Proximate and ultimate causes of brood reduction in Brown pelicans (Pelecanus occidentalis)

PROXIMATE AND ULTIMATE CAUSES OF BROOD REDUCTION IN BROWN
PELICANS (PELECANUS OCCIDENTALIS)

BY

BONNIE JEAN PLOGER

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

1992

ACKNOWLEDGEMENTS

I thank Jane Brockmann for her tremendous support

through all phases of my research. Jane has been a

wonderful advisor, treating me simultaneously as a

colleague, student and friend. Thanks also go to Peter

Feinsinger for his friendship, inspired teaching and helpful

criticisms of research proposals. I also thank Peter

Feinsinger, along with Carol Aubspurger, for getting me

excited about "siblicide" and "brood-reduction" in plants.

In addition to Peter, I also thank Marty Crump and Mike

Collopy for their ideas and advice during the planning of my

research. I wish that they had remained in Florida and thus

been able to remain on my committee.

I thank Jack Kaufmann, Doug Levey and John Sivinski for

generously coming on board as committee members at the 11th

hour. I thank them for their helpful comments on earlier

drafts of the dissertation. Doug Levey provided

particularly detailed comments on drafts of the

dissertation, for which I am grateful. I thank Lou

Guillette for all his ideas and advice about my project on

embryonic development in cattle egrets. I also thank Lou

for remaining on my committee after I changed projects and

strayed far from developmental biology.

Doug Mock has acted as an unofficial outside member of

my committee through periodic idea-full conversations and

through detailed critiques of my grant proposals. Doug also

provided valuable comments on earlier drafts of Chapter 2.

I thank Doug whole-heartedly for his continuing support.

Thanks also go to Trish Schwagmeyer for her suggestions and

ideas about the design of my research on brown pelicans.

Several people provided me with information on brown

pelicans that was either hard to obtain or unpublished. I

am grateful to D. Pinzon and Hugh Drummond for sending me

their unpublished manuscript. I thank Jim Rodgers for

providing hard-to-find references. Stephen Nesbitt provided

me with an annual performance report on brown pelicans in

nesting in Florida. He also provided helpful tips about

colony locations. I thank Carolina Murcia providing

information about a Panamanian pelican colony. I am also

indebted to Mark Shields for not only providing me with

unpublished data from his North Carolina population, but

also allowing me to collect blood from some of his birds for

a post-doctoral project.

The National Audubon Society provided invaluable

support during my 1990 brown pelican research. Special

thanks go to Rich Paul, manager of the Tampa Bay

Sanctuaries, for his generous support throughout the 1990

research season.. Rich not only provided me with permission

to work in the colonies, but also to stay in a cabin on one

iii

of the islands. I am particularly grateful to Rich for

loaning me a boat, and continuing to let me use it even

after I sank it. I also thank the Gardenier Mining Company

for permission to work on their islands and for providing

parking for research vehicles. Pam Phelps and Steve McGehee

provided outstanding assistance while living in primitive

conditions and working extremely long hours. I also thank

them both for their continued friendship.

I thank Jim Johnson, Refuge Manager of the Lower

Suwannee National Wildlife Refuge of the U.S. Fish and

Wildlife Service, for permission to work on Seahorse Key in

1989. Thanks also go to Frank Maturo for facilitating my

use of the biological station on the island, and to Chuck

Haven, Henry Coulter and especially K. C. Brown for

maintenance of the boats and research station. I thank

Laurie Eberhardt and Jane Brockmann for keeping track of

events in key focal nests on a few days when I needed

several hours on break. I also thank my mother, Eleanore

Ploger, for assisting with some data entry while enduring

extreme heat and humidity. Thanks also go to my brother,

Jim Ploger, for working on Jane Brockmann's horseshoe crabs

for a week so that I could continue my pelican observations

uninterrupted.

A total of 11 months of continuous observations created

a tremendous mountain of data. I thank Joan Binkley, Ron

Clouse, James ("J. P.") Jouver, and Mark Stowe for helping

iv

me with this arduous task. Thanks go to Mark Stowe for the

use of his computer, printer and modem for an extended

period. I also thank Ming Lee Prospero for assistance with

manuscript preparation and the literature search.

Financial support in 1990 was provided by an Elizabeth

Adams Fellowship from Mount Holyoke College and awards from

the Frank M. Chapman Memorial Fund and the Joseph Henry Fund

of Sigma Xi Grants-in-Aid of Research. The Department of

Zoology at the University of Florida provided me with a

research assistantship, use of a boat, and housing at the

Seahorse Key Biological Station for the 1989 season. Thanks

again go to H. J. Brockmann for providing supplemental

financial support in the 1989 season. Financial support in

both years was also provided by the University of Florida

Foundation. I also thank my husband, Don Allan, for

providing additional financial support during this period

when we were still engaged to be married.

En route to my work with brown pelicans, I conducted a

variety of research projects. I wish to thank all those who

assisted me during these earlier projects. My first

potential dissertation project involved communal roosting

behavior of Heleconius butterflies. I thank Lincoln Brower,

Tom Emmel, and Allan Masters for their helpful advice on

this project. Peter May introduced me to good field sites

in Florida. I also thank Larry Gilbert for showing me his

Costa Rica field site and discussing the details of

Heleconius biology. I did not stay with this project. The

lure to work with creatures with feathered rather than

scaled wings was too great.

My second potential dissertation project investigated

the proximate causes of hatching asynchrony in cattle

egrets. Numerous people helped me with this two-year study.

This project required placing floating blinds in willow-

swamps. The following people helped to build and/or launch

two floating platforms: Rich Buchholz, Kazuo Horikoshi,

Carlos Martinez del Rio, Paul Andreadis, Doug Levey, Laurie

Eberhardt, Alan Pounds, Jane Brockmann, Mike Miamoto, Mark

Stowe and Cathy Sehley. One of my experiments required

construction of 30 artificial cattle egret nests. I thank

Vas Demas, Laurie Eberhardt, Peter Martin, Cathy Langtimm,

Paul and Debbie Andreadis, John Scott Foster, Alehondro

Grehal, Rob and Linda Garrett, Brent and Sylvia Palmer, and

Mark Stowe for participating in a "nest fest" and building

artificial nests. Martha Groom, Bob Podolsky, Carlos

Martinez Del Rio, Paul Andreadis, Linda Fink, Doug Mock and

Jane Brockmann all provided valuable comments on various

drafts of my cattle egret research proposal. Lou Guillette,

Vince Demarco, Brent Palmer and Vicki McDonald provided

helpful advice about embryonic development. Laurence

Alexander provided information about potential colony

sites.I also thank Peter Frederick, Marilyn Spalding, Ken

Meyer, Naomi Edelson and other members of the "Wading Bird"

vi

group for valuable suggestions about the research design.

The cattle egret project would have been impossible

without assistance in the field. Russ Sanderson was

tremendously enthusiastic about splashing hip-deep in

alligator-infested water at 4:00 a.m. every morning. I also

thank Ken Kroel, Ilse Barube, Cathy Sahley and Pam Wexler-

Rubin for enduring difficult field conditions and a

stressed and sometimes testy supervisor.

Financial support for cattle egret project was provided

by an Animal Behavior Research Grant, a Herbert and Betty

Carnes Research Fund Award from the American Ornithologists'

Union, an E. Alexander Bergstrom Memorial Research Fund

Award from the Association of Field Ornithologists, a

Florida Ornithological Society Research Grant, a Grant-in-

Aid of Research from Sigma Xi and a Zoology Department

Research Assistantship.

Several people deserve thanks both for their friendship

and for their helpful comments about my research. These

people include: Rich Buccholtz, Laurie Eberhardt, Bob

Podolsky, Martha Groom and Dustin Penn. I also thank other

members of Jane Brockmann's "Behavior Group" for their

ideas.

Special thanks go to Mark Stowe for his friendship,

helpful criticisms of drafts of grant proposals, fruitful

discussions of research ideas, and his inventive electronic

wizardry. Mark has been a great help from the beginning of

vii

my work in Florida. Special thanks also go to Abby Palmer

and Jane Morris for their wonderful emotional support.

I can not thank my husband, Don Allan, enough for all

of his logistic support in 1990, assistance with manuscript

preparation, and tremendous emotional support throughout all

phases of my pelican research. Don has kept me healthy and

happy through a highly stressed period of my life.

Finally, I want to express my deepest gratitude to my

parents, Eleanore and William Ploger, and to my brother Jim

Ploger, for all of their support and encouragement. We have

shared many wonderful hours exploring backroads and

wilderness areas. I thank them for stimulating and

supporting my interest in natural history which brings me

such joy.

viii

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

PROXIMATE AND ULTIMATE CAUSES OF BROOD REDUCTION IN BROWN
PELICANS (PELECANUS OCCIDENTALIS)

By

BONNIE JEAN PLOGER

December, 1992

Chairman: Dr. H. Jane Brockmann
Major Department: Zoology

Brown pelican (Pelecanus occidentalis) parents produce

more offspring than they usually raise to independence

because brood-members that hatch last starve or are killed

by their older siblings. This pattern is puzzling because

offspring are produced that seem to provide no reproductive

value to their parents. But these "marginal" offspring

(most likely to be brood-reduction victims) may contribute

reproductive value by surviving as replacements for senior

siblings that die unexpectedly. Marginal offspring may also

have value as additional survivors during periods of food

abundance. I found that brood reduction by starvation and

siblicide was common. Eggs also failed to hatch and chicks

fell from nests and were killed by strange adults and

neighboring nestlings. Some second-hatched ("B-") chicks

replaced dead first-hatched ("A-") chicks and others

survived along with their seniors. All third-hatched ("C-")

chicks died.

Food shortages probably were not a proximate cause of

siblicidal aggression, except possibly when broods were 13-

17 days old. Fighting during the first week of life was

independent of nestling growth and food supplies, and

probably served to establish a dominance hierarchy. When

broods were 13-17 days old, fighting rates were best

predicted by the difference in the rate of bill growth of A-

chicks relative to their B-siblings. But when broods were

17-21 days old, fighting rates were best predicted by

accelerated B-chick growth, as could be expected if fast-

growing B-chicks threatened the dominance of A-chicks.

All adaptive explanations for brood reduction assume

that parents deliver a fixed amount of food and so survivors

gain extra food after the death of a sibling. I tested this

assumption by removing or adding a chick to three-chick

broods. Parents delivered similar amounts to enlarged,

control and reduced broods during the first 6 days post-

treatment. By 9 days post-treatment, parents brought less

food to reduced than to control broods. Seniors did not

gain more food in reduced broods during these periods. A

feeding hierarchy was evident, with A-chicks gaining more

food than their B-siblings who gained more than their C-

siblings. The fitness interests of parents may conflict

with those of one senior but not with the other senior

offspring.

TABLE OF CONTENTS

ACKNOWLEDGEMENTS ......................................... ii

ABSTRACT .................... ............................ viii

CHAPTERS

1 GENERAL INTRODUCTION.................................... 1

2 ROLE OF JUNIOR SIBLINGS IN RESOURCE TRACKING AND AS
INSURANCE FOR SENIOR LOSS ........................... 10

Introduction........................................... 10
Methods................................................ 15
Study Site .......................................... 15
General Procedure ................................... 16
Focal Nests ......................................... 19
Visual Census Nests ................................. 20
Measures of Hatching Success ........................ 21
Measures of Fledging Success ........................ 22
Determining Hatching Dates .......................... 22
Reproductive Value of Junior Siblings ............... 24
Analyses of Fates Based on Chick Ranks .............. 25
Analyses of Fates Independent of Chick Ranks ....... 26
Causes of Chick Mortality ........................... 26
Supplemental Census Nests ........................... 33
Estimating Effects of Colony Disturbance ............ 36
Statistical Analyses ................................ 37
Results .............................................. 38
Survival ............................................ 38
Reproductive Value of Junior Siblings ............... 43
For What Causes of Senior Death do Juniors Provide
Insurance? ....................................... 46
Comparisons of Food-dependent and Food-independent
Mortality ........................................ 46
Timing of Food-dependent Deaths ..................... 48
Discussion............................................. 49
Survival ............................................ 50
Partitioning the Reproductive Value of Junior
Chicks ........................................... 52
For What Causes of Senior Death do Juniors Provide
Insurance? ...................................... 57
Brood Reduction in Pelican Species .................. 62

xii

3 HUNGER AS A PROXIMATE CAUSE OF FIGHTING............

Introduction ........................
Methods .............................
Study Site .......................
Observation and Censusing Methods
Nest Observations ................
Feeding Behavior .................
Fighting Behavior ................
Analyses .........................
Results..............................
Discussion...........................
The Food-amount Hypothesis .......
Energetic Costs of Fighting ......

Size Hierarchies and Sibling Rivalry Reduction..

... 101

Sibling Aggression in Related Species ...............102

4 EFFECT OF BROOD SIZE MANIPULATIONS ON FOOD DELIVERIES

AND APPORTIONMENT TO SENIOR SIBLINGS ...

Introduction...............................
Methods...................................
Study Site .............................
Brood-size Manipulations ...............
Census and Observation Methods ........
Feeding Behavior .......................
Fighting Behavior ......................
Animal Care Considerations .............
Results ..................................
Discussion.................................
Do Seniors Gain a Food Bonus from Brood
Proximate Costs of Maintaining C-chicks
Parent-offspring Conflict .............

............. 114

...........
...........
...........
...........
...........
...........
...........
...........
...........
...........
Reduction?
...........
...........

5 SUMMARY AND CONCLUSIONS ..............................

Parent-Offspring Conflict and Food-Dependent Fighting..
Brood Reduction as a Product of Parent-Offspring
Conflict .......................... ............. ......

APPENDICES

A DETERMINING CHICK AGES .................................

B DETERMINING CLUTCH SIZES ...............................

C WHY DO OLDER CHICKS AND ADULTS ATTACK NESTLINGS?.......

D NEST TAKEOVERS .............................. ......... ..

LIST OF REFERENCES .......................................

xiii

114
119
119
119
123
125
127
127
128
131
133
139
142

148

151

154

165

172

173

177

179

..... 73

.................

.................
.................
.................
.................
.................
.................
.................
.................
.................
.................
.................

