the rate of cannibalism is a function of the number of rare

mutants in the population (Am) and then number of eggs the mutant has (Em). If the mutant does

not cannibalize, /= 0. The mutant is assumed to be rare in the population, and thus, density-

dependence operating on adult mutant reproduction occurs through competition with the

resident.

To evaluate the life-history characteristics and trade-offs affecting the invasion of a rare,

novel strategy, I calculated the fitness of individuals exhibiting the mutant strategy relative to the

fitness of individuals exhibiting the resident strategy. The stage-structured nature of the model









and the time delay representing the juvenile stage makes it impossible to compute relative fitness

from the differential equations above (eqns. 6-7 and 6-8). Instead, the lifetime fitness of the

mutant can then be found from the determinant of the matrix describing the mutant's invasion

dynamics:

A*
A+ d_ + mm, +8 -* 1---
S dE (6-9)
mEexp(-A r) c, +dAm
Hence, while some life-history parameters (e.g., fertilization rate of eggs, r) will be

correlated with lifetime fitness under some scenarios, the real measure of fitness in this model is

the eigenvalues of the invasion matrix. The eigenvalues represent the fitness (and hence the

invasibility) of the mutant strategy relative to the resident strategy when both evolutionary

factors (e.g., trade-offs between current and future reproduction) and ecological factors (e.g.,

resident population density, the intensity of competition amongst adults) are considered. To

evaluate the invasion and replacement dynamics of a rare mutant that provides parental care

and/or practices filial cannibalism, I used the fitness function of the mutant to calculate the

evolutionary stable states) (i.e., when the rate of change in fitness is zero). I then performed

mutual invasion analyses by evaluating when the fitness function is greater than zero (using a

Newton-Raphson algorithm with the resident dynamics (A*) set at equilibrium) for different

values of a life-history trait (see Table 6-1 and Online Appendix in Klug and Bonsall 2007). I

evaluated and present pairwise invasion boundaries for different values of maturation rate of

eggs. Comparing the invasion potential with regard to maturation rate of eggs is ideal because: 1)

it allows me to represent a wide range of life-history strategies, including faster and slower

reproducers, and 2) preliminary results suggest that lifetime fitness is highly sensitive to

maturation rate. Specifically, I illustrate the conditions for which 1) the mutant would invade and

out-compete the resident, 2) the boundaries for which the resident would invade and out-compete









the mutant, 3) the putative coexistence range, in which the strategies have the potential to

coexist, 4) a region of non-persistence, where neither strategy will persist (i.e., a region of

extinction), and 5) a region in which neither strategy will persist or initial conditions of the

model determine the strategy that invades. Local stability analyses were performed and are

described in the Online Appendix of Klug and Bonsall (2007), and numerical simulations were

performed to confirm that strategy coexistence occurs when the dynamics are stable and that

regions of parameter space exist (labeled NP/IC and NP in Figures 6-2 through 6-6) where either

neither strategy persists or the outcome is based on initial conditions. I evaluated the invasion

potential of the rare mutant for several biologically relevant scenarios by changing the values)

of a single life-history parameter of interest for the mutant and/or resident populations.

Biologically Relevant Comparisons

In addition to the fixed trade-offs reflecting varying life-history strategies (Table 6-1), I

explicitly considered the effects of varying selective regimes (e.g., differential mating success

associated with a particular strategy, effects of care or cannibalism on population resources and

hence carrying capacity) on the invasion dynamics. To do this, I used pairwise comparisons in

which I altered the magnitude of one (or more) parameters) to reflect a biological scenario of

interest.

First, I evaluated the importance of offspring survival benefits of care on the invasion

patterns of varying strategies. Empirically, parental care has been shown to reduce the death rate

of offspring (discussed in Clutton-Brock 1991), and thus I compared the invasion patterns of the

caring mutant for a range of cases in which care was effective (i.e., dE < dE ) to those in which

it was ineffective (i.e., dE, = dE ).









Second, sexual selection has been hypothesized to be a major force in the evolution and

fixation of parental care (Andersson 1994; Baylis 1981). In some systems mate choice for a

partner who will provide care is thought to affect the reproductive rate of the non-limiting sex

(e.g., the number of eggs a caring parent receives per reproductive bout or over the course of the

breeding season is correlated with parental care: Jordanellaflordiae, St. Mary et al. 2001;

Pomatoschistus minutus, Pampoulie et al. 2004). Likewise, filial cannibalism has been shown to

increase the attractiveness of a caring parent's nest in some cases (e.g., Sikkel 1994) and might

be preferred during mate choice if there are benefits of cannibalism to remaining offspring (e.g.,

through density-dependent egg survivorship). With regard to the model, if care or filial

cannibalism is a trait that is preferred by one sex, we would expect the mutant exhibiting care or

cannibalism to receive more fertilizations per time period (e.g., a breeding season or lifetime)

than a resident who does not exhibit care. In this sense, rm (i.e., egg fertilization rate in eqn. 6-7)

of a mutant who exhibits a preferred trait would be expected to be greater on average than that of

a resident who does not exhibit the preferred trait. To incorporate this aspect of sexual selection,

I compared invasion patterns for cases in which caring and/or cannibalism increased the

reproductive rate of the caring mutant relative to the resident (i.e., rm > r) to those in which the

magnitude of the reproductive rate did not differ between the mutant and resident (i.e., r, = r).

Likewise, it is possible that filial cannibalism creates reproductive conflict between

parents. In this case one would expect non-cannibalistic individuals to be favored during mate

choice (Kraak 1996; Lindstrom 2000). To assess the importance of mate preference for a non-

cannibalistic partner, I compared cases in which the mutant has a reduced reproductive rate

relative to the resident (i.e., rm < r), with the case in which the mutant and resident have equal

reproductive rates (i.e., r, = r).









Parental care, no care, and filial cannibalism might affect population-level resources, and

hence the competitive regime that individuals experience, in different ways. For example,

providing care might necessitate greater per capital resources (e.g., increased energetic need and

nesting resources per individual) than not providing care (i.e, K > Km for a caring mutant

invading a resident without care). Similarly, the ability to cannibalize while providing care might

represent a more efficient use of resources by individuals. Such an individual-level change in

resource use would be reflected in the carrying capacity of a population of individuals exhibiting

a particular strategy. Specifically, if a strategy allows individuals to use resources more

efficiently (i.e., increase their productivity), we would expect an increase in the carrying capacity

of a population of individuals exhibiting that strategy (relative to individuals who do not exhibit

the strategy of interest, i.e., K < Km for a mutant who can cannibalize and care invading a

resident who can only care). In this sense, the population-level carrying capacity K associated

with a strategy is a proxy for the effect that strategy has on individual-level resource use. To

begin to evaluate the importance of such resource-related effects of varying strategies, I compare

cases in which carrying capacity varies between the mutant and the resident population.

Likewise, for cases in which the mutant and residents have equal carrying capacities, I compared

patterns for a range of carrying capacities to determine if a relatively productive ecosystem (i.e.,

system with a large carrying capacity) or unproductive ecosystem (i.e., system with a relatively

small carrying capacity) favors the invasion of a particular strategy.

Finally, to evaluate further patterns of cannibalism evolution, I compared the fitness

boundaries for cases in which parents were allowed to selectively cannibalize eggs with reduced

future survivorship (i.e., dEm < dE and ca, > a, ) with those of parents that could not

selectively cannibalize (i.e., dEm = dE and ac = U ).









Results

All of the strategies considered (i.e., parental care, no care/total offspring abandonment,

filial cannibalism) evolved over a range of parameter space in all analyses. While the evolution

of parental care and/or filial cannibalism were favored by benefits to adults and/or offspring

(discussed below), such benefits were not essential for the invasion of a particular strategy,

highlighting the plausibility of a range of non-mutually exclusive alternative hypotheses (Table

6-2). In all cases considered, the coexistence dynamics were stable (Online Appendix of Klug

and Bonsall 2007). While incorporating non-linear trade-off functions (Table 6-1) into the model

altered the results quantitatively, there were no qualitative effects of these functions (i.e., the

patterns were the same), and thus, I only present results in which linear trade-offs were used.

Invasion of Parental Care

Effects of egg maturation rate, egg death rate, adult reproductive rate, and carrying

capacity: A mutant with parental care invaded or coexisted with a resident population lacking

care over a wide range of life-history parameters (Figure 6-2 A), particularly when care was

effective at decreasing the death rate of eggs (i.e., when dEm < dE), when it increased

survivorship through the juvenile stage (i.e., when OJm > J ), and when caring increased

maturation rate of the eggs (i.e, when mEm > mE, Figure 6-2 A) relative to the non-caring

strategy. Similarly, the range over which care invaded or coexisted with no care increased when

parental care was associated with an increased rate of egg fertilization (e.g., if it was a preferred

trait, such that rm > r; Figure 6-2 A versus 6-2 B) and when care was associated with a decreased

carrying capacity relative to the resident population (Figure 6-2 A versus 6-2 C). The evolution

of parental care was relatively insensitive to changes in carrying capacity for cases in which the

resident and mutant had equal carrying capacities.









Effect of cannibalism on the evolution of care: To evaluate whether the ability to

practice filial cannibalism affects the evolution of parental care, I compared the case in which a

mutant with only parental care was allowed to invade a resident population with no parental care

and no cannibalism, with the scenario in which the mutant could care and cannibalize (Figure 6-2

A versus 6-2 D). Indeed, filial cannibalism facilitated the evolution of care (Figure 6-2 D). When

the caring mutant was allowed to cannibalize (Figure 6-2 D), parental care (and filial

cannibalism) evolved over a wider range of parameter space and coexisted more often with no

care than when the mutant was not allowed to cannibalize (Figure 6-2 A).

Invasion of Filial Cannibalism (With and Without Parental Care)

Effects of Egg Maturation Rate, Reproductive Rate, and Selective Cannibalism:

Parental care with filial cannibalism was more likely to invade and/or coexist with a state of only

care when practicing filial cannibalism increased the maturation rate of eggs over what would be

achieved by only providing care (i.e., when mEm > mE, Figure 6-3 A), and when filial

cannibalism allowed a parent to improve the quality of care provided for remaining offspring

(i.e., cannibalism decreases dEm relative to dE). Care and cannibalism invaded and coexisted

more often when filial cannibalism increased the reproductive rate of the caring parents) (e.g.,

care and/or cannibalism are preferred, such that rm > r, Figure 6-3 A versus B), but invaded less

often if it decreased the reproductive rate of adults (e.g., non-cannibalism is preferred, such that

rm < r; Figure 6-3 A versus C). Similarly, care with cannibalism evolved more often when

parents could selectively cannibalize their offspring. Specifically, if parents cannibalized

offspring with a higher egg death rate and a lower juvenile survival rate (Figure 6-3 D),

cannibalism invaded more often than in cases in which parents were not capable of selectively

cannibalizing (Figure 6-3 A). These patterns were consistent when we considered parental care









and filial cannibalism evolving from a state of care or a state of no care. Likewise, filial

cannibalism (without care) invaded and/or coexisted with no care/no cannibalism over a greater


range of parameter space if filial cannibalism improved offspring survival (i.e., dEm < dE, or

Jm, > ), increased the maturation rate of eggs (mEm > mE), or when parents were able to

practice selective filial cannibalism of offspring that had reduced survival during the egg and/or

juvenile stage.

Effects of Density-Dependent Egg Survivorship

When egg survivorship was density-dependent, parental care and/or filial cannibalism

evolved and co-existed over a wide range of parameter space. However, in the absence of any

other benefits of filial cannibalism, density-dependent egg survivorship alone did not facilitate

the evolution of cannibalism. In fact, allowing a mutant that provides care to cannibalize

decreased the range over which care and cannibalism evolved, in comparison to the case in

which the mutant could not cannibalize (Figure 6-4 A versus B). However, when the mutant

provided care and cannibalized, parental care and cannibalism invaded over a greater range of

parameter space as the strength of density-dependence (i.e., o) ) increased (Figure 6-4 B versus

C). In other words, the evolution of filial cannibalism was not facilitated by density-dependent

egg survivorship per se, but relatively intense density-dependence (i.e., a relatively large increase

in egg mortality with increasing egg density) allowed cannibalism to evolve more often in

comparison to weaker density-dependence (i.e., a relatively small increase in egg mortality with

increasing egg density; as o in eqn. 6, increased, the range over which care with cannibalism

and no care could invade and/or coexist increased) These patterns were the same for both

density-dependent functions considered.









Effects of Energetic Benefits of Consuming Offspring

Energetic benefits of filial cannibalism (i.e., benefits associated with dAm and r,)

increased the range over which care and cannibalism could invade and coexist with no care

(Figure 6-5 A versus B). However, in this scenario (i.e., considering care and cannibalism

invading from a state of no care or cannibalism) it is possible that cannibalism could be thought

of as simply hitchhiking in with care. Thus, I considered the case in which cannibalism and care

invaded a resident who already provides care. While energetic benefits of cannibalism to adult

reproduction and survival increased the range of invasion and coexistence (Figure 6-5 C), filial

cannibalism (which in this case is equivalent to simply killing offspring, or abandoning offspring

that have no chance of surviving alone, during the course of care) was still able to invade in the

absence of benefits to adults (Figure 6-5 D) over the range of parameters considered.

Effects of Carrying Capacity

For cases in which cannibalism alters the efficiency with which individuals use resources

(and hence, population carrying capacity), filial cannibalism with parental care was more likely

to evolve from a state of only care if cannibalism increased the resource-use efficiency of a

population of individuals exhibiting that strategy (i.e., the carrying capacity). In other words,

filial cannibalism was more likely to invade if it somehow increased the productivity of the

system (Figure 6-6 A versus B). In contrast, for the case of filial cannibalism with parental care

evolving from a state of no care, cannibalism and care were more likely to evolve if they were

associated with a decrease in the population carrying capacity (i.e., if care with cannibalism

decreased the productivity of the system and the efficiency with which individuals use resources,

Figure 6-6 C versus D). Likewise, for the case of filial cannibalism (without parental care)









invading a resident state of no care/no cannibalism, filial cannibalism invaded and coexisted over

a greater range of parameter space if it increased carrying capacity (i.e., resource-use efficiency).

If carrying capacities were equal for the resident providing parental care and the mutant

providing care and practicing filial cannibalism (i.e., the efficiency of resource use was

equivalent for individuals exhibiting mutant and resident strategies), cannibalism invaded and/or

coexisted more often when carrying capacity was relatively low (i.e., in relatively unproductive

systems, Figure 6-6 A versus E). This trend (i.e., more invasion and coexistence of the mutant at

lower carrying capacities) was consistent for the case in which filial cannibalism (with no

parental care) invaded a resident with no care/no cannibalism. However, when I considered the

case in which cannibalism with care invaded a state of no care (and assumed the carrying

capacities were equal for populations exhibiting the resident and mutant strategies), cannibalism

and care were more likely to invade or coexist with no care/no cannibalism when carrying

capacity was relatively large (i.e., when the system was relatively more productive, Figure 6 C

versus F).

Discussion

I have shown that parental care, filial cannibalism, and no care/total offspring

abandonment can evolve over a wide range of life-history parameters. My results suggest that the

ability to abandon or consume offspring during the course of parental care can actually facilitate

the evolution of parental care, and that offspring abandonment/no care, parental care, and filial

cannibalism often have the potential to coexist. Even in the absence of direct benefits of filial

cannibalism, such as energetic gain or increased survival of remaining offspring, filial

cannibalism invaded (and coexisted with) non-cannibalistic strategies in multiple contexts (i.e.,

with or without care, across varying resident strategies, over a range of life-history parameters).

In the absence of such benefits, cannibalism is simply equivalent to killing (or abandoning









offspring that will subsequently die) during the course of care. My results suggest that the

evolutionary dynamics of filial cannibalism are likely comparable to those of simple offspring

abandonment (which provides no immediate benefits, such as energetic gain, to parents).

However, my results suggest that the evolution and fixation of filial cannibalism is favored by a

variety of evolutionary and ecological factors. While no single benefit of consuming eggs was

essential for the invasion of filial cannibalism to occur, several potential benefits facilitate the

evolution of filial cannibalism.

In particular, my model highlights the plausibility of several non-mutually exclusive

alternative hypotheses favoring the evolution filial cannibalism (Table 6-2). The ability to

selectively cannibalize eggs facilitated the evolution of cannibalism in all contexts. The ability to

cannibalize offspring selectively allows parents to alter the phenotypes of the offspring they

produce after fertilization and on a relatively fine time scale, which might be particularly

beneficial in a variable environment (although I didn't explicitly consider environmental

variability in this model). Selective cannibalism of clutches of lower reproductive value has been

demonstrated in relation to uncertainty of paternity (i.e., possible cuckolding events) in some

fishes (Lepomis macrochirus, Neff 2003; Telmatherina sarasinorum, Gray et al. 2007;

Gasterosteus aculeatus, Frommen et al. 2007), and the elimination of low quality offspring has

been focused on in other contexts (e.g., allowing lower quality offspring to be eliminated by

siblicide, Stearns 1987; spontaneous and selective abortion in humans, sex ratio adjustment in

red deer; Stearns 1987; Kozlowski and Stearns 1989). However, selective filial cannibalism of

viable offspring in relation to other aspects of offspring quality has received little empirical

attention (but see Chapters 4 and 5). In particular, I hypothesize that in some contexts filial









cannibalism of offspring with (1) reduced expected future survival, or (2) slower maturation rates

during the period in which care is being provided can be an adaptive strategy.

Alternatively, it is possible that filial cannibalism itself increases the development rate of

eggs. If filial cannibalism increases the maturation rate of eggs relative to those of non-

cannibalistic parents, filial cannibalism evolves over a greater range of parameter space (Figure

6-3 A). Indeed, for cases in which parent-offspring conflict exists over the optimal duration of

parental care, filial cannibalism might be a way in which parents speed-up the developmental

rate of their eggs, thereby allowing them to reduce per offspring costs of care or re-enter the

mating pool faster. According to this hypothesis, caring parents potentially benefit by providing

care for a shorter duration of time if filial cannibalism creates an environment in which offspring

are eager to escape the egg stage (e.g., because of increased risk of death; see also work on non-

parent predators increasing egg development rate, e.g., Warkentin 2000). To my knowledge, this

idea of filial cannibalism speeding-up egg development has not previously been considered, and

as mentioned previously, is likely to be relevant for cases in which parents and offspring differ in

the optimal amount of care they provide/receive.

Incorporating an energetic benefit of cannibalism facilitated the invasion of filial

cannibalism. This finding is consistent with previous theoretical and empirical work suggesting

that energetic need affects filial cannibalism (e.g., Rohwer 1978; Sargent 1992; Kraak 1996;

reviewed by Manica 2002). However, some empirical work suggests that the effects of energetic

need on filial cannibalism are not always straightforward-- in some species cannibalism increases

as parental energetic need increases (e.g., Thomas and Manica 2003), whereas in other species an

opposite pattern is observed (e.g., Klug et al. 2006). Moreover, in other systems, there appear to

be no effects of parental condition on filial cannibalism under some conditions (e.g., Lindstrom









and Sargent 1997), and in other species the relationship between energetic need and cannibalism

differs in varying contexts (Klug and Lindstrom, unpublished data). Furthermore, some have

suggested that the energetic benefits of cannibalism are not sufficient to explain the prevalence

of filial cannibalism (Smith 1992). In my model, filial cannibalism invaded over a range of

parameter space even when we removed benefits of cannibalism, suggesting that substantial

energetic benefit of cannibalism is not necessarily essential for the evolution of cannibalism.

That said, there is little doubt that filial cannibalism provides a caring parent with energy and/or

nutrients and such benefits are likely critical for adult survival and successful nest defense in

systems where parents are unable to feed during the course of providing parental care (Manica

2002, 2004). Indeed, energetic benefits certainly favor the evolution of filial cannibalism (Figure

6-5; previous work by Rohwer 1978; Sargent 1992; reviewed in Manica 2002).

Likewise, increasing the strength of density-dependent egg survivorship increased the

parameter space over which filial cannibalism evolved. However, density-dependent egg

survivorship alone did not facilitate the evolution of filial cannibalism. Indeed, it seems unlikely

that density-dependent egg survivorship per se would lead to the evolution of filial cannibalism

in the absence of other trade-offs associated with egg number. If animals can track their

environment, they would simply be expected to adjust the number of eggs they produce

according to expected egg survivorship (i.e., they should lay at densities that maximize survival).

Further work is needed to evaluate the importance of density-dependent egg survivorship when

other trade-offs are associated with the number of offspring produced or when the environment is

variable. Spatial and temporal variation in the environment has been hypothesized to influence

patterns of cannibalism observed in nature (e.g., Payne et al. 2004) and non-cannibalistic brood









reduction (e.g., Forbes and Mock 1998), but additional work is needed to understand more fully

the importance of such stochasticity at varying scales.

Sexual selection via mate choice and/or sexual conflict also affected the invasion and

fixation of filial cannibalism and/or parental care. My model suggests that the evolution and

fixation of parental care from a state of no care can be facilitated by differential reproductive

success if parental care or filial cannibalism increases the reproductive rate of individuals

exhibiting care or cannibalism (e.g., if parental care or cannibalism is preferred during mate

choice). This finding is consistent with some previous work. For example, Pampoulie et al.

(2004) and Lindstrom et al. (2006) recently demonstrated mating preferences for parental care,

suggesting a potentially larger role for sexual selection in the evolution of care than previously

thought. Additionally, filial cannibalism is possibly favored by sexual selection if cannibalism

directly benefits a choosing mate or when it makes a caring parent more attractive in some other

way (Sikkel 1994; Lindstrom 2000). Likewise, if a mating preference exists for non-cannibals,

the parameter space over which filial cannibalism evolves decreases. Interestingly, the role of

sexual conflict has received relatively little theoretical or empirical attention previously (but see

Kraak and van den Berghe 1992; Kraak 1996; Lindstrom 2000). In fishes, where filial

cannibalism is typically practiced by caring fathers, the focus of almost all work has been on

costs and benefits of cannibalism to caring males. One must also wonder if benefits to non-

cannibalistic females exist, and if such benefits are absent, why do females tolerate filial

cannibalism? Additionally, sexual conflict is also likely to exist when both parents practice filial

cannibalism, but this idea has received no attention. More empirical work is needed to better

understand costs and benefits of filial cannibalism to a parent who's mate practice filial

cannibalism.









Finally, population-level resource competition likely plays a role in the evolution of both

parental care and filial cannibalism. When care and/or cannibalism affected the efficiency with

which individuals exhibiting a given strategy use resources, parental care was more likely to

evolve if caring was associated with a reduction in the carrying capacity (e.g., when caring

decreased the efficiency with which individuals use resources), whereas, filial cannibalism was

more likely to invade if it increased carrying capacity (e.g., if cannibalism increased the

resource-use efficiency of individuals). Additionally, the evolution of filial cannibalism (with or

without parental care) was affected by the population carrying capacity, even for the case in

which the carrying capacity of the mutant and residents were equal. It is unclear how parental

care and filial cannibalism potentially alter population-level dynamics and resulting carrying

capacities in nature, but this idea warrants further attention. For example, it is possible that the

ability to cannibalize increases resource availability to caring parents, thereby freeing-up other

resources and increasing the productivity of a system. Regardless, understanding the ecological

dynamics of a system (i.e., intensity of resource competition and population growth parameters

such as carrying capacity) is likely to be critical for understanding the evolution of parental care

and filial cannibalism across animal taxa. While previous work has sometimes incorporated

population-level growth dynamics in parental care theory (e.g., McNamara et al. 2000), this is

not a common approach.

In summary, my results suggest that parental care and filial cannibalism can evolve over

a range of life-history patterns and ecological conditions, and that multiple strategies often have

the potential to coexist. Coexistence, while not well-studied (but see Webb et al. 1999), is

prevalent in nature (e.g., maternal- or paternal-only care in many taxa, reviewed in Clutton-

Brock 1991; care and no-care with total offspring abandonment following egg fertilization:









Jordanellafloridae, Hale pers. comm., the white stickleback Gasterosteidae spp., Blouw 1996;

care and care followed by abandonment, Hypoptychus dybowskii, Narimatsu and Munehara

2001). Likewise, there are many cases in which caring parents never or rarely consume or

abandon their offspring. Even in fishes, where care with filial cannibalism has been well-

documented, there are still many species exhibiting parental care in which filial cannibalism is

absent (e.g., Micropterus dolomieui, Gillooly and Baylis 1999). For species exhibiting filial

cannibalism, there is a great deal of variation in the patterns of cannibalism observed among

species and within and between individuals (e.g., how many eggs are consumed, who practices

cannibalism and when; Petersen and Marchetti 1989; Okuda and Yanagisawa 1996; Lindstrom

and Sargent 1997; Lissdker et al. 2002; Klug et al. 2005; Klug and St. Mary 2005).

Understanding such within- and between-species variation in filial cannibalism and parental care

will require more detailed theoretical and empirical work that simultaneously considers multiple

factors (such as variation in offspring quality, energetic needs of parents, mating preferences and

sexual conflict, general resource competition). Additionally, it will also be important to assess

the importance of environmental heterogeneity in the evolution of filial cannibalism. From this

study, my approach and results provides a novel basis for further developing this theme of

whether to care for or consume one's own offspring.










Table 6-1. Trade-off functions. The following trade-off functions were used to reflect the unique life histories of individuals who
provide parental care and/or practice filial cannibalism. The death rate of eggs is assumed to be a function of the parental
care provided (i.e., as de decreases, care is presumed to increase), and thus egg death rate is the proxy for care.

Parameter Trade-offs Strategy
No parental Parental care only Filial cannibalism Parental care & filial cannibalism
care & no filial only
cannibalism


Juvenile
survival rate; a,
and ajm


Adult death
rate; dA and dAm


1) Reproductive rate decreases as
caring increases (i.e., r or rm
decreases as de or dem decreases).
2) Reproductive rate increases as
maturation rate of eggs increases
(for a carer only) (i.e., r or rm
increases as mE or mEm increases).
3) Reproductive rate increases as
cannibalism increases (i.e., rm
increases as/3 increases).

1) Juvenile survival rate increases
as care increases (i.e., a, or Ujm
increases as dE or dEm decreases).


1) Adult death rate increases as
caring increases (i.e., dA or dAm
increases as dE or dEm decreases).

2) Adult death rate decreases as
cannibalism increases (i.e., dA or
dAm decreases as f increases).


Linear &
Non-linear:
r = r


Linear &
Non-linear:
oJ = oJ 0


Linear &
Non-linear:
dA = d


Linear:
rm =rm o (l+dEm +mEm)

Non-linear:

rm r (dE+mE)
m0 (l+dEm Em)


Linear:
m = j m (1 dEm)
Non-linear:

Jm (1 + dEm)
J m Jmo
dEm


Linear:
dAm


Non-linear:

iAm Am


dAm (1- dEm)


(1+ dEm)
dEm


Linear:
rm =rmo (1+6)
Non-linear:

rm m" (1+ )
(1+,))


Linear & Non-linear:
CJm = 'jm o


Linear: Linear:
dm = dA (1- ) d =


Non-linear:

dAm = dAmo (
P8


Linear:
=, (1+ dEm +mEm +/)
Non-linear:
(dEm + mEm + 8)
m mo (1 + dEm + mEm + fl)


Linear:
CJm = Jm0 *(1- dEm)
Non-linear:

S (1 + dEm)
'Jm ='Jm dE
Em


dA m (1- dEm


Non-linear:
(1+ dEm +/)
dAm = dAmo dEm /6
Em A-


Reproductive
rate; r and rm










Table 6-2. Alternative hypotheses regarding the evolutionary significance of filial cannibalism (FC). Here, I present several, non-
mutually exclusive hypotheses and briefly describe the findings of our model and those of some previous work in relation
to these hypotheses.


Hypothesis
1. Selective Filial
Cannibalism



2. Filial
cannibalism
speeds-up egg
development

3. Energy-Based
Filial
Cannibalism

4. Density-
Dependent-Egg-
Survivorship-
Mediated Filial
Cannibalism

5. Mate Choice-
Mediated Filial
Cannibalism
6. Sexual Conflict-
Mediated Filial
Cannibalism

7. Filial
Cannibalism
Driven by
Resource
Competition


Description
Offspring with particular
characteristics (e.g., reduced
survival, decreased maturation rate)
are preferentially consumed.

By increasing costs associated with
remaining in the egg stage, filial
cannibalism increases maturation
rate of eggs (i.e. FC decreases the
time it takes for eggs to develop).
FC provides energy that offsets
costs of care and is re-invested into
current and/or future reproduction.

Density-dependent egg survival
mediates FC: by consuming some
young, parents increase survival of
remaining offspring.


FC is preferred in mate choice,
thereby increasing relative
reproductive rate.
FC is a non-preferred trait and
decreases relative reproductive rate.


FC is driven by population-level
resource competition among adults.


Model findings
Evolution of FC facilitated by selective
cannibalism of offspring with lower
maturation rates, lower egg survival,
and/or lower juvenile survival.

Evolution of FC more likely if
cannibalism increases egg maturation
rate.


Energetic benefit of eggs facilitated
evolution of FC.


Density-dependent egg survival alone
did not facilitate the evolution of FC;
more intense density-dependence
facilitated evolution of FC in
comparison to weaker density-
dependence.
If FC increases relative reproductive
rate of cannibals, FC evolves more
often.
If FC decreases reproductive rate, FC
evolves less often.


Evolution of FC sensitive to
population-level carrying capacity.


Related previous findings
FC affected by certainty of paternity in some
systems (Neff 2003; Frommen et al. 2007;
Gray et al. 2007) but not in others (Svensson
et al. 1998); effect of other aspects of
offspring quality on FC largely unknown.
Not previously examined; potentially relevant
for systems in which parent-offspring conflict
exists over the optimal amount of care
provided/received.

Substantial energetic benefit of FC and/or
effect of energetic need on FC found in several
systems (reviewed in Manica 2002).

FC is affected by density-dependent egg
survivorship in two species (Payne et al. 2002,
2004; Klug et al. 2006)



FC increases nest attractiveness (and
consequently eggs received) in some cases
(Sikkel 1994).
Sexual conflict can inhibit FC in some cases
(Lindstrom 2002); sexual conflict regarding
FC not well-studied empirically (but see Kraak
1996).
Mate availability (Kondoh and Okuda 2002)
and other resource competition (Creighton
2005) affects FC in some cases; effects of
general resource availability on FC not well-
known.









r.i -kA(t)
K


t----------------------- --------------- ---------


Figure 6-1. The model: individuals develop through an egg and juvenile stage and reproduce as
adults. Eggs either die (at rate dE), are consumed by their parents) (at rate,8), or
mature into juveniles (at rate mE). Individuals survive and pass through the juvenile
stage (at rate mE cUj), where r represents the time spent in the juvenile stage. As

adults, individuals either die (at rate dA) or reproduce (at rate r 1-j where K
K '
represents the population carrying capacity). Boxes represent life-history stages; solid
arrows represent death, reproduction, and maturation; the dashed line represents
consumption of eggs by adults.











- 1 I 1 *
S0.8 care I coexistence 0.8 coexistence
Z I care
o 0.6 y 0.6
S0.4 NPIC 0.4
a .0 no care
ncr 0.2 no care
Sno care
0. -^ ^ -_
SNPIIC
| 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1
C. D.
1U
5 1 / 1
z 0.8 care &
0 care coexistence 0.8 cannibalism coexistence
0.6 0.6 /
0.4 0.4 NPIC
0.2 no care no care or
0 \P 0.2 cannibalism
NPIlC NP
0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

EGG MATURATION RATE IN RESIDENT POPULATION (mE)
Figure 6-2. Invasion of parental care. Parental care invades and/or coexists with no care more
often when parental care: (A) increases the maturation rate of eggs, (B) increases
parental reproductive rate (r = 1.0, r = 1.2), (C) is associated with a decreased
carrying capacity (K = 20, K, = 15), and (D) when the caring mutant is able to
cannibalize (i.e., f = 0.01 for the mutant, f = 0 for the residents). Lines represent
invasion boundaries for the mutant (solid line) and the resident (dotted line). Invasion
boundaries are shown for the maturation rate of the eggs. The mutant invades the
resident in the regions labeled 'care' (A-C) or 'care & cannibalism' (D), the resident
invades the mutant in the region labeled 'no care' (A-C) or 'no care or cannibalism'
(D), and both strategies coexist in the region labeled 'coexistence'. Neither strategy
will persist (i.e., they go extinct) in the region labeled 'NP'. The region labeled
'NP/IC' is a region in which neither strategy will persist, or where the outcome is
dependent upon initial conditions of the model. Unless noted above, r = r, = 1.0, dE =
dEm = 0.9, dA = dAm = 0.5, aj= ojm = 0.5, K = Km = 20, f = 0, r= 0.1.











1 care& 1
cannibalism care & j coexistence
0.8 / 0.8 cannibalism
o 0.6 NP/Ic 0.6 >
0.4 0.4 do
S0.2 --' 0.2 care only
NP care only
0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1
z C. D.
SIcare & coexistence -
cannibalism i care &
z 0.8 N 0.8 cannibalism ,
S0.6 NP/IC0.6 NP/ ,O
0.4 0 0.4 -NI
0 004
S0.2 .-*' e are only 0 care only
0care only NP
W 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1


EGG MATURATION RATE IN RESIDENT POPULATION (mE)
Figure 6-3. Invasion of filial cannibalism. A mutant that provides parental care and practices
filial cannibalism invades and coexists with a resident that only provides care (A)
more often when cannibalism increases the maturation rate of eggs, (B) more often
when cannibalism increases the parent's reproductive rate (r = 0.5, r m = 0.6), (C)
less often when cannibalism decreases the parent's reproductive rate (r = 0.6, r m =
0.5), and (D) more often when parents are able to selectively cannibalize offspring
with reduced future survival (de = 0.2, oa = 0.9, dem = 0.1, j,, = 0.95). Lines represent
invasion boundaries for the mutant (solid line) and the resident (dotted line). Invasion
boundaries are shown for the maturation rate of the eggs, mE and mE,, and unless
otherwise noted, r = r, = 0.5, dE = dEm = 0.2, dA = dAm = 0.5, aj = aj, = 0.9, K = Km
=20, = 0.015, z= 0.1. The mutant invades the resident in the region labeled 'care &
cannibalism', the resident invades the mutant in the region labeled 'care only', and
both strategies coexist in the region labeled 'coexistence'. Neither strategy will persist
in the region labeled 'NP'. The region labeled 'NP/IC' is a region in which neither
strategy will persist, or where the outcome is dependent upon initial conditions of the
model.









A. B. E
Ai
S-car 1
S 0.8 carq 0.8 U l
z
S 0.6 coexi0.6 coexistence
Scoexistence
0.4 t 0.4
0 0.2 0.2 NP1IC C^
0 NPIC 0.2 NP no care or
.- NtlC no care cannibalism
? 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1
i c. 1I


I I

S 0.6 coexistence
E 0.4

S 0.2 NP/IC '""- o care or cannibalism
0.2 0.4 0.6 0.8 1

EGG MATURATION RATE IN RESIDENT POPULATION (mE)

Figure 6-4. Effect of density-dependent egg survivorship on the evolution of parental care and
filial cannibalism. Parental care and no care invade and/or coexist over a large range
of parameter space when (A) the rare mutant does not cannibalize (f = 0). The range
over which parental care invades decreases when (B) the rare mutant cannibalizes (/f
= 0.01). However, increasing the strength of density-dependence (co, eqn. 6) increases
the range over which care and cannibalism invades-- care with cannibalism invades
more often when (C) the strength of the density-dependence is greater (o = 0.9), in
comparison to B) the case in which it is relatively weak (co = 0.6). Lines represent
invasion boundaries for the mutant (solid line) and the resident (dotted line).
Invasions boundaries are shown for the maturation rate of the eggs, mE and mnEm, and
unless otherwise noted, r = rm = 3, dE = 0.9, dEm = 0.3, dA = dam = 0.5, o = ajm, =
0.5, K = Km =20, / = 0.01, r= 1, co = 0.6. The mutant invades the resident in the
region labeled 'care' (A) or 'care and cannibalism' (B-C), the resident invades the
mutant in the region labeled 'no care' (A) or 'no care or cannibalism' (B-C), and both
strategies coexist in the region labeled 'coexistence'. The region labeled 'NP/IC' is a
region in which neither strategy will persist, or where the outcome is dependent upon
initial conditions of the model.










1 1
Scacare
a cannibaisim cannibalism Icoexistence
0.8 care & 0.8 cannibalism ,coexistence
0 c b
0.6 coexistence 0.6
10.4 0.4 NP/iC ^
A NPIC A NPJIC
g 0.2 N no care or / no care or
S 0.2 cannibalism 02 cannibalism
yNP NP
S0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1


0 1 care & 1 care&
o 0.8 cannibalism 0.8 cannibalsm
0.6 NP[Nc / 0.6 NP/IC


o 0.2 cfP"-- care only 0.2 NP care only
0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

EGG MATURATION RATE IN RESIDENT POPULATION (mE)
Figure 6-5. Effect of energetic benefits on the evolution of filial cannibalism. Parental care with
filial cannibalism is more likely to invade and coexist with no care if filial
cannibalism is (A) beneficial to a parent's survival and reproduction versus (B) the
case where there are no benefits of cannibalism. Likewise, care with cannibalism is
more likely to invade a state of only care when (C) adult survival and reproductive
benefits of egg eating exist versus (D) the case where such benefits are absent. Lines
represent invasion boundaries for the mutant (solid line) and the resident (dotted line).
Invasions boundaries are shown for the maturation rate of the eggs, mE and mEm.
Unless otherwise noted, r = r, = 1.0, dE = dEm = 0.9, dA = da, = 0.5, oj= ojm = 0.5,
K = Km 20, = 0.01, = 0.1 for A and B, andr = r= 0.5,dE = d, = 0.2, d = dA,
= 0.5, aJ= ojn, = 0.9, K = Km =20, f = 0.015, r= 0.1 for C and D. The mutant invades
the resident in the region labeled 'care & cannibalism', the resident invades the
mutant in the region labeled 'no care or cannibalism' (A-B) or 'care only' (C-D), and
both strategies coexist in the region labeled 'coexistence'. Neither strategy will persist
in the region labeled 'NP'. The region labeled 'NP/IC' is a region in which neither
strategy will persist, or where the outcome is dependent upon initial conditions of the
model.