BIOGRAPHICAL SKETCH ....................................... 192

xiv

CHAPTER 1
GENERAL INTRODUCTION

Many organisms produce more offspring than they usually

rear to independence because they abort, eat or neglect some

of their offspring, or allow siblings to kill (and sometimes

eat) each other. Abortion of embryos is common in many

plant species (see reviews in Buchholz 1922, Lloyd 1980,

Stephenson and Bertin 1983, Haig 1986, 1987, Sutherland 1986

and Mazer 1987) and in some mammals (reviews in Diamond

1987, Stearns 1987). Parents may consume their progeny in

some insects (e.g. Wilson 1971, Masuko 1986, Bartlett 1987),

fish (e.g. Salfert and Moodie 1984, FitzGerald 1992) and

amphibians (Simon 1984). In species that provide their

offspring with food, partial brood loss is often due to

starvation of some offspring through parental neglect or

sibling competition, as is widespread in birds (Lack 1968,

Howe 1976, O'Connor 1978, more recent reviews in Clark and

Wilson 1981, Mock 1984a). Parents in many taxa including

insects (e.g. Eickwort 1973), fish (e.g. Springer 1948,

Gilmore et al. 1983, Valerio and Barlow 1986, review in

Dominey and Blumer 1984), amphibians (review in Simon 1984),

and birds (Ingram 1959, Bortolotti et al. 1991) tolerate

cannibalistic sibling competition (reviewed in Polis 1984).

Noncannibalistic siblicide is more common than cannibalistic

siblicide in birds (review in Mock et al. 1990) and mammals

(e.g. O'Gara 1969, Fraser 1990, Frank et al. 1991).

These phenomena pose a problem for evolutionary

biologists: why do parents "waste" their time and resources

to produce offspring that they fail to provision with

parentally controlled resources? That this is common in a

wide variety of disparate taxa suggests that the production

of doomed offspring probably conferred enhanced fitness to

at least some of the participants (parents and/or some

offspring) in the past, and may continue to enhance fitness

in the present. Alternatively, such "brood reduction" (the

overproduction and subsequent elimination of some offspring)

could be a negative consequence of selection acting on some

other factor such as large clutch size, per sg, or hatching

asynchrony (e.g. Clark and Wilson 1981). My dissertation

focuses on the proximate and ultimate causes of brood

reduction in brown pelicans (Pelecanus occidentalis). For

convenience, I refer to those brood-members most likely to

be the victims of brood reduction as "marginal" offspring

(Mock and Parker 1986) because their survival chances are

marginal and typically contingent on the death of a sibling

or unusually abundant food.

Four major hypotheses have been proposed to explain how

the production of "marginal" offspring may be beneficial to

parents (reviewed in Forbes 1990, 1991). First, "marginal"

offspring may be used as food for parents or offspring. This

is the "exploitation hypothesis" of Hrdy (1979), described

earlier by Ingram (1959) and called the "ice-box hypothesis"

by Alexander (1974). An obvious prediction of this

hypothesis is that cannibalism of offspring or siblings

occurs routinely (or at least during food shortages).

Second, "marginal" offspring may enable parents to select

offspring with the highest fitness expectations. I call

this the "progeny-choice hypothesis" (Forbes 1991). This

hypothesis was first called developmental selection by

Buchholz (1922) and also the "selective-abortion hypothesis"

by Kozlowski and Stearns (1989). The progeny-choice

hypothesis argues that offspring differ in quality and that

the brood-members that are eliminated are those that are

genetically or developmentally "inferior" to their siblings.

Elimination of these "inferior" siblings is predicted to

occur very early in the developmental period, as soon as

differences in offspring quality are detectable (Kozlowski

and Stearns 1989). Third, "marginal" offspring may function

as insurance for partial brood loss, serving as replacements

for siblings that die unexpectedly from accidental causes or

congenital defects. This is the "insurance hypothesis" of

Dorward (1962), reviewed by Forbes (1990, 1991). This

hypothesis predicts that accidents or congenital defects are

frequent causes of partial brood loss. For example, under

this hypothesis, hatching failure in birds is predicted to

be more common in brood-reducing species than in species

that do not produce "marginal" offspring (Anderson 1990).

Fourth, "marginal" offspring may enable parents to maximize

their reproductive success from a given clutch when

resources are unpredictable by laying as many eggs as they

could raise in a good year and reducing the brood if

resources turn out to be scarce. This is the "resource-

tracking hypothesis" formulated by Lack (1947), usually

called the "brood reduction hypothesis" since Ricklefs

(1968). More recently, this hypothesis has been called the

"resource availability hypothesis" and "resource tracking"

by Forbes (1990 and 1991, respectively), and the "bet-

hedging hypothesis" by Kozlowski and Stearns (1989). In the

dissertation, I will follow Forbes' (1991) and refer to this

as the resource-tracking hypothesis. These hypotheses are

not mutually exclusive, and indeed all may operate

simultaneously (Forbes 1991). But clearly, a key to

determining whether any or all of these hypotheses are valid

is to examine causes and patterns of egg and nestling

mortality.

Brown pelicans typically lay three eggs which hatch

asynchronously (Schreiber 1979). Nestling mortality is

biased toward last-hatched members of the brood (Schreiber

1979), who are frequently attacked by their elder siblings

(Pinzon and Drummond in press) and typically die of

starvation and siblicide (see Chapter 2). Parents generally

remove dead offspring by tossing them from the nest (unpub.

data). Cannibalism is rare and occurred in only one of the

122 nests that I kept under close observation during this

study. In this one nest, the parent ate its youngest

offspring after it died and the second-hatched chick ate a

dead nonsibling that had been experimentally added earlier

to enlarge the brood. The rarity of cannibalism makes the

exploitation hypothesis an unlikely explanation for brood

reduction in brown pelicans and I will not consider it

further. This leaves the progeny-choice, insurance and

resource-tracking hypotheses as possible explanations. In

Chapter 2 of this dissertation, I evaluate the insurance and

resource-tracking hypotheses for brown pelicans. I examine

in detail the role of the last-hatched member of brown

pelican broods as insurance for senior loss and as survivors

along with their senior siblings. This chapter also

includes some discussion of the validity of the progeny-

choice hypothesis for siblicidal pelican broods.

The resource-tracking hypothesis assumes that partial

brood losses increase directly with food scarcity. Where

siblicide is a major cause of brood reduction, as in brown

pelicans (see Chapter 2), the frequency of siblicide should

increase during periods of food shortage. One way that

siblicide could increase with food depletion is if sibling

aggression is proximately controlled by food supply, with

decreases in food to nestlings causing increased aggression

among brood-members. In brown pelicans, where older

siblings attack their juniors (Pinzon and Drummond in

press), the intensity of sibling aggression varies among

nests (Chapter 3). In Chapter 3, I investigate whether this

variation in sibling aggression depends on food supply, as

might be expected if the resource-tracking hypothesis is

operating.

The progeny-choice, insurance and resource-tracking

hypotheses all argue that the survival chances of young that

did not die increase when brood size decreases (O'Connor

1978) because the remaining young obtain more food after the

death of their competitor. This will only happen if parents

deliver the same amount of food to the brood before and

after brood reduction. These hypotheses consider the

fitness of both parents and surviving offspring to increase

similarly with brood reduction. But this need not be the

case. As Hamilton (1964) first argued, the fitness

interests of parents and offspring may differ. Indeed,

conflict is likely to be more common than congruence of

parent and offspring interests, because, as Trivers' (1974)

development of Hamilton's idea clarified, selection should

favor offspring that seek more investment from their parents

than their parents are selected to give. This insight

spawned many theoretical analyses of parent-offspring

conflict (see Godfray and Parker 1991 for review), including

many on parent-offspring conflict over brood size (O'Connor

1978, Godfray 1986, Lazarus and Inglis 1986, Parker and Mock

1987, Godfray and Harper 1990, Godfray and Parker 1991,

1992). These models predict parent-offspring conflict over

brood reduction. O'Connor's (1978) model predicted that the

threshold beyond which parents and senior offspring benefit

from brood reduction is lower for offspring than for

parents, creating conditions in which "siblings would gain

in net fitness by eliminating one young, albeit at the

expense of adult fitness" (O'Connor 1978:88). Thus, rather

than an adaptation favoring the fitness of both parents and

offspring, brood reduction may be the result of senior

siblings achieving their optimum to the detriment of

parents.

In Chapter 4, I explore some of the costs and benefits

to parents and offspring of eliminating the third-hatched

chick. Hypotheses proposing an adaptive value to brood

reduction all assume that senior survival increases

following brood reduction because seniors gain more food

following elimination of a competitor. Thus, parents and

seniors are assumed to benefit from brood reduction. But

this is possible only if parents do not decrease food

supplies following brood reduction. I experimentally

altered brood sizes to determine if parents fed reduced,

control and enlarged broods at the same level, as predicted.

I also determined if senior offspring gained more food in

reduced broods. Other outcomes would result if brood

reduction serves the interests of only the parents or the

offspring, or results from some compromise of parent-

offspring conflict.

Natural selection favors parents whose behavior

maximizes their lifetime reproductive success. From this,

Lack (1947) argued that selection would favor parental

behavior that maximizes the number of surviving offspring

produced from each individual clutch. The overproduction of

eggs and subsequent reduction in brood sizes through

starvation (and sometimes siblicide) seem paradoxical

because parents produce more offspring than they are able

(or willing) to raise, thereby wasting resources that could

otherwise be invested in the remaining offspring to enhance

their fitness prospects. The exploitation, progeny-choice,

insurance and resource-tracking hypotheses are all attempts

to explain how parents who practice brood reduction could

still be maximizing their short-term reproductive success

during a single breeding attempt. What these hypotheses

overlook is the potential for conflict over parental

investment among current offspring and between parents and

current offspring as parents attempt to balance tradeoffs

between their interests in current versus future offspring

in a way that maximizes lifetime reproductive success

(Williams 1966). My dissertation attempts to identify where

the interests of parents and offspring may conflict in

broods of brown pelicans and to lay the groundwork for

9

determining whose interests are being represented by brood

reduction, those of the parents or those of the offspring.

CHAPTER 2
ROLE OF JUNIOR SIBLINGS IN RESOURCE TRACKING AND AS
INSURANCE FOR SENIOR LOSS

Introduction

In many bird species, females lay more eggs than pairs

typically are able or willing to feed sufficiently. Among

species that hatch asynchronously, it is the youngest brood

members that usually die, typically of starvation or from

beatings delivered by siblings (siblicide). Two major

hypotheses, the "resource-tracking" and "insurance-

offspring" hypotheses, provide testable explanations for how

parents may benefit from producing offspring that are

usually doomed to die (see Forbes 1991 for additional

hypotheses). Mock and Parker (1986) pointed out that these

usually doomed "marginal" offspring may contribute to the

reproductive value of their parents in two ways. First,

"marginal" offspring may provide "extra-chick value" by

surviving along with their siblings when food is plentiful.

When food is scarce, "marginal" chicks starve quickly, which

presumably benefits surviving siblings who gain the

"marginal" chick's food share ("resource-tracking

hypothesis," Lack 1947, 1954, 1968). Under the resource-

tracking hypothesis, "marginal" chicks are "extra" in the

sense that they represent a bonus unit of reproductive

success to parents when food is abundant. Second,

"marginal" offspring may provide "insurance value" as

replacements for senior siblings that die (the "insurance-

offspring" hypothesis, Dorward 1962). Although originally

stated as a separate hypothesis, the insurance hypothesis is

actually a special case of the resource-tracking hypothesis

(Anderson 1990). Insurance operates when parents are faced

with unpredictable brood size (e.g. unpredictable levels of

hatching failure) and predictable resource shortages after

hatching. Resource-tracking operates when brood sizes are

predictable and resources are unpredictable. In both

situations, brood sizes are matched to current food supplies

(Forbes 1990). The insurance hypothesis is usually invoked

for the extreme case of obligate siblicide, in which the

youngest brood members have only "insurance value" (Mock and

Parker 1986) because they never survive along with their

seniors. The resource-tracking hypothesis is usually

invoked when partial brood loss is less frequent and the

youngest chick survives along with its siblings when

conditions are favorable, thereby providing its parents with

an "extra" unit of reproductive success (Mock and Parker

1986). In facultative brood reducers with intermediate

levels of junior chick mortality, parents may derive

additional reproductive success through both routes, with

last-hatched chicks contributing both "extra-chick" and

"insurance" value to their parents (Mock and Parker 1986).

The purpose of this study is to evaluate some of the

selective pressures that contribute to the laying of "extra"

eggs by brown pelicans (Pelecanus occidentalis). I used

three approaches to examine the importance of junior chicks

as "extra" survivors and as insurance against senior chick

loss. First, I examined the exact causes of nestling

mortality to evaluate two predictions: (1) if junior chicks

serve primarily as "extra" survivors when food is abundant,

then junior chick mortality should be food-dependent (e.g.

starvation and siblicide); (2) if juniors serve primarily as

insurance against senior death, then senior chick mortality

should be due to food-independent causes such as hatching

failure, deaths of hatchlings (Dorward 1962, O'Connor 1978,

Stinson 1979, Mock 1984a, Magrath 1990), predation (Nisbet

1975, Nisbet and Cohen 1975, Mock and Parker 1986, Drummond

1986), or ectoparasites (Bryant and Tatner 1990). Second, I

compared the reproductive value of junior chicks as

additional survivors and as replacements for senior siblings

that died. The junior chick's value as an "extra" survivor

was measured as that component of its survivorship that was

independent of its siblings' survival. The junior chick's

insurance value was measured as that component of its

survivorship that depended on the fate of its senior

siblings (Mock and Parker 1986, Ploger and Mock unpub. MS).

I could not use this direct approach for last-hatched chicks

in three-chick broods, because all of the last-hatched

nestlings died during my study. Instead, I used a third

approach, which was to inspect the order and timing of last-

hatched versus senior deaths. To have potential insurance

value, a last-hatched chick would need to survive into the

period of peak mortality risk to seniors.

Brown pelicans lay more eggs than they ordinarily

fledge; clutch sizes average three eggs which typically

yield one or two fledglings (Blus and Keahey 1978, Schreiber

1979). Nestling mortality has been generally attributed to

starvation, based on reports of light-weight nestlings

(Schreiber 1976, Keith 1978) and correlations between

fledging success and fish supplies (Anderson et al. 1977,

1982) or regurgitation frequencies (Schreiber 1979). All

past work with this species has relied on circumstantial

evidence from censusing nests to assign causes of mortality.

Such evidence is not sufficient and may lead to false

estimates of the relative frequencies of food-dependent

versus food-independent deaths. For example, Schreiber

(1976) attributed the selective mortality of last-hatched

brood members to starvation whenever a chick failed to grow

between the last censusing visits prior to the chick's

disappearance. But a thin chick that vanished between

censusing visits might not have starved to death. Instead,

such a chick might have been recovering when it was taken by

a predator, or might have died from beatings delivered by

siblings, as described for the first time for brown pelicans

by Pinzon and Drummond (in press). Pinzon and Drummond also

relied primarily on nest censusing to identify causes of

death, although they directly observed siblicidal aggression

in some nests. They attributed death to siblicidal

expulsion whenever a chick's body was found outside of the

nest. But some such chicks might have been accidently

knocked from the nest and subsequently attacked by

neighbors, or have fallen accidently from tree nests and

been injured by the fall. These alternative explanations

are quite possible for brown pelicans, as I will discuss.

I also present observations of infanticidal attacks

initiated by unrelated adults and chicks against brown

pelican nestlings. This chapter presents a detailed

examination of the causes of mortality of nestling brown

pelicans based on continuous observations of a large sample

of nests.

Pinzon and Drummond (in press) present the first

discussion and evidence that brown pelicans may be

facultative brood reducers, producing three-egg clutches and

selectively eliminating some chicks during food shortages.

But in three-chick broods of brown pelicans, the youngest

virtually always dies whereas survival of the second-hatched

is highly variable (Schreiber 1976). Thus, brood reduction

in brown pelicans may be obligate among last-hatched young

and facultative among those hatching second in the brood.

As a result, these chicks are likely to differ in their

roles as insurance or "extra" chicks. In addition, the

nature of the insurance value may differ between second and

third-hatched chicks. In most obligately brood-reducing

species, the insurance value of marginal chicks is

restricted to the first week or less of the nestling period

(Anderson 1990, Mock et al. 1990) and so covers only egg

failure and hatchling losses such as those caused by

developmental abnormalities. In contrast, facultative

brood-reducers like brown pelicans may retain all brood

members until food-supplies become limiting (perhaps later

in the nestling period). Thus, the insurance value of a

marginal chick can remain in effect for longer and may cover

a wider range of risks to the seniors than in obligate

brood-reducers (Mock and Parker 1986).

Methods

Study Site

In 1990, I studied brown pelicans nesting on Bird and

Sunken Islands, two spoil islands connected by a sandbar

that together are known as Alafia Banks in Hillsborough Bay,

near Tampa, Florida. Approximately 850-900 brown pelicans

nested on the islands in 1990. They nested primarily on top

of the canopy in black mangrove (Avicennia germinans), red

mangrove (Rhizophora mangle), and Brazilian pepper (Schinus

terebinthifolius). (See Lewis and Lewis 1978 for a detailed

description of vegetation on these islands.)

General Procedure

I censused a total of 107 nests from 15 March through 4

August to determine clutch sizes, hatching success and

fledging success. I separated these nests into three groups

according to the censusing methods used. The "focal nest"

group consisted of 27 nests that were continuously observed

with a spotting scope and binoculars from dawn to dusk

during the first 20 days of nestling life (see Focal Nests,

below, for details). The "visual census" group consisted of

51 nests that were adjacent to focal nests but were not

continuously observed or monitored for nestling growth (see

Visual Census nests, below, for details). The "growth nest"

group consisted of 29 nests containing chicks that were

weighed and measured every 4-12 days (see Handling schedule,

below, for details) but were never kept under continuous

observation.

Timing of census initiation. Censusing of 76 nests

began during the incubation period. Clutch sizes for these

nests were known precisely (see Appendix B for clutch-size

determination criteria). Hereafter, for convenience,

clutches that definitely contained two or three eggs will be

represented symbolically as C/2 and C/3, respectively.

Censusing of the 31 remaining nests began after hatching.

Clutch sizes were not known for these nests. The initial

brood size (number of chicks that hatched) was known

precisely for four of these 31 nests and for 65 of the nests

for which clutch sizes were also known. The initial brood

size was considered to be known if the number of nestlings

was first counted within 5 days after the first (A-) chick

hatched. Hereafter, broods that initially contained two or

three chicks (regardless of clutch size) will be represented

as B/2 and B/3, respectively.

Marking. In focal and growth nests, I individually

marked each chick to facilitate distinguishing A-chicks from

second-hatched (B-) and third-hatched (C-) chicks. All

newly hatched chicks aged 0 (day of hatching) to 4 days were

marked according to their hatching order with yellow and

black indelible marker pens. Older chicks were marked with

blue and yellow acrylic paint on their backs and heads, plus

the same color flagging tape squares glued with contact

cement flat onto the back and head. Paint and flagging had

to be reapplied frequently as the combination of water, fish

oils and guano acted as a solvent for both oil- and water-

based paints and glues.

Handling schedule. All chicks in focal and growth

nests were handled regularly to refresh their color-marks,

assess their physical condition, and monitor their growth.

I measured bill length (length of exposed culmen) to the

nearest mm with a clear plastic ruler. I weighed nestlings

with 100 g, 2.5 kg and 5 kg spring scales. All chicks in

focal and growth nests were handled, weighed and measured

every 2-4 days until the brood's A- chick was (on average)

29 days old (range 25-31 days). Thereafter, all members of

the brood were handled a minimum of once a week when the

brood's A-chick was from 30 through 40 days old, and at

least every 10-12 days when A-chicks were 40 through 70 days

old.

Definition of fledging. All nests were censused until

all brood-members had died or until the A-chick reached age

70 days (see Determining Hatching Dates, below, for details

about aging chicks). I discontinued censusing after 70

days. After this age, chicks were difficult to catch

because they made 3-10 m flights between perches in the

vicinity of the nest (unpub. data). These short flights

began once the juvenile plumage had developed. I therefore

operationally defined "fledging" as occurring at 70 days,

the average age by which eight chicks had developed their

brown juvenile plumage (range=57-82 days) in 1989 at

Seahorse Key (see Supplemental Census Nests, below). Pinzon

and Drummond (in press) used the same operational definition

of fledging for the population of brown pelicans that they

studied in Mexico. The first sustained flight was not

achieved until a mean age of 81.3 days (2.14) for seven

chicks at Seahorse Key in 1989. Sustained flight occurred

at an average age of 76 days over the 4 years of Schreiber's

1976 study, and at age 75 days in the chicks observed by

Pinzon and Drummond (in press).

Focal Nests

"Focal nests" were observed from two blinds, one on

Sunken Island and the other on Bird Island. Nests were

located from 1-33 m from the Bird Island blind and from 28-

42 m from the Sunken Island blind. Hatching peaked in focal

nests on Sunken Island about a month earlier than on Bird

Island. Therefore, observers (BJP plus an assistant) were

able to watch focal nests on Sunken Island from hatching

until chicks were at least 20 days old, then move to Bird

Island to watch another set of focal nests beginning at

hatching. We observed focal nests on Sunken Island from 2

April through 9 May and on Bird Island from 11 May through 8

July 1990. Focal nests on both islands were observed

continuously during daylight hours On Sunken Island,

observers alternated approximately 7-hour observation

periods, usually trading off around 1300. But during the 1

week of peak nestling activity, both observers were present

in the same blind throughout each day. On Bird Island,

observers alternated daily dawn to dusk observation periods.

Pelicans nesting on Alafia Banks initiated clutches

from the middle of February through the end of May, 1990.

Thus, on a given day the subcolony under observation might

have contained nests in the incubation through late nestling

stages. On any one day, we closely observed up to 16 focal

nests in a visual arc of 70-800, for a total of 27 focal

nests over the season.

Nests became part of the focal group as soon as their

broods were completed (i.e. no further eggs hatched). To

keep chicks individually marked and to measure nestling

growth, I visited focal nests following the schedule

described earlier (see Handling schedule in General

Procedure, above). In addition to visiting focal nests

periodically, I also kept focal nests under continuous

daylight observation until A-chicks reached the age of 20

days, by which time most younger siblings had already died.

When A-chicks were 20 through 30 days old, focal nests were

simply censused visually at least once each day and visited

every 2-4 days to monitor growth. Focal nests of these ages

were not kept under continuous observation. After A-chicks

reached a minimum of 30 days old, observations from the

blind were discontinued. But I continued to measure and

weigh chicks in these nests every 7-12 days depending on A-

chick age (see General Procedure, Handling schedule, above)

until all brood members had died or reached age 70 days.

Visual Census Nests

Nests in the "visual census" group were observed

opportunistically from the Sunken Island and Bird Island

blinds. In contrast focal nests whose contents were under

continuous observation, the number of eggs and chicks in

visual census nests were counted only 1-3 times per day.

Nestlings in visual census nests were never handled and so

remained unmarked and were not measured for growth.

Measures of Hatching Success

To calculate hatching success in all broods,

independent of brood size, I used the 65 nests for which I

knew both the clutch size and the initial brood size. These

nests included 19 focal, 24 visual census and 22 growth

nests. This calculation of hatching success included nests

in which all chicks failed to hatch.

Last-laid eggs may have value as replacements for older

eggs that fail to hatch. To assess this possibility, I

calculated hatching success a second time, restricting the

analysis to only those nests of known clutch and initial

brood sizes that also hatched at least one chick. Because I

also wanted to assess fledging success in the same sample, I

restricted this analysis further to include only nests in

which I also knew how many chicks fledged. A total of 45

nests met these criteria (17 focal, 7 visual census and 21

growth nests). In four of these nests, all of the chicks

died on the same day because the parents abandoned the brood

within a few days of brood completion. This left 41 broods

that were not abandoned, for which clutch sizes, initial

brood sizes and fledging success were known and in which at

least one chick hatched. These nests were used to compare

hatching failure in 21 C/2 versus 19 C/3 clutches. Because

eggs were not marked, I could not assess the relationship,

if any, between laying order and egg failure. Thus, I could

not determine if last-laid eggs ever replaced earlier-laid

eggs that failed to hatch. But low rates of hatching

failure would suggest that last-laid eggs have little egg-

insurance value, whereas high rates of hatching failure

would suggest the potential for last-laid eggs to replace

senior eggs that failed to hatch.

Measures of Fledging Success

I calculated fledging success per clutch from the 71

nests for which I knew both the clutch size and the number

of chicks that fledged. This sample included 18 focal, 29

visual census and 24 growth nests. These nests were also

used to calculate clutch sizes.

I also calculated fledging success per brood. For this

analysis, I used the 45 nests in which at least one chick

hatched and for which I was able to determine the initial

brood size and number of chicks that fledged..

Determining Hatching Dates

Eggs hatched asynchronously within clutches, with the

A-chick hatching 1.3 (+ 0.6) days (for three B/2) and 1.1 (+

0.7) days (for 16 B/3 broods) before the B-chick. B-chicks

hatched 1.6 ( 0.6) days before their C-siblings (N = 16

broods). These data were from pairs of chicks whose

hatching dates were known precisely. Whenever possible, I

determined the age and hatching order of siblings from

direct knowledge of hatching dates determined during daily

censuses (N = 41). This is the "known-age" group. I used a

growth curve for culmen lengths of 62 known-age chicks (r2 =

0.945, N=204 observations, Appendix A) to estimate the ages

and hatching order of an additional 51 "estimated-age"

chicks that were found in the first 10 days of life but

whose hatching dates were not known precisely. Schreiber

(1976) and Pinzon and Drummond (in press) also found a

correlation between age and culmen length. Chick ages when

first estimated had to be < 10 days to be included in the

estimated-age group, because after this age, culmen lengths

often reflected nutritional condition and would have

produced biased age estimates (Appendix A). In three

additional broods (eight chicks total) less than 10 days

old, culmens were not measured and ages were estimated from

skin colors and plumage development by comparing them to

known-age chicks (Appendix A). I thus determined the ranks

of a total of 100 chicks.

Thirty-two of the 62 known-age chicks that were used in

the regression were from nests that were later manipulated

for experiments (see Chapter 4) that may have affected chick

fates. These 32 nests were omitted from all analyses of

fledging success and causes of mortality.

Reproductive Value of Junior Siblings

I calculated the reproductive value of nestlings as the

total number of chicks that fledged divided by the total

that hatched. When chick ranks were known, I calculated

the reproductive value of chicks of each rank separately. I

partitioned the reproductive value of junior chicks of each

rank (B- and C-) and brood size (B/2 and B/3) into "extra-

chick" (RVe) and insurance (RVi) components (following

methods of Mock and Parker 1986, modified by Beissinger and

Waltman 1991 and Ploger and Mock unpublished MS). Thus, RVe

for B-chicks = the number of B-chicks that fledged along

with their seniors, divided by the total number of B-chicks

that hatched. The RVi for B-chicks = the number of B-chicks

that fledged after replacing a senior that died, divided by

the total number of B-chicks that hatched. Analogous

calculations were made for C-chicks. I could not calculate

RVe and RVi to include eggs that failed to hatch because I

did not know the order in which eggs were laid (see Measures

of Hatching Success, above, for details and discussion of

estimating egg-loss insurance).

Analyses of Fates Based on Chick Ranks

Known-age and estimated-age chicks were pooled for all

analyses of chicks according to rank. The assignment of

chick ranks thus reflect relative sizes of siblings when

less than 10 days old. I was able to determine the initial

brood size and the fates (lived or died of various causes;

see Causes of chick mortality, below) of all brood members

according to their size ranks for 14 B/3 and 20 B/2 broods.

These broods were used to compare the frequencies of food-

independent and food-dependent mortality for chicks of

different ranks (see Causes of Chick Mortality, below). I

also used these broods to analyze the reproductive values of

chicks according to their size ranks. Of the B/3 broods

used in these analyses, nine were focal nests and five were

growth nests. Of the B/2 broods used in these analyses, 11

were focal nests, 8 were growth nests, and 1 was a visual

census nest.

In an additional seven B/3 and 14 B/4 nests, chick

ranks were assigned according to relative sizes and plumage

development when chicks were more than 10 days old. These

"estimated-rank" chicks were used only when pooled with

chicks of known ranks to compare the fledging success among

chick ranks.

Analyses of Fates Independent of Chick Ranks

Restricting analyses to chicks whose ranks were known

would not have adequately represented all of the different

types of food-independent, food-dependent and unknown causes

of death that I observed in 1990 (see Causes of Chick

Mortality, below). For this reason, in addition to using

the 100 chicks whose ranks were known, I also determined the

fates of 72 "unranked" chicks. These "unranked" chicks were

added to the census sample when too old to determine

accurate hatching order and ages. Eight chicks of known

rank and two of the unranked chicks died when their nests

were abandoned. For fates analyses, I omitted these six

nests and all nests that were abandoned before hatching an

egg. I listed the fates of the remaining 177 eggs that

hatched 162 chicks (92 ranked and 70 unranked chicks).

These chicks hatched from the 81 nests for which I was able

to determine the fate of at least one egg or chick.