A. 1 care &
0.8 cannibalis m
0.8 ,
0.6 NPIIC ^/

0.2 ,-," care only
NP
0.2 0.4 0.6 0.8 1
C.
C are &
0.8 cannibalisnr1
0.6 'coexistence
0.4 N lICe
0.2 no care or
NP cannibalism
0.2 0.4 0.6 0.8 1
E.
coexiste ce
1 care &
cannibalis
0.8
0.6 / NP/IC /n
4 / -- only
0.2 s,--- care only


B.1
care& j
0.8 cannibalism coexistence
coexistence
0.6
0.4 /
0.2 / care only

0.2 0.4 0.6 0.8 1
D.
Scare &
cannibalisrv
0.8
0.6 / coexistence
0.4
0.2 / no care or
,^^ cannibalism
0.2 0.4 0.6 0.8 1
F.
care &
cannibalisn/
0.8 7
0.6 /I coexistence

no care or
0.2 .l cannibalism


0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1
EGG MATURATION RATE IN RESIDENT POPULATION (me)
Figure 6-6. Effect of carrying capacity on the evolution of filial cannibalism. Parental care with
filial cannibalism invades and coexists with care over a range of parameter space
when (A) cannibalism does not affect carrying capacity (K = K, = 20). However,
cannibalism invades over a greater range of parameter space when (B) cannibalism
increases carrying capacity (K = 10, K, = 20). Likewise, care with cannibalism
invades no care over a range of parameter space when (C) care and cannibalism do
not affect carrying capacity (K = K, = 20), but it invades or coexists more often when
(D) care and cannibalism decrease carrying capacity (K = 20, K, = 10). For cases in
which the resident and mutant have equal carrying capacities, parental care and
cannibalism are more likely to invade when (E) carrying capacity is relatively small
(K= K, = 10) versus the case in which it is relatively large (A). In contrast, care with
cannibalism is more likely to invade no care/no cannibalism when (F) carrying
capacity is relatively large (K = K, = 50) versus (C) the case in which it is relatively
small. Lines represent invasion boundaries for the mutant (solid line) and the resident
(dotted line). Invasions boundaries are shown for the maturation rate of the eggs, mE
andmEm. Unless otherwise noted, r = r, = 0.5, dE = dEm = 0.2, dA = dAm = 0.5, ,oj =
am = 0.9, P = 0.01,r= 0.1 for A, B and E, and, r= r, = 1.0, dE dEm = 0.9, dA =
dAm = 0.5, ,oj= am, = 0.5, / = 0.01,r= 0.1 for C,D, and F. The mutant invades the
resident in the region labeled 'care & cannibalism', the resident invades the mutant in
the region labeled 'care only' (A-B, E) or 'no care or cannibalism' (C-D, F), and both
strategies coexist in the region labeled 'coexistence'. Neither strategy will persist in
the region labeled 'NP'. The region labeled 'NP/IC' is a region in which neither
strategy will persist, or where the outcome is dependent upon initial conditions of the
model.









CHAPTER 7
GENERAL CONCLUSIONS AND SYNTHESIS

Introduction

Parental care typically increases offspring survival and/or quality, thereby increasing

parental fitness. Thus, it is surprising that filial cannibalism, the consumption of one's own

offspring, is prevalent in fishes exhibiting paternal care. Indeed, it's difficult to imagine many

situations in which regularly consuming one's own young is an adaptive strategy. Because

parental males often consume more eggs than die naturally (Manica 2002; Klug et al. 2006 and

Chapter 6), filial cannibalism does not solely serve to clean the nest of dead eggs. Currently,

filial cannibalism is thought to represent an adaptive trade-off (Manica 2002).

Prior to my dissertation work, there were two general hypotheses explaining the adaptive

significance of filial cannibalism: the energy-based hypothesis (Rohwer 1978; Sargent 1992) and

the oxygen-mediated hypothesis (Payne et al. 2004, 2004). The energy-based hypothesis

suggests that filial cannibalism is an adaptive strategy in which males gain energy or nutrients

from eggs that are then reinvested into current or future reproduction, thereby increasing net

reproductive success (Rohwer 1978). According to this hypothesis, filial cannibalism is expected

to increase when food availability is low and/or when the caring parent's condition is poor. There

has been mixed support for the energy-based hypothesis (Belles-Isles & Fitzgerald 1991; Smith

1992; Lindstrom and Sargent 1997), and at best, it can only explain filial cannibalism in some

systems (e.g., Manica 2004). The oxygen-mediated hypothesis of filial cannibalism (Payne et al.

2002) suggests that filial cannibalism is an adaptive strategy in which partial clutch cannibalism

improves the survival of remaining eggs by increasing oxygen availability to remaining eggs.

Specifically, Payne et al. (2002) suggested that caring males potentially improve overall clutch









survivorship by consuming some of their eggs. The oxygen-mediated hypothesis has received

support in one species (Payne et al. 2002), but has not been generally evaluated.

In addition to energy and oxygen availability, some studies suggest that mate choice or

sexual conflict (Sikkel 1994; Kraak 1996; Lindstrom 2000), egg age (Salfert and Moodie 1985;

Sikkel 1994), and certainty of paternity (Neff 2003; Gray et al. 2007; Frommen et al. 2007)

affects the occurrence of filial cannibalism (but see Svensson et al. 1997 and Svensson and

Kvarnemo 2007, who find that certainty of paternity does not affect filial cannibalism in the sand

goby). There has been relatively little theoretical or empirical examination of such factors, and

thus the general importance of mate choice, sexual conflict, egg age, and certainty of paternity

remains unknown (e.g., Takeyama et al. 2007).

Because of (1) the lack of support for any particular hypothesis (i.e., the energy-based or

oxygen-mediated hypothesis) and (2) a general lack of alternative hypotheses, the evolutionary

significance of filial cannibalism in fishes remains unclear. In Chapter 1, I argued that an

enhanced understanding of the evolutionary significance of filial cannibalism necessitates three

approaches: 1) a re-evaluation of current theory by explicitly focusing on fitness consequences of

filial cannibalism; 2) the development and examination of alternative hypotheses of filial

cannibalism; and 3) the development and evaluation of a synthetic model of filial cannibalism

that simultaneously considers the potential importance of a range of factors. In Chapters 2 and 3,

I evaluated predictions of the energy-based and the oxygen-mediated hypotheses in two species,

the flagfish (Jordanellafloridae) and the sand goby (Pomatoschistus minutus). In Chapters 4 and

5, I developed the novel hypothesis of selective filial cannibalism, and I evaluated this

hypothesis in the flagfish and the sand goby. In Chapter 6, I developed a model of filial

cannibalism. Using this model, I evaluated the plausibility of a range of alternative hypotheses of









filial cannibalism, and I concluded that a variety of factors can favor the evolution of filial

cannibalism.

In this final discussion, I will synthesize the findings of Chapters 2-5, and discuss my

findings in terms of previous and novel hypotheses.

Are the Current Energy-Based and Oxygen-Mediated Hypotheses Sufficient?

Until 2002, the energy-based hypothesis was the only adaptive hypothesis of filial

cannibalism and it remains the most widely accepted hypothesis of filial cannibalism (reviewed

in Manica 2002). However, as mentioned before, evidence regarding this hypothesis has been

mixed (Belles-Isles & Fitzgerald 1991; Smith 1992; Lindstrom and Sargent 1997).

In Chapter 2, I described an experiment in which I experimentally manipulated 1) the

ability of parental males to cannibalize eggs (i.e., males either had full access to eggs, or filial

cannibalism was prevented by a nest cover) and 2) diet (i.e., high quality versus low quality diet)

to evaluate the effect of filial cannibalism and diet on components of reproductive success in

male flagfish. According to the energy-based hypothesis, energy gained from eggs should be

translated into increased future reproduction, and I therefore predicted that males that were able

to practice filial cannibalism would receive more eggs and spawn more frequently during the 90

days of the experiment. Contrary to these predictions, I found that filial cannibalism did not

increase the total number of eggs males received or the frequency of spawning. Indeed, filial

cannibalism was always associated with a decrease in the total number of eggs received,

suggesting that energy or nutrients attained from eggs is not directly translated into future

reproduction in the flagfish. Furthermore, the energy-based hypothesis suggests that filial

cannibalism should increase when food availability is low. In contrast to this prediction, I found

that filial cannibalism decreased when food availability was low. Specifically, males on the low









quality diet consumed fewer of their eggs than males on the high quality diet. Thus, I found no

support for the energy-based hypothesis in the flagfish.

Similarly, I examined the relationship between parental male condition and filial

cannibalism in the sand goby (Chapter 3). The energy-based hypothesis predicts that males will

consume more eggs when parental condition is relatively poor. However, I found that males in

poorer condition consumed a smaller proportion (and fewer) of their eggs than males that were in

better condition. This finding is directly in contrast to predictions of the energy-based

hypothesis. Thus, I found no support for the energy-based hypothesis of filial cannibalism in

either the flagfish or the sand goby. Because of my findings (Chapters 2 and 3) and those of

other studies (Belles-Isles and Fitzgerald 1991; Smith 1992; Lindstrom and Sargent 1997), I

conclude that the energy-based hypothesis is not sufficient to explain the prevalence of filial

cannibalism. While eggs certainly provide some energy or/and nutrients, energetic benefits of

cannibalism cannot explain filial cannibalism in an adaptive context.

Likewise, I did not find support for the oxygen-mediated hypothesis of filial cannibalism

(Chapter 3). Specifically, the oxygen-mediated hypothesis of filial cannibalism predicts that 1)

filial cannibalism of some eggs in a nest increases oxygen to the remaining eggs, thereby

increasing total egg survival, 2) cannibalism will increase as oxygen decreases, and 3)

cannibalism will decrease as egg density increases. In the sand goby, I found that filial

cannibalism increased as oxygen decreased and as egg density increased. While both findings are

consistent with the oxygen-mediated hypothesis, it is possible that males increased cannibalism

at lower oxygen levels because the costs of providing care in low oxygen environments are

greater than the costs in high oxygen environments. I therefore directly evaluated the effect of

oxygen level and egg density by exposing eggs to two levels of simulated filial cannibalism (i.e.,









simulated cannibalism or no simulated cannibalism) and two levels of oxygen availability (i.e.,

high versus low oxygen). Eggs were reared in the absence of males and I quantified egg survival.

Indeed, I found that egg survival was density-dependent, but this density-dependence was not

mediated by oxygen. Specifically, there was no effect of oxygen on egg survival in this

experiment. Thus, I propose a more general hypothesis of filial cannibalism mediated by density-

dependent egg survival. I suggest that density-dependent egg survival might be due to a range of

factors (e.g., waste accumulation in the nest or disease transmission), and that the factors

affecting egg survival likely vary across species and environmental conditions.

In summary, current hypotheses of filial cannibalism (i.e., the energy-based and oxygen-

mediated hypotheses) are inadequate for generally explaining the adaptive significance of filial

cannibalism in fish species. While energetic need and oxygen might be important in some

contexts and in some species, neither energy nor oxygen alone can explain the prevalence of

filial cannibalism in fishes. Thus, my dissertation work aimed at re-evaluating current hypotheses

(Chapters 2 and 3) further supports the need for 1) the development of alternative hypotheses and

2) increased theoretical examination of the importance of a range of factors in the evolution of

filial cannibalism.

An Alternative Hypothesis: Selective Filial Cannibalism

Parental care is costly and leads to reduced future reproduction (reviewed in Smith and

Wootton 1995). Therefore, males should not waste energy caring for low quality eggs if the cost

to the male (i.e., reduced future reproduction) outweighs the current benefit in offspring

produced. Specifically, I hypothesize that males should preferentially cannibalize offspring of

reduced quality (i.e., offspring that have reduced expected future survival or reproductive

success) when there is some energetic benefit of consuming eggs or when offspring survival is

density-dependent. The elimination of lower quality offspring has been demonstrated in relation









to selective embryo abortion in humans and plants (Forbes 1997; Diamond 1987; Burd 1998;

Karkkainen et al. 1999), brood reduction (Mock and Forbes 1995; Forbes and Mock 1998), and

parents allowing or encouraging siblicide of low quality offspring (Stearns 1987), but this idea

has not been considered in relation to filial cannibalism. Indeed, the elimination of low quality

offspring is thought to play a central role in explaining the evolutionary significance of offspring

abandonment and brood reduction (e.g. Stearns 1987, 1992; Forbes and Mock 1998), and thus I

hypothesize that the ability to cannibalize offspring selectively might be an important factor in

explaining the adaptive significance of filial cannibalism.

I evaluated the hypothesis of selective filial cannibalism in the sand goby (Chapter 4) and

the flagfish (Chapter 5). In the sand goby, I examined within-clutch patterns of cannibalism

when males received eggs from either one or two females. I focused on the relationship between

filial cannibalism and egg size, which has been correlated with post-hatching survival in a range

of fishes (reviewed in Kamler 2005). In the single-female scenario, I found that males exhibited

no preferences with regard to egg size. In the multiple-female scenario, males preferentially

consumed the larger eggs of the second female, but they exhibited no size preferences for the

eggs of the first female they spawned with. To evaluate further patterns of egg survival and

hatching, I reared subsets of eggs in the absence of males. For the clutches reared in the absence

of males, there was no relationship between egg size and survival, but larger eggs took longer to

hatch than smaller eggs. Thus, the findings that 1) larger eggs take longer to hatch and 2) males

preferentially consume larger eggs of the second female whose eggs are already younger than

those of the first female. This pattern suggests that males preferentially consume eggs in a

manner that reduces the amount of time they spend caring for the current clutch of eggs (Chapter

4). Specifically, my results (e.g., Figure 4-1) suggest that males might be able to reduce the









duration of time spent caring by several days if they preferentially consume the largest eggs. In

Chapter 4, I hypothesize that reducing the duration of time spent caring for a given brood might

allow a male to re-enter the mating pool sooner. Specifically, if a male sand goby can reduce the

per-clutch time he spends providing parental care, it is possible that he can gain an additional

brood cycle, which might in turn increase his net reproductive success. This hypothesis is

supported further by the finding that whole clutch cannibalism tends to decrease as the breeding

season progresses (Chapter 3). Later in the breeding season, females become scarce and a male's

expected future reproduction decreases. Thus, I would expect benefits associated with decreasing

the duration of parental care to decrease later in the breeding season.

In the flagfish, I examined the relationship between filial cannibalism and mean egg

energetic content and female size (Chapter 5). Whole clutch cannibalism increased as mean egg

energetic content increased. In contrast, I found a negative relationship between partial clutch

cannibalism and mean energetic content of eggs and maternal size. Egg energetic content and

maternal size have been correlated with post-hatching survival in fishes (reviewed in Kamler

2005), and thus, it appears that when males practice whole clutch cannibalism, they

preferentially consume their higher quality offspring, which provide a relatively high energetic

benefit. However, when males practice partial clutch cannibalism, they preferentially cannibalize

offspring that are likely to have lower future survival (Chapter 5). This finding is consistent with

other work suggesting that filial cannibalism increases when a brood has relatively low expected

reproductive value. For example, whole clutch cannibalism increases when the initial number of

eggs present is relatively small (reviewed in Manica 2002) and when males have been cuckolded

(e.g., Frommen et al. 2007).









My experiments on selective filial cannibalism in the sand goby and flagfish (Chapters 4

and 5) highlight the potential importance of the hypothesis of selective filial cannibalism. I have

demonstrated that males preferentially consume eggs based on aspects of phenotype in some

cases. As mentioned previously, selective elimination of low quality offspring is hypothesized

to play a large role in the evolution of selective abortion, brood reduction, and offspring

abandonment. I therefore hypothesize that selective cannibalism might play a large role in

explaining the evolutionary significance of filial cannibalism, but additional theoretical and

empirical work in other species is needed to evaluate the relative importance of selective filial

cannibalism.

The Plausibility of Multiple Hypotheses

In Chapters 2, 3, 4, and 5, I demonstrated that food availability, paternal condition,

density-dependent egg survival, egg size, egg energetic content, and maternal size affects filial

cannibalism in the flagfish and/or sand goby. To begin to evaluate the relative importance of

these (and other) factors in understanding the evolutionary significance of filial cannibalism, I

developed and analyzed a general model of filial cannibalism (Chapter 6). The results of this

model suggest that no single benefit of filial cannibalism is essential for the evolution of filial

cannibalism. Indeed, my model suggests that the evolutionary dynamics of filial cannibalism

appear to be similar to those of other forms of offspring abandonment (Chapter 6). Specifically,

the ability to terminate parental care through filial cannibalism, infanticide, brood reduction, or

abandonment can represent an adaptive strategy under some conditions (i.e., under certain

environmental conditions, for organisms with particular life history characteristics) even when

obvious benefits (i.e., energetic gain or benefits to remaining offspring) are absent. That said,

benefits associated with energetic gain, offspring survival, and mate preferences certainly exist in









some species under some contexts (e.g., Manica 2004, Payne et al. 2002, Sikkel 1994) and can

directly facilitate the evolution of filial cannibalism.

Indeed, I found that the evolution of filial cannibalism was facilitated when (1) parents

could selectively cannibalize lower quality offspring or offspring with slower egg development

rates (Hypothesis 1, Table 6-2), (2) filial cannibalism increased egg maturation rate (Hypothesis

2, Table 6-2), (3) energetic benefits of eggs existed (Hypothesis 3, Table 6-2), (4) cannibalism

increased a parent's reproductive rate (e.g., through mate attractiveness; Hypothesis 5, Table 6-

3). Density-dependent egg survivorship alone did not favor the evolution of cannibalism

(Hypothesis 4, Table 6-2). However, when egg survival was density-dependent, filial

cannibalism invaded more often when the density-dependence was relatively more intense.

Additionally, sexual conflict potentially inhibits the evolution of filial cannibalism in some cases

(Hypothesis 6, Table 6-2). I also hypothesize that population-level resource competition can play

a large role in the evolution of filial cannibalism (Hypothesis 7, Table 6-2). Indeed, in my model,

the evolution of filial cannibalism was highly sensitive to population carrying capacity, and filial

cannibalism was more likely to evolve when it allowed individuals to utilize resources more

efficiently.

In summary, my modeling work (Chapter 6) highlights the plausibility of several non-

mutually exclusive alternative hypotheses. Additionally, I argue that attempting to explain the

evolutionary significance of filial cannibalism with any single benefit (e.g., energetic need) is

futile, and future work should consider the importance of a range of factors.

Future Directions

More research is needed to understand the evolutionary significance of filial cannibalism.

Below, I discuss six avenues of future research.









Determining the Relative Importance of Varying Factors

Future work should focus on determining the relative importance of energetic and

nutritional benefits of eggs, density-dependent egg survival, mate choice, sexual conflict, and

egg quality and size in the evolution of filial cannibalism. Specifically, it will be important to

assess the role of such factors in a range of organisms with diverse life histories and under

varying conditions. Doing so will help characterize the selection pressures that shape patterns of

filial cannibalism. Such an approach will also determine whether particular factors are more

likely to affect filial cannibalism than others. Additionally, it will continue to be important to

identify additional factors that affect filial cannibalism.

Role of Environmental Variation

Filial cannibalism raises a question that has been dealt with rarely: why do organisms

produce more offspring than can survive to maturity? This question has received some attention

in regard to selective embryo abortion (e.g., Burd 1988; Forbes and Mock 1998), but it has not

been dealt with in relation to filial cannibalism. Some have hypothesized that the over-

production of offspring can be favored when (1) the cost of producing additional offspring is

relatively small and (2) there is a relatively large benefit associated with the ability to screen and

weed out weaker offspring post-fertilization (Mock and Parker 1997; Forbes and Mock 1998). I

hypothesize that benefits of screening and weeding out particular offspring post-fertilization are

likely greatest when the environment is variable. Specifically, if the environment is static,

parents would be expected to accurately gauge and produce some optimal number of offspring of

an optimal quality. However, when the environment is highly variable, it presumably becomes

more difficult for parents to gauge the optimal number and optimal quality of offspring. The role

of environmental variation has not been explored directly in studies of filial cannibalism, but

warrants additional research. Specifically, I hypothesize that environmental variability plays a









large role in the evolution of filial cannibalism, and that when the environment is highly variable,

filial cannibalism is more likely to be selected for.

The Non-Cannibalistic Parent

As mentioned in Chapter 6, almost all of the focus of filial cannibalism has been on the

cannibalistic parent (but see Lindstrom 2000). Thus, the question remains: what role does a non-

cannibalistic parent play in the evolution of filial cannibalism? Lindstrom (2000) suggested that

a non-caring parent might benefit from filial cannibalism if cannibalism by the caring parent

increases the probability that the caring parent will successfully rear the clutch. However, the

benefits of filial cannibalism to a non-cannibalistic parent remain unknown. Indeed, more

empirical work that explicitly quantifies the costs and benefits of filial cannibalism to both

parents is needed.

Identification of Additional Species Practicing Filial Cannibalism

For many years, filial cannibalism in fishes was dismissed as a rare behavior with little or

no adaptive significance. Since beginning my dissertation work, I've had numerous people

mention that their study organisms (e.g., bears, wasps, skinks) exhibit filial cannibalism, but

because it was a relatively rare occurrence they didn't give it much thought. This view makes it

less likely that researchers will document and investigate filial cannibalism. In the future, it will

be particularly important to document filial cannibalism in other taxa. Only then can a truly

synthetic framework of filial cannibalism be developed.

A Comparative Framework of Filial Cannibalism

Once filial cannibalism is better documented, it will be important to consider filial

cannibalism from a comparative perspective. Using a comparative approach to better understand

filial cannibalism is an obvious next step. However, I would argue this approach is currently

impossible, in large part because filial cannibalism is not formally documented in many animals.









In particular, a comparative framework of filial cannibalism would facilitate better understanding

of the general life-history characteristics that are likely to be associated with filial cannibalism.

Why Don't All Parents Exhibit Filial Cannibalism?

My dissertation research suggests that filial cannibalism represents an adaptive strategy

in many contexts and in animals with varying life histories (Chapter 6). If this is the case, why

isn't filial cannibalism more common in animals? In fact, I would argue that filial cannibalism is

prevalent in animals and that it likely occurs at some level in the majority of animals. However,

as discussed previously, filial cannibalism likely isn't documented in species in which it is

difficult to detect or relatively infrequent. For animals that never or infrequently exhibit filial

cannibalism, it will be important to quantify the costs of filial cannibalism. Indeed, costs of

cannibalism in relation to disease transmission have been well-established in several species

(Rudolf and Antonovics 2007), and disease transmission as a possible cost of filial cannibalism

warrants further attention.









APPENDIX
ISOLATION AND CHARACTERIZATION OF MICROSATELLITE DNA MARKERS FOR
THE FLAGFISH JORDANELLA FLORIDAE

The flagfish, Jordanellafloridae, is a freshwater fish found throughout Florida. Flagfish

have been the focus of studies of behavior and evolution (Bonnevier et al. 2003; Klug et al.

2005; Klug and St. Mary 2005), population- and community-level ecology (Jordan and

McCreary 1996; Barber and Babbitt 2003; Ruetz et al. 2005), toxicology (Holdway and Sprague

1979; Rowe et al. 1983; Reinert et al. 2002), and conservation biology (McCormick and Leino

1999). Despite such wide-spread interest in the flagfish, published microsatellite DNA markers

are not yet available for this species. Such markers will be useful for paternity assays, estimating

heritability, and characterizing genetic population diversity in the flagfish. Here, I describe the

identification and characterization of 6 polymorphic microsatellite markers isolated from a

population of flagfish found in the Otter Creek/Waccasassa River drainage in northwest-central

Florida.

I isolated DNA from anal fin clippings using the DNeasy Blood and Tissue Kit (Qiagen).

A genomic DNA library from one individual was enriched for CA/GT microsatellite repeats

using the protocol described in Tools for Developing Molecular Markers (ICBR 2001; modified

from Kandpal et al. 1994). I then digested the genomic DNA with Sau3AI enzyme and then

fractionated the digested DNA using Chroma Spin columns (BD Biosciences) to capture

fragments in the size range of 400 bp and larger. The resulting DNA fragments were ligated to

Sau3AI linkers using T4 DNA ligase. I used fractionation using Chroma Spin columns to remove

excess Sau3AI linkers, and the fragments were polymerase chain reaction (PCR) amplified using

the Sau-Linker-A as the primer. I then denatured the entire PCR library (by heating to 980C) and

hybridized the library to a biotinylated repeat probe (5'-(CA)s5TATAAGATA-biotin) at 450C.

The hybridized DNA was recovered using VECTREX Avidin D (Vector Laboratories), and the









resulting DNA was again PCR-amplified using the Sau-Linker-A as the primer. The enriched

microsatellite libraries were cloned using a TOPO TA kit (Invitrogen) and transformed into

Escherichia coli cells (One Shot TOPO cells, Invitrogen). I screened the clones using a biotin-

labeled (CA)15 probe (Lifecodes) and the chemiluminescent substrate Lumi-Phos 480

(Lifecodes). Clones from positive colonies were grown overnight at 370C, purified using a

Miniprep Kit (Qiagen), and then sequenced using an ABI 377 sequencer (Applied Biosystems).

I designed primers for 23 clones identified to be of sufficient length using OLIGO 6.0

(Molecular Biology Insights). For the 10 clones that PCR-amplified consistently, I ordered and

optimized fluorescently-labeled primers (FAM upper primer; Biotech). PCR amplifications were

performed in an Eppendorf MasterCycler EP gradient thermocycler. Each 25 [tL reaction

contained 1 x PCR buffer (Sigma), 800 tM dNTPs, 3.0 mM MgC12, 0.26 tM of each primer, 1

U Taq polymerase (Sigma), and at least 50 ng template DNA. PCR conditions were as follows:

940C denaturation for 4 min followed by 30 cycles of 30 s at 940C, 30 s at the locus specific

annealing temperature Ta (Table A-i), and 30 s extension at 720C, followed by 5 min at 720C.

Samples were run on an ABI 377 Automated DNA Sequencer (Applied Biosystems) and

analyzed using GENESCAN and GENOTYPER (Applied Biosystems).

Of the clones that amplified consistently, six were polymorphic (Table A-i) and free of

extraneous bands after optimization. I determined microsatellite variability for 37 to 135

individual flagfish (see Table A-i for locus-specific sample sizes). I calculated observed and

expected heterozygosities using POPGENE 1.31 (Yeh et al. 1999). I performed tests for

deviations from Hardy-Weinberg expectations and linkage disequilibrium using GENEPOP 1.2

(Raymond et al. 1995). Expected heterozygosities ranged from 0.69 0.84 (Table A-i). Three of

the loci (JFJ4, JF511, and JFJ25; Table A-i) showed significant deviations from Hardy-









Weinberg expectations after sequential Bonferroni correction (Rice 1989), suggesting the

possibility of null alleles, non-random mating, or the Wahlund effect. None of the loci showed

significant linkage disequilibrium after sequential Bonferroni correction. The microsatellite

markers described herein will likely prove useful in studies characterizing population genetic

diversity, assessing paternity, and quantifying heritability.









Table A-1. Characteristics of flagfish microsatellite loci; shown here are the locus name, primer
sequences, repeat motif, optimum annealing temperature (Ta C), size range, number
of alleles (NA), the number of individuals tested (N), observed heterozygosity (Ho),
and expected heterozygosity (HE). ** indicates statistically significant deviation from
Hardy-Weinberg expectations after sequential Bonferroni correction.
Locus Primer sequences (5'-3') Repeat Ta Size NA N Ho HE
motif range
(bp)
JFJ25 GGAGGTCTCGAGGTGTTC (TG)22 58 296- 24 114 0.66 0.83**
AACCCTAAAACTCATCCTAAA 329

JFI3 GGAAAACACTGGAACCTCG (TG)19 58 228- 14 103 0.81 0.84
ATCATGCATGTGCCTCTAGC 252

JFJ4 GATAGAGGTGAGAAGGTGCAA (CA) 1 61 275- 10 135 0.60 0.72**
CTGGCTGCGTGCACTGA 299

JF511 CTCTGTTTGTCGCGTTTGTA (GT)21 64.9 316- 15 109 0.58 0.74**
AGAGGCCAAACATGCTACC 339

JF5121 AAGGGTCACGGTTAGGCT (GT)28 60.1 278- 4 44 0.59 0.69
AAATCTAACTCCCAATCCAA 288

JF515 GCCATGCGTCGTGAGTCAGA (CA)22 65.7 152- 8 37 0.62 0.75
GGAGGGAGGACATTGGG 180









LIST OF REFERENCES

Andersson, M. 1994. Sexual Selection. Princeton University Press, Princeton, NJ.

Anthony, C.D. 2003. Kinship influences cannibalism in the wolf spider, Pardosa milvina.
Journal of Insect Behavior 16: 23-36.

Baber, M.J., Babbit, K.J.. 2003. The relative impacts of native and introduced predatory fish on a
temporary wetland tadpole assemblage. Oecologia 136: 289 295.

Balshine, S., Kempenaers, B. & Szekely, T. (eds). 2002. Conflict and cooperation in parental
care. Papers of a Theme Issue. Philosophical Transactions of the Royal Society of London
B 357: 237-403.

Bartlett, J. 1987. Filial cannibalism in burying beetles. Behavioral Ecology and Sociobiology 21:
197-183.

Baylis, J.R. 1981 The evolution of parental care in fishes, with reference to Darwin's rule of
male sexual selection. Environmental Biology of Fishes 6: 223-251.

Belles-Isles, J.C. & Fitzgerald, G. J. 1991. Filial cannibalism in sticklebacks- a reproductive
management strategy. Ethology Ecology and Evolution 3: 49-62.

Bellows, T.S. 1981. The descriptive properties of some models for density dependence. Journal
of Animal Ecology 51: 139-156.

Blouw, D. M. 1996. Evolution of offspring desertion in a stickleback fish. Ecoscience 3: 18-24.

Bonnevier, K., Lindstrom, K., St. Mary, C.M. 2003. Parental care and mate attraction in the
Florida flagfish, Jordanellafloridae. Behavioral Ecology and Sociobiology 53: 358 363.

Browman, H.I., St-Pierre, J.-F., Skiftesvik, A.B. and Racca, R.G. 2003. Behaviour of Atlantic
cod (Gadus morhua) larvae: an attempt to link maternal condition with larval quality. In:
Browman, H.I. and Skiftesvik, A.B. (eds.), The Big Fish Bang. Proceedings of the 26th
Annual Larval Fish Conference, Bergen, 22-26 July 2002. Institute of Marine Research,
Bergen, pp. 71-95.

Burd, M. 1998. "Excess" flower production and selective fruit abortion: a model of potential
benefits. Ecology 79: 2123-2132.

Chesson, J. 1983. The estimation and analysis of preference and its relationship to foraging
models. Ecology 64: 1297-1304.

Clutton-Brock, T. H. 1991. The Evolution of Parental Care. Princeton, NJ: Princeton University
Press.

Creighton, J.C. 2005. Population density, body size, and phenotypic plasticity of brood size in a
burying beetle. Behavioural Ecology 16: 1031-1036.









Diamond, J.M. 1987. News and views: causes of death before birth. Nature 329: 487-488.

Dominey, W. J. 1981. Anti-predator function of bluegill sunfish nesting colonies. Nature 290:
586-588.

Elgar, M.A. and Crespi, B.J. (eds.) 1992. Cannibalism: Ecology and evolution among diverse
taxa. Oxford University Press, Oxford.

Fairbanks, L. A. & McGuire, M. T. 1986. Age, reproductive value, and dominance related
behaviour in vervet monkey females: cross-generational social influences on social
relationships and reproduction. Animal Behaviour 34: 1718-1721.

Forbes, L.S. 1997. The evolutionary biology of spontaneous abortion in humans. Trends in
Ecology and Evolution 12: 446-450.

Forbes, L.S. & Mock, D.W. 1998. Parental optimism and progeny choice: when is screening for
offspring quality affordable. Journal of Theoretical Biology 192: 3-14.

Forester, D.C. 1979. The adaptiveness of parental care in Desmognathus ochrophaeus. Copeia
1979: 332-341.

Foster, N.R., Cairns Jr., J., & Kaesler, R.L. 1969. The flagfish, Jordanellafloridae, as a
laboratory animal for behavioral bioassay studies. Proceedings of the Academy of Natural
Sciences of Philadelphia 121: 129-152.

Frommen, J.G., Brendler, C. and Bakker, T.C.M. 2007. The tale of the bad stepfather: male
three-spined sticklebacks Gasterosteus aculeatus L. recognize foreign eggs in their
manipulated nest by egg cues alone. Journal of Fish Biology 70: 1295-1301.

Gilbert, W.M., Nolan, P. M., Stoehr, A.M., and Hil, G.E. 2005. Filial cannibalism at a House
Finch nest. Wilson Bulletin 117: 413-415.

Gillooly, J. F. and Baylis, J. R. 1999. Reproductive success and the energetic cost of parental
care in male smallmouth bass. Journal of Fish Biology 54: 573-584.

Gray, S.M., Dill, L.M., McKinnon, J.S. 2007. Cuckoldry incites cannibalism: male fish turn to
cannibalism when perceived certainty of paternity decreases. American Naturalist 169:
258-263.

Gross, M. R. & Sargent, R. C. 1985. The evolution of male and female parental care in fishes.
American Zoologist 25: 807 822.

Gurney, W. S. C. and Nisbet, R. M. 1998. EcologicalDynamics. Oxford University Press,
Oxford.

Hale, R.E., St. Mary, C. M., & Lindstrom, K. 2003. Parental response to changes in costs and
benefits along and environmental gradient. Environmental Biology of Fishes 67: 107-116.









Hesketh, T. and Xing, Z.W. 2006. Abnormal sex ratios in humans populations: causes and
consequences. Proceedings of the National Academy of Sciences of the USA 103: 13271-
13274.

Hoelzer, G. A. 1992. The ecology and evolution of partial-clutch cannibalism by parental cortez
damselfish. Oikos 65: 113-120.

Holdway, D.A., and Sprague, J.B. 1979. Chronic toxicity of vanadium to flagfish. Water
Research 13: 905 910.

Interdisciplinary Center for Biotechnology Research. 2001. Tools for Developing Molecular
Markers Manual. University of Florida, Florida, USA.

Jones, J. and Reynolds, J. D. 1999. Costs of egg ventilation for male common gobies breeding in
conditions of low dissolved oxygen. Animal Behaviour 57: 181 188.

Jordan, F., and McCreary, A.C. 1996. Effects of an odonate predator and habitat complexity on
survival of the flagfish Jordanellafloridae. Wetlands 16: 583 586.

Kamler, E. 1992. Early life histories of fish: an energetic approach. Chapman Hall, London.

Kamler, E. 2005. Parent-egg-progeny relationships in teleost fishes: an energetic perspective.
Reviews in Fish Biology and Fisheries 15: 399-421.

Kandpal, R.P., Kandpal, G., and Weissman, S.M. 1994. Construction of libraries enriched for
sequence repeats and jumping clones, and hybridization selection for region-specific
markers. Proceedings of the National Academy of Sciences USA 91: 88-92.

Karkkainen, K. Savolainen, 0., and Koski, V. 1999. Why do so many plants abort so many
developing seeds: bad offspring or bad maternal genotypes? Evolutionary Ecology 13 305-
317.

Keckeis, H., Bauer-Nemeschkal, E., Menshutkin, V.V., Nemeschkal, H.L. and Kamler, E. (2000)
Effects of female attributes and egg properties on offspring viability in a rheophilic
cyprinid, Chondrostoma nasus. Can. J. Fish. Aquat. Sci. 57, 789-796.

Klemme, I., Eccard, J.A., and Ylonen, I. 2006. Do female bank voles (Clei'th iiuvy\ glareolus)
mate multiply to improve on previous mates? Behavioral Ecology and Sociobiology 60:
415-421.

Klug, H. and St. Mary, C. M. 2005. Breeding season fitness consequences of filial cannibalism in
the flagfish, Jordanellafloridae. Animal Behaviour 70: 685-691.

Klug, H. and Bonsall, M.B. 2007. When to care for, abandon, or eat your offspring: the evolution
of parental care and filial cannibalism. American Naturalist 170: 0000 0000.

Klug, H. M., Chin, A., & St. Mary, C. M. 2005. The net effects of guarding on egg survivorship
in the flagfish, Jordanellafloridae. Animal Behaviour 69: 661-668.









Klug, H., Lindstrom, K., and St. Mary, C.M. 2006. Parents benefit from eating offspring:
density-dependent egg survivorship compensates for filial cannibalsim. Evolution 60:
2087-2095.

Kondoh, M. and Okuda, M. 2002. Mate availability influences filial cannibalism in fish with
paternal care. Animal Behaviour 63: 227-233.

Kozlowski, J. and Stearns, S.C. 1989. Hypotheses for the production of excess zygotes: models
of bet-hedging and selective abortion. Evolution 43: 1369-1377.

Kraak, S.B.M. 1996. Female preference and filial cannibalism Aidablennius sphynx (Teleostei,
Blenniidae): a combined field and laboratory study. Behavioural Processes 36: 85-97.

Kraak, S.B.M. and van den Berghe, E. P. 1992. Do females assess paternal quality by means of
test eggs? Animal Behaviour 43: 865-867.

Kume, G., Yamaguchi, A., and Taniuchi, T. 2000. Filial cannibalism in the paternal
mouthbrooding cardinalfish Apogon lineatus: egg production by the female as the nutrition
source for the mouthbrooding male. Environmental Biology of Fishes 58: 233-236.

Kvarnemo, C., Svensson, 0., and Forsgren, E. 1998. Parental behaviour in relation to food
availability in the common goby. Animal Behaviour 56: 1285-1290.

Lindstrom, K. 1998. Effects of costs and benefits of brood care in the sand goby. Behavioral
Ecology and Sociobiology 42: 101-106.

Lindstrom, K. 2000. The evolution of filial cannibalism and female mate choice strategies as
resolutions to sexual conflicts in fishes. Evolution 54: 617-627.

Lindstrom, K. and Sargent, R. C. 1997. Food access, brood size, and filial cannibalism in the
fantail darter, Etheostomaflabellare. Behavioral Ecology and Sociobiology 40: 107-110.

Lindstrom, K. St. Mary, C.M, and Pampoulie, C. 2006. Sexual selection for male parental care in
the sand goby, Pomatoschistus minutus. Behavioral Ecology and Sociobiology 60: 46-51.

Lissiker, M., Kvarnemo, C. and Svensson, 0. 2003. Effects of a low oxygen environment on
parental effort and filial cannibalism in the male sand goby, Pomatoschistus minutus.
Behavioral Ecology 14: 374-381.

McCormick, J.H. and Leino, R.L. 1999. Factors contributing to first year recruitment failure of
fishes in acidified waters with some implications for environmental research. Transactions
of the American Fisheries Society 128: 265 277.

McEdward, L. R., and Carson, S. F. 1987. Variation in egg organic content and its relationship
with egg size in the starfish Solaster stimpsoni. Marine Ecology Progress Series 37: 159-
169.









McNamara, J.M., Szekely, T., Webb, J.N. and Houston, A.I. 2000. A dynamic game-theoretic
model of parental care. Journal of Theoretical Biology 205: 605-623.

Manica, A. 2002. Filial cannibalism in teleost fish. Biological Reviews 77: 261-277.

Manica, A. 2004. Parental fish change their cannibalistic behaviour in response to the cost-to
benefit ratio of parental care. Animal Behaviour 67: 1015-1021.

Manly, B.F.J., Miller, P. and Cook, L.M. 1972. Analysis of a selective predation experiment.
American Naturalist 106: 719-736.

Melser, C. and Klinkhamer, PGL. 2001. Selective seed abortion increases offspring survival in
Cynoglossum officinale (Boraginaceae). American Journal of Botany. 88: 1033-1040.

Mertz, J. C. and Barlow, G. W. 1966. On the reproductive behavior of Jordanellafloridae
(Pisces: Cyprinodontidae) with special reference to a quantitative analysis of parental
fanning. Zeitscrift fur Tierpsychologie 23: 537 554.