Causes of Chick Mortality

I separated causes of mortality into deaths due to

food-independent, deaths due to food-dependent and deaths

due to unknown causes. Food-independent deaths included

chicks that "died as hatchlings," "fell accidentally," were

"killed by invader chicks or adults," or died from an

"unknown accident" (see definitions in Food-independent

deaths, below). Food-dependent deaths included victims of

"siblicide," victims of "starvation" and chicks that died of

"starvation &/or siblicide" (see definitions in Food-

dependent deaths, below). "Unknown causes" of mortality

included deaths that could not be classified as either food-

independent or food-dependent.

Food-independent deaths. Food-independent deaths

included nestling deaths that were unlikely to be related to

food conditions within their nest. The most obvious cases

of food-independent deaths were those in which a nestling

was "killed by invader chicks or adults" (also referred to

as "infanticide"). I defined an "invader" to be a chick or

adult that moved its head and/or body into a nest that was

not its own. Invader chicks included recently fledged

young. Invader adults included birds at least 1 year old

that had subadult or adult plumage (see Schreiber et al.

1989 for plumages of subadults versus adults). Invaders

often attacked the chicks whose residence they invaded (see

Appendix C for discussion of why some pelicans invade nests

and attack residents). An "attack" involved one or more

"blows" delivered by one individual (the "attacker") against

another individual (the "victim"). To be counted, "blows"

had to be forceful enough to move the victim's head or neck

when struck. A chick that was "killed by an invader" was

one that (1) was seen being tossed (N = 3) or knocked (N =

1) from the nest by an invader or (2) was attacked by an

invader within 1 (N = 2) to 3 (N = 1) days of the victim's

death if the victim was not attacked by a sibling during

this period.

Another type of food-independent death occurred when a

chick "fell accidently." A chick was classified as dying

when it "fell accidently" only if an observer saw the fall,

the fall was not directly preceded by a sibling's attack,

and the fall involved an alert chick able to crawl or walk.

I also included in this category one chick that died after

its wing became inextricably entangled in branches near its

nest. Chicks that fell accidently did not include extremely

weak chicks that fell or became entangled in branches during

convulsions.

I classified as "unknown accidents" deaths involving

chicks older than 16 days that died at least 3 days after

the death of their siblings, and that continued to gain mass

prior to death. These deaths could not have been food-

dependent because the victims were gaining mass rather than

starving and because there were no siblings in the nest to

cause siblicide. These chicks were old enough to (1) stand

(and potentially fall from the nest) and (2) be left

unattended for hours (and potentially be attacked by

neighbors). Thus, these chicks either died because they

fell accidently or were killed by an invader, but I could

not determine which type of food-independent mortality was

the exact cause of death.

A final category of food-independent deaths included

chicks that "died as hatchlings." I classified a chick as

having "died as a hatchling" if (1) the victim died when 2

days old (N = 5), when chicks still had reserves of yolk,.or

(2) the victim died when more than 2 days but 5 days old

and it was well-fed (N = 1). To be classified as having died

as a hatchling, there also had to be no evidence of fighting

among siblings prior to the victim's death. I considered

the chicks that I classified as having died as hatchlings to

be victims of food-independent deaths because evidence

suggested that these chicks were not attacked by their

siblings and did not experience food shortages before their

deaths. I assumed that these chicks were killed accidently

by parents that crushed chicks or flipped them from the nest

when parents departed suddenly. Several lines of evidence

support this assumption. First, chicks 5 days old were

small enough to die of these causes. Second, I found one

chick that may have been crushed to death. Third, I have

seen hatchlings flipped off a parent's foot-webs to the edge

of the nest when parents moved to a perch (N = 3). The

discovery of live, healthy-looking hatchlings on the ground

(N = 2) suggests that chicks were sometimes flipped

completely out of the nest. Additional reasons for chicks

to have died as hatchlings include deaths due to

developmental abnormalities and improper incubation, which

are both likely to be food-independent sources of mortality.

One additional 3-day old chick was classified as having died

as a hatchling even though I did not know whether or not

this chick had received food. I included this chick because

it vanished from a nest in which the chicks were not brooded

properly because there was a deep pit in the center of the

nest. This chick probably died from exposure during

improper brooding.

Food-dependent deaths. I defined food-dependent deaths

to be nestling deaths that were probably related to

insufficient food being delivered to broods. The most

obviously food-dependent deaths involved chicks that died of

"starvation." I defined a chick's death as caused by

"starvation" if it was "emaciated" when last handled, and it

had not been attacked by another pelican (invader or

sibling) for 4 days preceding its death (N = 11).

"Emaciated" chicks were those which had at least two of the

following characteristics: (1) loose skin on the breast,

abdomen and/or back (areas with considerable tonus in well-

fed individuals); (2) an abdomen that felt flaccid when

palpated, or looked flat or concave when viewed laterally;

(3) very liquid, yellow feces; (4) listlessness and skin

temperature that felt cold to my touch; and (5) failure to

gain mass since my last visit to weigh the chick. Chicks

may have become emaciated and died from endoparasitic

infections or pesticide contamination rather than from

starvation. But endoparasitic infections are probably

lethal only to chicks that are already starving. Similarly,

pesticides stored in fats have greater effects when fats are

mobilized in response to starvation. Mortality from

endoparasitic infections and pesticide contamination were

not separated from starvation, and thus were assumed to be

food-dependent.

Another type of food-dependent death was "siblicide,"

which I defined to include the following cases. First,

siblicide included chicks that were seen being knocked from

the nest by a sibling (N = 2). Second, siblicide included

chicks whose siblings permanently drove their victims out of

the nest into the surrounding branches (N = 2). These

victims, prevented by their siblings from returning to the

nest, wandered around in the canopy until they succumbed to

starvation, exposure and the attacks of neighbors whose

nests they wandered near. Third, siblicide included chicks

whose deaths were not directly observed, but that received

at least 25 blows from their siblings (and none from

invaders) in the last 2 days of life (N = 13). I considered

siblicide to be a food-dependent cause of death because

sibling attacks limited the victim's access to food. The

victim's of repeated sibling attacks sometimes became so

intimidated that they remained in a submissive posture (see

Chapter 3) during an entire bout of feeding activity and so

failed to participate in feeding. Another reason for

considering siblicide to be food-dependent is that the

frequency and intensity of sibling fights may depend on food

supplies in chicks more than a week old (see Chapter 3).

Some deaths were clearly food-dependent but I could not

determine whether the death was caused by (1) starvation in

the absence of nestling aggression, (2) siblicidal attacks,

or (3) the combined effects of nestling aggression and

starvation. These deaths were classified as being due to

"starvation &/or siblicide." To be classified as

"starvation &/or siblicide," deaths had to meet at least one

of the following conditions. First, starvation &/or

siblicide included emaciated chicks that received less than

25 blows in nests where some sibling fighting (and no

invader attacks) occurred in the victim's last 4 days of

life (N = 5). Second, starvation &/or siblicide included

single deaths that occurred more than 5 days and < 16 days

after the A-chicks hatched in nests for which information on

sibling fights and nestling condition was not available for

the last 4 days of nestling life (N = 9). Deaths meeting

the latter criteria were considered to be food-independent

because chicks c 16 days were (1) too young to fall

accidently because they could not yet crawl from the nest,

and (2) still constantly attended by parents and so not

subject to invader attacks. I also classified one death of

a chick < 5 days old as being due to starvation &/or

siblicide because there was evidence of fighting in the nest

prior to the victim's death.

Unknown causes of death. Chicks whose deaths could not

be classified as either food-dependent or food-independent

were considered to have died of "unknown causes of death,"

unless the deaths were due to parental "abandonment." I

assumed that a clutch had been abandoned if eggs were

present on one visit, and damaged or gone on the next. I

assumed that a brood had been abandoned if all the chicks

were found dead or missing on the same day, five or fewer

days after the A-chick had hatched. Abandoned nests were

excluded from all analyses except where otherwise indicated.

Supplemental Census Nests

General procedure. I determined fates for eggs and

unmarked nestlings in 78 nests that I observed from 22 March

through 29 August 1989 on Seahorse Key, part of the Cedar

Keys National Wildlife Refuge, Levy County, Florida. I

censused these nests visually each day from a lighthouse

tower located within 6-10 m of the nests, which were 3-12 m

high in mixed hardwoods. Fates of chicks in nests on

Seahorse Key were determined by scanning the colony through

a spotting scope continuously from dawn to dusk for 4-6 hour

periods, with short breaks between, for a total of 11-16

hour/day.

As in 1990 on Alafia Banks, I classified 1989 deaths on

Seahorse Key as being food-independent, food-dependent, or

involving unknown causes. I defined food-independent

mortality and unknown causes of death using the 1990

criteria. Food-dependent deaths included siblicide (same

criteria as in 1990) and deaths that could have been due to

starvation &/or siblicide. I could not separate starvation

deaths from starvation &/or siblicide because I did not

handle chicks in 1989. The criteria used in 1989 for

placing a death into the starvation &/or siblicide category

were the same as those used in 1990, except that I defined

emaciation differently in 1989 because I did not handle

chicks. In 1989, a chick was classified as "emaciated" if

it got no food for the last 3 days of life and edema (common

in starving chicks) could be seen with a spotting scope. I

also classified as emaciated any chick whose bill was as

short or shorter than half the length of the bill of its

largest sibling. I compared the relative lengths of brood-

members' bills when the profiles of at least two chicks were

visible simultaneously through a spotting scope.

Measures of 1989 hatching success. I used two measures

of hatching success to compare 1989 Seahorse Key broods

versus 1990 Alafia Bank broods. First, I calculated

hatching success in all clutches, regardless of clutch size

and including nests in which all chicks failed to hatch.

For this analysis, I used the 54 Seahorse Key nests for

which I knew both the clutch and initial brood sizes (see

Measures of Hatching Success, above, for 1990 Alafia Bank

sample sizes used in hatching success analyses). Second, I

calculated hatching success in nests that hatched at least

one chick and that were not abandoned in the first days of

nestling life. Thirteen C/2 and 20 C/3 nests with known

clutch and initial brood sizes fit these criteria in the

Seahorse Key sample. These nests were used to compare

hatching failure by clutch size on Seahorse Key and Alafia

Banks.

Measures of 1989 fledging success. I calculated

Seahorse Key clutch sizes and fledging success per clutch

from the 53 nests for which I knew both the clutch size and

the number of chicks that fledged. Fledging success per

brood was calculated from the 36 Seahorse Key nests that

hatched at least one chick and for which I knew both the

initial brood size and number of chicks that fledged (see

Measures of Fledging Success, above, for Alafia Bank sample

sizes used in clutch size and fledging success analyses).

Analyses of Fates in 1989. Because nestlings were not

marked in 1989, I could not positively identify individuals

according to their hatching ranks. 1989 chicks are

therefore usually referred to as "unranked" chicks (but see

following paragraph). I determined the fates of 105 eggs

that hatched 94 chicks from the 44 nests that were not

abandoned and for which I was able to determine the fate of

at least one egg or chick. These nests were used for

analyses comparing fates regardless of chick ranks.

For analyses of fledging success according to chick

ranks, I estimated chick ranks in 16 B/3 and 15 B/2 broods.

These chicks are referred to as having "estimated ranks".

Ranks were initially determined by noting hatching dates and

skin colors of chicks within their first week of life. I

then tracked changes in the body sizes, culmen lengths and

plumage development by visually comparing each nestling to

its siblings every day until all chicks had died or fledged.

Size differences and ranks could not always be determined

until nestlings were more than 10 days old. Therefore, I

could not detect early reversals in dominance or size using

this method of rank estimation.

Estimating Effects of Colony Disturbance

To determine whether our activities in the colony

adversely affected nestling survival in 1990, I used a boat

to count nestlings and nests in subcolonies of the Alafia

Banks colony that faced different levels of researcher

disturbance. I compared subcolonies that contained focal

and growth nests ("highly" disturbed, visited every 2-4 days

during the first 30 days, as described earlier) to

subcolonies that we never entered ("never" disturbed) or

walked past while walking along the beach ("moderately"

disturbed). To determine productivity in each subcolony, I

used Schreiber's (1979) methods, described as follows.

Every 2 weeks throughout the nesting season, I counted the

number of nests and nestlings that were visible from the

boat in each subcolony. At the end of the season, I

determined for each subcolony which of these repeated

censuses had the largest number of nests. This count was

the "maximum number of nests" for each subcolony. I also

determined which of the repeated censuses in a subcolony had

the largest number of nestlings prior to the first

observation of fledging in that subcolony. This was the

"maximum number of nestlings" in a subcolony. I compared

the productivity of each subcolony by comparing each

subcolony's maximum number of nestlings per maximum number

of nests (called "maximum nestlings/maximum nests").

Statistical Analyses

Statistical analyses were performed using Statview SE+

Graphics (Feldman et al. 1988) on a Macintosh SE computer.

All statistical tests were 2-tailed unless otherwise stated.

Row-by-column (R X C) G-tests of independence were used to

compare survival frequencies among chick ranks whenever 20%

of cells had expected frequencies of five or more. Whenever

an R X C G-test was significant, I conducted 2 X 2 G-tests

between pairs of sibling ranks. For these pairwise

comparisons, I kept experimentwise error at P < 0.05 by

using critical values of the chi-square distribution based

on Sidak's multiplicative inequality (Sokal and Rohlf 1981).

When 20% of expected cell frequencies were less than five, I

cast data into 2 X 2 tables and performed Fisher exact tests

using the expanded tables of Finney et al. (1963).

Differences were considered significant when P< 0.05.

P-values are presented for all statistical tests including

those that were nonsignificant, except for nonsignificant

results of Fisher Exact Tests. I could not provide P-values

for nonsignificant Fisher Exact Tests because P-values

higher than 0.05 for 1-tailed and 0.1 for 2-tailed tests

were not given in statistical tables for 2 X 2 contingency

tables (Finney et al. 1963). Means are presented 1 SD.

Results

Survival

Clutch sizes and hatching success were similar in both

years of this study (Table 2-1). Fledging success, although

low in both years, was lowest in 1990 (Table 2-1). The

productivity of 1990 nests, as measured by the maximum

number of nestlings per maximum nests, was significantly

associated with degree of researcher disturbance, with

productivity increasing with disturbance (Table 2-2,

Kruskal-Wallis test H = 5.357, P < 0.03). Thus, researcher

activities were probably not responsible for the low

fledging success in 1990. The productivity values that I

obtained in 1990 fell in the range of those obtained over an

8 year period (0.33-1.79 fledglings per total nest per year)

on nearby islands by Schreiber (1979). That highly

disturbed nests were more productive than less disturbed

nests could have occurred because I was more familiar with

the location and brood sizes of focal nests. Thus, my

counts of total numbers of nestlings and nests might have

been higher (and more accurate) in the highly disturbed

relative to less disturbed subcolonies. In addition, our

activities in the highly disturbed colonies may have

decreased opportunities for nest takeovers and stick thefts,

which often contribute to partial and complete losses of

clutches and broods (see below).