Mock, DW and Forbes, LS. 1995. The evolution of parental optimism. Trends in Ecology and
Evolution 10: 130-134.

Mock, D.W. and G.A. Parker. 1997. The Evolution of Sibling Rivalry. Oxford University Press.

Mrowka, W. 1987. Filial cannibalism and reproductive success in the maternal mouthbrooding
cichlid fish Pseudocrenilabrus multicolor. Behavioral Ecology and Sociobiology 21- 257-
265.

Narimatsu, Y. and Munehara, H. 2001. Territoriality, egg desertion, and mating success of a
paternal care fish, Hypoptychusd Ybowskii. Behavior 138: 85-96.

Neff, B.D. 2003. Paternity and condition affect cannibalistic behavior in nest-tending bluegill
sunfish. Behavioral Ecology and Sociobiology 54: 377-384.

Neff, B.D. and Sherman, P.W. 2003. Nestling recognition via direct cues by parental male
bluegill sunfish (Lepomis macrochirus). Animal Cognition 6:87-92

Okuda, N. and Yanagisawa, Y. 1996. Filial cannibalism by the mouthbrooding cardinalfish,
Apogon doederlini, in relation to their physical condition. Environmental Biology of Fishes
45: 397-404.

Pampoulie C., Lindstrom K., and St. Mary C.M. 2004. Have your cake and eat it too: male sand
gobies show more parental care in the presence of female partners. Behavioural Ecology
15:199-204

Payne, A. G., Smith, C., and Campbell, A. 2002. Filial cannibalism improves survival and
development of beaugregory damselfish embryos. Proceedings of the Royal Society of
London Series B 269: 2095-2102.









Payne A. G., Smith, C., and Campbell, A. 2003. The effect of clutch size on whole clutch
cannibalism in the beaugregory damselfish. Journal of Fish Biology 62: 955-958.

Payne, A. G., Smith, C., and Campbell, A. C. 2004. A model of oxygen-mediated filial
cannibalism in fishes. Ecological Modelling 174: 253-266.

Petersen, C. W. and Marchetti, K. 1989. Filial cannibalism in the Cortez damselfish Stegastes
rectifraenum. Evolution 43: 158-168.

Polis, G. A. 1981. The evolution and dynamics of intraspecific predation. Annual Reviews in
Ecology and Systematics 12: 225-251.

Raymond, M. and Rousset, F. 1995. GENEPOP 1.2: Population genetics software for exact tests
and ecumenicism. Journal of Heredity 86: 248 249.

Reinert, K.H., Giddings, J.A., and Judd, L. 2002. Effects analysis of time-varying or repeated
exposures in aquatic ecological risk assessment of agrochemicals. Environmental
Toxicology and Chemistry 21: 1977 1992.

Rohwer, S. 1978. Parent cannibalism of offspring and egg raiding as a courtship strategy.
American Naturalist 112: 429-440.

Rice, W. 1989. Analysis tables of statistical tests. Evolution 43: 223 225.

Rosenblatt, J. S. and Snowdon, C. T. 1996. Parental care: evolution, mechanisms, and adaptive
significance. NY: Academic Press.

Rowe, D.W., Sprague, J.B., and Heming, T.A. 1983. Sublethal effects of treated liquid affluent
from a petroleum refinery: chronic toxicity to flagfish. Aquatic Toxicology 3: 149 159.

Rudolf, V.W. and Antonovics, J. 2007. Disease transmission by cannibalism: rare event or
common occurrence. Proceedings of the Royal Society B 274: 1205-1210.

Ruetz, C.R., Trexler, J.C., Jordan, F., Loftus, W., and Perry, S. 2005. Population dynamics of
wetland fishes: spatio-temporal patterns synchronized by hydrological disturbance?
Journal of Animal Ecology, 74, 322 332.

Salfert, I.G. and Moodie, G.E.E. 1985. Filial egg cannibalism in the brook stickleback, Culae
inconstans (Kirtland). Behaviour 93: 82-100.

Sargent, R. C. 1988. Paternal care and egg survival both increase with clutch size in the fathead
minnow, Pimephalespromelas. Behavioral Ecology and Sociobiology 23: 33 37.

Sargent, R. C. 1992. Ecology of filial cannibalism in fish: theoretical perspective. In:
Cannibalism: Ecology and Evolution Among Diverse Taxa (Ed. by M. A. Elgar & B. J.
Crespi), pp.38-62. Oxford: Oxford University Press.









Sargent, R. C. 1997. Parental care. In Behavioural Ecology of Teleost Fishes (Ed. by J.-G. J.
Godin), pp. 292-315.Oxford: Oxford University Press.

SAS Institute. 2001. SAS 8.2. Cary, NC: SAS Institute.

Sikkel, P. C. 1994. Filial cannibalism in a paternal-caring marine fish: the influence of egg
developmental stage and position in the nest. Animal Behaviour 47: 1149-1158.

Simon, M. P. 1983. The ecology of parental care in a territorial breeding frog from New Guinea.
Behavioral Ecology and Sociobiology 14: 61-67.

Smith, C. 1992. Filial cannibalism as a reproductive strategy in care-giving teleosts. Netherlands
Journal of Zoology 42: 607-613.

Smith, C. and Wootton, R. J. 1995. The costs of parental care in teleost fishes. Reviews in Fish
Biology and Fisheries 5: 7-22.

SPSS Inc. 2000. SYSTAT 9.0 for Windows. Chicago, IL: SPSS Inc.

St. Mary, C. M, Noureddine, C. G, and Lindstrom, K. 2001. Effects of the environment on male
reproductive success and parental care in the Florida flagfish, Jordanellafloridae.
Ethology 107: 1035-1052.

Stearns, S.C. 1987. The selection arena hypothesis, pp. 337-349. In. S.C. Steams (ed.) The
evolution of sex and its consequences. Birkhauser, Basel.

Stearns, S.C. 1992. The Evolution of Life Histories. Oxford University Press.

Svensson, O., Magnhagen, C., Forsgren, E. and Kvarnemo, C. 1998. Parental behaviour in
relation to the occurrence of sneaking in the common goby. Animal Behaviour 56: 175-
179.

Svensson, O. and Kvarnemo, C. 2007. Parasitic spawning in sand gobies: an experimental
assessment of nest-opening size, sneaker male cues, paternity, and filial cannibalism.
Behavioral Ecology 18: 410-419.

Takeyama, T., Okuda, N., and Yanagisawa, Y. 2002. Seasonal pattern of by Apogon doederleini
mouthbrooding males. Journal of Fish Biology 61: 633-644.

Takeyama, T., Okuda, N., and Yanagisawa, Y. 2007. Filial cannibalism as a conditional strategy
in males of a paternal mouthbrooding fish. Evolutinoary Ecology 21: 109-119.

Thomas, L.K. and Manica, A. 2003. Filial cannibalism in an assassin bug. Animal Behaviour 66:
205-210.

Trivers, R. L. 1972. Parental investment and sexual selection. In: Sexual Selection and the
Descent ofMan (Ed. by B. Campbell), pp. 136-179. Chicago: Aldine.









Vincent, T.L. and Brown, J.S. 2005. Evolutionary game theory, natural selection and darwinian
dynamics. Cambridge University Press, Cambridge.

Vinyoles, D., Cote, I., and de Sostoa, A. 1999. Egg cannibalism in river blennies: the role of
natural prey availability. Journal of Fish Biology 55: 1223-1232.

Warkentin, K.M. 2000. Wasp predation and wasp-induced hatching of red-eyed treefrog eggs.
Animal Behaviour 60: 503-510.

Webb, J.N., Houston, A.I., McNamara, J.M. and Szekely, T. 1999. Multiple patterns of parental
care. Animal Behaviour 58: 983-993.

Williams, G. C. 1975. Sex and Evolution. Princeton University Press, Princeton, NJ.

Williams, J.E. 2000. "The Coefficient of Condition of Fish". In Schneider, James C. (ed.) 2000.
Manual of fisheries survey methods II: with periodic updates. Michigan Department of
Natural Resources, fisheries Special Report 25, Ann Arbor.

Wittenberger, J. F. 1981. Animal Social Behavior. Duxbury Press, Boston:MA.

Yeh, F.C., Yang, R.C., and Boyle, T. 1999. POPGENE 1.31. Microsoft Windows-Based
Software for Population Genetics Analysis. University of Alberta and Centre for
International Forestry Research, Alberta, Canada.









BIOGRAPHICAL SKETCH

Hope Klug was born June 25, 1980, and grew up in Palm Harbor, FL. She graduated from

the International Baccalaureate Program at St. Petersburg High School in 1998, and received a

B.S. in zoology and psychology from the University of Florida in 2001. She began her Ph.D.

work in 2002, and plans to conduct post-doctoral research in Finland beginning in January 2008.





PAGE 1

EVOLUTIONARY SIGNIFICANCE OF FILIAL CANNIBALISM IN FISHES WITH PARENTAL CARE By HOPE KLUG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007 1

PAGE 2

2007 Hope Klug 2

PAGE 3

To my family, who has always supported me. 3

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank my advisor, H. Jane Brockmann, and my supervisory committee, Rebecca Kimball, Kai Lindstrm, Steve Phelps and Clive Wynne, for their guidance and feedback throughout my graduate career. Thei r constant support a nd thoughtfulness has facilitated this dissertation work and, more gene rally, my professional development as a scientist. I am indebted to the lab group of Ben Bolker Craig Osenberg, and Colette St. Mary for providing me with logistical support, as well as critical feedback dur ing the first four years of my graduate career. It was in their lab group that I learned to think and wr ite logically, and I am thankful for every bit of critical feedback I r eceived from this lab group. I thank Mike Bonsall, whose support, patience, and passion for science ma de chapter 6 possible. I am also grateful for the logistical support provided during two visits to Oxford. Similarly, Kai Lindstrm and his lab group provided me with space, supplies, and a great deal of assistance at Tvrminne Zoological Station. My work on the sand goby would have been impossible wit hout their support. My dissertation research was greatly imp roved by the funding I received throughout my graduate career. In particular, Im grateful for the funding provided by an NSF Graduate Research Fellowship, a SPICE fellowship (PI: Doug Levey), an NSF Dissertation Improvement Grant, and the Department of Zoology at the University of Florida. I also thank the Department of Zoology and the University of Helsinkis Tvrminne Zoological Station for additional logistical support. Finally, I thank my family and friends, w hose continuous love and support allowed me to pursue my interests in science. Without their encouragement, none of this would have been possible.

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS ...............................................................................................................4 LIST OF TABLES ...........................................................................................................................9 LIST OF FIGURES .......................................................................................................................10 ABSTRACT ...................................................................................................................................12 CHAPTER 1 INTRODUCTION................................................................................................................. .14 Background: Theory and Empirical Evidence ........................................................................14 Re-Evaluation of Current Theory with an Explicit Focus on Fitness Consequences ......17 Development and Evaluation of Alternative Hypotheses ...............................................17 Development of a Synthetic Model of Filial Cannibalism ..............................................18 Summary of Dissertation Objectives ......................................................................................18 Study Systems .........................................................................................................................19 Overview of Dissertation Chapters .........................................................................................20 Reproductive Fitness Consequences of Filial Cannibalism in the Flagfish ....................20 Parents Benefit from Eating Offspring: Density-Dependent Egg Survivorship Compensates for Filial Cannibalism ............................................................................21 Selective Filial Cannibalism in the Sand Goby ...............................................................21 Selective Filial Cannibalism in the Flagfish ....................................................................22 A Model of the Evolution of Pare ntal Care and Filial Cannibalism. ..............................23 Conclusions .....................................................................................................................24 2 REPRODUCTIVE FITNESS CONSEQUENC ES OF FILIAL CANNIBALISM IN THE FLAGFISH, JORDANELLA FLORIDAE ......................................................................25 Introduction .............................................................................................................................25 Methods ..................................................................................................................................28 Study Species ...................................................................................................................28 Experimental Design and Data Collection ......................................................................29 Statistical Analyses ..........................................................................................................31 Results .....................................................................................................................................32 Occurrence of Filial Cannibalism ....................................................................................32 Effect of Diet on Filial Cannibalism ...............................................................................32 Costs and Benefits of Filial Cannibalism for Reproduction ............................................32 Effects of Food and Access to E ggs on Components of Fitness .....................................33 Reproduction ............................................................................................................33 Male weight and length ............................................................................................33 Discussion ...............................................................................................................................34 5

PAGE 6

3 PARENTS BENEFIT FROM EATING O FFSPRING: DENSITY-DEPENDENT EGG SURVIVORSHIP COMPENSATES FOR FILIAL CANNIBALISM..................................42 Introduction .............................................................................................................................42 Methods ..................................................................................................................................46 Study Species and Experimental Site ..............................................................................46 Experimental Design .......................................................................................................46 Experiment 1: Effect of oxygen a nd egg density on filial cannibalism ...................46 Experiment 2: Effect of simulated filial cannibalism on egg survivorship ..............49 Data Analysis ...................................................................................................................50 Experiment 1: Effect of oxygen a nd egg density on filial cannibalism ...................50 Experiment 2: Effect of simulated filial cannibalism on egg survivorship ..............52 Results .....................................................................................................................................52 Experiment 1: Effect of Oxygen, Egg Density, and Male Condition on Filial Cannibalism .................................................................................................................52 Occurrence of whole clutch cannibalism .................................................................52 Egg survivorship ......................................................................................................53 Male condition ..........................................................................................................54 Experiment 2: Effect of Simulated Filial Cannibalism on Egg Survivorship .................55 Effect of oxygen and egg removal on remaining egg survivorship .........................55 Effect of oxygen and egg removal on total number of eggs surviving ....................55 Discussion ...............................................................................................................................56 4 SELECTIVE FILIAL CANNIBALISM IN THE SAND GOBY..........................................67 Introduction .............................................................................................................................67 Materials and Methods ...........................................................................................................68 Experimental Design .......................................................................................................68 Single-female set-up .................................................................................................70 Image Analysis .........................................................................................................70 Preference Calculation .....................................................................................................71 Statistical Analyses ..........................................................................................................72 Results .....................................................................................................................................73 Differences in Egg Size, Egg Density and Cannibalism Rates between Years ..............73 Egg Size, Survivorship, and Development Ti me in Eggs Reared in the Absence of Males ............................................................................................................................74 Cannibalistic Preferences by Males .................................................................................74 Discussion ...............................................................................................................................74 5 SELECTIVE FILIAL CANNIBALISM IN THE FLAGFISH..............................................80 Introduction .............................................................................................................................80 Methods ..................................................................................................................................82 Study Species ...................................................................................................................82 Experimental Design .......................................................................................................82 Energy Assays .................................................................................................................84 Statistics ...........................................................................................................................84 6

PAGE 7

Results .....................................................................................................................................85 Parental Condition and Size, Egg Energetic Content, and Egg Number .........................86 Whole Clutch Cannibalism ..............................................................................................87 Partial Clutch Cannibalism ..............................................................................................87 Discussion ...............................................................................................................................88 6 WHEN TO CARE FOR, ABANDON, OR EAT YOUR OFFSPRING: A MODEL OF THE EVOLUTION OF PARENTAL CA RE AND FILIAL CANNIBALISM.....................96 Introduction .............................................................................................................................96 Methods ..................................................................................................................................99 Model Dynamics ...........................................................................................................100 Resident and Mutant Trade-Offs ...................................................................................101 Invasion Dynamics and Fitness .....................................................................................103 Biologically Relevant Comparisons ..............................................................................105 Results ...................................................................................................................................108 Invasion of Parental Care ..............................................................................................108 Effects of egg maturation rate, egg de ath rate, adult reproductive rate, and carrying capacity .................................................................................................108 Effect of cannibalism on the evolution of care ......................................................109 Invasion of Filial Cannibalism (With and Without Parental Care) ...............................109 Effects of Egg Maturation Rate, Reproduc tive Rate, and Selective Cannibalism .109 Effects of Density-Depe ndent Egg Survivorship ..........................................................110 Effects of Energetic Benefits of Consuming Offspring ................................................111 Effects of Carrying Capacity .........................................................................................111 Discussion .............................................................................................................................112 7 GENERAL CONCLUSI ONS AND SYNTHESIS..............................................................127 Introduction ...........................................................................................................................127 Are the Current Energy-Based and O xygen-Mediated Hypotheses Sufficient? ..................129 An Alternative Hypothesis: Selective Filial Cannibalism ....................................................131 The Plausibility of Multiple Hypotheses ..............................................................................134 Future Directions ..................................................................................................................135 Determining the Relative Importance of Varying Factors ............................................136 Role of Environmental Variation ..................................................................................136 The Non-Cannibalistic Parent .......................................................................................137 Identification of Additional Speci es Practicing Filial Cannibalism ..............................137 A Comparative Framework of Filial Cannibalism ........................................................137 Why Dont All Parents Exhibit Filial Cannibalism? .....................................................138 APPENDIX ISOLATION AND CHARACTER IZATION OF MICROSATELLITE DNA MARKERS FOR THE FLAGFISH.....................................................................................139 7

PAGE 8

8 LIST OF REFERENCES .............................................................................................................143 BIOGRAPHICAL SKETCH .......................................................................................................151

PAGE 9

LIST OF TABLES Table page 6-1 Trade-off functions associated with parental care and filial cannibalism.. ..........................119 6-2 Alternative hypotheses regarding the evol utionary significance of filial cannibalism (FC).. ................................................................................................................................120 A-1 Characteristics of flagfish microsatellite loci.. ....................................................................142 9

PAGE 10

LIST OF FIGURES Figure page 2-1 Expected and observed benefit of filial cannibalism in eggs received by males.. .................40 2-2 Effect of filial cannibalism on components of fitness.. ..........................................................41 3-1 Effect of oxygen and egg density on the prevalence of whole clutch cannibalism by parental males.. ..................................................................................................................62 3-2 Effect of oxygen and egg density on the me an (+/SE) egg survivorship (i.e., proportion of the clutch that survived until hatching) when males were present with eggs, including cases of both whole a nd partial clutch cannibalism.. .........................................63 3-3 Relationship between A) male condition (i.e., K=100*g/cm ) and the proportion of the clutch consumed by parental males, and B) partial clutch cannibalism and change in male condition.3...................................................................................................................64 3-4 Effects of simulated filial cannibalism ...................................................................................65 3-5 Simulated filial cannibalism: The effect of oxygen and egg removal on the total number of eggs surviving, including cases of both whole and partial clutch death.. ......................66 4-1 Relationship between initial egg size a nd development time (i.e. the number of days from spawning until hatching) in eggs reared in the absence of males. ............................77 4-2 Preferences in egg consumption by parental males. ...............................................................78 4-3 Distribution of egg size init ially and in the male diet. pro portion of eggs in A) 2004 and B) 2006...............................................................................................................................79 5-1 Relationship between the mean energy per egg (J egg ) within a clutch and A) female weight and B) male weight.-1...............................................................................................91 5-2 Relationship between the frequency of whole clutch cannibalism and A) the mean energy per egg (J egg ) within a clutch and B) female weight..-1.......................................92 5-3 Relationship between male weight and A) the proportion and B) the number of eggs consumed for cases of pa rtial clutch cannibalism. .............................................................93 5-4 Relationship between female weight a nd A) the proportion and B) the number of eggs consumed for cases of pa rtial clutch cannibalism. .............................................................94 5-5 Relationship between the mean energy per egg (J egg ) within a clutch and the number of eggs consumed for cases of partial clutch cannibalism.-1................................................95 6-1 Diagram of the model.. .........................................................................................................121 10

PAGE 11

11 6-2 Invasion dynamics of parental care. .....................................................................................122 6-3 Invasion dynamics of filial cannibalism ...............................................................................123 6-4 Effect of density-dependent egg survivorship on the evolution of parental care and filial cannibalism. .....................................................................................................................124 6-5 Effect of energetic benefits on the evolution of filial cannibalism. ......................................125 6-6 Effect of carrying capacity on the evolution of filial cannibalism. ......................................126

PAGE 12

Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EVOLUTIONARY SIGNIFICANCE OF FILIAL CANNIBALISM IN FISHES WITH PARENTAL CARE By Hope Klug December 2007 Chair: H. J. Brockmann Major: Zoology Parental care typically increas es offspring survival, thereby increasing parental fitness. Thus, it is surprising that filial cannibalism, the consumption of ones own offspring, is prevalent in fishes exhibiting parental care. The most widely-accepted hypothesis of filial cannibalism suggests that males gain energy from eggs that they invest into future reproduction (energybased hypothesis). Recently, an alternative hypothesis suggested that partia l-clutch cannibalism increases oxygen availability to remaining eggs, which in turn increases overall egg survival (oxygen-mediated hypothesis). Evidence for both hypotheses is mixed and there are few alternative hypotheses. Thus, the evolutionary significance of filial cannibalism remains unclear. To enhance our understanding of filial cannibalism, I re-exami ned current theory (i.e., the energy-based and oxygen-mediated hypotheses), developed and evaluated an alternative hypothesis, and developed a mathematical model of filial cannibalism. I experimentally quantified the effect of f ilial cannibalism on mating success of parental males in the flagfish (Jordanella floridae ), and found that filial cannibalism always reduced lifetime reproductive suc cess. In the sand goby ( Pomatoschistus minutus ), males that were in poorer condition consumed less of their eggs than males that we re in better condition. These findings are contrary to predicti ons of the energy-based hypothesis. 12

PAGE 13

13 In the sand goby, I found that egg survival is density-dependent a nd filial cannibalism increases when egg density is high. However, this density-dependence is not mediated by oxygen. Therefore, I did not find support for the oxygen-mediated hypothesis. I suggest a more general hypothesis of filial cannibalism mediated by density -dependent egg survival. I hypothesize that the ability to preferential ly cannibalize offspring of reduced quality might play a large role in the evolution of filial cannibalism (selective filial cannibalism hypothesis). To begin to understand the importance of selective cannibalism, I evaluated whether males cannibalize selectively in the sand goby and the flagfish. Male sand gobies cannibalized selectively with regard to egg development rate, and male flagfi sh cannibalized selectively with regard to egg energy and maternal size. Thus, selective filial ca nnibalism occurs in at least two species and this hypothesis wa rrants further attention. I then developed a mathematical model of filial cannibalism to isolate factors affecting the evolution of filial cannibalism. The findings of this model highlight the plausibility of a range of alternative hypotheses. Specifically, the evolu tion of filial cannibalism is enhanced if (1) parents can selectively cannibalize lower quality offspring, (2) filial cannibalism increases egg maturation rate, (3) there are energetic benefits of eggs to cannibalizing males, (4) cannibalism increases a parents reproductive rate (e.g., through mate attractiveness) Density-dependent egg survivorship alone did not favor the evolution of cannibalism. Additiona lly, the results of the model suggest that population-level dynamics pot entially play a large ro le in the evolution. Finally, I isolated and characte rized six polymorphic microsatellite loci in the flagfish. These microsatellite markers will be useful in paternity assays and estimating heritability of flagfish behaviors, such as filial cannibalism.

PAGE 14

CHAPTER 1 INTRODUCTION Background: Theory and Empirical Evidence Adaptive theories of evolution suggest that pa rents should exhibit strategies that increase offspring survival. Parental care, which is comm on across taxa, is one way in which parents are thought to do this (reviewed in Clutton-Broc k 1991). Because parental care is assumed to increase offspring survival, it is surprising th at filial cannibalism, the consumption of ones own offspring, may co-occur with parental care. Indee d, filial cannibalism is prevalent in a range of taxa (Polis 1981) and is particul arly common in teleost fishes e xhibiting paternal care (reviewed in Manica 2002). Because parental males often consume more eggs than would die naturally, filial cannibalism does not solely serve to clean the nest of dead or diseased eggs (e.g. Klug et al. 2005). Likewise, egg cannibalism cannot be attributed to the removal of eggs that were fertilized by other males (DeWoody et al. 2001; but see Neff and Sherman 2003), and filial cannibalism remains common in the absence of possible cuckolding events (Kva rnemo et al. 1998; Kume et al. 2000; Lissker et al. 2002; Klug et al. 2005; Svensson ad Kvarnemo 2007). Despite some theoretical development, the evolutionary significance of filial cannibalism remains unclear. Early ethologists considered filial cannibalism a social pathology of little or no adaptive significance. However, more recent studies sugges t that filial cannibalism reflects an adaptive trade-off between current and future reproductiv e success, in which males gain energy and nutrients from consumed eggs that are then re invested into future reproduction (Rohwer 1978; Sargent 1992). According to this energy-based h ypothesis, whole clutch cannibalism (i.e. the consumption of all eggs present) is expected to be more frequent within a species when clutch size is relatively small because the energy requir ement of the males can be satisfied only by clutches larger than a certain size (i.e., when clut ch size is small, males must consume all eggs to 14

PAGE 15

15 satisfy their energy requirements) Empirical studies support this prediction (reviewed by Manica 2002). Theory also predicts that cannibalism should increase as food av ailability decreases and/or when male condition is poor (Rohwer 1978; Sargent 1992). However, evidence with regard to these predictions is equivocal. Supplementary feeding of parental males reduces partial clutch cannibalism (i.e., the consumption of only some eggs in a nest) in the common goby ( Pomatoschitus microps Kvarnemo et al. 1998), the scissortail sergeant ( Abudefduf sexfasciatus, Manica 2004), and the Cortez damselfish ( Stegastes rectifraenum Hoelzer 1992), but food availability and/or male condition did not a ffect filial cannibalism in fantail darters ( Etheostoma flabellare, Lindstrm and Sargent 1997) and the three-spined stickleback ( Gasterosteus aculeatus, Belles-Isles and Fitz gerald 1991). Still other studies have quantified the energetic value of eggs to determine if eggs can provide a caring male with sufficient energy to offset the direct fitness costs of care. In one species, energy attained from partial brood cannibalism was found to be sufficient to offset costs related to care ( Apogon lineatus, Kume et al. 2000), while in another, energy from eggs wa s found to be insufficient ( Gasterosteus aculeatus, Smith 1992). Thus, there is no consistent support for the energy-based explana tion of filial cannibalism. Payne et al. (2002) recently suggested that by co nsuming some eggs in their nest, parental males can improve oxygen availability of the re maining eggs, thereby improving overall egg survivorship (oxygen-mediated hypothesis; Payn e et al. 2002, 2004). However, there has been only limited empirical examination of this hypo thesis, which does not account for the common occurrence of filial cannibalism in high oxygen e nvironments. For example, filial cannibalism occurs when oxygen availability is relatively high in the sand goby ( Pomatoschistus minutus, Klug et al. 2006), the flagfish (Jordanella floridae, Klug et al. 2005), and fantail darters ( Etheostoma flabellare, Lindstrm and Sargent 1997). Add itionally, Lissker et al. (2002) found

PAGE 16

no differences in partial or whole clutch can nibalism across oxygen levels. Thus, neither the energy-based hypothesis nor the oxygen-mediated hypothesis can explain the occurrence of filial cannibalism. Furthermore, there has been a gene ral lack of alternative explanations for the widespread occurrence of filial cannibalism in fishes. As an alternative to the energy-based and oxygen-mediated hypotheses, I propose the hypothesis of selective filial cannibalism (i.e., non-random consumption of offspring with regard to some aspect of offspring phenotype). Specifi cally, I hypothesize that the ability to cannibalize offspring selectively in relation to aspects of offspring quality can directly favor the evolution of filial cannibalism (see also Klug and Bonsall 200 7 and Chapter 6). The idea of weeding out inferior offspring has been documented in ot her contexts (brood reduction: Forbes & Mock 1998; selective abortion in humans: Stearns 1987; Forbes 1997; Diamond 1987; Hesketh & Xing 2006; selective abortion in plants: Burd 1998; Karkkainen et al. 1999; Melser & Klinkhamer 2001), but it has not yet been explicitly evaluated in relation to filial cannibalism (but see Mrowka 1987 and Kraak 1996 for work on consumption of unfertilized or diseased eggs). Indeed, selective elimina tion of lower quality offspring is thought to play a large role in the evolution of select ive embryo abortion, brood reduction, a nd offspring abandonment (Stearns 1987; Forbes and Mock 1998; Stearns 1992). Howe ver, the relationship between offspring phenotype and filial cannibalism remains unknown ( but see Chapters 4, 5, and 6), and thus, the relative importance of selectiv e filial cannibalism is unclear. Despite mixed empirical evidence and a lack of alternative hypothe ses, the energetic hypothesis of filial cannibalism is typically accepted as a valid explanation for the prevalence of filial cannibalism in fishes (e.g. Vinyoles et al 1999; Kume at al. 2000; Manica 2002). Overall, the support for this hypothesis is unsatisfying a nd it remains unclear in many systems whether or 16

PAGE 17

not (and potentially how) filial cannibalism is an adaptive strategy. Despite nearly three decades of empirical and theoretical work on the subject filial cannibalism in fishes still remains an evolutionary conundrum. Understanding the evol utionary significance of filial cannibalism requires both additional theoretical de velopment and empirical research. An enhanced understanding of the evoluti on of filial cannibalis m necessitates three research strategies. Re-Evaluation of Current Theory with an Explicit Focus on Fitness Consequences Previous studies have failed to quantify fitness consequences of filial cannibalism (i.e., the parent and offspring survivorship and/or reproductive benefits of egg eating). Most studies of filial cannibalism have focused on the relationship between male condition and filial cannibalism, or the effect of supplemental feeding on filial canni balism. Simply showing that males with less food eat more eggs does not demons trate that an adaptive trade-off is occurring. Thus, there is limited support for the energy-base d hypothesis, and as mentioned previously, the oxygen-mediated hypothesis has received little empirical ev aluation. The first step in understanding the evolutionary significance of filia l cannibalism is to explicitly evaluate the energy-based and oxygen-mediated hypotheses by focu sing on potential fitness consequences of filial cannibalism (i.e., the parent and offspr ing survivorship and reproductive costs and benefits). Development and Evaluation of Alternative Hypotheses The energy-based hypothesis continues to be accepted despite mixed evidence. Thus the development and empirical examination of new hypotheses is critical. As mentioned above, I propose the alternative hypothesis of selective filial cannibalism (d iscussed further in Chapters 4, 5, and 6). Specifically, I hypothesi ze that 1) parental males mi ght preferentially cannibalize lower quality offspring and 2) the ability to canni balize selectively potentially plays a large role 17

PAGE 18

Development of a Synthetic Mo del of Filial Cannibalism The two current hypotheses (i.e., the ener gy-based and oxygen-medi ated) are currently presented as mutually exclusive ideas, leading to a debate that has resulted in very little resolution. Instead, I argue that the prevalence of filial cannibalism cannot be explained by single factors (i.e., energy or oxygen). A model incorporating a variety of trade-offs associated with current and future reproductive success (i.e., a multiv ariate approach) is essential to identify and evaluate the plausibility of a range of alternative hypotheses. Summary of Dissertation Objectives The goals of my dissertation are three-fol d, and consist of four experiments and one mathematical model. First, I re-evaluate current theory by focusi ng on fitness consequences of cannibalism. In my first experiment, I evaluate the reproductive fitness consequen ces of filial cannibalism. To evaluate the energy-based hypothe sis, I quantify the effect of filial cannibalism on long-term mating success of parental males (i.e., current and future reproductive success) and determine whether energy consumed from eggs is directly translated into future reproduction in the flagfish. In my second experiment, I examine the relati onship between partial cl utch cannibalism and egg survivorship in relation to oxygen availabilit y. I evaluate the oxygen-mediated hypothesis by quantifying the effect of oxygen and egg density on filial cannibalism and net egg survival. Second, I develop and evaluate an alternat ive hypothesis. In my third and fourth experiments, I assess the relative importance of aspects of egg phenotype on filial cannibalism in the sand goby and flagfish. To begin to evalua te the hypothesis of selective filial cannibalism proposed above, I assess whether parental males preferentially consume eggs with regard to 18

PAGE 19

some aspect of offspring phenotype. Specifical ly, I focus on the relationship between filial cannibalism and egg size or energeti c content, because egg size and energetic content have been correlated with post-hatching survival and growth in fishes (Kamler 2005). Finally, I fully develop a mathem atical model incorporating a range of costs and benefits associated with filial cannibalism (i.e., the multivariate approach). I use this model to isolate the pivotal factors affecting the evolutionary dynami cs of filial cannibalism, and I evaluate the plausibility of a range of non-mutua lly exclusive alternative hypotheses. Study Systems Filial cannibalism likely functions differen tly among systems, and because there are potentially multiple explanations for the occu rrence of filial cannibalism, I focused on two distinct systems: the flagfish (Jordanella floridae) and the sand goby ( Pomatoschistus minutus ). Both systems are exemplary of care in fishes: (i ) males provide care (i.e., nest guarding, fanning, and cleaning), (ii) filial cannibalism is prevalent, and (iii) they live approximately one year, allowing fitness measurements quantified over on e breeding season to be representative of lifetime fitness. However, the species have diffe rent life histories (e gg laying patterns, nest structure, rates of mating), thus making them su itable for testing particul ar hypotheses related to filial cannibalism. Flagfish males defend open nests consisting of filamentous algae or bare substratum. Male flagfish fan eggs but fanning is not necessary for egg survival, and flagfish eggs are not clumped together (Mertz & Barlow 1966). Thus, limited oxyge n at the nest is not a main source of egg mortality, and there is no a priori expectation in flagfish that filial cannibalism leads to increased survivorship of remaining eggs through increased oxygen availability. Unlike flagfish, sand goby males build nests under shells and cover the nest with sand, leaving only a small (~1cm) opening. Sand goby eggs are clumped and oxygenation of the eggs by male fanning is required 19

PAGE 20

for egg survival. Thus, the sand goby is an id eal system to test the hypothesis of oxygenmediated filial cannibalism. Males of both species sequentially receive eggs from multiple females, increasing the possibility that eggs vary in quality. Overview of Dissertation Chapters Reproductive Fitness Consequences of Filial Cannibalism in the Flagfish In Chapter 2, I investigate the effects of f ilial cannibalism on components of fitness in the flagfish. Specifically, I expect fi lial cannibalism to increase fu ture reproductive success. To quantify the reproductive fitness consequences of filial cannibalism, I experimentally manipulate (1) the diet of parental males (i .e., males either receive a high qua lity or low quality diet) and (2) the ability to practice filial cannibalism (i.e., males either have full access to eggs or access to eggs is prevented by a nest cover). Each male experiences a single experi mental treatment (i.e., high food, access to eggs; high food, no access to eggs; low food, access to eggs; low food, no access to eggs) for the duration of the experiment. Specifically, I follow males over 90 days, approximately one breeding season (i.e., the expected reproductive lifetime of flagfish), and consider three components of fitn ess: 1) the total number of e ggs received, 2) the number of spawnings, and 3) the frequency of spawning. Contra ry to predictions, filial cannibalism reduces male reproductive success. While an enhanced di et increases the number of eggs received, the number of spawnings, the freque ncy of spawning, and male weight gain, there is no effect of filial cannibalism on any component of reproducti ve success or male weight. Thus, in the flagfish there is no evidence that energy or nutrients gained from filial cannibalism are directly translated into increased future reproductive success. See also Klug and St. Mary (2005). 20

PAGE 21

Parents Benefit from Eating Offspring: De nsity-Dependent Egg Survivorship Compensates for Filial Cannibalism In Chapter 3, I evaluate the hypothesis of oxyge n-mediated filial cannibalism in the sand goby by 1) examining the effect of oxygen and egg density on the occurrence of filial cannibalism, 2) evaluating the effects of partia l clutch cannibalism on th e survivorship of the remaining eggs, and (3) comparing potential costs a nd benefits of filial cannibalism related to the net number of eggs surviving. I find that oxygen le vel and egg density aff ect the occurrence of cannibalism and that simulated partial clutch cannibalism improves survivorship of the remaining eggs. Additionally, because increased egg survivorship, stemming from partial egg removal, compensates for the cost of cannibalism (i.e., number of eggs removed) at a range of cannibalism levels, filial cannibalism potentially results in no net losses in reproductive success. However, oxygen does not affect eg g survivorship. Thus, in this chapter, I suggest a more general hypothesis of filial ca nnibalism mediated by density-dependent egg survivorship. See also Klug et al. 2006. Selective Filial Cannibalism in the Sand Goby In Chapter 4, I evaluate the novel hypothesis of selective filial cannibalism. Specifically, I examine the relationship between aspects of egg phenotype and partial clutch filial cannibalism in the sand goby. Males are either allowed to spawn with one or two females, and I then evaluate cannibalistic preferences in relation to the order in which eggs were spawned and egg size. I focus on egg size because egg size has been correlat ed with post-hatching su rvival and growth in a range of fishes (Kamler 2005). In this experiment I find that males selectively cannibalize eggs with respect to egg phenotype, but only in some cases. When males mated with two females sequentially, they preferentially consume the larger eggs of the second female. Because egg size is correlated with development time (and because female 2's eggs were slightly younger), the 21

PAGE 22

patterns of cannibalism appear to represent a st rategy that reduces the duration of care necessary for the current brood. By reducing the duration of care required for each br ood, it is possible that males can attain additional brood cycles duri ng their breeding season, th ereby increasing their reproductive success. This work highlights the pote ntial role of selectiv ity in understanding the adaptive significance of filial cannibalism. Selective Filial Cannibalism in the Flagfish In Chapter 5, I examine the relationship between maternal size and condition, egg energetic content, and filial canni balism. Both egg energetic conten t and maternal size have been correlated with post-hatching surviv al in a range of fishes (Kam ler 2005). In this experiment, I find that males exhibit preferential cannibalism with regard to mean energetic content per egg and maternal size. However, the patterns of selective cannibalism differ between whole and partial clutch cannibalism. Specifically, males cannibalize the whole clutch more often when the mean per-egg energetic content is relatively hi gh. Because I find no relationship between mean egg energy and egg number, the in crease in whole clutch cannibalism when mean egg energetic content is relatively great is not explained by any differences in e gg number. In contrast, there is a negative relationship between mean egg energy content and the number of eggs consumed for the case of partial clutch canniba lism. Similarly, maternal size is negatively correlated with the proportion of and the number of eggs consumed. Because female size has been positively related to post-hatching survival in other fishes, it seems that for the case of partial clutch cannibalism, males in this experiment preferentially canniba lize offspring that have lower survival posthatching. In contrast, when they practice whol e clutch cannibalism, they appear to be maximizing their energetic gain. The findings in this chapter (in combin ation with those of Chapter 4) suggest that males practice selectiv e filial cannibalism with regard to aspects of offspring phenotype. 22