Nest failures. Nest abandonment was a common cause of

nest failure, affecting 24% (19/78) of clutches (nests

containing eggs) and no broods (nests hatching some chicks)

in 1989 and 13% of clutches and 4% (4/107) of broods in the

nests observed in 1990. Clutch abandonment in 1990 was

probably underestimated because I did not begin observations

in this year until most nests were in the last week of the

incubation period. By contrast, I began the 1989

observations when the birds were building the first nests.

These abandonments appeared to be spontaneous, as contrasted

with one additional clutch that was abandoned in 1989, and

five clutches and two broods abandoned in 1990 following

human activities in the colony. An additional 9% (7/78) and

2% (2/107) of clutches were lost in 1989 and 1990,

respectively, when the incubating parent was driven from its

nest by a courting pair or single male that attacked the

resident, tossed out its eggs and took over its nest (a

"takeover," discussed in Appendix D). As with abandonments,

the frequency of clutch takeovers was probably

underestimated in 1990. In 1989, 8% (6/78) of all clutches

were either abandoned or victims of takeovers, but the exact

cause could not be determined.

Survival of chicks according to size ranks. To assess

the value of the youngest brood members as insurance or as

"extra" chicks requires examining the frequency and causes

of single losses of eggs and chicks, as contrasted with the

simultaneous loss of all nestlings due to nest abandonment

and clutch takeovers. I therefore excluded abandonments and

clutch takeovers, and examined the frequency of deaths from

all other causes combined.

In 1990 B/2 nests of chicks with known ranks, I found

that B-chicks died more often than A-chicks, but not

significantly so (Table 2-3, 1-tailed Fisher exact test

against expectation that the youngest do not die more often

than their seniors, P > 0.05 for 4/20 surviving A-chicks

versus 0/20 surviving B-chicks). When I added 11 B/2 broods

with chicks of estimated ranks to these 20 known-rank

broods, I found that B-chicks died significantly more often

than did their A-siblings (1-tailed G-test, G = 4.55, df =

1, P = 0.05 for 8/31 surviving A-chicks versus 2/31

surviving B-chicks). Similarly, in 1989 B/2 broods for

which I estimated chick ranks, B-chicks died significantly

more often than did A-chicks (1-tailed G-test, G = 9.11, df

= 1, P = 0.005 for 10/16 surviving A-chicks versus 2/16

surviving B-chicks).

In 1989 B/3 nests, survival was significantly

associated with chick rank (1-tailed G-test, G = 31.78, df =

2, P < 0.001 for 13/15 surviving A-chicks versus 3/15

surviving B-chicks versus 0/15 surviving C-chicks).

Pairwise comparisons revealed that A-chicks survived

significantly more often than did their B-siblings (G =

14.66, experimentwise error set at P < 0.05, df = 2 for one

of three comparisons). Similarly, A-chicks survived

significantly more often than their C-siblings (G = 29.27,

experimentwise error set at P < 0.05, df = 2 for one of

three comparisons). Survival of B-chicks was not

significantly better than that of C-chicks (G = 4.49,

experimentwise error P > 0.05, df = 2 for one of three

comparisons).

In 1990 B/3 nests, C-chicks died more often than did

their A-siblings, but not significantly so when only chicks

of known rank were included (Table 2-3, 1-tailed Fisher

exact test, P < 0.05). When chicks of estimated ranks were

added to the analysis, this relationship was significant (1-

tailed Fisher exact test, P = 0.05 for 6/20 surviving A-

chicks versus 0/20 surviving C-chicks). A- and B-chicks in

1990 B/3 nests survived with similarly low frequency when I

considered only broods with chicks of known ranks (Table 2-

3). When I added six broods with chicks of estimated ranks,

differences in survival were still not significant (1-tailed

Fisher exact test, P > 0.05 for 6/20 surviving A-chicks

versus 2/20 surviving B-chicks).

The preceding 1990 analyses provide estimates of the

effect of hatching rank on chick survival. But two B-chicks

grew larger than their A-siblings within the first 10 days

of life in 1990 B/3 nests. By gaining size-superiority,

these chicks may have gained a survival advantage. To

evaluate the effect of size-superiority (rather than

hatching rank) on chick survival, I re-classified the larger

B-chicks as "A-"chicks, and the smaller A-chicks as a

"B-"chicks in these two nests. When I compared B/3 chicks

according to their size-superiority at 10 days of age, I

found that "A-"chicks survived significantly more often than

did "B-"chicks (1-tailed Fisher exact test, P = 0.05 for

7/20 surviving "A-"chicks versus 1/20 surviving "B-"chicks).

Such size reversals within the first 10 days did not occur

in 1990 B/2 nests and did not involve any 1990 C-chicks.

Thus, preceding analyses involving 1990 B/2 broods and B/3

C-chicks that included chicks with estimated ranks reflect

the effects of chick size superiority.

Reproductive Value of Junior Siblings

Hatching failure was low in my study, affecting only 8%

to 12% of all eggs from C/2 and C/3 clutches that hatched at

least one egg (Table 2-4). I observed only two C/4 nests,

one in 1989 that hatched three and fledged one chick, and

the other in 1990 that hatched four chicks that died as

nestlings. Because hatching failure was so low, I assessed

mortality of successfully hatched chicks to partition the

reproductive value of junior chicks into "extra-chick" and

insurance components.

Two chicks fledged from 12% of all B/2 nests (N = 16

broods) and 20% of B/3 nests (N = 15 broods) in 1989. Thus,

junior chicks in B/2 and B/3 nests had some "extra-chick"

value in 1989 (Table 2-5). I could not partition the

reproductive values of 1989 chicks of different ranks into

insurance and "extra-chick" components because chicks were

not ranked in that year. No B/3 nests fledged three chicks

in either year. The total reproductive value of chicks

averaged over all ranks was similar for B/2 and B/3 nests

within each year (Table 2-5).

In 1990, the only chicks to survive were those whose

siblings all died (Table 2-5). The total reproductive value

(proportion of survivors of each rank) of A- and B-chicks

from B/3 nests was identical, and about three-fourths that

of A-chicks in B/2 nests (Table 2-5). When I partitioned

the reproductive value of B-chicks in 1990 B/3 broods into

"extra-chick" and insurance components, I found that the

entire reproductive vale of these chicks lay in their

insurance value. Similar analysis for the last-hatched

nestlings in B/2 and B/3 broods was impossible because they

all died.

B-chicks survived better in B/3 than in B/2 nests in

1990 (Table 2-5). But the only B-chicks to survive in B/3

nests in 1990 were those whose A-siblings died. Thus, in

1990 the reproductive value of these chicks lay entirely in

their value as insurance against the demise of their

seniors. B-chicks in B/3 nests replaced their seniors quite

often; 38% of the B-chicks outlived their seniors and two of

these (15% of all B-chicks) fledged (Table 2-5). Most (83%)

of the B-chicks died during the first 24 days of the

nestling period (defined as beginning with the A-chick's

hatching, Figure 2-1). This was the period in which 77% of

the A-chicks died (Figure 2-1). Overall mortality peaked

during the same period for both A- and B-chicks (U = 63.5, P

= 0.40 for 13 ranks-certain A-chicks vs. 12 ranks-certain

B-chicks). Deaths occurred an average of 16.2 ( 12.3) days

into the nestling period for A-chicks and after 22.8 (.

18.9) days for B-chicks

Although all C-chicks died in 1990, 23% of them lived

longer than one of their seniors (Table 2-5). Mortality of

seniors peaked during the same period as that of C-chicks,

averaging 19.5 (. 16.1) and 16.6 ( 10.1) days for seniors

and C-chicks, respectively (U = 275.5, P = 0.80 for 24

seniors vs. 15 C-chicks plus 9 unranked chicks that were

first to die in their nests). In this analysis, unranked

chicks that were first to die were included (Figure 2-1)

because the comparison of interest is whether these chicks

lived long enough to be present during the period of peak

risk to their seniors. By including these unranked first

deaths as C-chicks, the time of death would be

underestimated in the event that some of these first deaths

actually involved seniors.

In B/2 nests, although all of the B-chicks eventually

died in 1990, 15% of them still outlived their seniors, one

of which did not die until it was 69 days old (Table 2-5).

B-chick mortality peaked during the same period as that of

A-chicks (Figure 2-1), with B-chicks dying an average of

21.0 (+ 14.2) days and A-chicks dying 19.4 (. 12.2) days

into the nestling period (U = 102.5, P = 0.80 for 11 A-

chicks and 15 B-chicks plus 5 unranked chicks that were the

first to die, see above discussion for B/3 C-chicks for an

explanation of why unranked chicks were included in this

analysis). Both chicks survived together for a maximum of

36 days into the nestling period of B/2 broods.

For What Causes of Senior Death do Juniors Provide
Insurance?

Juniors replaced A-chicks that died of food-

independent, food-dependent and unknown causes. Of the five

A-chicks that were outlived by their juniors in B/3 nests,

three were survived by both their B- and C-siblings. Two of

these A-chicks died as hatchlings, while the remaining one

died either of starvation and/or siblicide. The two A-

chicks that were survived by the B-chick alone probably fell

from their nests, although the exact causes of death were

not certain, and were classified as unknown. In B/2 nests,

of the three A-chicks that died before their juniors, one

died as a hatchling, one was killed by its B-sibling, and

one died of unknown causes.

Comparisons of Food-dependent and Food-independent Mortality

When chicks of all ranks from broods of all sizes were

considered together, food-dependent causes of death were

more common than food-independent deaths in both 1989 and

1990 (Table 2-6). Overall, siblicide was the most common

source of death from known causes in 1990, affecting 10% of

all 162 chicks observed (Table 2-6). Food-independent

deaths affected 12% of all 94 chicks observed in 1989 and

15% of those observed in 1990 (Table 2-6). Infanticide

(which affected 4% of all chicks observed in both 1989 and

1990, Table 2-6) was the most common of the food-independent

deaths that affected chicks older than hatchlings.

Infanticidal attacks were probably not an artifact of

frequent colony disturbance. While using a boat offshore to

census parts of the island that we never entered, I observed

birds in subadult (N = 4) and adult (N = 1) plumage

attacking downy young, and also saw two adults grappling in

apparent takeover attempts (N = 2).

Both food-dependent and food-independent deaths

occurred among chicks of all ranks in 1990 (Table 2-3). But

the relative importance of these sources of mortality varied

with chick ranks and brood sizes. A-chicks in B/2 and B/3

nests died of food-independent and food-dependent causes

with similar frequency (Table 2-3, Binomial test for four

food-dependent versus four food-independent A-chick deaths

in B/2 nests, P = 1.0, and Binomial test for two food-

dependent versus six food-independent A-chick deaths in B/3

nests, P = 0.40). Food-dependent and food-independent

deaths also affected a similar number of B-chicks in B/3

nests (Table 2-3, Binomial test for three food-dependent

versus two food-independent B-chick deaths, P = 1.0). By

contrast, in B/2 nests, B-chicks died significantly more

often of food-dependent than of food-independent causes

(Table 2-3, Binomial test for 10 food-dependent versus one

food-independent death, P = 0.01). C-chicks also died more

often of food-dependent causes, but not significantly so

(Table 2-3, Binomial test for six food-dependent versus two

food-independent C-chick deaths, P = 0.30), although the

difference was significant when I omitted C-chicks that died

as hatchlings (Table 2-3, Binomial test for six food-

dependent versus no food-dependent C-chick deaths, P =

0.03).

Timing of Food-dependent Deaths

In 1990, most starvation and siblicide of junior chicks

(B and C chicks in B/3 nests and B-chicks in B/2 nests)

occurred when these chicks were 10 to 20 days old (Figure 2-

1). The mean age of food-dependent deaths was 16.6 (+ 2.5)

days for eight B-chicks in B/2 nests, 18.0 (. 2.6) days for

three B-chicks in B/3 nests, and 13.3 ( 7.6) for eight C-

chicks (including two unranked chicks that were first to

die). Last-hatched chicks in B/2 and B/3 nests died at

similar ages from food-dependent causes (U = 20, ni = 8 C-

chicks, n2 = 8 B-chicks from B/2 nests, P = 0.20).

In the few 1990 broods in which two chicks died of

food-dependent causes, juniors died an average of 5.4 days

( 7.5, N = 3 C-chicks in B/3 broods and 4 B-chicks in B/2

broods) before their seniors (Wilcoxon signed-rank test, 5/6

juniors died before their senior siblings, P = 0.07). No

B/3 brood lost all members to food-dependent causes in 1990.

Siblicides in 1990 occurred an average of 19.4 (+ 9.4)

days into the nestling period, whereas starvation deaths

were delayed until 23.7 (. 8.9) days into this period

(Figure 2-1). These differences were not significant (U =

11.5, P = 0.10, nI = 5 siblicides and n2 = 9 starvation

deaths of chicks of all ranks whose fates were known, from

B/2 and B/3 broods).

Discussion

This study provides a mix of evidence for and against

both the resource-tracking and insurance hypotheses.

Starvation and siblicide were common and concentrated on the

last-hatched brood-members, as predicted by the resource-

tracking hypothesis (Lack 1954, O'Connor 1978, Mock 1984a,

Drummond 1987, Magrath 1990). As predicted by the insurance

hypothesis (Dorward 1962, Stinson 1979, Nisbet and Cohen

1975, Mock and Parker 1986, Bryant and Tatner 1990), seniors

faced a high risk of dying from food-independent causes

including hatching failure, accidental deaths and

infanticide. Furthermore, both B- and C-chicks frequently

replaced seniors that died from such causes. Survival was

associated with nestling size-ranks, as predicted by both

hypotheses. But this association was not significant when I

restricted analyses to chicks of known hatching ranks (see

Survival, below).

B-chicks provided only insurance value in 1990, but in

1989, some B-chicks in B/2 and B/3 nests survived along with

their siblings. Thus, the reproductive value of B-chicks

was divided between their value as "extra" chicks and as

insurance against senior death. No C-chicks survived.

Thus, their potential for having "extra" chick and insurance

value could not be evaluated in this study.