PAGE 23

A Model of the Evolution of Parental Care and Filial Cannibalism In Chapter 6, I evaluate the importance of a range of factors on the evolution of parental care and filial cannibalism using an evolutionary ecology approach (see also Klug and Bonsall 2007). Parental care, no care/total abandonment, and filial cannibali sm evolve and often co-exist over a range of life-history strategi es (i.e., strategies associated w ith a wide range of adult death rates, egg death rates, egg ma turation rates, population carrying capacities, juvenile survival rates). While no single benefit is essential for the evolution of filial cannibalism, benefits associated with adult or offspring survival a nd/or reproduction facili tate the evolution of cannibalism. This model highlights the plausibi lity of a range of alternative hypotheses. Specifically, the evolution of filial cannibalism is enhanced if 1) parents can selectively cannibalize lower quality offspri ng, 2) filial cannibalism increases egg maturation rate, 3) there are energetic benefits of eggs to cannibalizi ng males, 4) cannibalism increases a parents reproductive rate (e.g., through mate attractiveness). Density-dependent egg survivorship alone does not favor the evolution of cannibalism. Wh en egg survivorship is density-dependent, parents are expected to simply lay eggs at densities that ma ximize offspring survival. While density-dependent egg survival does not directly favor the evoluti on of filial cannibalism in the model, density-dependent egg su rvival does not preclude the e volution of filial cannibalism entirely. For the case of density-dependent egg survival, filial cannibalism evolves more often when the density-dependence is relatively more in tense (i.e., when there is a relatively large increase in offspring mortality with increasing density). Additionally, th e evolution of filial cannibalism and/or parental care is highly sensitive to population ca rrying capacity in the model, and care and/or cannibalism are more likely to evolve if they allow an organism to use resources more effectively. These results suggest that population-level resource competition potentially plays an important role in the evolution of both parental care and filial cannibalism. 23

PAGE 24

Conclusions In this final chapter, I summarize the major conclusions of my dissertation. I discuss my findings in relation to previous theory, and highlight future directions of research on filial cannibalism. Isolation and Characterization of Mi crosatellite Markers in the Flagfish In the Appendix (Klug, St. Mary and Clark, in preparation), I describe six polymorphic microsatellite loci in the flagfish. These microsate llite markers will be useful in future behavioral research on the flagfish for estimating pa ternity, reproductive success, and heritability. 24

PAGE 25

CHAPTER 2 REPRODUCTIVE FITNESS CONSEQUENCES OF FILIAL CANNIBALISM IN THE FLAGFISH, JORDANELLA FLORIDAE Introduction Parental care, which is assumed to increase the fitness of a parent by increasing survival or quality of the offspring (Clutton-Brock 1991), is common across animal taxa (reviewed in Clutton-Brock 1991; Rosenblatt & Snowdon 1996). In fishes care is typically paternal, and includes territory guarding and nest care (Gross and Sargent 1985) Because parental care is assumed to increase the survival of offspring, it is surprising that f ilial cannibalism, the consumption of ones own offspring, is prevalent in fishes exhibiting parent al care (reviewed in Manica 2002). Since parental males often consume more eggs than would die naturally, filial cannibalism does not solely serve to clean the nest of dead or di seased eggs (e.g., Klug et al. 2005). While early ethologists considered filial cannibalism a social pathology with little or no adaptive significance, filial cannibalism is cu rrently thought to reflect an adaptive trade-off between current and future reproductive success (Rohwer 1978). Whole clutch cannibalism (i.e., the consumpti on of all eggs present) and partial clutch cannibalism (i.e., the consumption of some eggs present) are thought to occur for different reasons. Whole clutch cannibalism theoretically represents a termination of parental care (Rohwer 1978; reviewed in Manica 2002). Because the net reproductive gain associated with caring for a small clutch is expected to be le ss than that of a larg e clutch, whole clutch cannibalism is expected to be more frequent when clutch size is relativel y small. Indeed, in many systems with paternal care, smaller clutches are subject to whole clutch cannibalism more frequently than larger clutch es (reviewed in Manica 2002). The most widely accepted explanation for pa rtial-clutch filial cannibalism is that it provides a male with energy or limited nutrient s allowing him to care for the remaining brood or 25

PAGE 26

to increase future reproduction (Rohwer 1978; Sa rgent 1992). According to this energy-based hypothesis of filial cannibalism, food availabi lity and male condition should affect the occurrence of filial cannibalism (Rohwer 1978; Sargent 1992; Sargent 1997). Evidence regarding these predictions is mixed. Contrary to the theory, neither initial male condition nor food availability predicted th e number of eggs consumed in fantail darters, Etheostoma flabellare (Lindstrm and Sargent 1997). Similarly, food availa bility was not related to the occurrence of filial cannibalism in the three-spined stickleback Gasterosteus aculeatus (Belles-Isles and Fitzgerald 1991). However, evidence in other speci es suggests food availability does affect filial cannibalism. Kvarnemo et al. ( 1998) found that starved male common gobies consumed more of their eggs than males whose diet was supplemente d with either mussel meat or both mussel meat and the eggs of conspecifics. Similarly, Ma nica (2004) found that supplementary feeding significantly reduced partial clutch filial cannibalism in the scissortail sergeant ( Abudefduf sexfasciatus ). In the Cortez damselfish ( Stegastes rectifraenum ), supplementation of a males diet with conspecific eggs redu ced cannibalism but did not fully inhibit it (Hoelzer 1992). Other studies have quantified the energetic value of eggs to determine if eggs can provide a caring male with sufficient energy to offset the direct fitn ess costs of cannibalism. In one species energy attained from partial brood cannibalism was found to be sufficient to offset costs related to care ( Apogon lineatus, Kume et al. 2000), while in another, energy from eggs was found to be insufficient ( Gasterosteus aculeatus, Smith 1992). Thus, there is a lack of consistent support across fish species for the energy-based hypothe sis of filial cannibali sm. Despite the mixed empirical support, the energy-based hypothesis of filial cannibalism is generally accepted as valid (e.g. Manica 2002), and it remains unclear in many systems whether (and potentially how) filial cannibalism is an adaptive strategy. While prev ious studies have focused on testing some of 26

PAGE 27

the assumptions of energy-based models, simply showing that males with less food eat more eggs does not demonstrate that an adaptive tr ade-off is occurring. A thorough examination of current theory instead necessitates an explicit ex amination of the fitness consequences of filial cannibalism (i.e., the survivorship and/or reproductive consequen ces of egg eating). According to the energy-based explanation of filial cannibalism (Rohwer 1978), we would expect cannibalism to increase lifetime reproductiv e success, and in a short-lived species we might expect filial cannibalism to increase reproductive success over the course of one breeding season. The present study quantified reproductive fitness consequences of filial cannibalism over approximately one breeding season in a short-live d species of fish exhibiting filial cannibalism. Specifically, I was interested in whether male s that eat eggs experience an increase in reproduction. The flagfish, Jordanella floridae, is an ideal subject for such a study because 1) males show care behavior, i.e., nest guarding, fanning, and cleani ng (Mertz and Barlow 1966), 2) filial cannibalism is prevalent (Mertz and Barlow 1966; Foster et al. 1969; Klug et al. 2005), and 3) they live approximately one year (H. Klug, unpub lished data), allowing fitness measurements quantified over one breeding season to be representative of lifetime fitness. In order to evaluate the effect of filial cannibalism on reproductiv e output, it is necessary to compare the reproductive success of males who are able to consume eggs and males that are prevented from consuming eggs. To do so effectively, males should be the only egg consumers and thus the experimental design must limit the ac cess of potential egg predators, even females, to the eggs. Males not allowed to cannibalize must be excluded from their eggs yet still encouraged to provide parental care. Thus, these males must have access to their nest (with eggs covered) and yet not be allowed to continue spawning (as they would then have the opportunity to eat some eggs). Additionally, such a study must be designed so that males are energy-limited 27

PAGE 28

and consume a substantial number of eggs in or der to adequately eval uate benefits of egg consumption. The evaluation of the effect of f ilial cannibalism on male survivorship requires additional environmental controls; such a st udy must be done as above (to control egg consumption) but also in the presence of predators of the male and with other natural stressors. Because of the difficulty of evaluating the survivor ship effects of filial cannibalism, I focus here on the reproductive effects of filial cannibalism. I explore the reproductive fitness consequences of filial cannibalism in the flagfish by comparing two cannibalism treatments, one in which males have the opportunity to consume their eggs and another in which filial cannibalis m is prevented, over 90 days, the approximate duration of a flagfish breeding season in north central Florida (Hale, unpublished data). Three components of reproductive fitness were considered : 1) total number of e ggs received, 2) total number of spawnings, and 3) the frequency of spawning. Specifically, I evaluated whether energy gained from the consumption of eggs was dire ctly translated into an increase in the total number of eggs received by a male. In addition, I examined the effect of filial cannibalism on components of male condition. While I only fo cus on reproduction here, systematically quantifying specific effects of filial cannibalism on components of fitness will provide key insight into the validity of current models of filial cannibalism. Methods Study Species Flagfish live approximately one year, and flag fish males care for their nest by guarding, cleaning, and fanning. The incubation period of eggs is typically less than one week, and males are able to feed on vegetation a nd invertebrates, whic h are common near nesting sites. Flagfish females spawn with multiple males, and brood cycling (i.e., the alteration of courtship and mating periods with periods of full brood care) is absent in this species. Nesting males 28

PAGE 29

potentially receive eggs continuous ly but may also care for only a single brood. In north central Florida, I have observed flagfi sh breeding approximately May through September. Flagfish have been successfully used in laboratory studies be fore (e.g., Mertz and Barlow 1966; Hale et al. 2003; St. Mary et al. 2001; Klug et al. 2005) and readily adapt to being housed in aquaria. Experimental Design and Data Collection The experiment was conducted in Gainesville Florida beginning in May and ending in November 2002. Due to space restrictions, two blocks of the experiment were completed, the first beginning in May and the second beginning in September. To replicate peak breeding season conditions in both blocks, the experi ment was conducted in an environmentallycontrolled room where temperatur e and lighting could be held c onstant. The two blocks did not differ significantly (i.e., there were no significant block effects in a ny of the analyses) so the data were pooled for all analyses. Males were used on ly once in the experiment and were euthanized after the experiment. Flagfish were collected from the Otter Creek/ Waccasassa River drainage just prior to use and both sexes were housed in separate 150 l freshwater holding tank maintained at approximately 28C prior to use. During this time the fish were fed ad libitum a diet consisting of algae tablets and frozen brine shrimp. In all cases the experiment began by placi ng one male and three females in a 36 l freshwater aquarium (measuring 26 cm x 30 cm x 30 cm) equipped with air-driven, activated carbon and Dacron floss filtration, a sp awning mat, and three artificial Ludwigia plants The mat consisted of a 100 cm2 tile covered with heavy, green acrylic felt carpet. The fish experienced a regular 14-hour daylight period and temperature was maintained at a constant 29oC. All males were randomly assigned to a treatment. Each male was allowed to spawn 90 days from his initial spawning date (i.e., date of first spawning for each male). I crossed tw o filial cannibalism treatments (access and no access to eggs) with two feeding regi mes (low and high food). Thus, 29

PAGE 30

there were four treatments: 1) high food and no filial cannibalism (HF NFC), 2) high food and filial cannibalism (HF FC), 3) low food and no f ilial cannibalism (LF NFC), and 4) low food and filial cannibalism (LF FC). Fish in a low f ood treatment were fed one algae tab weighing approximately 0.29 g every second day. Fish in a high food treatment received one algae tab weighing approximately 0.52 g and 1.2 g (wet weight ) of frozen brine shrimp daily. During the experiment, fish were fed at approximately 1500 hrs daily. As mentioned above, males either had access to their eggs or access was denied by covering the eggs with a screen nest cover. The nest covers used in this experiment were 182.5 cm2 and consisted of plastic netting with 1mm mesh. Males continue to actively care for the ne st despite the addition of the cover. I checked the nests for eggs four times each day at approximately 8am, 11am, 3pm, and 7pm. They were frequently checked in excess of four times daily, and if it was not possible to check them at least three times on a given day, a clear acrylic partition was placed in each tank separating males and females to prevent any un detected spawning. There were six days during the first block of the experiment and five days during the second block during which I was unable to check the nests at least three times per day. When eggs were discovered (day 0), I rem oved the nest from the tank, counted the eggs, and recorded the developmental st age to ensure eggs were discovered soon after spawning. I am confident that in all cases eggs were discovered immediately after spawning. In all treatments, I then placed a clear acrylic divide r containing nine holes in the re ar half of the tank immediately after eggs were discovered, physically separating ma les and females while still allowing them to remain in visual and chemical contact, thus allowing males to continuously court females. Physically separating males and females with th e partition (which was do ne in all treatments) was necessary to prevent females from consuming eggs. I then returned the eggs to the male, and 30

PAGE 31

if the male was in a no filial cannibalism treatme nt (i.e., NFC, no access to eggs), I immediately placed the mesh nest cover over the eggs. In all tr eatments, nests with eggs were briefly removed from the tank each day in order to count the eggs. On day five (all eggs either hatched or were consumed by day 5), I removed the nest c over and/or the partit ion, allowing males the opportunity to spawn again. This procedure wa s repeated each time a male spawned for the duration of the 90 days. Males remained in the same aquaria during the entire experiment. Females were used repeatedly throughout the experiment, but were frequently and randomly moved amongst the tanks to ensure that all females 1) experi enced similar feeding regimes and 2) spent approximately equal amounts of time with each male (to limit any female-specific effects). Males were weighed at several intervals during the experiment. Statistical Analyses I used a 2-way ANOVA to examine the effect of food and filial cannibalism on correlates of fitness and male weight and length. These analyses were performed in SYSTAT 9.0 (SPSS, Inc.). I used logistic regressi on to evaluate the effect of f ood on proportion of eggs consumed; this analysis was performed us ing SAS 8.2 (SAS institute). I pe rformed all analyses including and excluding data from males that died during the 90 days. Since the findings did not differ for any of the comparisons, I only present results ex cluding dead males. Becau se filial cannibalism treatment could not have affected the first clutch received, I evaluated the effect of filial cannibalism on total number of eggs received both including and excluding the first clutch received. The findings did not differ qualitatively, so results including all clutches are presented. In order to explicitly evaluate whether ca nnibalistic males experienced a net gain in reproductive success, I compared the observed number of eggs cannibalistic males received with the number of eggs they were expected to re ceive to achieve, as a minimum, no net loss in 31

PAGE 32

offspring produced. Specifically, I expected males that consumed eggs to receive, on average, the same number of eggs as non-cannibalistic male s plus at least the number of eggs that were consumed. Thus, the expected egg benefit nece ssary for males to receive no reduction in reproductive output is defined as the mean number of eggs consumed by FC males, and the observed egg benefit is defined as the difference between the mean number of eggs received by FC males and the mean number of eggs received by NFC males. I used two-sample, one-tailed ttests to compare the expected and observed egg benefit for high food and for low food males. Results Occurrence of Filial Cannibalism Filial cannibalism was prevalent when males had access to eggs. Indeed, when food was high males consumed approximately 95% +/3.9% ( X +/SE) of their eggs and when food was low males consumed approximately 84% +/3.6% ( X +/SE) of their eggs. Such high rates of cannibalism gave me confidence that in all cas es where access to eggs was allowed, I would expect to see benefits of filial cannibalism if they exist. Effect of Diet on Filial Cannibalism Diet significantly affected the proportion of eggs consumed (logistic regression food effect: 2 1X = 7.11, P = 0.008). Specifically, males in the low food treatment consumed a smaller proportion of eggs than males in the high food treatment. Costs and Benefits of Filial Cannibalism for Reproduction I compared the difference between the mean number of eggs received for FC males and NFC males (i.e. the observed egg benefit) and th e mean number of eggs consumed by FC males (i.e., the minimum expected egg benefit) (Figure 2-1). For high food males, the egg benefit of filial cannibalism necessary for f ilial cannibalism to result in no reduction in reproductive output 32

PAGE 33

was 205 eggs; surprisingly, high f ood FC males received approximately 40 fewer eggs than high food NFC males (Figure2-1). Consequently, high food males rece ived significantly fewer eggs than what was necessary to achieve no reduction in net eggs received (2-sample t test: t10 = 7.889, P < 0.0001). Similarly, the egg benefit necessary for low food males to break even was 107 eggs, but they received only 38 more eggs than low food NFC males. Again, low food males received significantly fewer eggs than was neces sary for no reduction in fitness (2-sample t test: t10 = 2.504, P = 0.016). Effects of Food and Access to Eggs on Components of Fitness Reproduction: There was a significant effect of food on all components of fitness I measured. In comparison to males receiving the low food diet, high food males received significantly more eggs (2-way ANOVA, food: F = 10.047, P = 0.005), spawned more times over the 90 day period (F = 11.861, P = 0.002), and had a great er frequency of spawning (F = 11.250, P = 0.003). In contrast, there was no eff ect of filial cannibalism on any of these variables (2-way ANOVA, FC: F = 0.001, P = 0.975; F= 0.533, P = 0.474; F = 1.103, P = 0.306, respectively) a nd no interaction (F = 0.911, P = 0.351; F = 0.035, P = 0.853; F1,23 = 0.559, P = 0.463, respectively) (Figure 2-2). 1,23 1,23 1,23 1,23 1,23 1,23 1,23 1,23 Male weight and length: There was no significant differe nce in initial male weight between the treatment gr oups (2-way ANOVA: food, F = 0.200, P = 0.659; FC, F = 0.743, P = 0.398; food x FC, F1,23 = 0.416, P = 0.526). During the experiment, high food cannibalistic males gained 1.73 +/0.73 g (1,23 1,23 X +/SD), high food non-cannibalis tic males gained 1.75 +/0.47 g ( X +/SD), low food cannibalistic males gained 0.79 +/0.60 g ( X +/SD), and low food noncannibalistic males ga ined 1.15 +/0.77 g ( X +/SD). Although males in the high food treatment gained significantly more weight than low food males (F = 8.538, P = 0.008), there was no 1,23 33

PAGE 34

34 effect of filial cannibalism (F1,23 = 0.532, P = 0.474) and no interaction between diet and cannibalism (F1,23 = 0.429, P = 0.520) on weight gained. Additionally, there was no significant differen ce in initial m ale standard length between the treatment groups (2-way ANOVA: food, F1,25 = 0.075, P = 0.786; FC, F1,25 = 0.402 P = 0.533, food x FC, F1,25 = 0.288, P = 0.597), and there was no significan t effect of f ood or filial cannibalism on change in ma le standard length (food, F1,23 = 1.881, P = 0.185; FC, F1,23 = 0.105, P = 0.749; food x FC, F1,23 = 0.105, P = 0.749). Discussion Parental investment, and parental care in particular, is expected to increase offspring survival (Trivers 1972; Sargent 1988; Clutton-Brock 1991; Balshine et al. 2002). Indeed, care has been shown to increase parental and offspri ng fitness in various animal taxa (Forester 1979; Dominey 1981; Simon 1983; Fairba nks and McGuire 1986; discussed in Balshine et al. 2002). Thus, the ambiguity regarding th e adaptive significance of filia l cannibalism coupled with the prevalence of filial cannibalism in fishes exhibiti ng parental care is surprising. Indeed, previous research in the flagfish suggests that the offspring survivorship be nefits of parental care may be offset by the occurrence of filial cannibalism (K lug et al. 2005). If filial cannibalism is an adaptive trade-off between current and futu re reproductive success (Rohwer 1978), filial cannibalism should lead to an in crease in net reproductive success. One way that an increase in net reproductive output could occur is through an increase in number of eggs fertilized (e.g., if males reinvest energy from cannibalism into increased courtship behavior). Contrary to the expectation that filial ca nnibalism will increase male mating success, I found that the number of additiona l eggs received by males who cannibalized never compensated for the loss resulting from cannibalism. Among males on the high food diet, those that were allowed to cannibalize eggs actually received fewer eggs from fema les than those that were not

PAGE 35

allowed to cannibalize eggs. For males on the low food diet, those th at were allowed to cannibalize eggs received more eggs from fe males than those that were not allowed to cannibalize eggs, but the number of extra eggs received never compensated for the number lost to cannibalism. Since the flagfish life span is approximately one year in the wild, such a reduction in the number of eggs over the course of one breeding season likely has a large impact on lifetime reproductive success. In order to explai n these findings in the context of the current energy-based explanation of f ilial cannibalism (Rohwer 1978; Sargent 1992), egg cannibalism would need to have a substantial positive effect on male survivorship, the survivorship of the eggs that remain in the nest, or on the survivorsh ip of any future eggs that a male receives. In contrast to cannibalism, food availability gr eatly affected correlate s of fitness. Indeed, an enhanced diet was related to an increase in eggs received, spawni ng frequency, number of clutches received, and increased weight gain. Thus, I have eviden ce that males in my study were energy-limited and that food and/or nutritional level affects reproductive success. The finding that food availability affects reproductive success coupled with the lack of benefits of filial cannibalism suggests that eggs do not have substan tial energetic or nutrient content relative to the costs of reproduction. This finding is also in consistent with the energy-based hypothesis of filial cannibalism (Rohwer 1978; Sargent 1992). Furt hermore, I found that males in the low food treatment consumed significantly fewer eggs than males in the high food treatment, which further contradicts the predictions of en ergy-based hypothesis (Rohw er 1978; Sargent 1992). However, reduced consumption of eggs by males in the low food treatment is consistent with life history theory if low food males have a reduc ed expectation of futu re reproduction and thus invest more in current reproduction. A similar tre nd (i.e., increased investment in current clutch as expected future reproducti on declines) has been associated with seasonal patterns of 35

PAGE 36

cannibalism in the cardinal fish ( Apogon doederleini ). Takeyama et al. (2002) found that 1-year old males cannibalised less at th e end of the breeding season. There were several limitations stemming from the controlled nature of the present study. In order to experimentally manipulate cannibalism while still allowing males to care for eggs I used a nest cover. After eggs were discovered, males and females were separated for 4 days in all treatments, thus preventing continuous spawning Such separation was necessary to prevent NFC males and females from consuming eggs. Not al lowing males to receive eggs continuously possibly led to clutch sizes that were small in comparison to t hose found in nature, and relatively small clutches, which have been associated with increased whole clutch cannibalism in other species (reviewed in Manica 2002), likely contributed to the hi gh rates of cannibalism observed. Indeed, I do not believe that such rates of canni balism are necessarily representative of natural filial cannibalism rates. Due to the nature of th e spawning substrate in the field, its currently impossible to accurately measure clutch size in the wild, and thus, I do not have reliable estimates of natural rates of f ilial cannibalism. Regardless, such high rates of cannibalism should result in even greater fitness effects of cannibali sm and thus gave me confidence that I would be able to detect any benefits of filial canniba lism. Yet, I found no net benefits related to reproduction and I have no reason to believe that th ere would be any benefit at lower levels of filial cannibalism. Additionally, I only evalua ted one major component of reproductive success and therefore, cannot evaluate alternative ways in which filial cannibalism could affect lifetime fitness. For example, separating males and fema les after spawning prevented me from assessing the effect of filial cannibalism on the number of simultaneous broods a male would subsequently receive, which may be an important component of reproductive success. Similarly, I did not measure hatching success, predation, and other potentially important components of reproductive 36

PAGE 37

success. Further work is clearly needed to evaluate the adaptive significance of filial cannibalism in the flagfish. Nonetheless, experimentally ma nipulating cannibalism while still allowing males to care for eggs allowed me to evaluate key pr edictions of energy-based explanations of filial cannibalism. Indeed, this study is the first to experimentally ma nipulate filial cannibalism while still allowing males to care for eggs, thus al lowing for the specific examination of some longerterm fitness consequences of filial cannibalism in relation to the en ergetic hypothesis (Rohwer 1978). The hypothesis that filial cannibalism in fish es reflects an adaptive trade-off in which energy or nutrients gained from eggs is invested into future reproduction is widely accepted as valid (e.g. Manica 2002) and has rarely been qu estioned since it was first proposed 25 years ago (but see Smith 1992). While whole clutch cannibalism may be explained as the termination of care, it does not appear that we have an adequate explanation fo r the widespread occurrence of partial clutch cannibalism. In the present study, fl agfish males consumed a large number of their eggs and no benefits related to increased re production or physical condition were observed. Thus, with regard to the flagfish, there is no evidence that Rohwers (1978) hypothesis provides adequate explanation for partial clutch filial cannibalism. More generally, there is mixed support for Rohwers theory of partial cl utch cannibalism, as is evident from the inconsistent results of studies examining the effect of food availability and parental condition on filial cannibalism (as discussed above). In the case of flagfish, food sources other than eggs are available and the incubation period is relatively short (less than 4 days at 29oC), making it even more difficult to understand why males would consume eggs purely for caloric or nutritional purposes. Thus, future experiments should evaluate further spec ific ways in which energy gained from eggs could be translated into net fitness benefits (i .e., male survivorship a nd increased quality of 37

PAGE 38

parental care). Furthermore, I suggest alternatives to Rohw ers (1978) energy-based hypothesis should be considered. Currently, I can envision several adaptive explanations for why parental flagfish males would consume eggs; 1) energy or nutrients from egg may increase male survival (consistent with Rohwer 1978), 2) partial cl utch cannibalism may improve surv ival of the remaining clutch (e.g. Payne et al. 2002), and 3) ma les may selectively cannibalize e ggs with reduced survivorship or quality. These alternatives ha ve received relatively little attention. Currently, the effect of filial cannibalism on male survival remains untested and further research is needed to evaluate this hypothesis. The effect of cannibalism on survival should be evalua ted in the presence of predators and other natural stressors. In the pres ent study, I did not measure the effect of filial cannibalism on offspring survival, and this idea should also be explicitly examined in separate experiments. Recently, Payne et al (2002) suggest ed that partial clutch cannibalism increases survivorship of remaining eggs through incr eased oxygen availability. This idea has received support in one system ( Stegastes leucostictus, Payne et al. 2002) but not in another ( Pomatoschistus minutus Lissker et al. 2002). However, it is possible that partial clutch cannibalism improves survivorship of remain ing offspring through other mechanisms. For instance, if density-depe ndent egg predation exists, partial clutch cannibalism might reduce the males risk of losing some or all of his eggs to egg predators. It is also possible that parental males use energy attained from eggs to increase the quality and/or quantity of parental care, thereby increasing remaining offspring survivorship or quality. This idea is still consistent with Rohwers hypothesis (1978) but has not yet been evaluated. The idea that males may selectively cannibalize eggs has rarely been considered. Ind eed, males of many species eat more eggs than would die naturally (e.g. Klug et al. 2005), but survival is typically measured only to hatching. It 38

PAGE 39

is possible that males consume eggs with decr eased post-hatching survivorship. Finally, it is possible that filial cannibalism is maladaptive. This idea has been dismissed in recent literature, but the validity of this dismissal remains unclear. In general, future work should focus on developing and examining alternatives to R ohwers energy-based explanation of filial cannibalism. 39

PAGE 40

Figure 2-1. Expected and observed benefit of filial cannibalism in eggs received by males. The minimum expected benefit (hatched bars) of filial cannibalism ne cessary to overcome the loss resulting from the consumption of eggs is defined as the number of eggs consumed by FC males; the observed benefit (solid bars) of filial cannibalism is the difference in the mean number of eggs rece ived by FC males and the mean number of eggs received by NFC males. Bars represent mean number of eggs and error bars are standard error. 40

PAGE 41

41 Figure 2-2. Effect of filial cannibalism on components of fitness. A) The mean total number of eggs received over 90 days for males in each of the four treatments. The treatments were high food and FC (i.e. access to eggs), high food and NO FC (i.e. no access to eggs), low food and FC, & low food and NO FC. B) The total number of clutches received over the 90 days across treatments. C) The frequency of spawning during the 90 days across treatments. Bars represent means and error bars are standard error.

PAGE 42

CHAPTER 3 PARENTS BENEFIT FROM EATING O FFSPRING: DENSITY-DEPENDENT EGG SURVIVORSHIP COMPENSATES FOR FILIAL CANNIBALISM Introduction Filial cannibalism is an evolutionary conundrum How is eating ones own offspring ever an adaptive strategy? Indeed, it is hard to imagine many circumstances in which regularly consuming ones own offspring leads to increased net reproductive success. Thus, it is surprising that filial cannibalism, which o ccurs in a range of taxa (Polis 1981), is particularly common in fishes exhibiting paternal care (reviewed in Manica 2002), a behavior assumed to increase an individuals fitness th rough increased offspring survivorsh ip or quality (Clutton-Brock 1991). While early ethologists considered filial cannibalism to be a ra re behavior with little or no evolutionary significance, filial cannibalism in fi shes has now been well documented in both the laboratory and the field (reviewed in Manica 2002), and currently, fili al cannibalism in fishes is thought to represent an adaptive strategy in whic h males maximize lifetime reproductive success. Specifically, filial cannibalism is thought to re flect a trade-off between current and future reproductive success, in which males gain energy and nutrients from eggs that are reinvested into current and future reproduction (the energy-based hypot hesis; as articulated by Rohwer 1978 and Sargent 1992). According to this hypothesis, whole clutch cannibalism (i.e., the consumption of all eggs present) is expected to be more freque nt when clutch size is relatively small because the energy requirements of caring males can be satisfi ed only by clutches larger than a certain size (Rohwer 1978). Specifically, this hy pothesis suggests that males should consume some specific number of eggs to satisfy their energetic needs, a nd when initial clutch size is smaller than this critical number, males should consume the whole clutch. Consistent with this prediction, several studies have found that whole clut ch cannibalism is more frequent when clutch size is relatively small (reviewed in Manica 2002; but see Payne et al. 2003, who found that smaller clutches were 42

PAGE 43

not preferentially eaten). The energy-based hypothesis also predicts that cannibalism should increase as food availability decreases and/ or when male condition is poor (Rohwer 1978; Sargent 1992). Evidence regardi ng these predictions is equivocal Consistent with the energybased hypothesis, supplementary feeding parental males reduced filial cannibalism in the common goby ( Pomatoschitus microps Kvarnemo et al. 1998), the scissortail sergeant ( Abudefduf sexfasciatus Manica 2004), and the Cortez damselfish (Stegastes rectifraenum Hoelzer 1992), and in some cases (e.g., Manica 2004) males do appear to simply be cleaning the nest of dead eggs (i.e., mortality resulting fr om filial cannibalism is similar to background mortality) when food is abundant. Contrary to predictions of the energy-based hypothesis, there was no relationship found between cannibalism a nd food availability and/or male condition in fantail darters ( Etheostoma flabellare, Lindstrm and Sargent 1997) and the three-spined stickleback ( Gasterosteus aculeatus, Belles-Isles & Fitzge rald 1991). Furthermore, male flagfish ( Jordanella floridae ) with reduced food availability actually consumed fewer eggs than males with high food availability (Klug and St. Mary 2005 and Chapter 2). Other studies have taken a different tack and examined the energetic content of eggs-one study suggested that energy from partial clutch ca nnibalism could potentially offset costs related to care (Apogon lineatus Kume et al. 2000), while another claimed that energy from eggs would be insufficient (Gasterosteus aculeatus, Smith 1992). Also inconsistent with the energy-based hypothesis, Payne et al. ( 2002) found that filial cannibalism incr eased in the later stages of egg development, when egg energetic value is much lower. Thus, there is a lack of general support for the energy-based explanation of filial canni balism (Rohwer 1978; Sargent 1992) and at best current theory can only explain cannibalism in some cases. Despite such mixed evidence and a lack of many alternative hypotheses, the energy-based hypothesis of filial cannibalism is often 43

PAGE 44

accepted as valid, and filial cannibalism is co mmonly considered an adaptive strategy (e.g., Vinyoles et al. 1999; Kume at al 2000; Manica 2002). Overall, the support for this hypothesis is unsatisfying and it remains unclear in many systems if (and potentially how) filial cannibalism is an adaptive strategy. Recently, Payne et al. (2002) proposed an alte rnative hypothesis suggesting that filial cannibalism is an adaptive strategy in which partial clutch cannibalism improves survivorship of remaining eggs by increasing oxygen availability to remaining eggs. In several systems low dissolved oxygen levels have been related to increased egg mortality (Kamler 1992), and according to the oxygen-mediated hypothesis of fili al cannibalism (as articulated by Payne et al. 2002, 2004), males potentially improve overall clutch survivorship by removing some of their eggs. Through a reduction in egg density, cannibalism can increase the surface area of the developing embryos exposed to the water, thereby improving oxygen exchange and overall survivorship of the remaining eggs. In other wo rds, when egg density is relatively low, each individual egg is expected to have greater oxyge n availability than when egg density is high. This hypothesis has received relatively little empirical examination. In their initial paper proposing the idea of oxygen-mediated filial cannibalism, Payne et al. (2002) found that reducing egg density in the beaugregory damselfish ( Stegastes leucostictus ) increased developmental rate of embryos and that part ial clutch cannibalism wa s significantly reduced when oxygen levels were high. However, the reef s inhabited by beaugregory damselfish have undergone significant environmental changes, particularly in rela tion to oxygen levels, over the past twenty years, and beugre gory damselfish do not oxygenate th eir eggs by fanning; thus, with respect to other systems, it is unclear how general we would expect oxygen-mediated cannibalism to be, particularly in species that are thought to o xygenate their eggs by fanning. In 44

PAGE 45

the sand goby ( Pomatoschistus minutus ), a system in which males fan eggs, Lissker et al. (2002) found no differences in whole clutch or partial clutch cannibalis m across oxygen levels, although the aim of that study was no t to explicitly evaluate oxyge n-mediated filial cannibalism. Thus, the importance of oxygen-mediated cannibalism remains unclear. Effectively evaluating the importance of oxygenmediated filial cann ibalism necessitates both an evaluation of predictions of the oxygenmediated hypothesis of filial cannibalism (i.e., does cannibalism vary across oxygen/egg density leve ls) and an examination of potential fitness consequences of filial cannibalism (i.e., does the potential benefit of cannibalism in terms of additional number of eggs surviving outweigh th e cost in number of eggs consumed by the male). Thus, the goals of my study were two-fo ld. First, I evaluated the effect of oxygen, egg density, and male condition on the occurrence of filial cannibalism and compared my findings to predictions of both the energy-based and oxyge n-mediated hypotheses of filial cannibalism. Specifically, the oxygen-mediated hypothesis predic ts that: 1) cannibalis m will increase as oxygen decreases (because of reduced egg survivorship when oxygen is relatively low); 2) cannibalism will increase as egg density increases (because of d ecreased oxygen availability and survivorship when egg density is high); and 3) there is no a priori expectation that male condition will affect the occurr ence of filial cannibalism. Like wise, the energy-based hypothesis predicts that 1) cannibalism w ill increase as oxygen decreases (because energetic costs increase when oxygen is low due to increased male fanning and increased metabolic costs to the males, Jones and Reynolds 1999); 2) there is no a priori expectation that egg density will affect cannibalism; and 3) cannibalism is expected to increase as initial male condition decreases. Secondly, by experimentally manipulating canniba lism, I evaluated the effect of simulated partial clutch cannibalism on egg su rvivorship and quantified the net benefits of partial clutch 45

PAGE 46

cannibalism. If partial clutch cannibalism is an adaptive mechanism in which removing some eggs improves overall egg survivorsh ip, we would expect partial clutch cannibalism to be related to no net reduction in offspring produced. The sand goby, Pomatoschistus minutus is an ideal system for evaluating the oxygen-mediated hypothesis of filial cannibalism because eggs are laid in a dense layer, nests often have relatively low oxygen availa bility, and egg development is dependent on male fanning (and therefore though t to be highly oxygen dependent); thus, if oxygen-mediated filial cannibalism occurs generall y, I would expect to find evidence of it in a system such as the sand goby. Methods Study Species and Experimental Site Sand goby males care for their eggs through guarding, cleaning, and fanning. Males build nests under suitable substrates and cover the nest with sand, leaving only a small (~1cm) opening. Sand goby eggs are clumped and male fa nning is required for egg development and survival. I conducted this study at Tvrminne Zoological Station, University of Helsinki, in southern Finland. Sand gobies were collected in shallow brackish water using a seine, and males and females were housed in separa te holding tanks (100 l) with con tinuous seawater flow prior to use. During this time the fish were fed ad libitum live Mysid shrimp and frozen Chironomidae larvae. The experiment was conducted during th e sand goby breeding season (June and July) of 2004. Experimental Design Experiment 1: Effect of oxygen and egg density on filial cannibalism: I crossed two oxygen treatments (high and low oxygen concentrati on) with two egg density levels (high and low egg density) to evaluate the effect of oxygen and egg density on the occurrence of filial cannibalism. Thus, there were four treatment s: 1) high oxygen, high e gg density, 2) high oxygen, 46

PAGE 47

low egg density, 3) low oxygen, high egg density, 4) low oxygen, low egg density. I used natural variation in initial male condition (measured as K=100*weight/length3;Williams 2000) to evaluate the relationship between parental condi tion and the occurrence of filial cannibalism. Each experimental tank was 60 l and equipped with continuous seaw ater flow through. The tanks contained either a large (8 cm diameter) or a small (4 cm diameter) half-flowerpot, which served as an artificial nest site. The two nest size treatments corresponded to my two experimental egg density levels: high and low. Because females spawn their eggs in a monolayer on the ceiling of the half-flowerpot and because egg number is approximately equal amongst females, egg density in small nests can be much greater than egg density in large nests (K. Lindstrm, personal observation). I fitted the inside of each nest with a transparent piece of plastic onto which females spawn their eggs; the transparent plastic allowed me to remove and photograph the clutch, when necessary, without di sturbing the eggs. Specifically, there were 1.8 +/0.05 eggs per mm2 ( X +/SE) in small, high density nests and 1.4 +/0.12 eggs per mm in the large, low density nests, and this difference was significan t (2-tailed t-test, t=-3.09, df=38, p=0.004). These egg density measurements are comp arable to those observed in the wild (egg density of nests in nature: range = 0.79 2.8 eggs per mm, 2 2X +/SE = 1.77 +/0.26 eggs per mm N = 9). 2I began each experimental by placing one male and one female in a tank with the randomly assigned egg density treatment (small nest or larg e nest). After spawning occurred, I removed the nest from the tank and photographed the eggs using a digital camera such that individual eggs could be counted. I subsequently quantified initial egg numbers usi ng these digital images. I then transferred the clutch, on its plas tic sheet, to a nest of intermed iate size (6 cm diameter) and returned the nest with eggs to the male. Transf erring the eggs to a nest of intermediate size 47

PAGE 48

ensured that males in both egg density treatmen ts fanned nests of similar size and thus had similar costs of care, which allowed me to isolate effects of egg density from effects of nest size. Immediately after return ing the eggs to the male, I randomly assigned an oxygen treatment (high or low). In the low oxygen treatment, low oxygen water constantly flowed directly into the males nest, reducing the oxygen concentration in side the nest to a pproximately 32.2 +/10.1 % (i.e., 3.2 +/1.0 mg/L at 14.3C, X +/SE) of fully air-satura ted water. This was done by continuously bubbling nitrogen gas into a c overed holding tank; using airline tubing, the reduced-oxygen seawater was then allowed to flow continuously in thr ough the rear of the males nest. In the high oxygen treatment, high oxygen water (from a flow through seawater system) was continuously allowed to flow into th e rear of the males nest; in this case oxygen concentration in the males nest was maintain ed at approximately 96.1 +/1.4% (i.e., 9.3 +/0.18 mg/L at 14.7C, X +/SE ) of fully air-saturated wate r. Because oxygen concentration was only manipulated in the nest and because all tanks had continuous flow through of sea water, oxygen concentration outside of the nest was approxi mately the same for both high and low oxygen treatments. I quantified the oxygen levels in the high and low oxygen nests for a sample of the males in the experiment using an ISO2 oxygen meter equipped with an OXELP oxygen electrode (World Precision Instruments); oxygen levels in side the high oxygen treatment nests were significantly greater than those in the low oxygen treatments (two-tailed t-test, t = -4.7, df = 9, p=0.001). I followed eggs until hatching and visually insp ected nests daily by shining a light into the nest. When eye shine was visible in the nest (approximately 1-3 days prior to hatching), I removed the nest, photographed the plastic transpar ency with eggs, and later counted the number of eggs using the digital photograph. Male and female weight and standard length were measured 48