Survival

Both the insurance and the resource-tracking hypotheses

predict that nestling survival will be higher for first-

hatched than for last-hatched siblings. As predicted, A-

chicks survived more often than did B-chicks in B/2 broods

in both 1989 and 1990, at least when I included chicks with

estimated ranks. In B/3 broods, A-chicks survived more

frequently than their C-siblings in both 1989 and 1990 when

I included chicks with estimated ranks. A-chicks in B/3

broods survived more frequently than their B-siblings only

in 1989. Inclusion of chicks with estimated ranks may have

caused me to miss some early size reversals and thus to

underestimate B-chick survival relative to that of A-chicks.

But restricting my 1990 analysis to chicks of known ranks

may have underestimated A-chick survival relative to that of

B- and C-chicks. This is because mortality was so high for

chicks of all ranks in 1990 that sample sizes may have been

too small to reject the null even if A-chicks did survive

more frequently than their juniors in the Alafia banks

population. Because of small sample sizes for chicks of

known hatching ranks, the importance of hatching rank to

chick survival was not clear from my study.

But chicks that gained size superiority over their

siblings clearly gained a survival advantage. Evidence for

this was provided by analyses that included chicks of

estimated ranks because these chick ranks were based on size

differences after the first 10 days of life. Thus, largest

("A-") chicks survived more frequently than did their

smaller ("B-") siblings in 1989 B/3 broods and in B/2 broods

in both years. These "A-" chicks also survived more often

than their smallest ("C-") siblings in both years. In 1990,

"A-"chicks also survived more often than their "B-" siblings

in B/3 broods when I re-classified chicks according to their

relative sizes at age 10 days in two nests where the B-

chicks grew larger than their A-siblings.

Results in the literature also indicate that survival

is associated with chick size and/or hatching order.

Schreiber (1976) reported that all A-chicks survived in the

4 years of his study, whereas B-chick survival varied among

years and C-chicks virtually always died. Similar annual

variation is likely in my Alafia banks site, which was

located about 30 miles from Schreiber's (1976) study colony.

Pinzon and Drummond (in press) also found that nestling

survival was highest among A-chicks and lowest among C-

chicks in a Mexican population of brown pelicans.

Partitioning the Reproductive Value of Junior Chicks

Assuming hatching success was independent of laying

order, as in white pelicans (O'Malley and Evans 1980), only

8-12% of A-eggs in C/2 and C/3 nests failed to hatch in 1989

and 1990 (Table 2-4). Clearly, some "extra" eggs had value

in replacing unhatched clutch-mates. But the replacement

value of junior chicks was primarily for loss of senior

chicks rather than eggs. Cash and Evans (1986) came to the

same conclusion for white pelicans, which had a higher

hatching failure rate of 15-18%. In contrast, Pinzon and

Drummond (in press) concluded that the primary advantage of

laying second and third eggs in their brown pelican

population was as insurance for hatching failure, which

affected 15-32% of the eggs. But junior eggs also replaced

7% of all A-chicks in B/2 and B/3 nests (Pinzon and Drummond

in press). Pinzon and Drummond's estimates of hatching

failure slightly overestimate the values for which juniors

could serve as insurance, because they included cases of

total clutch loss and eggs lost due to researcher

disturbance. Their higher hatching failure rates may be

because some of their nests were on the ground. Ground

nests may experience greater overheating and consequent

partial clutch loss than occurs in tree nests (Anderson

1990). The relative importance of junior chicks as

insurance for egg loss rather than chick loss may be greater

in ground-nesting than tree-nesting populations of brown

pelicans.

The pattern of B-chick deaths in B/3 broods was

consistent with their having value both as insurance and as

"extra-chicks." Some B-chicks replaced A-chicks that died,

so serving as insurance for A-chick death. In addition,

most B-chicks lived into the period when A-chicks faced a

high probability of dying, suggesting that they were at

least available as potential replacements in the event of A-

chick death. B-chicks clearly had value as "extra" chicks

in 1989 when 36% of B/3 broods fledged two chicks (Table 2-

5). Although no B-chicks provided "extra-chick" value in

1990, one B-chick almost fledged with its A-sibling in that

year (Figure 2-1). These data suggest that there may be

considerable variation the "extra-chick" value of B-chicks

among years and colonies. Schreiber's (1976) data also show

high annual variance in the survival of B-chicks when their

A-siblings also lived; the frequency with which both A- and

B-chicks survived ranged from 0-100% of B/3 broods per year,

with 11 B-chicks surviving out of the total of 16 broods

censused in the 4 years of his study. B-chicks in

Schreiber's (1976) study apparently provided no insurance

value; all A-chicks fledged.

No C-chicks fledged in this study. Thus, the C-chicks

in my study provided no reproductive value to their parents.

This is probably typical for C-chicks in most populations.

For example, Schreiber observed broods fledging all three

young in only 6% of the 16 B/3 nests that he monitored for

growth from 1969-1972 (Schreiber 1976) and in only 4% of all

nests that he observed from 1969-1976 (Schreiber 1979).

Three-chick survival was also low in a North Carolina

colony, where only 13% of 38 B/3 broods fledged all young

(M. Shields, pers. comm.). Similarly, in Mexico none of 11

B/3 nests fledged three chicks (Pinzon and Drummond in

press). When combined with my data, these results suggest

that C-chick survival is rare, but that C-chicks do

occasionally provide "extra-chick" value by surviving along

with their siblings in some years. Brood reduction

generally fitted an obligate pattern for C-chicks. By

contrast, in one Panamanian colony, C-chicks survived along

with their siblings in 19% of all B/3 nests observed

(Montgomery and Martinez 1984). This population fed on fish

that were predictably abundant because of upwelling that

reliably occurred during the breeding season. C-chicks in

this population may have "extra-chick" value more frequently

than in the other populations discussed, which rely on less

predictable food supplies.

Whether C-chicks ever provide insurance value awaits

future study. No C-chicks survived as replacements for dead

A-chicks in the populations studied by Schreiber in 1976 and

by Pinzon and Drummond (in press). But the remaining

studies mentioned in the preceding paragraph did not provide

sufficient information to determine whether C-chicks ever

survived as replacements for A-chicks that died. The timing

of C-chick deaths in my study suggests that they could have

served as insurance against the death of a senior (A- or B-)

sibling. In general, when juniors provide "insurance,"

seniors should refrain from killing their siblings and

parents should avoid starving their youngest offspring until

after the survival of the eldest seems secure (Mock et al.

1990). Elimination of the youngest chick should quickly

follow after this, because food consumed by the doomed

youngest would be wasted and the costs of killing it may

increase as the youngest grows larger (Mock et al. 1990).

These patterns were observed for C-chicks in B/3 broods,

with most C-chicks living into the period of peak senior

mortality 17-20 days after the A-chick hatched, and dying

quickly of starvation or siblicide if their seniors survived

this risk period. That some C-chicks outlived their A-

siblings before dying further suggests that C-chicks may

have potential insurance value as replacements for lost

seniors.

Alternatively, C-chicks may not make a significant

contribution to their parents' lifetime reproductive

success, either as "extra" or as insurance chicks. The

presence of C-chicks may reflect selection pressures in the

recent past (e.g. Boag and Grant 1981) rather than current

adaptive value. For example, C-chicks may have insurance

value in ground-nesting populations where hatching failure

due to overheating may be more common (Anderson 1990). C-

chicks may no longer have this value in tree-nesting

populations such as Schreiber (1976) and I studied. The

production of C-eggs could also be a recent response to

pesticide-induced egg-failures (see below).

In the B/2 broods that I studied, the youngest chick

had value as an "extra" survivor, with both chicks living in

25% of B/2 nests in 1989 (Table 2-5). Similarly, Schreiber

(1976) found that both survived in 62% of the 21 two-chick

broods that he monitored in his 4 year study, and Pinzon

and Drummond (in press) observed both chicks surviving in

19% of the 16 broods that they observed. B-chicks in B/2

nests may also have insurance value in some years, as

suggested by my 1990 data showing that some B-chicks lived

longer than their A-siblings in 1990 and many lived into the

period of peak A-chick mortality. Indeed, one B-chick

fledged after its A-sibling died in the 16 B/2 nests that

Pinzon and Drummond (in press) observed.

To properly assess the value of junior chicks as

insurance versus "extra" chicks will require long-term study

of a single population. The relative frequency with which

junior chicks survived as "extra" versus insurance chicks

could be totalled over each parent's lifetime. This could

lead to insight into the relative importance of these two

types of reproductive value. Similarly, the total value of

junior chicks could be evaluated by examining how much A-

versus B- and C-chicks contribute to the lifetime

reproductive success of their parents. If junior chicks

make a negligible contribution to the lifetime reproductive

success of their parents, then this would suggest that the

resource-tracking and insurance hypotheses do not explain

the production of junior chicks in this species.

For What Causes of Senior Death do Juniors Provide
Insurance?

The low incidence of hatching failure, the coincidence

of the peak periods of mortality for all chick ranks and the

delay of this mortality until about a week after the

hatchling period (Figure 2-1) suggests that in both brood

sizes, B- and C-chicks served as insurance for seniors dying

of causes other than egg or hatchling failure. Thus,

juniors probably served primarily as replacements of seniors

dying of infanticide, accidental deaths or food-dependent

causes. I saw no evidence of predation in this study.

I had expected that the insurance value of junior

chicks would be restricted to replacing seniors that died of

food-independent causes, because the size-based competitive

superiority of seniors should ensure that their younger

siblings would precede them in succumbing to starvation or

siblicide during food shortages. But the similarity in the

timing of food-dependent deaths of C-chicks and their

seniors suggests that C-chicks had at least some potential

for replacing a senior that died of such causes. Indeed, in

the few cases where juniors actually outlived a senior,

although most replaced A-chicks that died as hatchlings or

of unknown causes, one B- and C-chick in a B/3 brood and one

B-chick in a B/2 brood replaced an A-chick that died of

food-dependent causes. This suggests that juniors may

sometimes serve as insurance against a defective (Mock

1984a, Mock and Parker 1986, Mock et. al. 1990) or

competitively inferior (Cash and Evans 1986, Drummond 1986)

senior by dominating an inferior senior and reversing the

usual feeding advantages so that the subordinated senior

would starve or be killed by its dominant junior in the

event of a food shortage. The subordination of an A-chick

by its B-sibling could also occur if the B-chick was

unusually well-developed (a "superchick," Mock 1984a).

Dominance reversals with losers dying of food-dependent

causes also would be expected if the production and

subsequent elimination of "extra" offspring is a parental

ploy to selectively raise the fittest genotypes (Buchholz

1922, called "progeny choice" by Kozlowski and Stearns 1989)

by allowing siblings to eliminate inferior competitors

(Simmons 1988). In synchronously hatching species, this

could allow parents to sort out the "best" among offspring

with relatively similar competitive abilities, although at

potentially high costs in energy (Mock and Ploger 1987) and

risk to surviving offspring. In asynchronously hatching

species, the progeny-choice hypothesis is effectively the

same as producing an "extra" chick as insurance for an

inferior senior, because reversals are likely only if a

junior is sufficiently superior that it is able to overcome

its age and size disadvantage. Thus, the progeny-choice

hypothesis becomes a special case of the insurance

hypothesis when hatching is asynchronous. Alternatively,

even potentially superior seniors may die of food-dependent

causes in the first few days after hatching if there is a

temporary food shortage during this period when young chicks

are vulnerable to starvation after using up their yolk

reserves. A junior may provide insurance for early senior

starvation (Forbes 1990) if the food shortage that kills the

senior occurs when the junior is still an egg (Nelson

1978:895) or hatchling with sufficient yolk reserves to

survive the brief shortage. In brown pelican broods, the 0-

2 day intervals between hatching of A- and B-chicks is too

short to facilitate this type of insurance, although the

hatching of C-chicks up to 5 days after their A-siblings

suggests that C-chicks could potentially provide such

insurance at least for A-chicks.

In other populations of brown pelicans, juniors may

have insurance value that covers other causes of senior

mortality. For example, predation may be a significant

source of partial brood loss in some brown pelican

populations. Most reports of nest predation against brown

pelicans involve abandoned eggs or nestlings taken by avian

predators, including fish crows, Corvus ossifragus

(Schreiber and Risebrough 1972, Ploger pers. obs.), common

ravens, a. corax, western gulls, Larus occidentalis (Keith

1978) and Heerman's gulls, L. heermanni (Keith 1978, Pinzon

and Drummond in press). But western gulls and ravens also

take eggs from nests while a parent is in attendance (Keith

1978), and black vultures (Corafvps atratus) take downy

nestlings that are left unattended for extended periods (C.

Murcia pers. comm.), as is usual among these young which are

old enough to thermoregulate (Bartholomew and Dawson 1954).

Predation by vultures could cause partial brood loss,

although vultures usually took the entire broods of very

young chicks (C. Murcia pers. comm.). The only predation of

brown pelican eggs that I observed occurred when parents

temporarily abandoned a nest to bathe and were absent from

the nest for only 1-2 minutes. This usually resulted in

fish crows taking the entire clutch, but sometimes only one

egg was killed before the parent returned, and the remaining

eggs survived (Ploger unpub. data).

Brown pelican nestlings are also subject to tick

infestations (King et al. 1977a, b, Keith 1978, M. Shields

pers. comm.), endoparasitic infections (Courtney and

Forrester 1974) and death from exposure to temperature

extremes (Keith 1978). Heavy tick infestations usually lead

to total brood failure from parental abandonment, which may

occur throughout an entire colony (King et al. 1977a) or

sub-colony (M. Shields pers. comm.). Thus, tick infestation

is not a likely source of mortality for which youngest

chicks could serve as insurance. The value of juniors as

replacements for seniors with endoparasitic infections is

likely to be indistinguishable from their value as

replacements for apparently inferior, starving seniors.

This is because endoparasitic infections are probably lethal

only to starving chicks, whose exact cause of death (from

infection, starvation or their combined effects) can not be

determined. Temperature extremes are most likely to lead to

death of junior rather than senior nestlings, particularly

during the period when nestlings are first left unattended,

when seniors are able to thermoregulate but juniors may

still need protection from temperature extremes. If the

youngest brown pelican brood members are insurance for

losses of senior chicks, they are probably insurance for

losses due to senior inferiority, infanticide and/or

predation.