PAGE 49

at the beginning and end of each replicate, and the condition measure K (Williams 2000) was used as an indicator of male condition. Experiment 2: Effect of simulated filial cannibalism on egg survivorship: I examined the effect of two levels of simulated filial cannibalism (egg removal and no egg removal) in high and low oxygen environments. In this experiment, one male and one or two females were placed in an aquarium (described above ) with an intermediate-size ar tificial nest (described above) containing a transparent plastic sheet for spawni ng; after spawning occurred, I removed the nest from the tank. I then cut the plastic sheet into four approximately equal pieces and each quarter was randomly assigned to one of the four tr eatments: 1) high oxygen, egg removal, 2) high oxygen, no egg removal, 3) low oxygen, egg removal, 4) low oxygen, no egg removal. In the egg removal treatments, I simulated filia l cannibalism by using a pair of fine forceps to haphazardly remove some proportion of the e ggs corresponding to actual levels of filial cannibalism observed in experiment 1 (simulated cannibalism: 10 66% of eggs removed, 39.6 2.5 % = mean SE; observed partial cannibalism: 6 93%, 37.5 7.9%). In the no egg removal treatments, I removed a trivial number of eggs (approximately 5-10 eggs) from the nest and gently touched a proportion of the remaining eggs (corresponding to the approximate number of eggs that were removed in the egg removal treatment) with the forceps. Digital images were then taken of the transparencies and I later counted th e eggs. Each transparency with eggs was then placed in either a high oxygen water or low oxyge n water container, each approximately 18 cm x 13 cm x 3.5 cm. Specifically, using a pin, I attached the transparencies with eggs to a styrofoam lid designed to snuggly fit on each of the containers. I then placed th e lid on the container so that the eggs were always in water. Because eggs are dependent on fanning, an air stone continuously bubbled air approximately 5 cm directly below the eggs at all times in both treatments. This 49

PAGE 50

method allows eggs to develop at rates comparable to those observed when males fan the eggs (Maria Jrvi-Laturi, personal communication). The high oxygen water contained a second air stone that continuously bubbled air into the water. This air stone was placed in the corner of the container so that the bubbles did not directly hit the eggs, and the oxygen concentration was maintained at 93.0 +/3.1% (i.e., 8.9 +/0.28 mg/L, X +/SE) of fully satura ted water. Likewise, in the low oxygen treatment, I placed a second air stone in the corner of the tank and nitrogen was continuously bubbled into this container th rough this air stone. Again, the bubbles from the second air stone did not directly flow onto the e ggs, and thus, I eliminated any potential effects of direct contact of nitrogen gas with the eggs Oxygen concentration in this tank was maintained at approximately 24.9 +/6.4% (i.e., 2.4 +/0.61 mg/L, X +/SE ) of fully saturated water when the probe was placed 5cm above the air stone bu bbling air (i.e., where the eggs were placed and in direct flow of the air bubbles). I checked eggs and oxygen concentration daily by briefly removing the lids of the containers. When m easuring oxygen level, the Styrofoam lid was carefully raised and the probe was placed against the Styrofoam near the eggs, directly in the flow of the air bubbles. Despite bubbling air underneath the eggs in all treatments, oxygen concentration was significantly greater in the high oxygen containers than in the low oxygen containers (two-tailed t-test, t = -8.5, df = 5, p<0.001), and I am thus confident that eggs in the two oxygen treatments experienced very different oxygen levels. I r ecorded the proportion of the clutch surviving until eye shine was present, the day on which eye shine was visible, and the proportion of the clutch covered in fungus. Three to six clutches we re placed in a container at one time and four replicates were completed. Data Analysis Experiment 1: Effect of oxygen and egg density on filial cannibalism : I analyzed the effect of oxygen and egg density on the occurrence of whole clut ch cannibalism using a stepwise 50

PAGE 51

logistic regression (remove if p > 0.15); oxygen and egg density were treated as categorical variables, and date of spawning, initial male c ondition, change in male condition, and initial egg number were used as covariates in the analysis I analyzed the effect of oxygen and egg density on the day of whole clutch cannibalism, egg su rvivorship, and male condition using stepwise ANCOVA (remove if p > 0.15), and thus, for simplic itys sake, I only present significant effects of the final model, unless otherw ise stated. In all analyses, I initially evaluated all two-way interactions between factors and covariates and found them to be non-significant. I then removed these factor-covariate interactions and proceeded with the stepwise ANCOVA. For day of whole clutch cannibalism and egg survivorship anal yses, oxygen and egg density were treated as categorical fixed factors, and initial egg number, date of sp awning, initial male condition, and change in male condition were used as cova riates. The interacti on between oxygen and egg density was also included in the model. For ma le condition analyses (i.e., change in condition and final condition), oxygen and egg density were again treated as categorical fixed factors, and date of spawning, initial male c ondition, and number of eggs consumed were used as covariates, and again, the interaction betw een oxygen and egg density was included. For analyses of survivorship, proportion surviving was firs t arcsin square root transformed. Additionally, whole clutch canniba lism and partial clutch cannibalism likely have distinct biological significance; whole clut ch cannibalism is a termination of care and benefits of whole clutch cannibalism can only be seen in future re productive success, wherea s benefits of partial clutch cannibalism are potentially related to cu rrent and/or future reproductive success. Thus, I performed analyses both includ ing and excluding cases of w hole clutch cannibalism, when applicable. 51

PAGE 52

Experiment 2: Effect of simulated filial cannibalism on egg survivorship: I evaluated the effect of egg removal and oxygen on survivorship of remaining eggs using a completely randomized block design. For the analysis of surv ivorship, the proportion of the clutch surviving was first arcsin square root transformed. I treat ed oxygen and egg removal as fixed factors, and clutch (i.e., the original clut ch eggs were from) was treate d as a random factor. Blocking by clutch allowed me to account for inherent differe nces among clutches (e.g ., effects of parents) and also implicitly accounts for any effect of ti me (i.e., replicate); thus, blocking by replicate (i.e., time) is unnecessary (G. Wa llace, University of Florida Statistics Department, pers. comm.). I compared the number of eggs surviv ing with and without si mulated cannibalism in order to explicitly evaluate whether the potential benefit of cannibalism (i.e., increased survivorship of remaining eggs stemming from egg removal) out weighed the associated cost (i.e., number of eggs removed). For cannibalism to result in no net reduction in current reproductive success, the total number of eggs su rviving with cannibalism must, on average, be greater than or equal to the total number of eggs surviving without cannibalism. As above, I analyzed this data as a comp letely randomized block design, a nd treated oxygen and egg removal as fixed factors and clutch as a random factor All analyses were performed both including and excluding cases of whole clutch death. Results Experiment 1: Effect of Oxygen, Egg Dens ity, and Male Condition on Filial Cannibalism Occurrence of whole clutch cannibalism: Whole clutch cannibalism was more prevalent when oxygen was low and/or egg density was high (stepwise bina ry logistic regression, oxygen effect, = 3.81, p = 0.05; egg density effect, = 3.78, p = 0.05; Figure 3-1). Similarly, smaller clutches were su bject to whole clutch cannibalism more frequently than larger 2 1X2 1X 52

PAGE 53

clutches (initial egg number effect, = 4.43, p = 0.035). It is important to note that there was no significant difference in initial egg number between the oxygen or egg density treatments (2way ANOVA, oxygen effect, F= 0.1, p = 0.75, egg density effect, F = 0.001, p = 0.98). While there was a trend for whole clutch can nibalism to decrease as the breeding season progressed, there was no significant effect of the date of spawni ng on whole clutch cannibalism ( = 3.20, p = 0.07). Also noteworthy in relation to the energy-ba sed hypothesis, there was no effect of initial male condition or change in male condition on the occu rrence of whole clutch cannibalism (p > 0.15). Specificall y, the initial conditi on (K) of males that cannibalized their whole clutch was 0.64 +/0.014 g/cm3 (2 1X1,39 1,39 2 1X X +/SE), and the initial condi tion of males that did not practice whole clutch cannibalism was 0.63 +/0.024 g/cm3 ( X +/SE). To further analyze patterns of whole clutch cannibalism, I examined the day on which whole clutch cannibalism occurred. The timing of whole clutch cannibalism occurred earlier as the breeding season progressed (F1,24 = 9.85, p= 0.005) and when initial egg number was relatively small (F1,24 = 4.559, p = 0.05). Additionally, males in poorer condition cared for their clutches longer before cannibalizing them entirely (initial condition effect, F1,24 = 7.27, p = 0.01). There was significant interaction between the date spawning occurred and the initial number of eggs received (F1,24 = 17.6, p < 0.001). Egg survivorship: High oxygen tended to increase egg survivorship (oxygen effect, F = 3.62, p = 0.06; Figure 3-2), and egg survivorship al so increased as initial egg number increased (initial egg number effect, F = 7.96, 0.008); however, there was no significant effect of initial condition, egg density, or spawning date on egg survivorship (p > 0.15 for all). Because the trend for oxygen and initial egg number to affect egg survivorship was almost certainly the result of whole clutch cannibalism (which was affected by oxygen and initial egg number), I then 1,37 1,37 53

PAGE 54

repeated the analysis excluding cases of whole clutch cannibalism. In this case, I found a significant effect of oxygen (F = 9.9, p = 0.02) and spawning date (F = 6.1, p = 0.05). Additionally, egg survivorship was higher for males that were initially in poorer condition (initial condition effect, F = 15.0, p = 0.008; Figure 3-3 A), and ther e was also a relationship between change in male condition and egg survivorship (F = 15.5, p = 0.008). Specifically, males that were in poorer condition to begin with ate a smal ler proportion of (and fewer) eggs than males in better condition (Figure 3-3 A), and males that di d eat eggs had less of a decline in condition (Figure 3-3 B). Again, there was no effect of egg density. However, its important to note this analysis is, in a sense, not performed on a random sample of eggs; males have already decided whether or not to cannibalize the whole clutch, and potentially, they have d ecided to continue to care for eggs that they expect to have the great est survivorship. For example, only one male in the low oxygen, high egg density did not cannibaliz e his entire clutch. We have no idea what survivorship of the other low oxygen, high egg de nsity clutches would have been. Effectively analyzing the direct effects of oxygen and egg dens ity on egg survivorship necessitates that this is done in the absence of whole clutch cannibalism (as in Experiment 2). 1,12 1,12 1,12 1,12 Male condition: When males consumed their entire cl utch (i.e., including only cases of whole clutch cannibalism), there was a significant effect of in itial condition on final condition (F = 49.5, p < 0.001) but not on change in male condition (p> 0.15). Additionally, final male condition worsened as the breeding seas on progressed (spawning date effect, F = 4.5, p = 0.05). 1,24 1,24 When I considered only partial clutch cannibalism, providing care in low oxygen nests resulted in a poorer final c ondition and a greater decrease in condition, in comparison to providing care in high oxygen nests (final condition, F1,12 = 19.4, p = 0.003; change in condition, 54

PAGE 55

F1,12 = 16.4, p = 0.007). Additionally, condition worsen ed as the breeding season progressed (final condition, F1,12 = 10.6, p = 0.01; change in condition, F1,12 = 9.3, p = 0.02). The number of eggs a male consumed was positively related to condition, suggesting that eggs provide energetic benefits to caring male s (final condition, F1,12 = 17.2, p = 0.004; change in condition, F1,12 = 11.0, p = 0.02). Experiment 2: Effect of Simulated Filial Cannibalism on Egg Survivorship Effect of oxygen and egg removal on remaining egg survivorship: Egg removal significantly improved egg survivorship (F = 4.64, p=0.04; Figure 3-4 A). Additionally, there were significant differences among clutches in egg survivorship (i.e., eggs from particular clutches had higher egg survivorship than eggs from other clutches, regardless of treatment; F = 4.02, p = 0.004) and significant interaction between the clutch that eggs were from and oxygen (i.e., some clutches did better in the low oxygen e nvironment, while others did better in the high oxygen environment; F = 5.06, p < 0.001). Surprisingly, there was no effect of oxygen on egg survivorship. This analysis was performed in cluding cases in which the whole clutch died. In all cases in which the whole cl utch died, fungus had taken over the nest. However, in the presence of the male, fungus is never observed to attack the entire clutc h. Thus, I repeated this analysis excluding whole clutch death, which might be more representative of natural conditions. The results were unchanged; egg rem oval increased egg survivorship (F = 13.83, p =0.004; Figure 3-4 B) and there was a significant interaction between clutch and oxygen (F = 9.36, p = 0.007), but again, there was no effect of oxygen on egg survivorship. 1,33 1,16 16,33 1,10.1 8,6 Effect of oxygen and egg removal on total number of eggs surviving: I compared the number of eggs surviving with and without cannibalism to evaluate potential fitness consequences of cannibalism. Regardless of wh ether I included or excl uded cases of whole clutch death, I found no effect of oxygen or egg removal on the number of eggs surviving 55

PAGE 56

(Figure 3-5). In both cases, there is a significant interaction betw een oxygen and the clutch that the eggs were from (including whole clutch death, F = 4.60, p <0.001; excluding whole clutch death, F = 3.89, p =0.004). 17,33 18,16 Discussion Here, I present evidence that filial cannibali sm can potentially be an adaptive mechanism associated with density-dependent egg survivorship. When I simula ted partial clutch cannibalism by removing eggs, the survivorship of the remain ing eggs increased, and more importantly, egg removal at a range of cannibalism levels (10 66 % of clutch) did not re duce the total number of offspring produced. Indeed, my results show that males can consume, on average, 40% of their eggs with no reduction in current reproductive suc cess! Thus, partial clut ch filial cannibalism is potentially a mechanism by which males improve survivorship of remaining eggs. This finding is consistent with the theoretical predictions of Payne et al. (20 04), whose modeling results suggest that under some conditions, males can consume up to 80% of their clutch without a reduction in reproductive output. Under this scenario, males potentially also gain energy from eggs with no net loss in reproductive success. My results and previous work in the sand goby (Lindstrm 1998) suggest that consuming eggs is energetically beneficial. Additionally, actual partial clutch cannibalism is likely to be much more precise and selective th an my simulated cannibalism and males can potentially track the conditions in thei r nest, suggesting that pa rtial clutch cannibalism may be even more efficient at enhanci ng the survival of the remaining brood. Consistent with previous theory (Rohwer 1978; Sargent 1992), I found that whole clutch cannibalism in the sand goby is more prevalent when initial clutch size is relatively small. However, male condition did not affect the occurrence of whole clutch cannibalism; this finding is inconsistent with the current energy-based hypothesis, which suggests that whole clutch cannibalism should be more frequent when clut ch size is relatively small because the energy 56

PAGE 57

requirement of males can be satisfied only by cl utches larger than a certain size (Rohwer 1978). My results suggest that whole cl utch cannibalism depends on costs a nd expected benefits of care, as suggested by Manica (2004). Pare ntal care in fishes is ofte n assumed to be shareable among offspring in a nest (i.e., a unit of parental ca re may be given to one or several offspring; Wittenberger 1981, Williams 1975). Thus, the costs of care for small and large clutches is assumed to be comparable, whereas the benefit (i.e., offspring produced) of caring for a large clutch is assumed to be much greater than that of a small cl utch. In my study, whole clutch cannibalism increased as the expected benefit of care decreased (i.e., as initial egg number decreased) and as the cost of care increased (i.e., with decreasing oxygen). This finding that low oxygen led to more whole clutch cannibalism is c onsistent with both the previous energy-based hypothesis (Rohwer 1978; Sargent 1992) (because cost of care to the male increases as oxygen decreases) and with the oxygenmediated hypothesis (Paynet et al 2002; 2004) (because the expected benefit would be less when oxygen is lo w, according to this hypothesis). However, the finding that males altered thei r whole clutch cannibalism acco rding to egg density is only consistent with the oxygen-mediated hypothesis (s ince the benefit from a high density clutch would be less than that of a low density clutch, assuming egg numbers are equal, Figure3-4) and is not predicted by the energy-based hypothesis. The observed patterns of filial cannibalism, and in particular partial clutch cannibalism, are thus consistent with cannibalism mediated by density-dependent egg su rvivorship (e.g., the oxygen-mediated hypothesis, Payne et al 2002) and inconsiste nt with the energy-based hypothesis (Rohwer 1978; Sargent 1992). For ex ample, the energy-based hypothesis assumes that there is an adaptive trad e-off occurring between current and future reproductive success, in which males gain energy for future reproduction at a cost to current repr oductive success. In the 57

PAGE 58

sand goby, I found that there is no cost to current reproduction (i.e., no trad e-off between current and future reproductive success) because densit y-dependent egg survivorship compensates for filial cannibalism. Furthermore, the energy-ba sed hypothesis predicts a negative relationship between male condition and filial cannibalism. Contra ry to this prediction, I found that males in poorer condition actually consumed fewer of their eggs than males in better condition. A similar finding in the flagfish was reported by Klug and St Mary (2005; Chapter 2), who suggested that males in poor condition possibly have reduced exp ected future reproduction and thus invest more into their current clutch. While inconsistent with the current energy-based hypothesis, my findings are similar to those in some other systems (e.g., Belles-Isles and Fitzgerald 1991; Lindstrm and Sargent 1997; Klug and St. Mary 2005), but contrary to results in others (e.g., Hoelzer 1992; Kvarnemo et al. 1998; Manica 2004) However, my results are also, in part, contrary to oxygen-mediated filial cannibalism. Although whole cl utch cannibalism was affected by oxygen and egg density, and egg removal improve d egg survivorship, oxygen did not affect egg survivorship in the simulated cannibalism experiment. Thus, it does not appear that partial clutch cannibalism in the sand goby improves egg survivorship by increas ing oxygen availability of remaining eggs. Indeed, part ial clutch cannibalism does appear to be a mechanism in which the consumption of some eggs improves survivorship of remaining eggs ; however, the cause of density dependent egg survivorship is unknown in th is system. Additionally, the lack of an effect of oxygen on egg survivorship also indicates th at male fanning might not primarily serve to oxygenate eggs, as is typically assumed. I therefore suggest a general e xplanation of filial cannibalism that is mediated by densitydependent egg survivorship, but that is not solely relate d to oxygen, as in current theory (Payne et al. 2002, 2004). There are many ways in which de nsity-dependent egg survivorship can occur. 58

PAGE 59

For example, in systems such as the sand goby in which eggs are clumped, it is plausible that waste from developing embryos negatively affects the development and/ or survival of other embryos, and that decreased egg density reduces th e negative effects of such waste. In species whose eggs are not highly clumped (e.g., flagfish ) other density-dependent factors, such as density-dependent egg predation might affect egg survivorship. For instance, if male nest size has some limit and predators are attracted to nests with more eggs, a male might reduce the probability of losing some or all of his eggs to predators by consuming some proportion of them. In systems such as beaugregory damselfish, the system on which Payne et al. (2002) based their hypothesis of oxygen-mediated cannibalism, whic h lack male fanning of eggs, it does seem likely that density-dependent egg survivorship is due to limited oxygen av ailability. Specifically, the reefs inhabited by beaugregory damselfish recently underwent changes in oxygen levels, and thus it seems likely that males cannibalize their e ggs to increase oxygen availability of remaining eggs. More work is needed to assess the relative importance of the evolution of filial cannibalism and fanning, and future work should focus on syst ems with different fanning and parental care strategies. Indeed, the actual costs and benef its of fanning and filial cannibalism are unknown in many systems. More generally, other hypotheses a ssociated with filial ca nnibalism mediated by density-dependent egg survivorship should be explored further. Regardless of the mechanism, filial cannibalism mediated by density-dependent egg survivorship raises an intere sting question: why dont female s just lay fewer eggs? If egg survivorship is density dependent and layi ng tightly packed eggs reduces overall egg survivorship, why havent females been selected to lay less dense clut ches? In the sand goby, it seems unlikely that laying dispersed egg batches w ould be an evolutionarily stable strategy. If a female were to lay fewer, more dispersed eggs it is likely that the remaining space would be 59

PAGE 60

taken up by another females eggs, resulting in the same amount of cannibalism. If all nests (or nests of high quality males) were full of sparse ly laid eggs, a female would likely increase her fitness by laying a denser clutch in a nest already containing eggs as opposed to not laying at all or laying eggs with a low quality male. Additionally, it is important to note that females do, to some extent, mediate the egg density of their cl utch; in the present study, egg density in larger nests was significantly less than egg density in smaller nests. Thus, females do reduce the egg density of their clutch when it is possible to do so. Why females lay many, densely packed eggs is a complex problem, and necessitates considera tion of many factors aff ecting egg survivorship, including the costs associated with density-depende nt egg survivorship and the costs and benefits associated with mate quality. However, it does not appear that selection will necessarily favor females laying clutches with fewer, less tightly packed eggs when egg su rvivorship is densitydependent. In conclusion, alternative explanations of filial cannibalism need to be explored further, and new hypotheses should be developed and evaluated. For example, selective filial cannibalism seems plausible. Numerous studies show that males consume healthy and viable eggs (discussed in Manica 2002). However, if vari ation in offspring quality exists and if care is not entirely shareable among offspr ing, or if egg survivorship is density-dependent, males could potentially benefit by consuming viable eggs of reduced quality (e.g., eggs with reduced posthatching survival; see also liter ature on selective embryo abortion in plants, e.g., Burd 1998). To more explicitly evaluate the potential fitness cons equences of cannibalism, it would be useful to estimate the average offspring fitness in nests in which males have removed eggs and nests in which experimenters have removed eggs, to dete rmine if males are eating poor-quality eggs. Furthermore, the future benefit of eating eggs in this system remains unclear and should be 60

PAGE 61

evaluated in an experiment specif ically designed to quantify futu re reproductive and survivorship benefits of cannibalism. Eating e ggs clearly provides a male with some energetic gain, but the relative importance of this benefit for future re production is unknown in th is system. Finally, the evolutionary significance of filial cannibalism is likely not due solely to one factor (e.g., energetic benefit of eggs, density -dependent egg survivorship), and once the current and future costs of filial cannibalism are be tter understood, a more synthetic model of filial cannibalism will be necessary. 61

PAGE 62

Figure 3-1. Effect of oxygen and egg density on the prevalence of whole clutch cannibalism by parental males. Bars represent the proporti on of clutches in which the male consumed the entire clutch, and error bars are standard error based on a binomial distribution. 62

PAGE 63

Figure 3-2. Effect of oxygen and egg density on the mean (+/SE) egg survivorship (i.e., proportion of the clutch that survived until hatching) when males were present with eggs, including cases of both whole and partia l clutch cannibalism. It is important to note that in the low oxygen, high egg density treatment, only 1 male cared for his clutch until hatching ; the other 8 males in this tr eatment cannibalized their whole clutch. 63

PAGE 64

A 0.50 0.60 0.70 0.80Initial Condition (K) 0.00 0.20 0.40 0.60 0.80 1.00Proportion of Clutch Consumed B 0.00 0.20 0.40 0.60 0.80 1.00Proportion of Clutch Consumed -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15Change in Male Condition (Final K Initial K) Figure 3-3. Relationship between A) m ale condition (i.e., K=100*g/cm3) and the proportion of the clutch consumed by parental males, and B) partial clutch cannibalism and change in male condition. 64

PAGE 65

A B Figure 3-4. Effects of simulated filial cannibalism. A) The eff ect of oxygen and egg removal on egg survivorship (i.e., the mean +/SE proportion of the clutch surviving), including cases of both whole and partial clutch death. B) The effect of oxygen and egg removal on egg survivorship (i.e., the mean +/SE proportion of the clutch surviving), excluding cases of whole clutch death. 65

PAGE 66

66 Figure 3-5. Simulated filial cannibalism: The ef fect of oxygen and egg removal on the total number of eggs surviving, in cluding cases of both whole a nd partial clutch death. Egg removal did not significantly reduce the total number of eggs surviving. Bars represent means and error bars are standard error.

PAGE 67

CHAPTER 4 SELECTIVE FILIAL CANNIBALISM IN THE SAND GOBY Introduction Filial cannibalism is an evolutionary mystery. It is difficult to imagine how regularly consuming ones own viable young represents an adaptive strate gy, yet filial cannibalism is prevalent in a range of animals, particularly fish es exhibiting paternal care of eggs (discussed in Manica 2002; Klug and Bonsall 2007). Typically, f ilial cannibalism is viewed as an adaptive trade-off in which energy gained from eggs is us ed to better care for remaining offspring, or for increasing future reproduction (Rohwer 1978; Sargent 1992; Manica 2002). Because energy is such an obvious and direct be nefit of filial cannibalism, mu ch of the work aimed at understanding the adaptive significance of filia l cannibalism has focused on energetic benefits (reviewed in Manica 2002). However, some have suggested that energeti c benefits alone are unlikely to explain the preval ence of filial cannibalism in natural systems (e.g. Smith 1992; Payne et al. 2002; Klug et al. 2006 and Chapter 3). Recent theoretical work suggests that the abil ity to cannibalize offspring selectively in relation to aspects of offspring phenotype (e.g., quality or egg maturation rate) can directly favor the evolution of filial cannibalism (Klug and Bonsall 2007 and Chapter 6). While the idea of weeding out inferior offspring has been documen ted in other contexts (Forbes & Mock 1998; e.g. selective abortion in humans : Stearns, 1987; Forbes 1997; Diamond 1987; Hesketh & Xing 2006; selective abortion in plants: Burd 1998; Karkkainen et al. 1999; Melser & Klinkhamer 2001), it has not yet been explicitly evaluated in relation to filial canniba lism (but see Mrowka 1987 and Kraak 1996 for work on the consumption of unfertilized or diseased eggs, and Neff and Sherman 2003 regarding egg cannibalism of non-kin) Indeed, the relationship between offspring phenotype and filial cannibalism remains unknown. One study found preferential cannibalism of 67

PAGE 68

younger eggs when within-brood variation in egg age existed (Salfert and Moodie 1985), but in general, little is known about the specifics of which eggs ar e consumed when a parent does decide to cannibalize. Hence, the importance of sele ctive filial cannibalism remains unknown. To begin to understand the potential importa nce of selectivity in filial cannibalism, I evaluated patterns of within-br ood cannibalism in the sand goby, Pomatoschistus minutus a fish in which males alone provide parental care and practice filial cannibalis m during the egg stage. My primary goal was to determine if males practice selective canni balism (i.e., non-random consumption of eggs with regard to some asp ect of egg phenotype). Specifically, I focused on the relationship between egg size and partial clut ch filial cannibalism (i.e., the consumption of some eggs present during a given reproductive bout) by males. Egg size is correlated to larva size in this species (H.K., unpublished data) a nd has been correlated with post-hatching survivorship in a range of fishes (discussed in Kamler 2005). I considered cases in which males received eggs from one or two females. For th e case in which males had eggs from only one female, I asked whether males exhibited a preferen ce for eggs of a particular size range. When males had eggs from multiple females, I was interested in whether males 1) preferentially consumed eggs of the first or the second female that they spawned with, and 2) showed any preference with regard to egg size for each of the clutches in a nest (i.e., female 1s and female 2s clutches). I also compared the size of fema le 1s and female 2s eggs to determine if any differences in egg size existed between the female s, and if so, whether these differences might explain the patterns of cannibalism observed. Materials and Methods Experimental Design I evaluated selective filial cannibalism in the sand goby using data from two years. In both years adult sand gobies were collected in shallo w brackish water using a seine near Tvrminne 68

PAGE 69

Zoological Station (University of Helsinki) in southern Finland. All fish were fed ad libitum live Mysid shrimp and frozen Chironomidae larvae throughout the studies. Multiple-female set-up: In 2006, I initiated each replicate by placing a single male and a single female (female 1) in an aquarium equipped with continuous flow-through seawater system. Each nest contained a half-flowerpot (8 cm diameter) which served as the nesting site. The inside of each nest was fitted with a tran sparent piece of plastic onto which females spawn their eggs. The transparent plastic allowed me to remove and photograph the eggs, when necessary. The male-female pair was allowed to spawn, and immediately after spawning, I removed the nest with eggs, di gitally photographed the eggs, and cut out and removed a small subset of eggs (20-30 eggs) from the plastic tran sparency. I reared the subset of eggs in the absence of the male (described below) to determ ine if any size-specific patterns of egg mortality existed. I then returned the nest with eggs to the male, and a second female (female 2) was placed in the tank. The male and second female were then allowed 24 h to spawn (only clutches in which the second female spawned within 24 h of the first spawning were used in this experiment). After spawning, I again removed and photographed the eggs, and I removed a small subset of eggs. The nest with eggs was then retu rned to the male. Only cases of partial-clutch cannibalism were considered in th e present study, as I was interest ed in within-brood patterns of cannibalism. I followed all eggs un til hatching and visual ly inspected the nests daily by shining a light into the nest. Just prior to hatching (i.e., 1-3 days before hatc hing), eye shine (i.e., reflection from a flashlight) becomes visible in the devel oping embryos, and this is an indication that the eggs are about to begin hatching. When eye shine was visible in the eggs, I removed the nest and photographed the plastic tran sparency with eggs. 69

PAGE 70

The subsets of removed eggs were reared in individual plastic cont ainers, each equipped with an airstone. Water in the plastic containers was changed daily using water from the flowthrough system, and the water temperature in thes e containers was maintained at approximately 16C. Single-female set-up: For the single male-single female trials, I retrospectively analyzed digital images collected from a previous e xperiment conducted in 2004 (Klug et al. 2006 and Chapter 3) in the same location using the same population of fish. In this case, I placed a male and female in an aquarium equipped with c ontinuous sea water flow-through and a half flowerpot nest (8 cm diameter) equipped w ith a plastic transparen cy. Immediately after spawning, the nest and eggs were removed and photographed. In this case, I then transferred the clutch, on its plastic sheet, to a nest of intermediate size (6 cm diameter) and returned the nest with eggs to the male (see Chapter 3 for additio nal details). I only used clutches from the low density, high oxygen treatments (Klug et al. 2006 a nd Chapter 3) in the present study, as this setup was most comparable to the design used in 2006 (described above). Again, only cases of partial clutch cannibalism were considered. I fo llowed all eggs until hatching, and I visually inspected the eggs daily by shining a light into the nest. When eye shine was visible in the nest, I removed the nest and photographed the plastic transparency with eggs. Image Analysis For each males eggs in 2006 (the multiple-fem ale scenario), I superimposed the image immediately following female 1s spawning and the image immediately following females 2. I then identified all eggs and labeled them as bel onging to either female 1 or female 2. The image following female 2s spawning (in which all e ggs have now been identified) was then superimposed with the final image taken just before hatching. I then determined the specific eggs that had been consumed. Using Sigma Scan Pr o 5.0 (SPSS, Inc.) and the image containing the 70

PAGE 71

spawn of female 1 and 2, I quantified the initial diameter of 1) a random subset of female 1s eggs and female 2s eggs (range: 25-75 eggs of each female) and 2) a subset of the specific eggs consumed (range: 5-45 eggs of each female). For each males eggs in 2004 (the single-female scenario), I superimposed the initial and fi nal image to determine which eggs had been consumed. I then used Sigma Scan Pro 5.0 (SPSS, In c.) to quantify the initial diameter of 1) a random subset of the eggs (range: 35-80 eggs) and 2) a subset of the specific eggs consumed (range: 10-38 eggs). These data allowed me to quan tify the initial size distri butions of 1) all eggs in a nest and 2) the eggs that were consumed. Th is, in turn, allowed me to estimate cannibalistic preference for eggs of varying size classes (des cribed below), while taking into account the initial abundance of e ggs of varying sizes. Preference Calculation I was interested in whether males preferentially consumed 1) eggs of female 1 or female 2 (for the multiple-female data), and/or 2) eggs of a particular size class (for both the multipleand single-female data). To assess whether males consume eggs in some non-random way, I used the preference measure (with the ith component i ; Manly et al. 1972; Chesson 1983). This preference measure has been used widely in studi es of foraging (discussed in Chesson 1983). It is an ideal measure of preference for my purpos es because it allowed me to account for 1) the initial abundance of eggs of varying sizes and 2) the depletio n of eggs of varying size due to cannibalism. I calculated measures of egg preference in relation to preference for female 1s eggs and for female 2s eggs, and for eggs of give n size classes. Preference for each egg type ( i ) was calculated as follows: m i nrn nrnm j jjj ioii i 1 0 0 0)/()ln(( )/()ln(( where i = 1,,m (4-1) 71

PAGE 72

where ni0 is the number of eggs of type i present initially, ri is the number of eggs of type i consumed by the male, and m is the total number of different egg types present (modified from Manly et al. 1972 and Chesson 1983). For this meas ure of preference, 0 indicates no preference (i.e. consumption is equivalent to what is expected if males randomly consume eggs), a positive value reflects consumption that is greater than would be expected from random consumption (i.e. a preference for that egg type exis ts), and a negative value suggests that consumption is less than what would be expected from random consumption. Based on my data (i.e., the ra nge in observed egg size and pa tterns of consumption), I had sufficient resolution to identify 4 size classes fo r each brood of eggs. I calculated four equal size classes (i.e., small, small-medium, medium-large, and large) for each brood by dividing the range in egg diameter for a given brood by f our. For the single-female data, I calculated preference for each of the 4 size classes (small medium small elmedium arg and el arg ) In this case, if there was no preference for any part icular size class of eggs, I would expect 0 arg arg elelmedium medium small small For the 2006 data, I estimated preference for 4 size classes for each of the two females, and thus, ther e were a total of 8 egg types. Again, if there was no preference for eggs of a particular female or egg size, I would expect 0 arg2. arg 2. 1. 1. 1. 2.arg1.arg 2. elfem elmedium fem lmedium fem medium smallfem smallfem medium small fem smallfem elfem e Statistical Analyses I analyzed all preference data using non-para metric Friedman ANOVA. I used t-tests to examine differences in mean egg diameter between female 1 and female 2 in 2006, and to compare mean egg diameter in 2004 versus 2006. Given the differences in experimental set-up 72

PAGE 73

between years, I compared mean egg density between 2004 and 2006 using a t-test. Linear regressions were used to examine the relationshi p between mean egg diameter and egg survival, and between mean egg diameter and egg development rate (i.e. time from spawning until hatching) for the subsets of eggs reared in the absence of males. For the regression between egg diameter and development time, one clutch cont ained visible fungus and was excluded from this analysis, as fungus increases th e rate of egg development in this species (H. Klug, personal observation). In both cases, means were taken for the subsets from female 1 and female 2 for a given male to avoid pseudoreplication. Results Differences in Egg Size, Egg Density, and Cannibalism Rates between Years There was no significant difference between th e mean egg diameter of female 1 and of female 2 in 2006 (paired t-test, t = -0.51, df = 6, p = 0.62; female 1 mean +/SE egg diameter: 0.61 +/0.013 mm; female 2 mean +/SE egg di ameter: 0.62 +/0.011 mm). Eggs tended to be slightly larger in 2006 in co mparison to 2004 (2004 mean +/SE egg diameter: 0.58 +/0.018 mm; 2006 mean +/SE egg diameter: 0.62 +/0.0 11 mm), but this differe nce was not significant (independent samples t-test, t = -1.79, df = 10, p = 0.10). Additionally, egg density did not differ significantly between years (t = -0.12, df = 10, p = 0.91; 2004 mean +/SE: 1.42 +/0.27 eggs/mm2; 2006 mean +/SE: 1.45 +/0.12 eggs/mm2). In 2004 males consumed 32.1 +/0.12 % (mean +/SE) of their eggs, and in 2006 males consumed 36.2 +/0.079 % (mean +/SE) of their eggs. There was no significant difference in the proportion of eggs cannibalized between the years (t-test, t = -0.30, df = 10, p = 0.77). 73

PAGE 74

Egg Size, Survivorship, and Development Time in Eggs Reared in the Absence of Males There was no relationship between egg size a nd egg survivorship (linear regression, F1,8 = 0.14, p =0.72). However, egg size was positivel y correlated with develo pment time, suggesting that larger eggs take longer to develop (linear regression, F1,5 = 13.85, p = 0.02; Figure 4-1). Cannibalistic Preferences by Males In the single-female scenario, males e xhibited no significant size preferences ( 2 = 2.25, df = 3, p = 0.522; Figure 4-2 A). In other words, fo r each given size class (i.e. small, small-medium, medium-large, and large), the rela tive abundance of eggs consumed was comparable to the initial relative abundance of eggs of that size class (Figure 4-3 A) and this pattern of cannibalism is consistent with random consumption of eggs w ith regard to size. However, males exhibited significant preferences in the multiple-female scenario ( 2 = 15.13, df = 7, p = 0.034; Figure 4-2 B), and specifically, males exhib ited a significant preference for the larger eggs of female 2 (Figure 4-2 B). In this case, the relative abundance of female 2 s medium-large and large eggs that were consumed was greater than the relative abundance of those eggs that were present initially (Figure 4-3 B). Discussion Male sand gobies cannibalized eggs selectiv ely, but only in some cases. When males received eggs sequentially from two females, they preferentially consumed the larger eggs from the second female only. Thus, my results s uggest that sand goby ma les exhibit non-random preferences for eggs when they spawn sequentia lly with two females. These patterns raise several questions. First, why do ma les prefer larger eggs in some cases, but not others, and in particular, what distinguishes female 2s larger eggs from those of female 1? It is possible that energetic benefits play a role in cannibalistic preferences-larger eggs likely provide a male with more energy (Kamler 2005), and filial cannibalism is thought to be a 74

PAGE 75

way in which caring parents attain energy to offset costs of care (Rohwer 1978; Manica 2002). Thus, it wouldnt be surprising if males maximized their per-offs pring energetic gain. However, males didnt always consume larg er eggs-they exhibi ted no size preference for female 1s eggs in either 2004 or 2006. If males were attempting to maximize their per-of fspring energetic gain, we would have expected them to consume larger eggs in all cases. This was not the case, and thus, it does not appear that energetic gain alone can explai n the patterns of cannibalism observed in the sand goby. Alternatively, it is possible that the preference fo r the larger eggs of female 2 is associated with decreased duration of care and the ability to re-enter the mating pool sooner. Larger eggs took longer to develop, and the eggs of female 2 were already several hours (i.e., up to 24 h) behind those of female 1. Thus, the larger eggs of female 2 would likely hatch later and require a longer duration of care than female 1s eggs and the smaller eggs of female 2. Perhaps consuming the larger eggs of female 2 allows a male to decrease time spent caring for the current brood, thereby allowing him to re-enter th e mating pool sooner. This hypothesis seems particularly relevant for sand gobies, which live only one year and have multiple brood cycles over a six-week period. During a given brood cycle, males receive eggs for just a few days and then enter a care-only phase during which they do not receive additional eggs until their current brood hatches (typically 7 15 days from spawning). Indeed my results suggest that preferentially consuming larger eggs potentially reduces the duration of care for a given clutch by several days (Figure 4-1). It is easy to imagin e how even a small reduc tion in the duration of care over multiple brood cycles might allow a male an additional brood cycle, which in turn might increase the total number of eggs he receives over the breeding season. This hypothesis is consistent with some theoretical work (Chapt er 6 and Klug and Bonsa ll 2007), which suggests 75

PAGE 76

that parental fitness is highly sensitive to the maturation rate of eggs. Specifically, I suggest (Chapter 6 and Klug and Bonsall 2007) that males potentially be nefit by consuming eggs that take longer to develop, and that the consumption of slower developing eggs can directly facilitate the evolution of filial cannibalism. However, consuming larger eggs likely comes at a cost, as egg size is correlated with larva size in the sand goby and larger eggs have been shown to have higher post-hatching survival in a range of fishes (Kamler 2005). These ideas warrant further attention, and in particular, more work is n eeded to understand factor s affecting the optimal duration of care in this species and others. In summary, I have demonstrated that male sand gobies can cannibalize eggs selectively with regard to the order in which those eggs ar e laid and the size of eggs in some contexts. However, in this case, size per se does not appear to be the f actor influencing cannibalism. Rather, males seem to be sensitive to the expe cted development rate of eggs. More work is needed to further understand specific costs and benefits of consuming eggs of a particular size and expected developmental rate. Additionally, it will be important to assess the relationship between other aspects of egg phenotype and filial cannibalism in this and other species. 76

PAGE 77

Figure 4-1. Relationship between initial egg size and development time (i.e. the number of days from spawning until hatching) in eggs reared in the absence of males. 77

PAGE 78

A B Figure 4-2. Preferences ( ) in egg consumption by parental males. Male preferences for A) 4 size classes of eggs (labeled here as sma ll, small-medium, medium-large, and large) when each male mated with a single female in 2004 and B) 4 size classes of eggs from two females when males mated sequentially with 2 females in 2006. Bars represent means and error bars are standard error. 78

PAGE 79

A B Figure 4-3. Distribution of egg size initially and in the male diet. The mean (+/S.E.) proportion of eggs that were either small, small-medium, medium-large, or large initially, i.e. on Day 1 of each replicate (filled bars), a nd the mean (+/S.E.) proportion of eggs consumed by males (open bars) that were either small, small-medium, medium-large, or large in A) 2004 and B) 2006. 79

PAGE 80

CHAPTER 5 SELECTIVE FILIAL CANNIBALISM IN THE FLAGFISH Introduction Filial cannibalism commonly co-occurs with pa rental care in many animals and has been particularly well-documented in fishes exhibiting paternal care during the egg stage (Polis 1981; Manica 2002). While parental care typically incr eases offspring survival (discussed in CluttonBrock 1991), filial cannibalism involves the killing of ones own young. It is difficult to imagine how such a behavior could represent an adap tive strategy. Indeed, prior to the 1970s filial cannibalism was dismissed as a rare behavior with little or no evolutionary significance (discussed in Manica 2002). However, filial cannibalism has now been well-documented in nature, and in many species caring parents consum e more offspring than would die naturally. In recent years, much empirical and theoretical work has focused on understanding the adaptive significance of filial cannibalism (reviewed in Manica 2002). Curre ntly, filial cannibalism is thought to be an adaptive strategy, and specificall y, some have suggested that filial cannibalism involves an adaptive trade-off in which parents gain energy or nutrients from eggs, which they then use to better care for their remaining offs pring or re-invest in future reproduction (energybased hypothesis: Rohwer 1978; Sa rgent 1992; reviewed in Mani ca 2002). According to this hypothesis, whole clutch canniba lism (i.e., the consumption of all offspring present during a given reproductive bout) re presents an investment in future reproduction, whereas partial clutch cannibalism (i.e., the consumption of only some offspring pr esent) can either be an investment in future or current reproduction. Al ternatively, others have sugges ted that by reducing egg density in the nest, partial clutch filial cannibalism can improve egg survi vorship of remaining offspring, thereby increasing overall egg su rvival (oxygen-mediated hypothesi s: Payne et al. 2002, 2004; density-dependent egg survival hypothesis: Klug et al. 2006). However, neither energetic need 80

PAGE 81

nor density-dependent egg survival can explain th e prevalence of filial cannibalism in natural systems. Indeed, filial cannibalism is not affected by energetic need in some species (Belles-Isles and Fitzgerald 1991; Lindstrm and Sargent 1997) and continues to occur when egg density is relatively low in others (Klug et al. 2006 and Chapter 3). Thus, the evolutionary significance of filial cannibalism remains unclear in many systems. Alternatively, recent theoretical work suggest s that the ability to selectively cannibalize offspring that have redu ced phenotypic quality can independently facilitate and play a key role in the evolution of filial cannibalism (Klug and Bonsall 2007 and Chapter 6) The elimination of lower quality offspring has been demonstrated in relation to sele ctive embryo abortion in humans and plants (Forbes 1997; Diamond 1987; Burd 1998; Karkkainen et al. 1999), brood reduction (Mock and Forbes 1995; Mock a nd Parker 1997; Forbes and Mock 1998), and parents allowing or encouraging siblicide of low quality offs pring (Mock and Parker 1997; Stearns 1987). Because the elimination of low qua lity offspring is thought to play a central role in explaining the evolutionary signif icance of offspring abandonment and brood reduction (e.g., Stearns 1987, 1992; Forbes and Mock 1998), it is surprising that the relationship between offspring quality and filial cannibalism has received little attention. While some studies have found a relationship between filial cannibalism and un certainty of paternity (Neff 2003; Gray et al. 2007; Frommen et al. 2007) or egg age (Salfert and Moodie 1985; Sikkel 1994), little is known about the relationship between offspring quality and filia l cannibalism of viable young (see also Kraak 1996, for discussion of cannibalism of diseased eggs). Thus, the general importance of selective filial cannibalism remains unclear. The first step in understanding the potential impo rtance of selective filial cannibalism is to determine whether parents that provide care prefer entially consume eggs with regard to some 81

PAGE 82

aspect of offspring quality. Here, I examine th e relationship between one aspect of offspring quality, egg energy content, and fili al cannibalism in the flagfish (Jordanella floridae). Egg energy content was used as a proxy for quality because energy content and size have been strongly and positively correlated with post-hatching survival and growth (and hence fitness) in a range of fish species (reviewe d in Kamler 1992, 2005; Keckeis et al. 2000; Brownman et al. 2003). In addition, I examined the relationship be tween filial cannibalism and maternal condition and size, because maternal effects on egg qua lity have been well-documented (Kamler 2005), and specifically, because a positive relationship between female size and offspring quality has been found previously in fishes (reviewed in Kamler 2005). Methods Study Species Flagfish males alone provide parental care of eggs (including nest guarding, cleaning, and fanning), filial cannibalism is prevalent (Klug & St. Mary 2005), and pa rental males are known to consume more eggs than die naturally (i.e., the rate of filial cannibali sm is greater than the mortality rate; Klug et al. 2005). Thus, filial canniba lism does not function soley to clean the nest of dead or diseased eggs. Indeed, egg survival in the absence of parental males and predators is very high (> 90% egg survival), and thus, most egg mortality can be attributed to parental males. Flagfish typically live only one year in the wild, and both males and females mate multiply during a several month breeding season. Additionally, eggs are spawned and fertilized individually, and sneaking is not thou ght to occur in this species. Experimental Design The study was conducted April July 2005 in Gaines ville, Florida. All fish were collected from the Otter Creek/Waccasassa River drainage in Levy County, Florida within 20 days of the experiment. Fish of both sexes were housed in separate freshwater holding tanks. All 82

PAGE 83

experimental aquaria were 36 L and equipped with air-driven filtration, a spawning mat (i.e. a 100 cm2 ceramic tile with heavy, green acrylic felt ca rpet glued to the top of the tile), and three artificial plants. Throughout the experiment, all fish were fed ad libitum a diet consisting of algae tablets and frozen brine shrimp. During the expe riment, all fish experi enced a 14h:10h light:dark cycle and temperature was maintained at 26C. I initiated each replicate by randomly selecting and placing one male and one female in an aquarium. The male and the female were al lowed to spawn, and immediately following spawning, I briefly removed the ne st with eggs from the tank. I counted the number of eggs, and for a subset of the clutches (N = 18), I removed three eggs from the nest and used them in subsequent energy assays. No eggs were removed from the nests of six males, which allowed me to evaluate whether there was an effect of egg removal on filial cannibalism. In all cases, a clear acrylic divider containing multiple holes was used to physically separate the male and female following spawning, which ensured that all cannibalism was done by the male. After counting the eggs, I returned the nest w ith eggs to the male, who was allowed to care for the eggs until hatching. I counted the eggs daily by visually inspecting the nest. Eggs usually began to hatch on day four, and thus I measured egg survivorship through day three of each replicate. Eggs never became diseased or infected with fungus during the course of the study. I weighed and measured all experimental fish just prior to the start of each replicate. Fo r five replicates, I did not obtain reliable weight measurements, and thus these fish were excluded fr om analyses involving parental condition or size. I used the condi tion measure K (where K=100*weight/(length)3; Williams 2000), which provides a size-independent estimate of physical condition, to evaluate the relationship between parental condition and the occurrence of filial cannibalism. I also 83

PAGE 84

estimated the relationship between parental size and filial cannibalism, and weight and standard length (which were highly correlated) were used as estimates of size. Energy Assays I used dichromate oxidation technique (modified from McEdward and Carsons 1987) to quantify total energy content (i.e. J egg-1) of each sampled egg. Egg energy content was compared to a glucose standard (1 4 J mL-1). Specifically, I incubated each egg in 0.5 mL 70% phosphoric acid at 105C for 15 min. I allowed the solution to cool to ro om temperature, and then oxidized the sample with 1 mL of 0.3% pot assium dichromate in concentrated sulphuric acid at 105C for 15 min. Samples were then dilute d with 3.5 mL distille d water and I measured absorbance using a spectrophotometer ( 440 nm). I calculated tota l energy by comparing the absorbance of each sample with that of the glucose standards. I performed energy assays on eggs from 18 clutches. Only intact eggs were used for the energy assays, and I was able to quantify the total energy of one egg in 12 clutches, two eggs in three cl utches, and three eggs in two clutches. While it would have been ideal to measure a larger sample from each clutch, this was impossible because flagfish spawn relatively few eggs (typically < 100) and I wanted to minimize any effects of egg removal. Additionall y, because there was greater variation in the per-egg energy content between clutches than with in clutches (discussed in Results), a small within clutch sample of egg en ergetic content provide d an estimate of the mean energy content of eggs within a given clutch. Statistics I used linear regression to evaluate the re lationship between male and female condition (i.e. K) and size (i.e. weight a nd length); the relationship betw een the mean energy content per egg within a clutch and the number of eggs sp awned; the relationship between the number of 84

PAGE 85

eggs spawned or received and female and male condition and size; and the relationship between mean egg energy content and male and female condition and size. Whole clutch cannibalism represents a termin ation of current repr oduction, and therefore any benefit of whole clutch cannibalism is a ssociated with future reproductive success. In contrast, benefits of partial clutch cannibalism can be associated with either increased current or future reproductive success. Because whole and partial clut ch cannibalism likely represent different biological phenomena, I an alyzed these data separately. First, I used stepwise logistic regression (remove if p>0.15) to evaluate the relationship between whole clutch cannibalism and 1) male condition (i.e. K), 2) male size (i.e. weig ht), 3) female condition (i.e. K), 4) female size (i.e. weight), 5) the number of initial eggs presen t, and 6) mean energetic content per egg within a clutch. I then considered only cases of partia l clutch cannibalism and used stepwise linear regression (remove if p>0.15) to evaluate th e relationship between the proportion of eggs consumed and 1) male condition (i.e. K), 2) male size (i.e. weight), 3) female condition (i.e. K), 4) female size (i.e. weight), 5) mean egg energy, and 6) the initial numbe r of eggs present. To meet assumptions of normality, the proportion of eggs consumed was arcsin square root transformed. Data associated with the number of eggs consumed could not be transformed to meet assumptions of normality. Thus, I used spearman rank correlation tests to evaluate the relationship between the number of eggs consumed and male and female size (i.e. weight) and condition (i.e. K), mean egg energy, and the initial number of eggs present. Results Of the 24 males, 10 exhibited whole clutch cannibalism, 12 exhibited partial clutch cannibalism, and 2 males consumed no eggs. Ex cluding cases of whole clutch cannibalism, the mean (+/SE) percentage of eggs canniba lized was 38.7 +/0.097 % (or 45.2 +/0.10 % excluding males who didnt exhib it any cannibalism). As mentione d previously three eggs were 85

PAGE 86

removed from the nests of 18 males and no eggs we re removed from the nests of 6 males. There was no effect of egg removal (i.e., removal of 3 eggs for subsequent energy assays) on whole clutch cannibalism (logistic regression, 2 = 0.523, df = 1, p = 0.465) or partial clutch cannibalism (ANCOVA, F1,12 = 2.20, p = 0.164). On average, e ggs contained 1.94 +/0.57 J egg1 (mean +/SD). As mentioned previously, energy measurements were taken for multiple eggs for 5 clutches. In these cases, the mean energy content (+/SD) was 2.01 +/1.07 J egg-1. The standard deviation within a clutch ranged between 0.0094 to 0.089 J egg-1, and was on average 0.042 J egg-1. In all cases, the within clutch variation in energy content was much less than the between clutch variation in energy content. Parental Condition and Size, Egg Energetic Content, and Egg Number There was no significant relationship between male and female condition (F1,19 = 1.84, p = 0.19) or male and female size (weight: F1,19 = 0.91, p = 0.35; length: F1,19 = 0.97, p = 0.34). Additionally, there was no significant relationship between mean energy content of eggs and the number of eggs spawned (F1,16 = 0.84, p = 0.37), suggesting that there was not a clear trade-off between the number of offspri ng produced and the mean energy invested into those offspring within a given reproductive episode in this experiment. There was no significant rela tionship between female condition and mean egg energy content (linear regression, F1,25 = 0.37, p = 0.55) or the number of eggs spawned (F1,19 = 1.92, p = 0.18). Likewise, there was no si gnificant relationship between female size and the number of eggs spawned (weight: F1,19 = 1.69, p = 0.21; length: F1,19 = 0.88, p = 0.36). However, there was a significant relationship between female size and the mean egg energy content (weight: r2 = 0.41, F1,11 = 7.63, p = 0.02; length: r2 = 0.42, F1,11 = 7.92, p = 0.02; Figure 5-1 A). There was no significant relati onship between male condition and mean energetic content of eggs received (linear regression, F1,1l = 0.08, p = 0.78) or the number of eggs received (F1,19 = 86

PAGE 87

0.53, p = 0.53). Likewise, male size was unrelated to the number of eggs received (weight: F1,19 = 0.70, p = 0.41; length: F1,19 = 1.35, p = 0.26). However, larger males received eggs that were on average more en ergetic (weight: r2 = 0.27, F1,11 = 5.33, p = 0.041; length: r2 = 0.33, F1,11 = 7.29, p = 0.021; Figure 5-1 B). As mentioned above, there was no relationship between male and female size, and thus, assortative mating does not explain these patterns. Whole Clutch Cannibalism There was no significant effect of initial egg number, male condition, male size, or female condition on the occurrence of w hole clutch cannibalism, i.e., th e proportion of clutches that were entirely eaten (stepwise logistic regression, p > 0.15 in all cases). However, whole clutch cannibalism was more frequent when the mean en ergetic content of eggs was relatively great (logistic regression, 2 = 6.71, df = 1, p = 0.01; Figure 5-2 A) and when female size was greater ( 2 = 4.73, df = 1, p = 0.03; Figure 5-2 B). Partial Clutch Cannibalism When only cases of partial clutch cannibalism were considered, there was no relationship between male condition and the proportion of eggs surviving (p > 0.15), or the number of eggs consumed (df = 10, t = -0.41, p >0.05). However, la rger males tended to consume a smaller proportion of eggs (F1,8 = 3.90, p = 0.10; Figure 5-3 A), and they consumed significantly fewer eggs than smaller males (weight: df = 10, t = -13.54, p < 0.01; length: df = 10, t = -2.45, p < 0.05; Figure 5-3 B). Female condition was unrelat ed to the proportion of (p > 0.15) or the number of eggs consumed by males (df = 10, t = -0.72, p >0.05). However, female size was negatively correlated with the propor tion of eggs consumed (weight: F1,8 = 11.77, p = 0.009; Figure 5-4 A) and the number of e ggs consumed (weight: df = 10, t = -8.36, p <0.01; length: df = 10, t = -5.21, p <0.01; Figure 5-4 B). 87

PAGE 88

For cases of partial clutch ca nnibalism, egg energy was unrelated to the proportion of eggs consumed (p > 0.15), but there was a negative relationship between mean egg energy and the number of eggs consumed (df = 8, t = -6.49, p <0.01; Figure 5-5). There was no significant relationship between the number of eggs initially present and the pr oportion of eggs (p > 0.15) or the number of eggs (df = 12, t = 0.0266, p > 0.05) consumed. Discussion Male flagfish preferentially cannibalized eggs laid by females of larger body size and when the mean energy content of eggs was high for cas es of whole clutch cannibalism. Because egg size, egg energy, and maternal size are correlate d with post-hatching survival and growth in fishes (reviewed in Kamler 2005), it appears that males are sacrific ing high quality offspring for a relatively large energetic gain when they prac tice whole clutch canniba lism. With regard to partial clutch cannibalism, the number of eggs c onsumed increased as the mean energy content of eggs decreased. The energy-based hypothesis of filial cannibalism (Rohwer 1978; Sargent 1992) suggests that partial clutch cannibalism is a way in which males attain energy to offset costs of care (Manica 2002). If the function of filial cannibalism is to attain energy that can be reinvested in increased current or future reproduction (which I did not evaluate here), it seems likely that a males energetic need can be sati sfied by consuming a smaller number of eggs when those eggs have a relatively high energetic content. However, male s in this experiment received food (i.e., algae and brine shrimp) ad libitum, and thus energetic need alone cannot explain the filial cannibalism observed in the present study. Alternatively, because egg energy content is likely correlated with subseque nt offspring survival (Kamle r 2005), it is possible that the negative relationship between the nu mber of eggs consumed and mean energetic content of eggs suggests that males are investing more into offspr ing of relatively high quality. Similarly, males consumed a greater proportion of eggs spawned by relatively small females. While female size 88

PAGE 89

was correlated with mean egg energy content, egg energy alone did not explain the proportion of eggs cannibalized. Because female body size is pos itively correlated with egg size and resistance to starvation and predation in a range of fishes (reviewed in Kamler 2005), it appears that males consumed a greater proportion of relatively low quality eggs when they practiced partial clutch cannibalism. In summary, this experiment suggests that at least some aspects of offspring quality (i.e. egg energy, maternal size) affect both whole a nd partial clutch filial cannibalism, albeit in different ways. With regard to partial clutch cannibalism, males consume more of the lower quality offspring. This finding is consistent with previous work on the el imination of offspring via abandonment, siblicide, or infanticide, whic h focuses on the removal of low quality offspring (e.g., Mock and Parker 1997; Forbes and Mock 1998). Additionally, this finding is consistent with theoretical work suggesting that selective filial cannibalism of lo w quality offspring is beneficial to caring parents (Klug and Bonsall 2007 and Chapter 6). However, with regard to whole clutch canni balism, males are more likely to cannibalize higher quality offspring. This fi nding suggests that males are sens itive to the nutritional benefits of cannibalism, which is consistent with the id ea that cannibalistic parents use eggs as an alternative food source (Rohwer 1978; Manica 2002). However, the specific finding that males sacrificed their higher quality offspring for increa sed energetic gain is not explicitly predicted by current theory. Indeed, previous work in the flagfish (Klug and St. Mary 2006 and Chapter 2) suggests that the energetic benefi ts of consuming eggs to male size or reproduction are relatively small in comparison to those of food. Thus, add itional theoretical and empirical work is needed to better understand the expected trade-offs associated with filial cannibalism. In future studies, 89

PAGE 90

it will be important to consider within-clu tch patterns of cannibalism to better understand patterns of parental investment and filial cannibalism. Finally, there was no relationship between initial egg number and whole clutch cannibalism in the present study. This finding is in contrast to some previous theoretical predictions (Rohwer 1978) and em pirical findings (reviewed in Manica 2002) suggesting that whole clutch cannibalism is more common when clutch size is relatively small (but see Payne et al. 2003, who found that smaller clutches were no t preferentially consum ed). Similarly, there was no clear trade-off between the number of eggs spawned and the mean energetic content of those eggs. This trade-off is a key assumption of life-history evolution (S tearns 1987). However, it is likely that the scale of the present experiment as well as other source s of variation, made it difficult to detect such a trade-off if one does indeed exist in the flagfish. Indeed, better understanding of female investment will necessita te an experiment specifically designed to assess such trade-offs over a longer time frame. 90

PAGE 91

A B Figure 5-1. Relationship between th e mean energy per egg (J egg-1) within a clutch and A) female weight and B) male weight. 91

PAGE 92

A B Figure 5-2. Relationship between the frequency of whole clutch cannibalism and A) the mean energy per egg (J egg-1) within a clutch and B) female weight. Lines represent the predicted probability of whole clutch ca nnibalism as a function of A) mean egg energy or B) female weight, as determ ined by a logistic regression; for A) xe y 97.738.151 1 1 and (B) xe y 86.7158.31 1 1 where y is equal to the probability of whole clutch cannibalism and x is equal to either mean egg energy A) or female weight B). 92

PAGE 93

A B Figure 5-3. Relationship between male weight and A) the propor tion and B) the number of eggs consumed for cases of partial clutch cannibalism. 93

PAGE 94

A B Figure 5-4. Relationship between female weight and A) the proportion and B) the number of eggs consumed for cases of partial clutch cannibalism. 94

PAGE 95

95 Figure 5-5. Relationship between th e mean energy per egg (J egg-1) within a clutch and the number of eggs consumed for cases of partial clutch cannibalism.

PAGE 96

CHAPTER 6 WHEN TO CARE FOR, ABANDON, OR EA T YOUR OFFSPRING: A MODEL OF THE EVOLUTION OF PARENTAL CA RE AND FILIAL CANNIBALISM Introduction Adaptive theories of evolution typically suggest that parents should exhibit strategies that increase offspring survival, and parental care is one way in wh ich parents are thought to achieve this (reviewed by Clutton-Brock 1991). Although parental care is assumed to increase offspring survival, filial cannibalism, the consumption of ones own viable offspring, commonly co-occurs with parental care. Indeed, filial cannibalism is prevalent in a range of taxa exhibiting parental care (Polis 1981; Elgar and Cres pi 1992). For example, caring fe males consume some of their young in the bank vole ( Clethrionomys glareolus, Klemme et al. 2006), the house finch ( Carpodacus mexicanus, Gilbert et al. 2005), and the wolf spider ( Pardosa milvina, Anthony 2003), and both parents of the burying beetle ( Nicrophorus orbicollis ) are known to consume their offspring (Bartlett 1987). Filial cannibalism has been particularly well-documented in fish species with paternal care duri ng the egg stage (reviewed in Mani ca 2002). Indeed, because of its prevalence in fish systems, most theoretical a nd empirical work on filial cannibalism has focused on fish (but see Bartlett 1987, Thomas and Manica 2003, Creighton 2005). While early ethologists considered filial ca nnibalism a social pathology with little or no evolutionary significance, filial cannibalism is now typically th ought to reflect an adaptive trade-off between current and future reproductive success (e.g ., Manica 2002, 2004). However, despite much theoretical development and empirical work over the last few decades, the evolutionary significance of filial cannibalism re mains unclear in many systems. The most widely accepted hypothesis of filial can nibalism as an adaptive strategy suggests that energetic need is the primary factor leading to filial cannibalism, and that a caring parent gains energy and nutrients from consuming their o ffspring that are then reinvested into future 96

PAGE 97

reproduction, thereby increasing net reproduc tive success (Rohwer 1978; Sargent 1992). Specifically, whole-clutch cannibalism (i.e., the consumption of all offspring during a given reproductive bout) is assumed to be an investment in future repr oduction, whereas partial-clutch cannibalism (i.e., the consumption of only some o ffspring present) can represent an investment in either current or future re production. This energy-based hypot hesis predicts that cannibalism will increase as food availability decreases and when parental condition is poor (Rohwer 1978; Sargent 1992). While food availability and/or parental condition a ffect the amount of cannibalism in some species (e.g., Stegastes rectifraenum Hoelzer 1992; Pomatoschitus microps Kvarnemo et al. 1998; Abudefduf sexfasciatus, Manica 2004), it has no effect in others (e.g., Gasterosteus aculeatus, Belles-Isles and Fitzgerald 1991; Etheostoma flabellare, Lindstrm and Sargent 1997), and in two systems cannibalism declines as male condition or food availability decreases ( Pomatoschistus minutus, Klug et al. 2006 and Chapter 3; Jordanella floridae Klug and St. Mary 2005 and Chapter 2). Other studies have examined whether eggs can provide a caring parent with suffi cient energy to offset the costs of care. Again, the evidence is mixed-two studies concluded that energy attained from filial cannibalism is sufficient to offset costs related to care (Kume et al. 2000; Thomas and Manica 2003), while in another, energy from eggs was found to be insufficient (Smith 1992). Thus, parental energetic need alone cannot explain the prevalence of filial cannibalism. Alternatively, Payne et al. (2002) and Klug et al. (2006 and Chapter 3) suggested that filial cannibalism is mediated by density-dependent eg g survivorship, and that by consuming some eggs in their nests, caring parents can improve the survivorship of the remaining eggs and increase their net reproductive success. Such dens ity-dependent egg survivorship is potentially related to the physical environm ent (e.g., oxygen availability, Pa yne et al. 2002) or increased 97

PAGE 98

benefits of parental care to the remaining offspring. The hypothesis of filial cannibalism mediated by density-dependent egg survivorship has received suppor t in two marine fish species ( Stegastes leucostictus Payne et al. 2002; Pomatoschistus minutus, Klug et al. 2006 and Chapter 3), but has in general received littl e further empirical or theoretical examination (but see Payne et al. 2004). Likewise, some have suggested that filial cannibalism is a mechanism by which parents reduce brood size in response to anti cipated resource competition amongst their adult offspring (Bartlett 1987; Creighton 2005) or kill offspring of re duced quality (Forbes and Mock 1998; see also Kozlowski and Stearns 1989). While the former hypothesis has received some attention in the burying beetle (Creighton 2005), neither of these hypothese s of filial cannibalism has been explicitly evaluated. Because of the mixed empirical support for the energy-based hypothesis and the lack of empirical evidence regarding alternative hypotheses, filial cannibalism remains an evolutionary conundrum. Indeed, previous work suggests that a parents energetic need (Rohwer 1978; Sargent1992; Manica 2002), expectations regard ing offspring survival or reproductive value (Payne et al. 2002; Neff 2003; Kl ug et al. 2006 and Chapter 3), competition for mates (Sikkel 1994; Kondoh and Okuda 2002), and anticipated o ffspring resource competition (Creighton 2005) are potentially important factors for explaining the adaptive significance of filial cannibalism. However, previous theory has tended to focus on each of these factors in separate theoretical contexts (e.g., en ergetic benefits of consuming offspring: Rohwer 1978; Sargent 1992; expectations regarding offspring survivorship: Payne et al 2004; variation in offspring quality: Forbes and Mock 1998; mate availability: Kondoh and Okuda 2002), despite empirical evidence suggesting that it is unlik ely that any single factor alone can explain the prevalence of filial cannibalism (e.g., Manica 2004; Klug et al. 2006 and Chapter 3). 98

PAGE 99

Here, I develop a model of parental care, to tal offspring abandonment (i.e., no care), and filial cannibalism to begin to isolate the pivotal factors affecting the evolution of care and filial cannibalism. First, I determine the general conditions under which we would expect these strategies (i.e., care, no care/t otal abandonment, filial cannibalism) to evolve alone or in combination. I then evaluate the plausibility of multiple alternative hypo theses within a single theoretical context by asse ssing the importance of a range of pot ential costs and benefits of care and cannibalism. Specifically I focus on costs an d benefits related to energetics, offspring survival and quality, mate competition, and general resource competition. Methods The model is set up as an ecological problem in which a rare mutant with a unique lifehistory strategy is allowed to invade a resi dent population (e.g., Vincent and Brown 2005). Specifically, the resident strategy represents the st rategy that is currently exhibited by individuals in a population, and the mutant strategy is some alternative strategy not currently exhibited by individuals in the population. I assume that the resi dent strategy is in equilibrium (i.e., it is the strategy that currently pr evails in the population) and that an alternative mutant strategy invades from rare into the population. Because the general characteristics of an organism in one life-history stage (e.g., maturation rate, juvenile survival, adult mortal ity during the egg, juvenile, and adult stages) potentially affect the costs and benefits of stra tegies occurring in another life-history stage, I assume a stage-structured system in which individuals develop thr ough an egg stage and a juvenile stage, and then mature and reproduce as adults. While in the egg stage, individuals can either be abandoned by parents, receive parent al care, suffer filial cannibalism, or receive parental care and suffer filial cannibalism (Fi gure 6-1). Below, I outline the dynamics of a system in which a mutant with care and/or ca nnibalism invades a resident population that either 99

PAGE 100

lacks or provides parental care. I then use mutual invasion analysis to explore the effects of costs and benefits of varying strategi es on lifetime fitness and the e volution (i.e., invasion from rare and subsequent fixation) of parental care and/ or filial cannibalism. Model Dynamics I consider a stage-structured system (which is appropriate for many fish, bird, and insect systems) in which individuals pass through an egg (E ), juvenile, and adult stage ( A ). The number of eggs increase as adults reproduce and decrease as eggs mature and as eggs die, such that: () ()1 ()()EEdE At rAt dEtmEt dt K (6-1) where r represents the rate of egg fe rtilization (i.e., a proxy for mean reproductive rate of adults), represents death rate of eggs, and is the rate at which eggs mature. I assume logistic population growth, where K represents population carrying capacity, a nd density-dependence associated with resource competition affects adul t reproduction (i.e., the ra te of fertilization). Adults in the population increase as eggs mature and survive the juvenile stage, and decrease as adults die, such that: EdEm dA dt mEtdAtEJ A()(), (6-2) where is a time delay representing the juvenile stage, J is survival rate through the juvenile stage, and is the death rate of adults (Figure 6-1). Ad As mentioned above, I assume that the populat ion exhibiting the resi dent strategy is in equilibrium. Specifically, the resident strategy is assumed to be fixed in the population (and on average, all individuals therefore have relative fitness equal to 1). Because 1) the resident strategy is fixed and 2) the resi dent population (which exhibits the strategy of interest) is regulated by density-dependence, the resident population is assumed to be in ecological 100

PAGE 101

equilibrium (i.e., the population density is not in creasing or decreasing). Because we know that a population in equilibrium is not increasing or decreasing (i.e. A(t) and E(t) both equal zero), I can analytically solve for the equilibrial densities by setting A(t) equal to E(t) and then algebraically solving for the adult and egg densities at equilibrium ( A* and E* ) The equilibrial densities in this model are thus E d A A mEJ* (6-3) AK dd m d rEA EJ A J* 1. (6-4) Resident and Mutant Trade-Offs To explore the fixation of different strategies, I allowed rare mutants with different life histories to invade a resident population. I considered the follow ing cases: 1) a rare mutant who provides parental care invades a resident population with no care (a nd no cannibalism), 2) a rare mutant who practices filial cannibalism inva des a resident populati on with no care (and no cannibalism), 3) a rare mutant who provides pa rental care and practices filial cannibalism invades a resident population with no care and no cannibalism, and 4) a rare mutant who provides parental care and practices filial cannibalism invades a resi dent population that provides parental care (but does not canni balize). The different life-history strategies are represented through the incorporation of appropriate trade-offs into the model (described below and in Table 6-1), and the model was analyzed using linear additive trade-offs and non-linear trade-offs (Table 6-1). In cases in which parental care was provided (either by individuals in the resident population and/or by the rare mutant), I assumed th at parental care increase s the survivorship of eggs (i.e., as Ed decreases, parental care increases) and that receiving parental care during the 101

PAGE 102

egg stage increases an individua ls likelihood of surviving through the juvenile stage (i.e., the level of care receiv ed as an egg affects quality such that J increases as decreases). Providing parental care is assumed to be costly to the parent providing it, and t hus, I assumed that the reproductive rate of adults (i.e., their rate of producing fertilized eggs) decreases and that the death rate of adults increases as care increases (i.e., r decreases and increases as decreases). Furthermore, in all cases in which care is provided, a decrease in the maturation rate of the eggs was associated with an increase in the reproduc tive rates of adults: the less time that an individual has to spe nd caring for a clutch of eggs, the greater that individuals reproductive rate will be (i.e., as decreased, r increased). EdAdEdEm When I considered the case of a rare mutant practicing filial canni balism, I assumed some energetic benefit of cannibalism, su ch that the death rate of adults decreased and the reproductive rate of adults increased as cannibalism increased (i.e., decreases and r increases as the rate of cannibalism, increases). A further goal of these an alyses was to determine if cannibalism could evolve in the absence of a substantial bene fit of cannibalism. In these analyses, I assume no direct benefit of cannibalism (i.e., there is no effect of on or r ). AdAd I analyzed the model both assuming that e gg survivorship was density-independent and for the case in which egg survivorship was de nsity-dependent. For the cases in which egg survivorship was assumed to be density-dependent, the death rate of eggs follows an increasing function in E and I considered two functions: 2EdE, and (6-5) 1)1(E dE, (6-6) where is the strength of density -dependence (following Bellows 1981). I considered these two particular density-dependent functions because they are very general forms of density 102

PAGE 103

dependence that are common in nature (Bellows 1 981). The first function (eqn. 6-5) assumes that egg mortality increases exponentially as egg dens ity increases; the sec ond equation (eqn. 6-6) assumes that egg mortality increases with in creasing egg density, but the increase is not exponential and the precise nature of this relationship is determines by Specifically, a relatively large represents more intense density-depende nce (i.e., a relatively great increase in mortality as density increas es) and a relatively small represents relatively weak densitydependence (i.e., a relatively small increas e in mortality as density increases). Invasion Dynamics and Fitness Under the density-independent egg survivor ship scenario and the assumptions given above, the dynamics of the rare mutant are given by the following equations and by incorporating relevant trade-offs (Table 6-1): )()()( )( 1)( tAtEtEmtEd K A tAr dt dEm m mEm m m E m mm m (6-7) )( )( tAd tEm dt dmAm Jm mEm Am, (6-8) where is the mutants rate of cannibalism ( equals an average rate of cannibalism, which could represent some combination of w holeand partial-clutch cannibalism; = 0 if the mutant does not cannibalize). Specifically, the rate of ca nnibalism is a function of the number of rare mutants in the population ( Am) and then number of eggs the mutant has ( Em). If the mutant does not cannibalize, = 0. The mutant is assumed to be rare in the population, and thus, densitydependence operating on adult mutant reprodu ction occurs through competition with the resident. To evaluate the life-history characteristics a nd trade-offs affecting the invasion of a rare, novel strategy, I calculated the fitn ess of individuals exhibiting the mutant strategy relative to the fitness of individuals ex hibiting the resident strategy. The st age-structured nature of the model 103

PAGE 104

and the time delay representing the juvenile stag e makes it impossible to compute relative fitness from the differential equations above (eqns. 6-7 and 6-8). Inst ead, the lifetime fitness of the mutant can then be found from the determinant of the matrix describing the mutants invasion dynamics: Am Jm Em m m Em Emd m K A r md )exp( 1 (6-9) Hence, while some life-history parameters (e.g., fertilization rate of eggs, r ) will be correlated with lifetime fitness under some scenarios, the real measure of fitness in this model is the eigenvalues of the invasion matrix. The eigenvalues represent the fitness (and hence the invasibility) of the mutant strategy relative to the resident strategy when both evolutionary factors (e.g., trade-offs between current and fu ture reproduction) and ecological factors (e.g., resident population density, the intensity of competition amongs t adults) are considered. To evaluate the invasion and repla cement dynamics of a rare mutant that provides parental care and/or practices filial cannibalism, I used the fitness function of the mutant to calculate the evolutionary stable state(s) (i.e., when the rate of change in fitness is zero). I then performed mutual invasion analyses by evaluating when the fitness function is greater than zero (using a Newton-Raphson algorithm with the resident dynamics (A*) set at equilibrium) for different values of a life-history trait (see Table 6-1 and Online Appendix in Klug and Bonsall 2007). I evaluated and present pairwise invasion boundaries for different values of maturation rate of eggs. Comparing the invasion potential with regard to maturation rate of e ggs is ideal because: 1) it allows me to represent a wide range of lifehistory strategies, includ ing faster and slower reproducers, and 2) preliminary results suggest that lifetime f itness is highly sensitive to maturation rate. Specifically, I il lustrate the conditions for which 1) the mutant would invade and out-compete the resident, 2) the boundaries for wh ich the resident would invade and out-compete 104

PAGE 105

the mutant, 3) the putative coexistence range, in which the strategies have the potential to coexist, 4) a region of non-persistence, where ne ither strategy will pers ist (i.e., a region of extinction), and 5) a region in which neither st rategy will persist or initial conditions of the model determine the strategy that invades. Local stability analyses were performed and are described in the Online Appendix of Klug and B onsall (2007), and numerical simulations were performed to confirm that strategy coexistence occurs when the dynamics are stable and that regions of parameter space exist (labeled NP/IC and NP in Figures 6-2 through 6-6) where either neither strategy persists or the outcome is ba sed on initial conditions. I evaluated the invasion potential of the rare mutant for several biologic ally relevant scenarios by changing the value(s) of a single life-history parameter of interest for the mutant and/or resident populations. Biologically Relevant Comparisons In addition to the fixed trade-offs reflecting varying life-history stra tegies (Table 6-1), I explicitly considered the eff ects of varying selective regimes (e.g., differential mating success associated with a particular strategy, effects of care or ca nnibalism on population resources and hence carrying capacity) on the invasion dynamics. To do this, I used pairwise comparisons in which I altered the magnitude of one (or more) pa rameter(s) to reflect a biological scenario of interest. First, I evaluated the importance of offspring survival benefits of care on the invasion patterns of varying strategies. Empirically, parental care has been shown to reduce the death rate of offspring (discussed in Clutton-Brock 1991), and thus I compared the invasion patterns of the caring mutant for a range of cases in which care was effective (i.e.,
PAGE 106

Second, sexual selection has been hypothesized to be a major force in the evolution and fixation of parental care (A ndersson 1994; Baylis 1981). In some systems mate choice for a partner who will provide care is thought to aff ect the reproductive rate of the non-limiting sex (e.g., the number of eggs a caring parent receives per reproductive bout or over the course of the breeding season is correlated with parental care: Jordanella flordiae St. Mary et al. 2001; Pomatoschistus minutus Pampoulie et al. 2004). Likewise, fi lial cannibalism has been shown to increase the attractivenes s of a caring parents nest in some cases (e.g., Sikkel 1994) and might be preferred during mate choice if there are benefits of cannibalism to remaining offspring (e.g., through density-dependent egg survivorship). W ith regard to the mode l, if care or filial cannibalism is a trait that is pref erred by one sex, we would expect the mutant exhibiting care or cannibalism to receive more fertilizations per time period (e.g., a breedin g season or lifetime) than a resident who does not exhibit care. In this sense, rm (i.e., egg fertilizati on rate in eqn. 6-7) of a mutant who exhibits a prefer red trait would be expected to be greater on average than that of a resident who does not exhibit the preferred trait. To incorporate th is aspect of sexual selection, I compared invasion patterns for cases in which caring and/or cannibalism increased the reproductive rate of the caring mutant relative to the resident (i.e., rm > r ) to those in which the magnitude of the reproductive rate did not differ between the mutant and resident (i.e., rm = r ). Likewise, it is possible that filial cannibalism creates reproductive conflict between parents. In this case one woul d expect non-cannibalistic individuals to be favored during mate choice (Kraak 1996; Lindstrm 2000). To assess the importance of mate preference for a noncannibalistic partner, I compared cases in which the muta nt has a reduced reproductive rate relative to the resident (i.e., rm < r ), with the case in which the mutant and resident have equal reproductive rates (i.e., rm = r ). 106