Partial clutch loss could also be due to egg shell

thinning. I saw no evidence of shell thinning in the 2

years of my study. However, the presence of "extra" eggs

may be due to recent past selective pressures (e.g. Boag and

Grant 1981) rather than current selection acting to maintain

larger clutch sizes. If there were within nest variance in

shell thinning due to pesticide contamination, then the

"extra" egg might have had value as replacement for a thin-

shelled sibling. This replacement value might have been

important in the 1960's and 1970's when, prior to its ban,

DDT contamination was common (reviewed by Anderson and Gress

1983). This is not a likely explanation of my results

because C/3 clutches have been the norm in Florida since the

1920's (Bent 1922). Furthermore, when examined in 1969 and

1970, Florida eggs experienced only moderate levels of

pesticide contamination and egg shell thinning, with only 0-

3% of all eggs breaking during incubation (Schreiber and

Risebrough 1972).

Brood Reduction in Pelican Species

The occurrence of facultative brood reduction of B-

chicks along with obligate C-chick deaths in broods of brown

pelicans contrasts with the pattern seen in most other

pelican species. The pelican species for which the most

complete information is available all have modal clutch

sizes of two eggs and are obliaatelv siblicidal: American

white pelican, Pelecanus ervthrorhvnchos (Johnson and Sloan

1978, Knopf 1981, Cash and Evans 1986), white pelican,

Pelecanus onocrotalus (Vesey-Fitzgerald 1957, Cooper 1980),

and pink-backed pelican Pelecanus rufescens (Din and

Eltringham 1974). Drummond (1987) speculated that in

Australian pelicans, Pelecanus conspicillatus, which also

usually lay two-egg clutches (Vestjens 1977), brood

reduction might be facultative, but further information is

needed for this species. Virtually nothing is known about

whether partial brood loss is obligate or facultative in the

initially three-chick broods of Philippine pelican

(Pelecanus philiDDensis, Neelakantan 1949) and the Dalmatian

pelican (Pelecanus crisnus, Dementiev and Gladkov 1966,

Cramp and Simmons 1977, Crivelli and Vizi 1981).

Death occurs in the first week of life in most

obligately siblicidal species (Mock et al. 1990). But the

timing of obligate siblicide in pelican species varies

considerably, occurring within the first 3 days in white

pelicans (Cooper 1980), ranging from the first 3-10 days

(Johnson and Sloan 1978, Cash and Evans 1986) through the

first 3 weeks in American white pelicans (Knopf 1981), and

peaking in the 8-9th week in pink-backed pelicans (Din and

Eltringham 1974). The apparently obligate deaths of brown

pelican C-chicks occurred at an intermediate age relative to

those of their congeners, peaking in the second week of life

in this study and in the third week in a Mexican population

(Pinzon and Drummond in press). These differences in the

timing of obligate brood reduction could arise if the

youngest chicks serve as insurance for different causes of

senior mortality that come into effect at different points

in the nestling period. The rapid siblicide of white

pelican B-chicks suggests that they serve primarily as

insurance for egg loss or death as hatchlings of inferior A-

chicks. In contrast, the occurrence of siblicide later in

the nestling period may be because junior chicks are

insurance primarily for predation (Nisbet 1975, Nisbet and

Cohen 1975, Mock and Parker 1986, Bryant and Tatner 1990),

accidental deaths and infanticide (this study) or other

deaths that peak after the first week of nestling life.

The role of the youngest chick as insurance or as an

"extra" chick has been evaluated only for American white

pelicans, whose last-hatched chicks serve primarily as

insurance against hatching failure and early death of the A-

chick (Cash and Evans 1986). For other species, the

division of reproductive value into "extra-chick" and

insurance elements awaits further study. Because Dalmatian

and Philippine pelicans may have brood sizes similar to

brown pelicans, the question arises of whether junior chicks

have both insurance and "extra" chick value, as apparently

is the case for brown pelican B-chicks in B/3 broods.

Direct experimental tests of the insurance value of junior

chicks have been carried out only for white pelicans (Cash

and Evans 1986). Similar experiments on brown pelicans and

other species with C/3 clutches are necessary to clarify the

value of C-chicks.

Table 2-1. Comparison of clutch sizes, hatching success and
fledging success (mean + SD) of brown pelican nests at Seahorse
Key in 1989 and at Alafia Banks in 1990.

1989 1990

Mean N Mean Na t P

Clutch size 2.4+0.7 53 2.5+0.6 71

Hatchlings/clutch 1.6+1.2 54 1.7+1.2 65 -0.305 --

Fledglings/clutch 0.6+0.7 53 0.3+0.4 71 3.488 <0.001

Fledglings/brood 0.9+0.6 36 0.30.5 45 4.467 <0.001

aSamples included
or hatching. See

nests that were abandoned during incubation
text for discussion of nest abandonment.

Table 2-2. Brown pelican productivity (maximum
nestlings/maximum nests) in subcolonies that were highly
(High), moderately (Mod.) or never (None) disturbed by
researcher activities.

Disturbance Maximum Max. nestlings

Site level nests Max. nests

Bird Island focal nests High 34 0.90

Sunken Island focal nests High 38 0.87

Sunken Island growth nests High 105 0.63

Bird Island Cove Mod. 112 0.39

Sunken Island South Mod. 135 0.38

Sunken Island Extension None 106 0.32

Sunken Island North None 223 0.37

Note: see Methods for definitions
nestlings/maximum nests.

of maximum nests and maximum

67

Table 2-3. Fates of hatchlings of known ranks from two- and three-chick
broods in 1990.

Chicks in
Three-chick broods

Chicks in
Two-chick broods

Chick Rank

Lived

Food-Dependent Deaths

Starvation

Siblicide

Starvation &/or Siblicide

Total Food-Dependent Death

2(14%) 2(14%) 0 (0%)

1 (7%)

0 (0%)

1 (7%)

2(14%)

1 (7%)

2(14%)

0 (0%)

3(21%)

3(21%)

1 (7%)

2(14%)

6(43%)

4(20%) 0 (0%)

2(10%)

1 (5%)

1 (5%)

4(20%)

1 (5%)

4(20%)

5(25%)

10(50%)

Food-Independent Deaths

Died as Hatchling

Fell Accidently

Killed by Invader Chick

Killed by Invader Adult

Unknown Accident

Total Food-Independent Death

Unknown Causes of Death

2(14%)

0 (0%)

0 (0%)

3(21%)

1 (7%)

6(43%)

4(29%)

14

0 (0%)

0 (0%)

0 (0%)

1 (7%)

1 (7%)

2(14%)

7(50%)

14

2(14%)

0 (0%)

0 (0%)

0 (0%)

0 (0%)

2(14%)

6(43%)

14

1 (5%)

1 (5%)

1 (5%)

0 (0%)

1 (5%)

4(20%)

8(40%)

20

0 (0%)

1 (5%)

0 (0%)

0 (0%)

0 (0%)

1 (5%)

9(45%)

20

Note: Abandoned nests are not included in samples presented in this and
all remaining tables and figures in this chapter.

Table 2-4. Hatching success and chick survival (to 70 days) in
nests for which clutch and brood sizes and fledging success
could be determined. Includes only nests hatching at least one
chick.

Fledglings

Clutch Total % eggs per

size clutches hatched clutch

a. 1989

C/2 13 92% 0.69

C/3 20 88%a 1.00

b. 1990

C/2 21 90% 0.29

C/3 19 89%a 0.42

aIncludes one brood in which only one egg hatched

Table 2-5. Brown pelican nestling survival and reproductive values (RVe =
"extra-chick" component, RVi = "insurance" component, Total RV = RVe + RVi).
Junior chicks that fledged along with their seniors were called "extra"
chicks. Junior chicks that "replaced" a senior chick were those that lived
longer than a senior sibling. N = total number of chicks in each brood-size
and chick-rank category that fledged or died.

Number of Number of chicks Total

Initial "extra" that replaced a chicks

brood Chick chicks that senior chick and: that Total

size rank fledged fledged died fledged N RVe RVi RV

a. 1989

B/2 Ba 2 ? ? 2 16 0.12 ? ?

B/2 Combined -- -- -- 12 32 -- -- 0.38

B/3 B or Cb 3 ? ? 3 15 0.20 ? ?

B/3 Combined -- -- -- 16 45 -- -- 0.36

b. 1990

B/2 A -- -- -- 4 20 -- -- 0.20

B/2 B 0 0 3c 0 20 0 0 0

B/2 Combined -- -- -- 4 40 -- -- 0.10

B/3 A -- -- -- 2 14 -- -- 0.14

B/3 B 0 2 3d 2 14 0 0.14 0.14

B/3 C 0 0 3d 0 14 0 0 0

B/3 Combined -- -- -- 4 42 -- -- 0.10

Note: RVe = (the number of "extra" chicks of a particular junior rank (B or
C) / N chicks of that rank). RVi = (the number of juniors of a particular
rank that fledged after replacing a senior that died / N chicks of that
rank). Total RV = (the number of juniors of a particular rank that
fledged / N chicks of that rank).
Chicks were not marked in 1989 so data are presented only for two nests that
fledged both chicks and therefore had surviving B-chicks.
b
Data are presented only for three nests that fledged two chicks and
therefore had surviving B-chicks (or C-chicks if there were reversals).
c
One of these chicks died when 69 days old.
d
In one reversal, the B- and C-chicks may have died on the same day, 3-5
days after the A-chick died.

70

Table 2-6. Fates of eggs and chicks in all nests in which at
least one chick hatched, including nests with unknown chick
ranks, clutches and/or brood sizes. These eggs and chicks came
from 44 nests in 1989 and 81 nests in 1990.

Lived

Never Hatched

Food-Dependent Deaths:
Starvation

Siblicide

Starvation &/or Siblicide

Total Food-Dependent Deaths

Food-Independent Deaths:

Died as Hatchling

Fell Accidently

Infanticide

Killed by Invader Chick

Killed by Invader Adult

Unknown Accident

Total Food-Independent Deaths

Deaths from Unknown Causes

2

-6

105

aStarvation deaths could not be separated from Starvation &/or
Siblicide in 1989 because chicks were not handled in 1989.

177

Figure 2-1. Frequency of mortality through the nestling
period (defined as beginning with hatching of the A-chick)
for A-, B- and C-chicks in B/3, and A- and B-chicks in B/2
broods in 1990. The graphs for C-chicks and B/2 B-chicks
include chicks of these ranks, plus unranked 1990 chicks
that were the first to die in B/3 and B/2 broods,
respectively. Only chicks from broods hatching all eggs
were included. Numbers above bars indicate for each time
period the number of A-chicks that died before their juniors
(A-chick graphs), or the number of B-chicks or C-chicks that
died after a senior sibling (B- and C-chick graphs,
respectively).

B/3 Nests

A-chicks
2

22 1

IL...1......

A-chicks

S.n. .in.i.n.

2 7 1217

222732 374247 52576267

1

BB-chicks
1
1

1

2 7 121722273237424752576267

Days in nestling period

* Food-independent deaths
Q Food-dependent deaths
O All deaths

B/2 Nests

4

2

0

6

=4
C"
0-
(U
LL

B-chicks

1

1 1 Jl
.. . H. n. -

2

11

C-chicks

1j
1 1

I I I. .n I .

0 19 A---- __

"'

CHAPTER 3
HUNGER AS A PROXIMATE CAUSE OF FIGHTING

Introduction

Fierce fighting among nestlings is common in a variety

of avian taxa (reviews in O'Connor 1978, Stinson 1979, Mock

et al. 1990). In obligatelyy" siblicidal species, death is

the virtually inevitable result of nestling aggression (Mock

et al. 1990). In facultativelyy" siblicidal species, the

lethality of sibling fighting varies and may depend on food

supplies to the brood (Mock et al. 1990).

The ultimate cause of both obligate and facultative

siblicide is presumably food insufficiency (Mock 1984a, Mock

et al. 1987, 1990, Drummond and Garcia Chevelas 1989);

nestlings fight to eliminate competitors when food proves

inadequate for the full brood. Obligate siblicide is

usually explained as a way of eliminating a brood-member

because food is certain to become inadequate for raising the

full brood when nestlings become older (Stinson 1979, Mock

et al. 1990). Facultative siblicide usually is explained as

a form of resource tracking, whereby parents attempt to

match brood size to unpredictable resources by producing an

extra chick that survives if food is abundant but is

eliminated if food becomes scarce (Lack 1947, 1954, called

"Lack's brood reduction hypothesis" by Ricklefs 1968).

Theoretically, unpredictable food supplies ultimately favor

facultative siblicide (Mock et al. 1990) in species where

nestling starvation is brood-size dependent. When such

species face food shortages routinely, selection should lead

to reductions in clutch (rather than brood) size to match

resources. But routine shortages can favor overproduction

followed by obligate siblicide when the designated victim

has some value as a potential replacement for a sibling that

dies (Dorward 1962, reviews in Stinson 1979, Anderson 1990,

see also Chapter 2).

Most species accomplish brood reduction without overt

nestling aggression (see reviews in Howe 1976, Clark and

Wilson 1981). Several conditions must be met for the

evolution of siblicidal aggression to be favored. First,

nestlings must possess potentially lethal weaponry. Second,

they must experience spatial confinement that precludes

escape from sibling attacks. Third, nestlings must engage

in competition for food that is provisioned in small units

that can be defended easily through aggression (Mock et al.

1990). Siblicidal species are also characterized by

competitive disparities among siblings. These disparities,

which are usually initiated by hatching asynchrony, may

function to reduce siblicidal aggression (Drummond and

Garcia Chavelas 1989, Hahn 1981, Fujioka 1985, Mock and

Ploger 1987, but see Hussell 1972, Clark and Wilson 1981;

Magrath 1990 reviewed causes of hatching asynchrony). Large

size disparities, however, may be a consequence rather than

cause of nestling aggression (Mock et al. 1990).

The degree of nestling aggression varies among broods

within and between populations of facultatively siblicidal

species (Mock et al. 1990). The degree of within-brood

aggression can not be predicted by simply determining that a

species possesses all of the attributes (e.g. weaponry,

monopolizably delivered food) that favor siblicide. To

explain the variance in nestling aggression requires

examination of its proximate causes.

One frequently invoked hypothesis is that hunger is the

proximate mechanism that triggers fighting ("food-amount

hypothesis," Ingram 1959, Lack 1966, Procter 1975, Nelson

1978:565, Poole 1979, 1982, Braun and Hunt 1983, Fujioka

1985). This intuitive hypothesis derives from the

hypothesis that food limitation is the ultimate selective

pressure favoring siblicide (Drummond and Garcia Chavelas

1989, Mock et al. 1987, 1990). Most evidence for the food-

amount hypothesis is correlational. This evidence includes

(1) a temporal association between fighting and meals, (2) a

disinclination of recently fed chicks to attack, and (3) an

association between junior chick death with reduced parental

feeding rates during protracted inclement weather (see

review in Mock et al. 1987). There is also an inverse

relationship between parental feeding rates and sibling

aggression in oystercatchers (Haematopus ostraleaus, Safriel

1981), ospreys (Pandion haliaetus, Poole 1979, 1982,

summarized in Mock et al. 1990) and some other raptors

(Newton 1977).