PAGE 107

Parental care, no care, and filial cannibalism might affect population-level resources, and hence the competitive regime that individuals experience, in different ways. For example, providing care might necessitate greater per capita resources (e.g ., increased energetic need and nesting resources per individual) than not providing care (i.e, K > Km for a caring mutant invading a resident without care). Similarly, th e ability to cannibalize while providing care might represent a more efficient use of resources by in dividuals. Such an individual-level change in resource use would be reflected in the carrying capacity of a population of individuals exhibiting a particular strategy. Specifica lly, if a strategy allows indivi duals to use resources more efficiently (i.e., increase their pr oductivity), we would expect an increase in the carrying capacity of a population of individuals exhi biting that strategy (r elative to individual s who do not exhibit the strategy of interest, i.e., K < Km for a mutant who can cannibalize and care invading a resident who can only care). In this se nse, the population-le vel carrying capacity K associated with a strategy is a proxy for the effect th at strategy has on indivi dual-level resource use. To begin to evaluate the importance of such resource -related effects of varyi ng strategies, I compare cases in which carrying capacity varies betw een the mutant and the resident population. Likewise, for cases in which the mutant and resi dents have equal carrying capacities, I compared patterns for a range of carrying capacities to dete rmine if a relatively productive ecosystem (i.e., system with a large carrying cap acity) or unproductive eco system (i.e., system with a relatively small carrying capacity) favors the invasion of a particular strategy. Finally, to evaluate further patterns of cannibalism evolution, I compared the fitness boundaries for cases in which parents were allowe d to selectively cannibalize eggs with reduced future survivorship (i.e., dEm < dE and Jm > J ) with those of parents that could not selectively cannibalize (i.e., dEm = dE and Jm = J ). 107

PAGE 108

Results All of the strategies consid ered (i.e., parental care, no car e/total offspring abandonment, filial cannibalism) evolved over a range of parame ter space in all analyses. While the evolution of parental care and/or filial cannibalism were favored by benef its to adults and/or offspring (discussed below), such benefits were not esse ntial for the invasion of a particular strategy, highlighting the plausibility of a range of non-mutually exclusive altern ative hypotheses (Table 6-2). In all cases considered, the coexistence dynamics were stable (Online Appendix of Klug and Bonsall 2007). While incorpora ting non-linear trade-off functions (Table 6-1) into the model altered the results quantitatively, there were no qualitative effects of these functions (i.e., the patterns were the same), and thus, I only present results in which linear trade-offs were used. Invasion of Parental Care Effects of egg maturation rate, egg death rate, adult reproductive rate, and carrying capacity: A mutant with parental care invaded or coexisted with a resident population lacking care over a wide range of life-hi story parameters (Figure 6-2 A) particularly when care was effective at decreasing the death rate of eggs (i.e., when < ), when it increased survivorship through the j uvenile stage (i.e., when EmdEdJm > J ), and when caring increased maturation rate of the eggs (i.e, when > Figure 6-2 A) relative to the non-caring strategy. Similarly, the range over which care invaded or coexis ted with no care increased when parental care was associated with an increased ra te of egg fertilization (e.g., if it was a preferred trait, such that rm > r; Figure 6-2 A versus 6-2 B) and when care was associated with a decreased carrying capacity relative to th e resident population (Figure 6-2 A versus 6-2 C). The evolution of parental care was relatively insensitive to changes in carrying capacity for cases in which the resident and mutant had equal carrying capacities. EmmEm 108

PAGE 109

Effect of cannibalism on the evolution of care: To evaluate whether the ability to practice filial cannibalism affect s the evolution of parental care, I compared the case in which a mutant with only parental care was allowed to in vade a resident population with no parental care and no cannibalism, with the scenario in which the mutant could care and cannibalize (Figure 6-2 A versus 6-2 D). Indeed, filial cannibalism facilitat ed the evolution of care (Figure 6-2 D). When the caring mutant was allowed to cannibalize (Figure 6-2 D), parental care (and filial cannibalism) evolved over a wider range of pa rameter space and coexiste d more often with no care than when the mutant was not allowed to cannibalize (Figure 6-2 A). Invasion of Filial Cannibalism (W ith and Without Parental Care) Effects of Egg Maturation Rate, Reproductive Rate, and Selective Cannibalism: Parental care with filial cannibalism was more like ly to invade and/or coex ist with a state of only care when practicing filial cannibalism increased the maturation rate of eggs over what would be achieved by only providing care (i.e., when > Figure 6-3 A), and when filial cannibalism allowed a parent to improve the qua lity of care provided for remaining offspring (i.e., cannibalism decreases relative to ). Care and cannibalism invaded and coexisted more often when filial cannibalism increased th e reproductive rate of the caring parent(s) (e.g., care and/or cannibalism are preferre d, such that rm > r, Figure 6-3 A versus B), but invaded less often if it decreased the reproductive rate of adults (e.g., non-canni balism is preferred, such that rm < r; Figure 6-3 A versus C). Similarly, care with cannibalism evolved more often when parents could selectively canni balize their offspring. Specifi cally, if parents cannibalized offspring with a higher egg death rate and a lo wer juvenile survival rate (Figure 6-3 D), cannibalism invaded more often than in cases in which parents were not capable of selectively cannibalizing (Figure 6-3 A). Thes e patterns were consistent when we considered parental care EmmEdEmEmd 109

PAGE 110

and filial cannibalism evolving from a state of care or a state of no care. Likewise, filial cannibalism (without care) invade d and/or coexisted with no care/no cannibalism over a greater range of parameter space if filial cannibalism improved offspring survival (i.e., < or EmdEdJm > J ), increased the maturation rate of eggs ( > ), or when parents were able to practice selective filial cannibalism of offspring that had reduced survival during the egg and/or juvenile stage. EmmEmEffects of Density-Depe ndent Egg Survivorship When egg survivorship was density-dependent parental care and/or filial cannibalism evolved and co-existed over a wi de range of parameter space. However, in the absence of any other benefits of filial cannibalism, density-dependent egg survivorship alone did not facilitate the evolution of cannibalism. In fact, allowing a mutant that provides care to cannibalize decreased the range over which car e and cannibalism evolved, in comparison to the case in which the mutant could not cannibalize (Figure 6-4 A versus B). Howeve r, when the mutant provided care and cannibalized, parental care and cannibalism invaded ove r a greater range of parameter space as the strength of density-dependence (i.e., ) increased (Figure 6-4 B versus C). In other words, the evoluti on of filial cannibalism was not f acilitated by density-dependent egg survivorship per se, but relatively intense density-dependence (i.e., a relatively large increase in egg mortality with increasing egg density) allowed cannibalism to evolve more often in comparison to weaker density-dependence (i.e., a relatively small increase in egg mortality with increasing egg density; as in eqn. 6, increased, the ra nge over which care with cannibalism and no care could invade and/or coexist increased) These patterns were the same for both density-dependent functions considered. 110

PAGE 111

Effects of Energetic Benefits of Consuming Offspring Energetic benefits of filial canniba lism (i.e., benefits associated with and rm Adm) increased the range over which care and cannibali sm could invade and coexist with no care (Figure 6-5 A versus B). However, in this sc enario (i.e., considering care and cannibalism invading from a state of no care or cannibalism) it is possible th at cannibalism could be thought of as simply hitchhiking in with care. Thus, I considered the case in which cannibalism and care invaded a resident who already provides care. While energetic be nefits of cannibalism to adult reproduction and survival increased the range of invasion and coexistence (Figure 6-5 C), filial cannibalism (which in this case is equivalent to simply killing offspring, or abandoning offspring that have no chance of surviving alone, during the course of care) was still able to invade in the absence of benefits to adults (Figure 6-5 D) over the range of parameters considered. Effects of Carrying Capacity For cases in which cannibalism alters the efficiency with which individuals use resources (and hence, population carrying capacity), filial canni balism with parental care was more likely to evolve from a state of only care if canniba lism increased the resource-use efficiency of a population of individuals exhibiti ng that strategy (i.e., the carryi ng capacity). In other words, filial cannibalism was more likely to invade if it somehow increased the productivity of the system (Figure 6-6 A versus B). In contrast, for the case of filial cannibalism with parental care evolving from a state of no care, cannibalism and care were more likely to evolve if they were associated with a decrease in the population carrying capacity (i.e., if care with cannibalism decreased the productivity of the system and the efficiency with which individuals use resources, Figure 6-6 C versus D). Likewise, for the case of filial cannibalism (without parental care) 111

PAGE 112

invading a resident state of no care/no cannibalis m, filial cannibalism invaded and coexisted over a greater range of parameter space if it increased carrying capacity (i .e., resource-use efficiency). If carrying capacities were equal for the resi dent providing pa rental care and the mutant providing care and practicing f ilial cannibalism (i.e., the efficiency of resource use was equivalent for individuals exhibiting mutant and resident strategies), can nibalism invaded and/or coexisted more often when carry ing capacity was relatively low (i.e., in relatively unproductive systems, Figure 6-6 A versus E). This trend (i.e., more invasion and coexistence of the mutant at lower carrying capacities) was consistent for the case in which filial cannibalism (with no parental care) invaded a resident with no care/no cannibalism. However, when I considered the case in which cannibalism with care invaded a state of no care (and assumed the carrying capacities were equal for populations exhibiting the resident and mu tant strategies), cannibalism and care were more likely to invade or coexist with no care/no cannibalism when carrying capacity was relatively large (i.e., when the syst em was relatively more productive, Figure 6 C versus F). Discussion I have shown that parental care, fili al cannibalism, and no care/total offspring abandonment can evolve over a wide range of life-history parameters My results suggest that the ability to abandon or consume offspr ing during the course of parental care can actually facilitate the evolution of parental care, and that offspr ing abandonment/no care, pa rental care, and filial cannibalism often have the potential to coexist. Ev en in the absence of direct benefits of filial cannibalism, such as energetic gain or increas ed survival of remain ing offspring, filial cannibalism invaded (and coexisted with) non-canniba listic strategies in multiple contexts (i.e., with or without care, across varyi ng resident strategies, over a range of life-history parameters). In the absence of such benefits, cannibalism is simply equivalent to killing (or abandoning 112

PAGE 113

offspring that will subsequently die) during the course of care. My results suggest that the evolutionary dynamics of filial cannibalism are like ly comparable to those of simple offspring abandonment (which provides no immediate benef its, such as energetic gain, to parents). However, my results suggest that the evolution and fixation of filial cannibalism is favored by a variety of evolutionary and ecological factors. While no single benefit of consuming eggs was essential for the invasion of filial cannibalism to occur, several potential benefits facilitate the evolution of filial cannibalism. In particular, my model highlights the plau sibility of several non-mutually exclusive alternative hypotheses favoring th e evolution filial cannibalism (Table 6-2). The ability to selectively cannibalize eggs facilitated the evolution of cannibalism in all contexts. The ability to cannibalize offspring selectively allows parents to alter the ph enotypes of the offspring they produce after fertilizatio n and on a relatively fine time scale, which might be particularly beneficial in a variable environment (alt hough I didnt explicitly consider environmental variability in this model). Sele ctive cannibalism of clutches of lower reproductive value has been demonstrated in relation to uncer tainty of paternity (i.e., possible cuckolding events) in some fishes ( Lepomis macrochirus, Neff 2003; Telmatherina sarasinorum Gray et al. 2007; Gasterosteus aculeatus, Frommen et al. 2007), and the elimin ation of low quality offspring has been focused on in other contexts (e.g., allowi ng lower quality offspring to be eliminated by siblicide, Stearns 1987; spontaneous and selectiv e abortion in humans, sex ratio adjustment in red deer; Stearns 1987; Kozlowski and Stearns 1 989). However, selective filial cannibalism of viable offspring in relation to other aspects of offspring qualit y has received little empirical attention (but see Chapters 4 and 5). In partic ular, I hypothesize that in some contexts filial 113

PAGE 114

cannibalism of offspring with (1) reduced expected future survival, or (2) slower maturation rates during the period in which care is being provided can be an adaptive strategy. Alternatively, it is possible that filial canni balism itself increases the development rate of eggs. If filial cannibalism incr eases the maturation rate of eggs relative to those of noncannibalistic parents, filial cannibalism evolves over a greater ra nge of parameter space (Figure 6-3 A). Indeed, for cases in which parent-offspr ing conflict exists over the optimal duration of parental care, filial cannibalism might be a way in which parents speed-up the developmental rate of their eggs, thereby allowing them to re duce per offspring costs of care or re-enter the mating pool faster. According to this hypothesis, caring parent s potentially benefit by providing care for a shorter duration of time if filial cannibalism creates an environment in which offspring are eager to escape the egg stage (e.g., because of increased risk of death; see also work on nonparent predators increasing egg development ra te, e.g., Warkentin 2000). To my knowledge, this idea of filial cannibalism speeding-up egg development has not previously been considered, and as mentioned previously, is likely to be relevant for cases in whic h parents and offspring differ in the optimal amount of care they provide/receive. Incorporating an energetic benefit of cannibalism facilitated the invasion of filial cannibalism. This finding is consistent with pr evious theoretical and em pirical work suggesting that energetic need affects filial cannibalism (e.g., Rohwer 1978; Sargent 1992; Kraak 1996; reviewed by Manica 2002). However, some empirica l work suggests that the effects of energetic need on filial cannibalism are not always straight forward-in some species cannibalism increases as parental energetic need incr eases (e.g., Thomas and Manica 2003) whereas in other species an opposite pattern is observed (e.g., Klug et al. 2006). Moreover, in other systems, there appear to be no effects of parental condition on filial cannibalism under some conditions (e.g., Lindstrm 114

PAGE 115

and Sargent 1997), and in other species the relatio nship between energetic need and cannibalism differs in varying contexts (Klug and Lindstr m, unpublished data). Furthermore, some have suggested that the energetic bene fits of cannibalism are not suffi cient to explain the prevalence of filial cannibalism (Smith 1992). In my model, filial cannibalism invaded over a range of parameter space even when we removed benefits of ca nnibalism, suggesting that substantial energetic benefit of cannibalism is not necessar ily essential for the evolution of cannibalism. That said, there is little doubt that filial cannibalism provide s a caring parent with energy and/or nutrients and such benefits are likely critical fo r adult survival and succ essful nest defense in systems where parents are unable to feed during the course of providing parental care (Manica 2002, 2004). Indeed, energetic benefits certainly favor the evolution of filial cannibalism (Figure 6-5; previous work by Rohwer 1978; Sarg ent 1992; reviewed in Manica 2002). Likewise, increasing the stre ngth of density-dependent e gg survivorship increased the parameter space over which filial cannibalism evolved. However, density-dependent egg survivorship alone did not facilitate the evolution of filial cannibalism. Indeed, it seems unlikely that density-dependent egg survivorship per se would lead to the evolution of filial cannibalism in the absence of other trade-offs associated with egg number. If animals can track their environment, they would simply be expected to adjust the number of eggs they produce according to expected egg survivorship (i.e., they should lay at densities that maximize survival). Further work is needed to evaluate the importa nce of density-dependent egg survivorship when other trade-offs are associated with the number of offspring produced or when the environment is variable. Spatial and tem poral variation in the environment has been hypothesized to influence patterns of cannibalism observed in nature (e.g., Payne et al. 2004) and non-cannibalistic brood 115

PAGE 116

reduction (e.g., Forbes and Mock 1998), but additi onal work is needed to understand more fully the importance of such stocha sticity at varying scales. Sexual selection via mate choice and/or se xual conflict also aff ected the invasion and fixation of filial cannibalism and/or parental care. My model suggests that the evolution and fixation of parental care from a state of no care can be facili tated by differential reproductive success if parental care or f ilial cannibalism increases the reproductive rate of individuals exhibiting care or cannibalism (e.g., if parental care or cannibalism is preferred during mate choice). This finding is consistent with some previous work. For example, Pampoulie et al. (2004) and Lindstrm et al. (2006) recently demonstrated mating pr eferences for parental care, suggesting a potentially larger ro le for sexual selection in the e volution of care than previously thought. Additionally, filial cannibalism is possibl y favored by sexual selection if cannibalism directly benefits a choosing mate or when it make s a caring parent more attractive in some other way (Sikkel 1994; Lindstrm 2000). Likewise, if a mating prefer ence exists for non-cannibals, the parameter space over which filial cannibalism evolves decreases. Intere stingly, the role of sexual conflict has received relatively little theore tical or empirical attent ion previously (but see Kraak and van den Berghe 1992; Kraak 1996; Lindstrm 2000). In fishes, where filial cannibalism is typically practiced by caring fathers, the focus of almost all work has been on costs and benefits of cannibalism to caring ma les. One must also wonder if benefits to noncannibalistic females exist, and if such benefits are absent, why do females tolerate filial cannibalism? Additionally, sexual co nflict is also likely to exist when both parents practice filial cannibalism, but this idea has received no attention. More empirical work is needed to better understand costs and benefits of filial cannibalis m to a parent whos mate practice filial cannibalism. 116

PAGE 117

Finally, population-level resource competition li kely plays a role in the evolution of both parental care and filial cannibalism. When care a nd/or cannibalism affected the efficiency with which individuals exhibiting a give n strategy use resources, parent al care was more likely to evolve if caring was associated with a reduc tion in the carrying capacity (e.g., when caring decreased the efficiency with which individuals use resources), whereas, filial cannibalism was more likely to invade if it increased carryi ng capacity (e.g., if cannibalism increased the resource-use efficiency of individuals). Additionally, the evolution of filial cannibalism (with or without parental care) was affected by the popul ation carrying capacity, even for the case in which the carrying capacity of the mutant and resi dents were equal. It is unclear how parental care and filial cannibalism pot entially alter population-level d ynamics and resulting carrying capacities in nature, but this idea warrants further attention. For example, it is possible that the ability to cannibalize increases re source availability to caring pa rents, thereby freeing-up other resources and increasing the productivity of a system. Regard less, understanding the ecological dynamics of a system (i.e., intensity of res ource competition and population growth parameters such as carrying capacity) is like ly to be critical for understanding the evolution of parental care and filial cannibalism across animal taxa. While previous work has sometimes incorporated population-level growth dynamics in parental care theory (e.g., McNamara et al. 2000), this is not a common approach. In summary, my results suggest that parent al care and filial canni balism can evolve over a range of life-history patterns and ecological co nditions, and that multiple strategies often have the potential to coexist. Coexistence, while not well-studied (but see Webb et al. 1999), is prevalent in nature (e.g., maternalor patern al-only care in many taxa, reviewed in CluttonBrock 1991; care and no-care with total offspr ing abandonment following egg fertilization: 117

PAGE 118

118 Jordanella floridae Hale pers. comm., the white stickl eback Gasterosteidae spp., Blouw 1996; care and care followed by abandonment, Hypoptychus dybowskii Narimatsu and Munehara 2001). Likewise, there are many cases in which caring parents never or rarely consume or abandon their offspring. Even in fishes, where care with filial cannibalism has been welldocumented, there are still many species exhibiti ng parental care in which filial cannibalism is absent (e.g., Micropterus dolomieui Gillooly and Baylis 1999). For species e xhibiting filial cannibalism, there is a great de al of variation in the patter ns of cannibalism observed among species and within and between individuals (e.g ., how many eggs are consumed, who practices cannibalism and when; Petersen and Marchetti 1989; Okuda a nd Yanagisawa 1996; Lindstrm and Sargent 1997; Lissker et al. 2002; Klug et al. 2005; Klug and St. Mary 2005). Understanding such withinand between-species va riation in filial cannibalism and parental care will require more detailed theoretical and empiri cal work that simultaneously considers multiple factors (such as variation in o ffspring quality, energetic needs of parents, mating preferences and sexual conflict, general resource competition). Ad ditionally, it will also be important to assess the importance of environmental heterogeneity in the evolution of filial cannibalism. From this study, my approach and results provides a nove l basis for further deve loping this theme of whether to care for or consume ones own offspring.

PAGE 119

Table 6-1. Trade-off fu nctions. The following trade-off functions were used to reflect the unique lif e histories of individual s who provide parental care and/or practice filial cannibalism. The death rate of eggs is assumed to be a function of the parental care provided (i.e., as de decreases, ca re is presumed to increase), and thus egg death rate is the proxy for care. Strategy Parameter Trade-offs No parental care & no filial cannibalism Parental care only Filial cannibalism only Parental care & filial cannibalism Reproductive rate; r and rm 1) Reproductive rate decreases as caring increases (i.e., r or rm decreases as de or dem decreases). 2) Reproductive rate increases as maturation rate of eggs increases (for a carer only) (i.e., r or rm increases as mE or mEm increases). 3) Reproductive rate increases as cannibalism increases (i.e., rm increases as increases). Linear & Non-linear: 0rr Linear: ) 1(0m EEm mmmdrr Non-linear: ) 1( )(0 EmEm Em m E mmmd md rr Linear: )1(0 mmrr Non-linear: )1(0 mmrr Linear: ) 1(0 Em Em mmmdrrNon-linear: ) 1( ) (0 m EEm Em m E mmmd md rr Juvenile survival rate; j and jm119 1) Juvenile survival rate increases as care increases (i.e., j or jm increases as dE or dEm decreases). Linear & Non-linear: 0JJ Linear: )1(0Em m JJmd Non-linear: m E Em Jm m Jd d )1(0 Linear & Non-linear: 0Jm Jm Linear: )1(0Em m JJmd Non-linear: m E Em Jm m Jd d )1(0 Adult death rate; dA and dAm 1) Adult death rate increases as caring increases (i.e., dA or dAm increases as dE or dEm decreases). 2) Adult death rate decreases as cannibalism increases (i.e., dA or dAm decreases as increases). Linear & Non-linear: 0AAdd Linear: )1(0Em m A Amddd Non-linear: Em m E m A m Ad d dd )1(0 Linear: )1(0 m A m Add Non-linear: )1(0 Am m Add Linear: )1(0 Em m A m Addd Non-linear: Em Em Am m Ad d dd )1(0

PAGE 120

120Table 6-2. Alternative hypotheses regarding the evolutionary si gnificance of filial cannibalism (FC). Here, I present several, nonmutually exclusive hypotheses and briefly describe the findings of our model and those of some previous work in relation to these hypotheses. Hypothesis Description Model findings Related previous findings 1. Selective Filial Cannibalism Offspring with particular characteristics (e.g., reduced survival, decreased maturation rate) are preferentially consumed. Evolution of FC facilitated by selective cannibalism of offspring with lower maturation rates, lower egg survival, and/or lower juvenile survival. FC affected by certainty of paternity in some systems (Neff 2003; Frommen et al. 2007; Gray et al. 2007) but not in others (Svensson et al. 1998); effect of other aspects of offspring quality on FC largely unknown. 2. Filial cannibalism speeds-up egg development By increasing costs associated with remaining in the egg stage, filial cannibalism increases maturation rate of eggs (i.e. FC decreases the time it takes for eggs to develop). Evolution of FC more likely if cannibalism increases egg maturation rate. Not previously examined; potentially relevant for systems in which parent-offspring conflict exists over the optimal amount of care provided/received. 3. Energy-Based Filial Cannibalism FC provides energy that offsets costs of care and is re-invested into current and/or future reproduction. Energetic benefit of eggs facilitated evolution of FC. Substantial energetic be nefit of FC and/or effect of energetic need on FC found in several systems (reviewed in Manica 2002). 4. DensityDependent-EggSurvivorshipMediated Filial Cannibalism Density-dependent egg survival mediates FC: by consuming some young, parents increase survival of remaining offspring. Density-dependent egg survival alone did not facilitate the evolution of FC; more intense density-dependence facilitated evolution of FC in comparison to weaker densitydependence. FC is affected by density-dependent egg survivorship in two species (Payne et al. 2002, 2004; Klug et al. 2006) 5. Mate ChoiceMediated Filial Cannibalism FC is preferred in mate choice, thereby increasing relative reproductive rate. If FC increases relative reproductive rate of cannibals, FC evolves more often. FC increases nest attractiveness (and consequently eggs received) in some cases (Sikkel 1994). 6. Sexual ConflictMediated Filial Cannibalism FC is a non-preferred trait and decreases relative reproductive rate. If FC decreases reproductive rate, FC evolves less often. Sexual conflict can inhibit FC in some cases (Lindstrm 2002); sexua l conflict regarding FC not well-studied empirically (but see Kraak 1996). 7. Filial Cannibalism Driven by Resource Competition FC is driven by population-level resource competition among adults. Evolution of FC sensitive to population-level carrying capacity. Mate availability (K ondoh and Okuda 2002) and other resource competition (Creighton 2005) affects FC in some cases; effects of general resource availability on FC not wellknown.

PAGE 121

EGGS E JUVENILES ADULTS A K tA r )( 1 Em JEm Ed Ad Figure 6-1. The model: individuals develop through an egg and j uvenile stage and reproduce as adults. Eggs either die (at rate are consumed by their parent(s) (at rate)Ed ), or mature into juveniles (at rate ). Individuals survive a nd pass through the juvenile stage (at rate EmJEm ), where represents the time spent in the juvenile stage. As adults, individuals either die (at rate ) or reproduce (at rate Ad K tA r )( 1 where K represents the population carrying capacity). Boxes repr esent life-history stages; solid arrows represent death, reproduction, and maturation; the dash ed line represents consumption of eggs by adults. 121

PAGE 122

Figure 6-2. Invasion of parental care. Parental care invades and/ or coexists with no care more often when parental care: (A) increases th e maturation rate of eggs, (B) increases parental reproductive rate ( r = 1.0, r m = 1.2), (C) is associated with a decreased carrying capacity ( K = 20, Km = 15), and (D) when the caring mutant is able to cannibalize (i.e., = 0.01 for the mutant, = 0 for the residents). Lines represent invasion boundaries for the mutant (solid line ) and the resident (dotted line). Invasion boundaries are shown for the maturation rate of the eggs. The mutant invades the resident in the regions labele d care (A-C) or care & cannibalism (D), the resident invades the mutant in the region labeled no care (A-C) or no care or cannibalism (D), and both strategies coex ist in the region labeled c oexistence. Neither strategy will persist (i.e., they go extinct) in th e region labeled NP. The region labeled NP/IC is a region in which neither strate gy will persist, or where the outcome is dependent upon initial conditions of the model. Unless noted above, r = rm = 1.0, dE = dEm = 0.9, dA = dAm = 0.5, J = Jm = 0.5, K = Km = 20, = 0, = 0.1. 122

PAGE 123

Figure 6-3. Invasion of filial cannibalism. A muta nt that provides parental care and practices filial cannibalism invades and coexists with a resident that only provides care (A) more often when cannibalism increases the maturation rate of eggs, (B) more often when cannibalism increases the parents reproductive rate ( r = 0.5, r m = 0.6), (C) less often when cannibalism decreases the parents reproductive rate ( r = 0.6, r m = 0.5), and (D) more often when parents are able to selectively cannibalize offspring with reduced future survival ( de = 0.2, j = 0.9, dem = 0.1, jm = 0.95). Lines represent invasion boundaries for the mutant (solid line ) and the resident (dotted line). Invasion boundaries are shown for the ma turation rate of the eggs, mE and mEm, and unless otherwise noted, r = rm = 0.5, dE = dEm = 0.2, dA = dAm = 0.5, J = Jm = 0.9, K = Km =20, = 0.015, = 0.1. The mutant invades the resident in the region labeled care & cannibalism, the resident invades the muta nt in the region labeled care only, and both strategies coexist in the region labeled coexistence. Neither strategy will persist in the region labeled NP. The region labeled NP/IC is a region in which neither strategy will persist, or where the outcome is dependent upon init ial conditions of the model. 123

PAGE 124

Figure 6-4. Effect of density-dep endent egg survivorship on the evolution of parental care and filial cannibalism. Parental care and no care invade and/or coexist over a large range of parameter space when (A) the rare mutant does not cannibalize ( = 0). The range over which parental care invades decreases when (B) the rare mutant cannibalizes ( = 0.01). However, increasing the strength of density-dependence ( eqn. 6) increases the range over which care and cannibalism invades-care with cannibalism invades more often when (C) the strength of the density-dependence is greater ( = 0.9), in comparison to B) the case in which it is relatively weak ( = 0.6). Lines represent invasion boundaries for the mutant (solid line) and the resident (dotted line). Invasions boundaries are shown for the maturation rate of the eggs, mE and mEm, and unless otherwise noted, r = rm = 3 dE = 0.9, dEm = 0.3, dA = dAm = 0.5, J = Jm = 0.5, K = Km =20, = 0.01, = 1, = 0.6. The mutant invades the resident in the region labeled care (A) or care and cannibalism (B-C), the resident invades the mutant in the region labeled no care (A) or no care or cannibalism (B-C), and both strategies coexist in the region labeled c oexistence. The region labeled NP/IC is a region in which neither strate gy will persist, or where th e outcome is dependent upon initial conditions of the model. 124

PAGE 125

Figure 6-5. Effect of ener getic benefits on the evolution of fili al cannibalism. Parental care with filial cannibalism is more likely to invade and coexist with no care if filial cannibalism is (A) beneficial to a parents survival and reproduction versus (B) the case where there are no benefits of cannibalism. Likewise, care with cannibalism is more likely to invade a state of only care when (C) adult survival and reproductive benefits of egg eating exist ve rsus (D) the case where such benefits are absent. Lines represent invasion boundaries fo r the mutant (solid line) a nd the resident (dotted line). Invasions boundaries are shown for the maturation rate of the eggs, mE and mEm. Unless otherwise noted, r = rm = 1.0, dE = dEm = 0.9, dA = dAm = 0.5, J = Jm = 0.5, K = Km =20, = 0.01, = 0.1 for A and B, and r = rm = 0.5, dE = dEm = 0.2, dA = dAm = 0.5, J = Jm = 0.9, K = Km =20, = 0.015, = 0.1 for C and D. The mutant invades the resident in the region labeled care & cannibalism, the resident invades the mutant in the region labeled no care or can nibalism (A-B) or care only (C-D), and both strategies coexist in the region labeled coexistence. Neither strategy will persist in the region labeled NP. The region labeled NP/IC is a region in which neither strategy will persist, or where the outcome is dependent upon init ial conditions of the model. 125

PAGE 126

126 Figure 6-6. Effect of carrying capacity on the evol ution of filial cannibali sm. Parental care with filial cannibalism invades and coexists w ith care over a range of parameter space when (A) cannibalism does not affect carrying capacity ( K = Km = 20). However, cannibalism invades over a gr eater range of parameter space when (B) cannibalism increases carrying capacity ( K = 10, Km = 20). Likewise, care with cannibalism invades no care over a range of paramete r space when (C) care and cannibalism do not affect carrying capacity (K = Km = 20), but it invades or co exists more often when (D) care and cannibalism decr ease carrying capacity (K = 20, Km = 10). For cases in which the resident and mutant have equa l carrying capacities, parental care and cannibalism are more likely to invade when (E) carrying capacity is relatively small ( K = Km = 10) versus the case in which it is relativ ely large (A). In contrast, care with cannibalism is more likely to invade no care/no cannibalism when (F) carrying capacity is relatively large ( K = Km = 50) versus (C) the case in which it is relatively small. Lines represent invasi on boundaries for the mutant (s olid line) and the resident (dotted line). Invasions boundaries are shown for the maturation rate of the eggs, mE and mEm. Unless otherwise noted, r = rm = 0.5, dE = dEm = 0.2, dA = dAm = 0.5, ,J = Jm = 0.9, = 0.01, = 0.1 for A, B and E, and r = rm = 1.0, dE = dEm = 0.9, dA = dAm = 0.5, ,J = Jm = 0.5, = 0.01, = 0.1 for C,D, and F. The mutant invades the resident in the region labeled care & cannibalism, the resident invades the mutant in the region labeled care only (A-B, E) or no care or can nibalism (C-D, F), and both strategies coexist in the region labeled coexistence. Neither strategy will persist in the region labeled NP. The region labeled NP/IC is a region in which neither strategy will persist, or where the outcome is dependent upon init ial conditions of the model.

PAGE 127

CHAPTER 7 GENERAL CONCLUSIONS AND SYNTHESIS Introduction Parental care typically increas es offspring survival and/or quality, thereby increasing parental fitness. Thus, it is surprising that filial cannibalism, the c onsumption of ones own offspring, is prevalent in fish es exhibiting paternal care. Indeed, its difficult to imagine many situations in which regularly consuming ones own young is an adaptive strategy. Because parental males often consume more eggs than die naturally (Manica 2002 ; Klug et al. 2006 and Chapter 6), filial cannibalism does not solely serv e to clean the nest of dead eggs. Currently, filial cannibalism is thought to represent an adaptive trade-off (Manica 2002). Prior to my dissertation work, there were two general hypotheses explaining the adaptive significance of filial cannibalism: the energybased hypothesis (Rohwer 1978; Sargent 1992) and the oxygen-mediated hypothesis (Payne et al 2004, 2004). The energy-based hypothesis suggests that filial cannibalism is an adaptive st rategy in which males gain energy or nutrients from eggs that are then reinvested into current or future reproduction, thereby increasing net reproductive success (Rohwer 1978). Acco rding to this hypothesis, fili al cannibalism is expected to increase when food availability is low and/or when the caring parents condition is poor. There has been mixed support for the energy-based hypot hesis (Belles-Isles & Fi tzgerald 1991; Smith 1992; Lindstrm and Sargent 1997), a nd at best, it can only explai n filial cannibalism in some systems (e.g., Manica 2004). The oxygen-mediated hypothesis of filial cannibalism (Payne et al. 2002) suggests that filial cannibalism is an adapti ve strategy in which part ial clutch cannibalism improves the survival of remaining eggs by in creasing oxygen availability to remaining eggs. Specifically, Payne et al. (2002) suggested that caring males poten tially improve overall clutch 127

PAGE 128

survivorship by consuming some of their eggs. The oxygen-mediated hypothesis has received support in one species (Payne et al. 2002), but has not been generally evaluated. In addition to energy and oxygen availability, some studies suggest that mate choice or sexual conflict (Sikkel 1994; Kraak 1996; Lindstrm 2000), egg age (Salfert and Moodie 1985; Sikkel 1994), and certainty of paternity (Neff 2003; Gray et al. 2007; Frommen et al. 2007) affects the occurrence of filia l cannibalism (but see Svenss on et al. 1997 and Svensson and Kvarnemo 2007, who find that certainty of patern ity does not affect filial cannibalism in the sand goby). There has been relatively li ttle theoretical or empirical examination of such factors, and thus the general importance of mate choice, sexual conflict, egg age, and certainty of paternity remains unknown (e.g., Takeyama et al. 2007). Because of (1) the lack of support for any pa rticular hypothesis (i.e., the energy-based or oxygen-mediated hypothesis) and (2) a general lack of alternative hypotheses, the evolutionary significance of filial cannibalism in fishes remains unclear. In Chapter 1, I argued that an enhanced understanding of the e volutionary significance of filial cannibalism necessitates three approaches: 1) a re-evalu ation of current theory by explicitly focusing on fitness consequences of filial cannibalism; 2) the development and examination of alternative hypotheses of filial cannibalism; and 3) the development and evaluation of a synthetic model of filial cannibalism that simultaneously considers the potential importance of a range of factors. In Chapters 2 and 3, I evaluated predictions of the energy-based and the oxygen-mediated hypotheses in two species, the flagfish ( Jordanella floridae ) and the sand goby ( Pomatoschistus minutus ). In Chapters 4 and 5, I developed the novel hypothesis of selectiv e filial cannibalism, and I evaluated this hypothesis in the flagfish and the sand goby. In Chapter 6, I devel oped a model of filial cannibalism. Using this model, I evaluated the plausibility of a ra nge of alternative hypotheses of 128

PAGE 129

filial cannibalism, and I concluded that a variety of factors can favor the evolution of filial cannibalism. In this final discussion, I will synthesize th e findings of Chapters 2-5, and discuss my findings in terms of previ ous and novel hypotheses. Are the Current Energy-Based and Oxyge n-Mediated Hypotheses Sufficient? Until 2002, the energy-based hypothesis was the only adaptive hypothesis of filial cannibalism and it remains the most widely accepted hypothesis of filial cannibalism (reviewed in Manica 2002). However, as mentioned before evidence regarding this hypothesis has been mixed (Belles-Isles & Fitz gerald 1991; Smith 1992; Lindstrm and Sargent 1997). In Chapter 2, I described an experiment in which I experimentally manipulated 1) the ability of parental males to cannibalize eggs (i.e ., males either had full access to eggs, or filial cannibalism was prevented by a nest cover) and 2) diet (i.e., high qua lity versus low quality diet) to evaluate the effect of f ilial cannibalism and diet on compon ents of reproductive success in male flagfish. According to the energy-based hy pothesis, energy gained from eggs should be translated into increased future reproduction, and I therefore predicted that males that were able to practice filial cannibalism would receive more eggs and spawn more frequently during the 90 days of the experiment. Contrary to these pred ictions, I found that fili al cannibalism did not increase the total number of eggs males received or the frequency of spawning. Indeed, filial cannibalism was always associated with a decrea se in the total number of eggs received, suggesting that energy or nutrients attained from eggs is not directly translated into future reproduction in the flagfish. Furthermore, the energy-based hypothesis su ggests that filial cannibalism should increase when food availability is low. In contrast to this prediction, I found that filial cannibalism decrea sed when food availability was low. Specifically, males on the low 129

PAGE 130

quality diet consumed fewer of their eggs th an males on the high quality diet. Thus, I found no support for the energy-based hypothesis in the flagfish. Similarly, I examined the relationship be tween parental male condition and filial cannibalism in the sand goby (Chapter 3). The energy-based hypothesis predicts that males will consume more eggs when parental condition is re latively poor. However, I found that males in poorer condition consumed a smalle r proportion (and fewer) of their e ggs than males that were in better condition. This finding is directly in contrast to predictions of the energy-based hypothesis. Thus, I found no support for the energybased hypothesis of filial cannibalism in either the flagfish or the sand goby. Because of my findings (Chapters 2 and 3) and those of other studies (Belles-Isles a nd Fitzgerald 1991; Smith 1992; Lindstrm and Sargent 1997), I conclude that the energy-based hypothesis is not sufficient to explain the prevalence of filial cannibalism. While eggs certainly provide some energy or/and nutrients, energetic benefits of cannibalism cannot explain filial cannibalism in an adaptive context. Likewise, I did not find suppor t for the oxygen-mediated hypothesis of filial cannibalism (Chapter 3). Specifically, the oxyge n-mediated hypothesis of filial cannibalism predicts that 1) filial cannibalism of some eggs in a nest increases oxygen to the remaining eggs, thereby increasing total egg survival, 2) cannibali sm will increase as oxygen decreases, and 3) cannibalism will decrease as egg density increases. In the sand goby, I found that filial cannibalism increased as oxygen decreased and as egg density increased. While both findings are consistent with the oxygen-mediated hypothesis, it is possible that males increased cannibalism at lower oxygen levels because the costs of providing care in low oxygen environments are greater than the costs in high oxyge n environments. I therefore dire ctly evaluated the effect of oxygen level and egg density by exposing eggs to tw o levels of simulated filial cannibalism (i.e., 130

PAGE 131

simulated cannibalism or no simulated cannibalis m) and two levels of oxygen availability (i.e., high versus low oxygen). Eggs were reared in the absence of male s and I quantified egg survival. Indeed, I found that egg survival was density-d ependent, but this density-dependence was not mediated by oxygen. Specifically, there was no e ffect of oxygen on egg survival in this experiment. Thus, I propose a more general hypothe sis of filial cannibalism mediated by densitydependent egg survival. I suggest that density-dependent egg survival might be due to a range of factors (e.g., waste accumulation in the nest or disease transmission), and that the factors affecting egg survival likely vary across species and environmental conditions. In summary, current hypotheses of filial cannibalism (i.e., the energy-based and oxygenmediated hypotheses) are inadequate for generally explaining the adaptive significance of filial cannibalism in fish species. While energetic need and oxygen might be important in some contexts and in some species, neither energy nor oxygen alone can explai n the prevalence of filial cannibalism in fishes. Thus, my dissertation work aimed at re-evaluating current hypotheses (Chapters 2 and 3) further supports the need for 1) the development of a lternative hypotheses and 2) increased theoretical examination of the importa nce of a range of factor s in the evolution of filial cannibalism. An Alternative Hypothesis: Selective Filial Cannibalism Parental care is costly and leads to redu ced future reproduction (reviewed in Smith and Wootton 1995). Therefore, males should not waste en ergy caring for low quality eggs if the cost to the male (i.e., reduced future reproduction) outweighs the current benefit in offspring produced. Specifically, I hypothesize that males s hould preferentially cannibalize offspring of reduced quality (i.e., offspring that have reduced expected future survival or reproductive success) when there is some energetic benefit of consuming eggs or when offspring survival is density-dependent. The elimination of lower quality offspring has been demonstrated in relation 131

PAGE 132

to selective embryo abortion in humans and pl ants (Forbes 1997; Diamond 1987; Burd 1998; Karkkainen et al. 1999), brood reduction (Mock and Forbes 1995; Forbes and Mock 1998), and parents allowing or encouraging siblicide of low quality offspr ing (Stearns 1987), but this idea has not been considered in relation to filial ca nnibalism. Indeed, the elimination of low quality offspring is thought to play a cen tral role in explaini ng the evolutionary significance of offspring abandonment and brood reduction (e.g. Stearns 1987, 1992; Forbes and Mock 1998), and thus I hypothesize that the ability to canni balize offspring selectively might be an important factor in explaining the adaptive significance of filial cannibalism. I evaluated the hypothesis of selective filia l cannibalism in the sand goby (Chapter 4) and the flagfish (Chapter 5). In the sand goby, I examined within-clutch patterns of cannibalism when males received eggs from either one or two females. I focused on the relationship between filial cannibalism and egg size, which has been correlated with post-hatching survival in a range of fishes (reviewed in Kamler 2005). In the singl e-female scenario, I found that males exhibited no preferences with regard to egg size. In the multiple-female scenario, males preferentially consumed the larger eggs of the second female but they exhibited no size preferences for the eggs of the first female they spawned with. To evaluate further patterns of egg survival and hatching, I reared subsets of eggs in the absence of males. For the clutches reared in the absence of males, there was no relationship between egg si ze and survival, but larger eggs took longer to hatch than smaller eggs. Thus, the findings that 1) larger eggs take longe r to hatch and 2) males preferentially consume larger eggs of the s econd female whose eggs are already younger than those of the first female. This pattern suggests that males preferentially consume eggs in a manner that reduces the amount of time they spend caring for the current clut ch of eggs (Chapter 4). Specifically, my results (e.g., Figure 4-1) s uggest that males might be able to reduce the 132

PAGE 133

duration of time spent caring by seve ral days if they preferentially consume the largest eggs. In Chapter 4, I hypothesize that reducing the duration of time spent caring for a given brood might allow a male to re-enter the mating pool sooner. Specifically, if a male sand goby can reduce the per-clutch time he spends providing parental care it is possible that he can gain an additional brood cycle, which might in turn increase hi s net reproductive success. This hypothesis is supported further by the finding that whole clutch cannibalism tends to decrease as the breeding season progresses (Chapter 3). Later in the bree ding season, females become scarce and a males expected future reproduction decreases. Thus, I woul d expect benefits associ ated with decreasing the duration of parental care to decrease later in the breeding season. In the flagfish, I examined the relationship between filial cannibalism and mean egg energetic content and female size (Chapter 5). W hole clutch cannibalism increased as mean egg energetic content increased. In contrast, I found a ne gative relationship betw een partial clutch cannibalism and mean energetic content of eggs and maternal size. Egg energetic content and maternal size have been correlated with post-ha tching survival in fishes (reviewed in Kamler 2005), and thus, it appears that when males practice whole clutch cannibalism, they preferentially consume their higher quality offspring, which provide a relatively high energetic benefit. However, when males practice partial cl utch cannibalism, they pr eferentially cannibalize offspring that are likely to have lower future surviv al (Chapter 5). This find ing is consistent with other work suggesting that filial cannibalism increases when a brood has relatively low expected reproductive value. For example, whole clutch cannibalism increases when the initial number of eggs present is relatively small (reviewed in Manica 2002) and when males have been cuckolded (e.g., Frommen et al. 2007). 133

PAGE 134

My experiments on selective fi lial cannibalism in the sand goby and flagfish (Chapters 4 and 5) highlight the pote ntial importance of the hypothesis of se lective filial cannibalism. I have demonstrated that males preferentially consum e eggs based on aspects of phenotype in some casess. As mentioned previously, selective elimination of low quality offspring is hypothesized to play a large role in the evolution of selective abortion, brood reduction, and offspring abandonment. I therefore hypothesize that selec tive cannibalism might pl ay a large role in explaining the evolutionary significance of filial cannibalism, but additional theoretical and empirical work in other species is needed to ev aluate the relative importance of selective filial cannibalism. The Plausibility of Multiple Hypotheses In Chapters 2, 3, 4, and 5, I demonstrated that food availability, paternal condition, density-dependent egg survival, egg size, egg energetic content, and maternal size affects filial cannibalism in the flagfish and/or sand goby. To begin to evaluate th e relative importance of these (and other) factors in understanding the evol utionary significance of filial cannibalism, I developed and analyzed a general model of filia l cannibalism (Chapter 6). The results of this model suggest that no single benefit of filial can nibalism is essential for the evolution of filial cannibalism. Indeed, my model suggests that th e evolutionary dynamics of filial cannibalism appear to be similar to those of other forms of offspring aban donment (Chapter 6). Specifically, the ability to terminate parental care through filial ca nnibalism, infanticide, brood reduction, or abandonment can represent an adaptive strate gy under some conditions (i.e., under certain environmental conditions, for organisms with partic ular life history charac teristics) even when obvious benefits (i.e., energetic ga in or benefits to remaining o ffspring) are absent. That said, benefits associated with energetic gain, offspring survival, and mate preferences certainly exist in 134

PAGE 135

some species under some contexts (e.g., Mani ca 2004, Payne et al. 2002, Sikkel 1994) and can directly facilitate the evolution of filial cannibalism. Indeed, I found that the evolution of filial cannibalism was facilitated when (1) parents could selectively cannibalize lower quality offspring or offspri ng with slower egg development rates (Hypothesis 1, Table 6-2), (2) filial cannibalism increased egg maturation rate (Hypothesis 2, Table 6-2), (3) energetic bene fits of eggs existed (Hypothesis 3, Table 6-2), (4) cannibalism increased a parents reproductive rate (e.g., through mate attrac tiveness; Hypothesis 5, Table 63). Density-dependent egg survivorship alone did not favor the evolution of cannibalism (Hypothesis 4, Table 6-2). However, when e gg survival was density-dependent, filial cannibalism invaded more often when the density-dependence was relatively more intense. Additionally, sexual conflict potentially inhibits the evolution of filial cannibalism in some cases (Hypothesis 6, Table 6-2). I also hypothesize th at population-level resource competition can play a large role in the evolution of filial cannibalism (Hypothesis 7, Table 6-2) Indeed, in my model, the evolution of filial cannibalism was highly sensitive to population carrying capacity, and filial cannibalism was more likely to evolve when it allowed individuals to utilize resources more efficiently. In summary, my modeling work (Chapter 6) highlights the plausibility of several nonmutually exclusive alternative hypotheses. Additionally, I argue that attempting to explain the evolutionary significance of filial cannibalism with any single benefit (e.g., energetic need) is futile, and future work should consider th e importance of a range of factors. Future Directions More research is needed to understand the e volutionary significance of filial cannibalism. Below, I discuss six avenues of future research. 135

PAGE 136

Determining the Relative Importance of Varying Factors Future work should focus on determining the relative importance of energetic and nutritional benefits of eggs, de nsity-dependent egg su rvival, mate choice, sexual conflict, and egg quality and size in the evolution of filial cannibalism. Specifically, it will be important to assess the role of such factors in a range of organisms with diverse life histories and under varying conditions. Doing so will help characterize the selection pressures that shape patterns of filial cannibalism. Such an approach will also determine whether particular factors are more likely to affect filial cannibalism than others. Additionally, it will continue to be important to identify additional factors that affect filial cannibalism. Role of Environmental Variation Filial cannibalism raises a question that ha s been dealt with rarely: why do organisms produce more offspring than can survive to maturi ty? This question has received some attention in regard to selective embryo abortion (e.g., Burd 1988; Forb es and Mock 1998), but it has not been dealt with in relation to filial cannibalism. Some have hypothesized that the overproduction of offspring can be favored when (1) the cost of producing additional offspring is relatively small and (2) there is a relatively large benefit associated with the ability to screen and weed out weaker offspring post-fertilization (Mock and Parker 1997; Forbes and Mock 1998). I hypothesize that benefits of screen ing and weeding out particular offspring post-fertilization are likely greatest when the environment is variab le. Specifically, if the environment is static, parents would be expected to accurately gauge an d produce some optimal nu mber of offspring of an optimal quality. However, when the environm ent is highly variable, it presumably becomes more difficult for parents to gauge the optimal num ber and optimal quality of offspring. The role of environmental variation has not been explored directly in studies of filial cannibalism, but warrants additional research. Specifically, I hypoth esize that environmental variability plays a 136

PAGE 137

large role in the evolution of filial cannibalism, a nd that when the environment is highly variable, filial cannibalism is more likely to be selected for. The Non-Cannibalistic Parent As mentioned in Chapter 6, almost all of the focus of filial cannibalism has been on the cannibalistic parent (but see Li ndstrm 2000). Thus, the question remains: what role does a noncannibalistic parent play in the evolution of f ilial cannibalism? Lindstrm (2000) suggested that a non-caring parent might benefit from filial can nibalism if cannibalism by the caring parent increases the probability that th e caring parent will su ccessfully rear the clutch. However, the benefits of filial cannibalism to a non-canniba listic parent remain unknown. Indeed, more empirical work that explicitly quantifies the costs and benefits of filial cannibalism to both parents is needed. Identification of Additional Species Practicing Filial Cannibalism For many years, filial cannibalism in fishes wa s dismissed as a rare behavior with little or no adaptive significance. Since beginning my dissertation work, Ive had numerous people mention that their study organisms (e.g., bears, wasps, skinks) exhibit filial cannibalism, but because it was a relatively rare occurrence they didnt give it much thought. This view makes it less likely that researchers will document and investigate filial cannibalism. In the future, it will be particularly important to document filial can nibalism in other taxa. Only then can a truly synthetic framework of filial cannibalism be developed. A Comparative Framework of Filial Cannibalism Once filial cannibalism is better documented, it will be important to consider filial cannibalism from a comparative pe rspective. Using a comparative approach to better understand filial cannibalism is an obvious next step. Howeve r, I would argue this approach is currently impossible, in large part because filial cannibali sm is not formally documented in many animals. 137

PAGE 138

138 In particular, a comparative framework of filial cannibalism would facilitate better understanding of the general life-history characteristics that are likely to be associated with filial cannibalism. Why Dont All Parents Exhibit Filial Cannibalism? My dissertation research suggests that filia l cannibalism represents an adaptive strategy in many contexts and in animals with varying life histories (Chapter 6). If this is the case, why isnt filial cannibalism more common in animals? In fact, I would argue that filial cannibalism is prevalent in animals and that it li kely occurs at some level in the majority of animals. However, as discussed previously, filial cannibalism likely isnt documented in species in which it is difficult to detect or relatively infrequent. For animals that never or infrequently exhibit filial cannibalism, it will be important to quantify the costs of filial cannibalism. Indeed, costs of cannibalism in relation to diseas e transmission have been well-e stablished in several species (Rudolf and Antonovics 2007), and di sease transmission as a possible cost of filial cannibalism warrants further attention.

PAGE 139

APPENDIX ISOLATION AND CHARACTERIZATION OF MICROSATELLITE DNA MARKERS FOR THE FLAGFISH JORDANELLA FLORIDAE The flagfish, Jordanella floridae is a freshwater fish found throughout Florida. Flagfish have been the focus of studies of behavior and evolution (Bonnevier et al. 2003; Klug et al. 2005; Klug and St. Mary 2005), populationand community-level ecology (Jordan and McCreary 1996; Barber and Babb itt 2003; Ruetz et al. 2005), toxi cology (Holdway and Sprague 1979; Rowe et al. 1983; Reinert et al. 2002), and conservation biol ogy (McCormick and Leino 1999). Despite such wide-spread in terest in the flagfish, publishe d microsatellite DNA markers are not yet available for this species. Such markers will be useful for paternity assays, estimating heritability, and characterizing genetic population di versity in the flagfish. Here, I describe the identification and characterization of 6 polymorphic microsatellite markers isolated from a population of flagfish found in the Otter Creek/W accasassa River drainage in northwest-central Florida. I isolated DNA from anal fin clippings us ing the DNeasy Blood and Tissue Kit (Qiagen). A genomic DNA library from one individual was enriched for CA/GT microsatellite repeats using the protocol described in Tools for Developing Molecular Markers (ICBR 2001; modified from Kandpal et al. 1994). I then digested the genomic DNA with Sau3AI enzyme and then fractionated the digested DNA using Chroma Spin columns (BD Biosciences) to capture fragments in the size range of 400 bp and larger The resulting DNA fragments were ligated to Sau3AI linkers using T4 DNA ligase. I used fractionation using Chroma Spin columns to remove excess Sau3AI linkers, and the fragments were polymera se chain reaction (PCR) amplified using the Sau-Linker-A as the primer. I then denatured the entire PCR library (by heating to 98C) and hybridized the library to a biot inylated repeat probe (5-(CA)15TATAAGATA-biotin) at 45C. The hybridized DNA was recovered using VECTRE X Avidin D (Vector Laboratories), and the 139

PAGE 140

resulting DNA was again PCR-amplified using the Sau-Linker-A as the primer. The enriched microsatellite libraries were cloned using a TO PO TA kit (Invitrogen) and transformed into Escherichia coli cells (One Shot TOPO cells, Invitrogen). I screened the cl ones using a biotinlabeled (CA)15 probe (Lifecodes) and the chemiluminescent substrate Lumi-Phos 480 (Lifecodes). Clones from positive colonies were grown overnight at 37C, purified using a Miniprep Kit (Qiagen), and then sequenced usi ng an ABI 377 sequencer (Applied Biosystems). I designed primers for 23 clones identified to be of sufficient le ngth using OLIGO 6.0 (Molecular Biology Insights). For the 10 clones that PCR-amplifie d consistently, I ordered and optimized fluorescently-labeled primers (FAM uppe r primer; Biotech). PCR amplifications were performed in an Eppendorf MasterCycler EP gradient thermocycler. Each 25 L reaction contained 1 x PCR buffer (Sigma), 800 M dNTPs, 3.0 mM MgCl2, 0.26 M of each primer, 1 U Taq polymerase (Sigma), and at least 50 ng temp late DNA. PCR conditions were as follows: 94C denaturation for 4 min followed by 30 cycles of 30 s at 94C, 30 s at the locus specific annealing temperature Ta (Table A-1), and 30 s extension at 72C, followed by 5 min at 72C. Samples were run on an ABI 377 Automated DNA Sequencer (Applied Biosystems) and analyzed using GENESCAN and GENOT YPER (Applied Biosystems). Of the clones that amplified consistently, si x were polymorphic (Table A-1) and free of extraneous bands after optimization. I determ ined microsatellite va riability for 37 to 135 individual flagfish (see Table A1 for locus-specific sample sizes). I calculated observed and expected heterozygosities using POPGENE 1.31 (Yeh et al. 1999). I performed tests for deviations from Hardy-Weinberg expectations and linkage di sequilibrium using GENEPOP 1.2 (Raymond et al. 1995). Expected heterozygosities ranged from 0.69 0.84 (Table A-1). Three of the loci (JFJ4, JF511, and JFJ25; Table A-1) showed significant deviations from Hardy140

PAGE 141

Weinberg expectations after sequential Bonferroni correction (Rice 198 9), suggesting the possibility of null alleles, non-random mating, or the Wahlund effect. None of the loci showed significant linkage disequilibrium after sequentia l Bonferroni correction. The microsatellite markers described herein will likely prove usef ul in studies characterizing population genetic diversity, assessing pa ternity, and quantifyi ng heritability. 141

PAGE 142

142 Table A-1. Characteristics of flagfish microsatellite lo ci; shown here are the locus name, primer sequences, repeat motif, optimum annealing temperature ( Ta C), size range, number of alleles ( NA), the number of individuals tested ( N ), observed heterozygosity ( HO), and expected heterozygosity ( HE). ** indicates statistically significant deviation from Hardy-Weinberg expectations after sequential Bonferr oni correction. Locus Primer sequences (5-3) Repeat motif TaSize range (bp) NAN HOHE JFJ25 JFI3 JFJ4 JF511 JF5121 JF515 GGAGGTCTCGAGGTGTTC AACCCTAAAACTCATCCTAAA GGAAAACACTGGAACCTCG ATCATGCATGTGCCTCTAGC GATAGAGGTGAGAAGGTGCAA CTGGCTGCGTGCACTGA CTCTGTTTGTCGCGTTTGTA AGAGGCCAAACATGCTACC AAGGGTCACGGTTAGGCT AAATCTAACTCCCAATCCAA GCCATGCGTCGTGAGTCAGA GGAGGGAGGACATTGGG (TG)22 (TG)19 (CA)18 (GT)21 (GT)28 (CA)22 58 58 61 64.9 60.1 65.7 296329 228252 275299 316339 278288 152180 24 14 10 15 4 8 114 103 135 109 44 37 0.66 0.81 0.60 0.58 0.59 0.62 0.83** 0.84 0.72** 0.74** 0.69 0.75

PAGE 143

LIST OF REFERENCES Andersson, M. 1994. Sexual Selection. Princeton University Press, Princeton, NJ. Anthony, C.D. 2003. Kinship influences cannibalism in the wolf spider, Pardosa milvina. Journal of Insect Behavior 16: 23-36. Baber, M.J., Babbit, K.J.. 2003. The relative impact s of native and introduced predatory fish on a temporary wetland tadpole assemblage. Oecologia 136: 289 295. Balshine, S., Kempenaers, B. & Szkely, T. (eds). 2002. Conflict and coop eration in parental care. Papers of a Theme Issue Philisophical Transactions of the Royal Society of London B 357: 237-403. Bartlett, J. 1987. Filial cannibalism in burying beetles. Behavioral Ecology and Sociobiology 21: 197-183. Baylis, J.R. 1981 The evolution of parental care in fishes, with reference to Darwins rule of male sexual selection. Environmental Biology of Fishes 6: 223-251. Belles-Isles, J.C. & Fitzgerald, G. J. 1991. Filial cannibalism in sticklebacksa reproductive management strategy. Ethology Ecology and Evolution 3: 49-62. Bellows, T.S. 1981. The descriptive properties of some models for density dependence. Journal of Animal Ecology 51: 139-156. Blouw, D. M. 1996. Evolution of offspring desertion in a stickleback fish. Ecoscience 3: 18. Bonnevier, K., Lindstrm, K., St. Mary, C.M. 2003. Parental care and ma te attraction in the Florida flagfish, Jordanella floridae Behavioral Ecology and Sociobiology 53: 358 363. Browman, H.I., St-Pierre, J.-F., Skiftesvik, A. B. and Racca, R.G. 2003. Behaviour of Atlantic cod (Gadus morhua) larvae: an attempt to li nk maternal condition with larval quality. In: Browman, H.I. and Skiftesvik, A.B. (eds.), The Big Fish Bang. Proceedings of the 26th Annual Larval Fish Conferen ce, Bergen, 22 July 2002. Institute of Marine Research, Bergen, pp. 71. Burd, M. 1998. Excess flower production and selective fruit abortion: a model of potential benefits. Ecology 79: 2123-2132. Chesson, J. 1983. The estimation and analysis of preference and its relationship to foraging models. Ecology 64: 1297-1304. Clutton-Brock, T. H. 1991. The Evolution of Pare ntal Care. Princeton, NJ : Princeton University Press. Creighton, J.C. 2005. Population density, body size, and phenotypic plasticity of brood size in a burying beetle. Behaviou ral Ecology 16: 1031-1036. 143

PAGE 144

Diamond, J.M. 1987. News and views: causes of death before birth. Nature 329: 487-488. Dominey, W. J. 1981. Anti-predat or function of bluegill sunfish nesting colonies. Nature 290: 586-588. Elgar, M.A. and Crespi B.J. (eds.) 1992. Cannibalism: Ecology and evolution among diverse taxa. Oxford University Press, Oxford. Fairbanks, L. A. & McGuire, M. T. 1986. Age, reproductive value, and dominance related behaviour in vervet monkey females: crossgenerational social in fluences on social relationships and reproduction. Animal Behaviour 34: 1718-1721. Forbes, L.S. 1997. The evolutionary biology of spontaneous abortion in humans. Trends in Ecology and Evolution 12: 446-450. Forbes, L.S. & Mock, D.W. 1998. Parental optim ism and progeny choice: when is screening for offspring quality affordable. Journa l of Theoretical Biology 192: 3-14. Forester, D.C. 1979. The adaptiveness of parental care in Desmognathus ochrophaeus Copeia 1979: 332-341. Foster, N.R., Cairns Jr., J., & Kaesler, R.L. 1969. The flagfish, Jordanella floridae, as a laboratory animal for behavioural bioassay stud ies. Proceedings of the Academy of Natural Sciences of Philadelphia 121: 129-152. Frommen, J.G., Brendler, C. and Bakker, T.C.M. 2007. The tale of the bad stepfather: male three-spined sticklebacks Gasterosteus aculeatus L. recognize foreign eggs in their manipulated nest by egg cues alone. Journal of Fish Biology 70: 1295-1301. Gilbert, W.M., Nolan, P. M., Stoehr, A.M., and Hil, G.E. 2005. Filial cannibalism at a House Finch nest. Wilson Bulletin 117: 413-415. Gillooly, J. F. and Baylis, J. R. 1999. Reproductive success and th e energetic cost of parental care in male smallmouth bass. Jour nal of Fish Biology 54: 573-584. Gray, S.M., Dill, L.M., McKinnon, J.S. 2007. Cuckol dry incites cannibalism: male fish turn to cannibalism when perceived certainty of pate rnity decreases. American Naturalist 169: 258-263. Gross, M. R. & Sargent, R. C. 1985. The evolutio n of male and female pa rental care in fishes. American Zoologist 25: 807 822. Gurney, W. S. C. and Nisbet, R. M. 1998. Ecological Dynamics. Oxford University Press, Oxford. Hale, R.E., St. Mary, C. M., & Lindstrm, K. 20 03. Parental response to changes in costs and benefits along and environmental gradie nt. Environmental Biology of Fishes 67: 107-116. 144

PAGE 145

Hesketh, T. and Xing, Z.W. 2006. Abnormal se x ratios in humans populations: causes and consequences. Proceedings of the National Academy of Sciences of the USA 103: 1327113274. Hoelzer, G. A. 1992. The ecology an d evolution of partial-clutch cannibalism by parental cortez damselfish. Oikos 65: 113-120. Holdway, D.A., and Sprague, J.B. 1979. Chronic toxicity of vanadium to flagfish. Water Research 13: 905 910. Interdisciplinary Center for Biotechnology Research. 2001. T ools for Developing Molecular Markers Manual. University of Florida, Florida, USA. Jones, J. and Reynolds, J. D. 1999. Costs of e gg ventilation for male common gobies breeding in conditions of low dissolved oxygen. Animal Behaviour 57: 181 188. Jordan, F., and McCreary, A.C. 1996. Effects of an odonate predator and habitat complexity on survival of the flagfish Jordanella floridae Wetlands 16: 583 586. Kamler, E. 1992. Early life histories of fish: an energetics approach. Chapman Hall, London. Kamler, E. 2005. Parent-egg-progeny relationships in teleost fishes : an energetics perspective. Reviews in Fish Biology and Fisheries 15: 399-421. Kandpal, R.P., Kandpal, G., and Weissman, S.M. 1994. Construction of libraries enriched for sequence repeats and jumping clones, and hybridization selection for region-specific markers. Proceedings of the National Academy of Sciences USA 91: 88-92. Karkkainen, K. Savolainen, O., and Koski, V. 1999. Why do so many plants abort so many developing seeds: bad offspr ing or bad maternal genotypes? Evolutionary Ecology 13 305317. Keckeis, H., Bauer-Nemeschkal, E., Menshutkin, V.V., Nemeschkal, H.L. and Kamler, E. (2000) Effects of female attributes and egg prope rties on offspring viability in a rheophilic cyprinid, Chondrostoma nasus. Can. J. Fish. Aquat. Sci. 57, 789. Klemme, I., Eccard, J.A., and Ylnen, I. 2006. Do female bank voles ( Clethrionomys glareolus ) mate multiply to improve on previous mate s? Behavioral Ecology and Sociobiology 60: 415-421. Klug, H. and St. Mary, C. M. 2005. Breeding season fitness consequences of filial cannibalism in the flagfish, Jordanella floridae Animal Behaviour 70: 685-691. Klug, H. and Bonsall, M.B. 2007. When to care for, abandon, or eat your offspring: the evolution of parental care and fili al cannibalism. American Na turalist 170: 0000 0000. Klug, H. M., Chin, A., & St. Mary, C. M. 2005. The net effects of guarding on egg survivorship in the flagfish, Jordanella floridae Animal Behaviour 69: 661-668. 145

PAGE 146

Klug, H., Lindstrm, K., and St. Mary, C.M. 2006. Parents benefit from eating offspring: density-dependent egg survivorship compensates for filial cannibalsim. Evolution 60: 2087-2095. Kondoh, M. and Okuda, M. 2002. Mate availability influences filial cannibalism in fish with paternal care. Animal Behaviour 63: 227-233. Kozlowski, J. and Stearns, S.C. 1989. Hypotheses for the production of excess zygotes: models of bet-hedging and selective ab ortion. Evolution 43: 1369-1377. Kraak, S.B.M. 1996. Female preference and filial cannibalism Aidablennius sphynx (Teleostei, Blenniidae): a combined field and laboratory study. Behavioural Processes 36: 85-97. Kraak, S.B.M. and van den Berghe E. P. 1992. Do females assess paternal quality by means of test eggs? Animal Be haviour 43: 865-867. Kume, G., Yamaguchi, A., and Taniuchi, T. 2000. Filial cannibalism in the paternal mouthbrooding cardinalfish Apogon lineatus : egg production by the female as the nutrition source for the mouthbrooding male. Environm ental Biology of Fishes 58: 233-236. Kvarnemo, C., Svensson, O., and Forsgren, E. 1998. Parental behaviou r in relation to food availability in the common goby. Animal Behaviour 56: 1285-1290. Lindstrm, K. 1998. Effects of costs and benef its of brood care in the sand goby. Behavioral Ecology and Sociobiology 42: 101-106. Lindstrm, K. 2000. The evolution of filial canniba lism and female mate choice strategies as resolutions to sexual conflicts in fishes. Evolution 54: 617-627. Lindstrm, K. and Sargent, R. C. 1997. Food acce ss, brood size, and filial cannibalism in the fantail darter, Etheostoma flabellare. Behavioral Ecol ogy and Sociobiology 40: 107-110. Lindstrm, K. St. Mary, C.M, and Pampoulie, C. 2006. Sexual selection for male parental care in the sand goby, Pomatoschistus minutus Behavioral Ecology and Sociobiology 60: 46-51. Lissker, M., Kvarnemo, C. and Svensson, O. 2003. Effects of a low oxygen environment on parental effort and filial ca nnibalism in the male sand goby, Pomatoschistus minutus Behavioral Ecology 14: 374-381. McCormick, J.H. and Leino, R.L. 1999. Factors c ontributing to first year recruitment failure of fishes in acidified waters with some implications for environmental research. Transactions of the American Fisheries Society 128: 265 277. McEdward, L. R., and Carson, S. F. 1987. Variation in egg organic content and its relationship with egg size in the starfish Solaster stimpsoni Marine Ecology Progress Series 37: 159169. 146

PAGE 147

McNamara, J.M., Szekely, T., Webb, J.N. and Houston, A.I. 2000. A dynamic game-theoretic model of parental care. Journal of Theoretical Biology 205: 605-623. Manica, A. 2002. Filial cannibalism in tele ost fish. Biological Reviews 77: 261-277. Manica, A. 2004. Parental fish change their cannibalistic behaviour in response to the cost-to benefit ratio of parental ca re. Animal Behaviour 67: 1015. Manly, B.F.J., Miller, P. and Cook, L.M. 1972. Analysis of a selective predation experiment. American Naturalist 106: 719-736. Melser, C. and Klinkhamer, PGL. 2001. Selective seed abortion increases offspring survival in Cynoglossum officinale (Bor aginaceae). American Journal of Botany. 88: 1033-1040. Mertz, J. C. and Barlow, G. W. 1966. On the reproductive behavior of Jordanella floridae (Pisces: Cyprinodontidae) with special reference to a quant itative analysis of parental fanning. Zeitscrift fur Ti erpsychologie 23: 537 554. Mock, DW and Forbes, LS. 1995. The evolution of parental optimism. Trends in Ecology and Evolution 10: 130-134. Mock, D.W. and G.A. Parker. 1997. The Evolution of Sibling Rivalry. Oxford University Press. Mrowka, W. 1987. Filial cannibalism and reprodu ctive success in the maternal mouthbrooding cichlid fish Pseudocrenilabrus multicolor. Behavioral Ecology and Sociobiology 21257265. Narimatsu, Y. and Munehara, H. 2001. Territori ality, egg desertion, and mating success of a paternal care fish, Hypoptychusd Ybowskii. Behavior 138: 85-96. Neff, B.D. 2003. Paternity and condition affect cannibalistic be havior in nest-tending bluegill sunfish. Behavioral Ecology and Sociobiology 54: 377-384. Neff, B.D. and Sherman, P.W. 2003. Nestling recognition via direct cues by parental male bluegill sunfish ( Lepomis macrochirus ). Animal Cognition 6:87-92 Okuda, N. and Yanagisawa, Y. 1996. Filial cannibalism by the mouthbrooding cardinalfish, Apogon doederlini in relation to their physical conditi on. Environmental Biology of Fishes 45: 397. Pampoulie C., Lindstrm K., and St. Mary C.M. 2004. Have your cake and eat it too: male sand gobies show more parental care in the pres ence of female partners. Behavioural Ecology 15:199 Payne, A. G., Smith, C., and Campbell, A. 2002. Filial cannibalism improves survival and development of beaugregory damselfish embr yos. Proceedings of the Royal Society of London Series B 269: 2095-2102. 147

PAGE 148

Payne A. G., Smith, C., and Campbell, A. 2003. The effect of clutch size on whole clutch cannibalism in the beaugregory damselfish Journal of Fish Biology 62: 955-958. Payne, A. G., Smith, C., and Campbell, A. C. 2004. A model of oxygen-mediated filial cannibalism in fishes. Ecological Modelling 174: 253-266. Petersen, C. W. and Marchetti, K. 1989. Filial cannibalism in the Cortez damselfish Stegastes rectifraenum Evolution 43: 158. Polis, G. A. 1981. The evolution and dynamics of intraspecific predati on. Annual Reviews in Ecology and Systematics 12: 225-251. Raymond, M. and Rousset, F. 1995. GENEPOP 1.2: Population genetics soft ware for exact tests and ecumenicism. Journal of Heredity 86: 248 249. Reinert, K.H., Giddings, J.A., and Judd, L. 2002. Effects analysis of time-varying or repeated exposures in aquatic ecologi cal risk assessment of agrochemicals. Environmental Toxicology and Chemistry 21: 1977 1992. Rohwer, S. 1978. Parent cannibalism of offspring and egg raiding as a courtship strategy. American Naturalist 112: 429-440. Rice, W. 1989. Analysis tables of statistical tests. Evolution 43: 223 225. Rosenblatt, J. S. and Snowdon, C. T. 1996. Pare ntal care: evolution, mechanisms, and adaptive significance. NY: Academic Press. Rowe, D.W., Sprague, J.B., and Heming, T.A. 1983. Sublethal effects of treated liquid affluent from a petroleum refinery: chronic toxici ty to flagfish. Aquatic Toxicology 3: 149 159. Rudolf, V.W. and Antonovics, J. 2007. Disease transmission by cannibalism: rare event or common occurrence. Proceedings of the Royal Society B 274: 1205-1210. Ruetz, C.R., Trexler, J.C., Jordan, F., Loftus W., and Perry, S. 2005. Population dynamics of wetland fishes: spatio-temporal patterns s ynchronized by hydrol ogical disturbance? Journal of Animal Ecology, 74, 322 332. Salfert, I.G. and Moodie, G.E.E. 1985. F ilial egg cannibalism in the brook stickleback, Culae inconstans (Kirtland). Behaviour 93: 82-100. Sargent, R. C. 1988. Paternal car e and egg survival both increase with clutch size in the fathead minnow, Pimephales promelas. Behavioral Ecology and Sociobiology 23: 33 37. Sargent, R. C. 1992. Ecology of filial canniba lism in fish: theoretical perspective. In: Cannibalism: Ecology and Evolution Among Diverse Taxa (Ed. by M. A. Elgar & B. J. Crespi), pp.38-62. Oxford: Oxford University Press. 148

PAGE 149

Sargent, R. C. 1997. Parental care. In Behavioural Ecology of Teleost Fishes (Ed. by J.-G. J. Godin), pp. 292-315.Oxford: Oxford University Press. SAS Institute. 2001. SAS 8.2. Cary, NC: SAS Institute. Sikkel, P. C. 1994. Filial cannibalism in a patern al-caring marine fish: the influence of egg developmental stage and position in the nest. Animal Behaviour 47: 1149-1158. Simon, M. P. 1983. The ecology of parental care in a te rritorial breeding frog from New Guinea. Behavioral Ecology and Sociobiology 14: 61-67. Smith, C. 1992. Filial cannibalism as a reproductive strategy in care-giving teleosts. Netherlands Journal of Zoology 42: 607-613. Smith, C. and Wootton, R. J. 1995. The costs of pa rental care in teleost fishes. Reviews in Fish Biology and Fisheries 5: 7-22. SPSS Inc. 2000. SYSTAT 9.0 for Windows. Chicago, IL: SPSS Inc. St. Mary, C. M, Noureddine, C. G, and Lindstrm, K. 2001. Effects of the environment on male reproductive success and parental care in the Florida flagfish, Jordanella floridae. Ethology 107: 1035-1052. Stearns, S.C. 1987. The selection arena hypothesi s, pp. 337-349. In. S.C. Stearns (ed.) The evolution of sex and its cons equences. Birkhuser, Basel. Stearns, S.C. 1992. The Evolution of Life Histories. Oxford University Press. Svensson, O., Magnhagen, C., Forsgren, E. and Kvarnemo, C. 1998. Parental behaviour in relation to the occurrence of sneaking in the common goby. Animal Behaviour 56: 175179. Svensson, O. and Kvarnemo, C. 2007. Parasi tic spawning in sand gobi es: an experimental assessment of nest-opening size, sneaker male cues, paternity, and filial cannibalism. Behavioral Ecol ogy 18: 410-419. Takeyama, T., Okuda, N., and Yanagisawa, Y. 2002. Seasonal pattern of by Apogon doederleini mouthbrooding males. Journal of Fish Biology 61: 633-644. Takeyama, T., Okuda, N., and Yanagisawa, Y. 2007. Filial cannibalism as a conditional strategy in males of a paternal mouthbrooding fish. Evolutinoary Ecology 21: 109-119. Thomas, L.K. and Manica, A. 2003. Filial cannibalism in an assassin bug. Animal Behaviour 66: 205-210. Trivers, R. L. 1972. Parental inve stment and sexual selection. In: Sexual Selection and the Descent of Man (Ed. by B. Campbell), pp. 136-179. Chicago: Aldine. 149

PAGE 150

150 Vincent, T.L. and Brown, J.S. 2005. Evolutionary game theory, natural selection and darwinian dynamics. Cambridge University Press, Cambridge. Vinyoles, D., Cote, I., and de Sostoa, A. 1999. Eg g cannibalism in river bl ennies: the role of natural prey availability. Journa l of Fish Biology 55: 1223-1232. Warkentin, K.M. 2000. Wasp predation and wasp-induced hatching of red-eyed treefrog eggs. Animal Behaviour 60: 503-510. Webb, J.N., Houston, A.I., McNamara, J.M. and S zekely, T. 1999. Multiple patterns of parental care. Animal Behaviour 58: 983-993. Williams, G. C. 1975. Sex and Evolution. Pr inceton University Press, Princeton, NJ. Williams, J.E. 2000. The Coefficient of Condition of Fish. In Schneider, James C. (ed.) 2000. Manual of fisheries survey methods II: with periodic updates. Mich igan Department of Natural Resources, fisheries Special Report 25, Ann Arbor. Wittenberger, J. F. 1981. Animal Social Behavior. Duxbury Press, Boston:MA. Yeh, F.C., Yang, R.C., and Boyle, T. 1999. POPGENE 1.31. Microsoft Windows-Based Software for Population Genetics Analysis. University of Alberta and Centre for International Forestry Research, Alberta, Canada.

PAGE 151

BIOGRAPHICAL SKETCH Hope Klug was born June 25, 1980, and grew up in Palm Harbor, FL. She graduated from the International Baccalaureate Program at St. Petersburg High School in 1998, and received a B.S. in zoology and psychology from the Univers ity of Florida in 2001. She began her Ph.D. work in 2002, and plans to conduct post-doctoral research in Finland be ginning in January 2008. 151