The best evidence that hunger triggers fighting comes

from an experimental study of blue-footed boobies (Sula

nebouxii, Drummond and Garcia Chavelas 1989). Brood

reduction in this species usually occurs shortly after the

senior chick's mass drops about 20% below that expected at

its current age in a good year (Drummond et al. 1986).

Senior nestlings whose necks were taped to prevent

swallowing pecked their siblings over three times more often

than before taping or after tapes were removed (Drummond and

Garcia Chavelas 1989). Rates of such aggressive pecking

rose most steeply when senior mass dropped to 20% below

potential. Similarly, experimental food deprivation also

seemed to cause elevated fighting among south polar skua

chicks (Catharacta maccormicki, Procter 1975), although

design problems caused inconclusive results. In the only

other experimental test of the food-amount hypothesis,

sibling fighting was not correlated with food ingestion in

broods of great egrets (Casmerodius albus, Mock et al.

1987). Provisioned broods in field experiments fought

slightly more than unprovisioned controls, and captive

broods fed high amounts fought more than did broods

receiving low food allotments. Experimental provisioning of

great blue herons (Ardea herodias) foster-parented by great

egrets also failed to depress fighting rates relative to

unprovisioned broods (Mock 1984b). In addition, field

observations of great egrets, great blue herons and cattle

egrets (Bubulcus ibis) failed to show increased aggression

with decreased food (results of various studies summarized

in Mock et al. 1987).

Although fighting occurred independently of food

amounts in these ardeid species, mortality, including

siblicide, was correlated with food shortage (Mock et al.

1987). A possible proximate mechanism to explain this

correlation in the absence of food-dependent fighting is

that the victim becomes more vulnerable to aggression during

food shortages even though levels of aggression are

invariant (Mock 1984b, Mock et al. 1990, Drummond and Garcia

Chavelas 1989). There are two ways that victim

vulnerability could be enhanced during food shortages.

First, parents may be absent more often on foraging trips

and thus, may rarely be able to suppress (fortuitously or

deliberately) nestling aggression (Newton 1977). Second,

competitive disparities among chicks may be exacerbated

during food scarcity, leading to malnourishment of the

younger one which thus succumbs more easily to the physical

abuse (Spellerberg 1971, Meyburg 1974, Edwards and Collopy

1983, Mock et al. 1987). Food abundance is also unlikely to

exert proximate control on nestling aggression in obligately

siblicidal species. In these species, fatal aggression is

the rule even during periods of food abundance (Mock et al.

1990).

I investigated the food-amount hypothesis as a

proximate explanation for aggression among nestling brown

pelicans (Pelecanus occidentalis). Brown pelicans hatch

their eggs asynchronously. Death is obligate for the last-

hatched members of three-chick broods and is facultative for

second-hatched chicks (Chapter 2). Nestlings frequently

fight (Pinzon and Drummond in press) and these attacks often

contribute to the death of junior brood-members (Chapter 2,

Pinzon and Drummond in press). For the food-amount

hypothesis to be a possible explanation for nestling

aggression in this species, there should be an inverse

relationship between nestling aggression and amount of food

consumed and growth of at least some brood members.

Methods

Study Site

I observed brown pelicans nesting in the canopy of

mangroves and other trees growing on Bird and Sunken

Islands, together known as Alafia Banks, in Hillsborough

Bay, Tampa, Florida (see Chapter 2 for further description

of the study site).

Observation and Censusina Methods

Observations were made with spotting scope and

binoculars from two blinds, one on Bird Island 1-33 m from

observation nests, and the other on Sunken island, 28-42 m

from observation nests. Two observers participated in

continuous daylight vigils on alternate days from 15 March

through 8 July 1990. Both observers were present

simultaneously during the 2 weeks of peak nestling activity.

A daily maximum of 16 focal nests were monitored

simultaneously in a visual arc of 70-800. These focal nests

included broods used in other studies (see Chapters 2 and

4), as well as the 13 nests used in this study. I used only

two-chick broods in this study, including 10 broods from

two-egg clutches and three broods that initially contained

three chicks but were reduced to two chicks by the age when

they were used for the analyses presented in this study.

Because the pelicans initiated clutches from the middle

of February through the end of May, 1990, on any given day

the subcolony under observation often consisted of a mix of

nests containing eggs through old nestlings. Upon brood

completion, we added nests to the focal group for continuous

daylight observation until A-chicks reached age 20 days,

after which nests were retired and only censused visually

each day to determine chick fates. Nests were censused and

growth was monitored until all residents had died or reached

age 70, my operational definition of fledging (see Chapter

2).

We marked hatchling (0-4 day old) chicks according to

hatching order with yellow and black indelible pens. Older

chicks received blue and yellow acrylic paint and same color

flagging tape squares glued with contact cement on their

backs and heads. Paint and flagging were reapplied

frequently (see Chapter 2). Chicks were weighed (to the

nearest g with spring scales) and the length of the culmen

was measured (to the nearest mm with a clear plastic ruler)

every 2-4 days until the brood's first-hatched (A-) chick

was (on average) 29 days old (range 25-31 days), a minimum

of once a week when A-age was between 30-40 days, and at

least every 10-12 days thereafter. Ages and hatching orders

of chicks were determined by direct observation of newly

hatched chicks whenever possible. For other chicks, ages

were estimated from a regression of culmen length on age of

known-age nestlings (see Chapter 2 and Appendix A for more

details).

Nest Observations

Focal nests were scanned in order following a pre-set

sequence. We continued to scan nests until we detected

feeding or fighting behavior (see below), at which point we

began to monitor all activities in the nest until we

terminated observations because all feeding and fighting

behavior at the target nest had ceased. After termination

of observations at a target nest, we resumed scanning of

nests starting with the next nest in the sequence. When

fighting was occurring in one nest while feeding was

occurring at another, we selected the nest with fighting as

our target for behavioral observations. If two or more

nests both had feeding or both had fighting behavior

occurring during a scan, we chose to watch whichever nest

was next in the scan sequence.

Feeding Behavior

Parents regurgitated fish onto the nest floor during

the first week of nestling life ("indirect" feeding, Pinzon

and Drummond in press and Ploger, unpub. data). They

gradually shifted to making deliveries directly as the

chicks got older (Pinzon and Drummond in press and Ploger,

unpub. data). When chicks fed "directly", they reached into

their parent's pouch to obtain food. "Feeding behavior"

included all direct and indirect deliveries of food to

nestlings, plus all cases in which parents opened their

bills over young chicks or had older chicks thrusting deep

into the base of the pouch without any evidence of food

being delivered. A period of feeding activity was defined

as ending when the parent began a nonfeeding activity

without resuming feeding activity within 1 minute. I

defined nonfeeding activities to include preening, wing-

flapping, nest-cleaning (tossing fish bones, skin, sticks

and various unidentifiable scraps from the nest), adjusting

sticks in the nest, adopting a "resting" posture in which

the parent held its closed bill out of reach of its chicks

(postures shown in figures 16, 18, and 19 of Schreiber

1977), nest relief behavior (Schreiber 1977), displaying to

or snapping at a neighbor, hopping to a perch or flying

away.

I estimated the amount of food delivered to chicks by

recording the longest linear dimensions of food boluses

swallowed by each chick. To estimate bolus sizes, I

expressed the length of the bulge in a chick's neck as a

percentage of the parent's bill length (based on Mock 1985,

units = "food-units"). Because parents sometimes blocked

the observer's view, prohibiting determination of whether

food deliveries had occurred, my data must be considered

minimum estimates of the amounts of food obtained by

nestlings.

Fighting Behavior

I counted as "fights" all cases where one chick

delivered at least one blow (see below) to the body of

another chick with enough force to move the victim's head

when struck. An individual fight continued until one of the

chicks adopted a "submissive" posture (sensu Pinzon and

Drummond in press, see definitions below), or no further

blows were delivered for at least 30 seconds. A total of

1646 fights were observed in 1990 in all of the nests used

in this and related studies (see Chapters 2 and 4 for

descriptions of these related studies).

Fights involved two types of blows, "Bites" and

"Shakes" (defined below) The first blow of a fight was a

Bite in 1255 fights and a Shake in 310 fights (of the

1565/1646 fights in 1990 for which we could identify the

type of blow that was delivered first). When Biting, the

attacker reached toward the head or body of its adversary,

closed its mandibles over some part of the victim's anatomy

such that the sharp nail at the tip of the upper mandible

depressed the victim's skin and then immediately released

its hold. Bites were delivered to the head and neck with

such speed and force that the victim's head was pushed

backward by the force. Bites to the back also pushed the

body away from the blow, but this movement was often small

because blows usually pushed the victim's body into the nest

floor which damped some of the movement. Bites were often

directed toward the eyes or base of the skull. Shakes

occurred when the attacker grasped its victim's head in a

scissor grip and forced the victim's head to strike against

the victim's body or the nest fabric. The attacker then

pulled the victim's head away from the object that it

struck. This was often followed by the attacker once again

slamming the victim's head against its body or the nest

floor. When several Shakes occurred in a continuous series,

the attacker did not relinquish its hold on the victim

between blows.

Fights occasionally resulted in puncture wounds to the

head and neck, but breaks in the skin were rare. The most

common evidence of injury was that after being beaten

repeatedly, the victim's skin often became puffed out in odd

shapes along its back, sides and neck, possibly from air-

sacs that had been broken during sibling attacks. Some

chicks near death had 5-10 small puncture wounds on their

abdomens or breasts. These wounds could not have been

caused by sibling bites, because victims almost always kept

their breasts pressed to the nest fabric during fights. But

a victim's breast might be punctured when an attacker

directed powerful blows to the victim's back, walked or sat

on the victim, or otherwise shoved the victim's body

forcefully into the often sharp sticks of the nest.

Nineteen percent of all fights observed in 1990 ended

in submissive postures (described below) that were visible

to the observer. Some additional fights also may have ended

in submissive postures that were not detectable because the

loser was hidden from the observer by the nest rim or the

body of a family member. Submissive postures included:

Crouch (N = 44 fights), Curl Neck (N = 105 fights), Turn Low

(N = 75 fights), Duck (N = 53 fights), Reverse Head (N = 1

fight), Lie Flat (N = 4 fights), and Hang Head (N = 36

fights). An additional submissive posture, Lie Back, was

observed one time in a population of brown pelicans that I

observed in 1989 as part of another study (study presented

in Chapter 2). A chick in the Crouch position (Figure 3-1A)

squatted on its heels with the back raised off the nest

floor at a 20-600 angle from horizontal while its throat was

pressed against its neck in such a way that the bill was

pressed against the abdomen, parallel to the angle of the

back. When in the Curl Neck posture (Figure 3-1B), a chick

lay with its abdomen pressed to the nest floor, its back

horizontal, the dorsal surface of its neck lying on its

shoulders and its throat resting on its breast such that the

bill was pressed against the front of the neck and breast

approximately perpendicular to the nest floor. If the bill

was not perpendicular to the nest floor, it was within 100

of perpendicular, such that the tip of the bill was

posterior to the forehead (position shown in Figure 3-1B).

The position of the body of chicks in the Turn Low, Duck,

Lie Back, Reverse Head and Lie Flat postures were all the

same as was just described for the Curl Neck position (see

Figure 3-1, B-G). These postures were distinguished by the

position of the chick's bill, head and neck. When adopting

the Turn Low posture (Figure 3-1C), a chick turned its head

to one side of its body while keeping the dorsal surface of

its neck pressed to its shoulders, its throat pressed to its

side just below its wing, and its bill pressed along its

side with the tip pointing posteriorly at a 0-450 angle to

the horizontal plane. In the Duck posture (Figure 3-1D), a

chick curled its head under its body such that the ventral

surface of the posteriorly pointing bill was pressed against

the abdomen and the dorsal surfaces of the bill and head

were pressed against the nest fabric. This posture

effectively shielded a chick's face from sibling attacks,

but left its nape exposed. When in the Lie Back posture

(Figure 3-1E), a chick pressed one cheek against its back

between its wings while its bill, resting on the back,

jutted out to the side horizontally, while being held

approximately perpendicular to the medial plane. A chick

adopting the Reverse Head position (Figure 3-1F) placed its

throat and the ventral surface of its neck and posteriorly

pointing bill against its back between the wings. In the

Lie Flat position (Figure 3-1G), a chick lay with the

lateral surface of its neck and head pressed against the

nest fabric while the dorsal surfaces of the distal and

medial halves of the neck remained folded against one-

another, and the throat pressed against the ventral surface

of the neck. The Hang Head posture (Figure 3-1H) was

defined by a chick hanging its head over the rim of the nest

or the edge of its perch such that the tip of the bill hung

parallel to or below the chick's feet. Body positions

during the Hang Head ranged from those described for the

Crouch to the Curl Neck postures.

	Title Page
	Acknowledgement
	Abstract
	Table of Contents
	Introduction
	Role of junior siblings in resource...
	Hunger as a proximate cause of...
	Effect of brood size manipulations...
	Summary and conclusions
	Appendix A: Determining chick...
	Appendix B: Determining clutch...
	Appendix C: Why do older chicks...
	Appendix D: Nest takeovers
	References
	Biographical sketch

MILLISECOND	CLASS.METHOD	MESSAGE
0	sobekcm_page_globals.constructor
0	sobekcm_page_globals.constructor	Application State validated or built
0	sobekcm_database.verify_item_lookup_object
0	sobekcm_page_globals.constructor	Navigation Object created from URI query string
0	sobekcm_database.verify_item_lookup_object
0	sobekcm_page_globals.display_item	Retrieving item or group information
0	sobekcm_page_globals.get_entire_collection_hierarchy	Retrieving hierarchy information
0	sobekcm_assistant.get_entire_collection_hierarchy
0	cached_data_manager.retrieve_item_aggregation
0	cached_data_manager.retrieve_item_aggregation	Found item aggregation on local cache
0	item_aggregation_builder.get_item_aggregation	Found 'all' item aggregation in cache
0	system.web.ui.page.page_load (ufdc.page_load)
0	sobekcm_page_globals.constructor.on_page_load
0	html_echo_mainwriter.add_style_references	Adding style references to HTML
0	html_echo_mainwriter.add_text_to_page	Reading the text from the file and echoing back to the output stream
47	html_echo_mainwriter.add_text_to_page	Finished reading and writing the file