Citation
Context-Dependent Natural and Sexual Selection on Male Nesting Activity in Flagfish (Jordanella floridae)

Material Information

Title:
Context-Dependent Natural and Sexual Selection on Male Nesting Activity in Flagfish (Jordanella floridae)
Creator:
HALE, REBECCA ELIZABETH
Copyright Date:
2008

Subjects

Subjects / Keywords:
Animal nesting ( jstor )
Brackish water ( jstor )
Clutches ( jstor )
Eggs ( jstor )
Female animals ( jstor )
Fish ( jstor )
Fresh water ( jstor )
Hatching ( jstor )
Mating behavior ( jstor )
Salinity ( jstor )
City of St. Marks ( local )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Rebecca Elizabeth Hale. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
12/31/2011
Resource Identifier:
659562747 ( OCLC )

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

1 CONTEXT-DEPENDENT NATURAL AND SEXUAL SELECTION ON MALE NESTING ACTIVITY IN FLAGFISH ( Jordanella floridae ) By REBECCA ELIZABETH HALE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006

PAGE 2

2 Copyright 2006 by Rebecca Elizabeth Hale

PAGE 3

3 I dedicate this to my family, w ho has supported me from the beginning.

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank my co -advisors, Colette St. Mary and Craig Osenberg, and my graduate committee, Jane Brockmann, Joseph Trav is, and Shirley Baker, for their guidance and input throughout the development and analysis of this work. Their diverse perspectives and suggestions greatly contributed to both my doctoral research and to my development as a scientist. The faculty, graduate students, and st aff of Zoology Department at the University of Florida have been supportive, have challenged me to think more broadly, and generally contributed to an excellent gradua te experience. In particular, the laboratories of Colette St. Mary, Craig Osenberg, Ben Bo lker, and Jane Brockmann created a fun and stimulating environment for conducting and evaluating science. I am particularly grateful for the friendship and support of James Vonesh, Sophia Balcomb, Jackie Wilson, Toshi Okuyama, Ben Miner, Billy Gunnels, Kavita Isvaran, Suhel Quader, and Laura Sirot. I am greatly indebted to Joseph Travis and hi s graduate students for allowing me to work in their laboratory at Florida State University and for their continued as sistance with field and laboratory work, with data analysis, and with manuscript preparation. Katie McGhee, Matt Schrader, Margaret Gunzburger, and Becky Fuller have been incredibly generous in aiding my work and have made my research really enjoyable. I am very appreciative of the help with st atistical analysis an d programming provided by Ben Bolker and Toshi Okuyama, with metabolic rate analysis provided by Erin Reardon and Frank Nordlie, with electrical engineering as sistance provided by Joel Abdullah, and with water sample analysis provided by Dano Fiore, Bi ll Landing, Cathy Levenson, and Cliff Buck. In addition, I would like to thank the many people who helped with field work: Mark Endries, Becky Fuller, Billy Gunnels, Margaret Gunzburge r, Rico Holdo, Nate Jue, Holly Kindsvater, Hope Klug, Andy Lane, Georgina McDowell, Katie McGhee, Ben Miner, Toshi Okuyama, Gigi

PAGE 5

5 Ostrow, Suhel Quader, Andrea Quaid, Matt Schr ader, Nat Seavy, Laura Sirot, Sheryl SoucyLubbell, Manjula Tiwari, Melissa Wilson, and Cedric Worman. Finally, this dissertation woul d never have come to fruition without the support of my husband, Mark Endries, my parent s, Peter and Susan Hale, and my sister, Jennifer Hale. I am forever grateful for the encouraging and suppor tive environment created by my family that helped me pursue my interest in science. I was supported throughout my graduate studies by teaching assistantships in the Departments of Zoology and Biology Sciences at the University of Florida. The U.S. Fish and Wildlife Service was generous in permitting my research in Merritt Island National Wildlife Refuge and St. Marks National Wildlife Refuge. This work was funded by NSF Doctoral Dissertation Improvement Grant number 0704904 to R. Hale and C. St. Mary.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........9 LIST OF FIGURES................................................................................................................ .......10 ABSTRACT....................................................................................................................... ............12 CHAPTER 1 INTRODUCTION................................................................................................................. .14 2 A RE-EXAMINATION OF THE INFLUENCE OF OFFSPRING REPRODUCTIVE VALUE ON PARENTAL EFFORT......................................................................................19 Introduction................................................................................................................... ..........19 The Models..................................................................................................................... ........22 Results........................................................................................................................ .............24 Brood Size..................................................................................................................... ..24 Offspring Vulnerability...................................................................................................25 Brood Age...................................................................................................................... .27 Discussion..................................................................................................................... ..........29 Brood Size..................................................................................................................... ..29 Offspring Vulnerability...................................................................................................30 Brood Age...................................................................................................................... .31 Summary........................................................................................................................ ..36 3 NEST TENDING INCREASES REPRODUCTIVE SUCCESS, SOMETIMES – ENVIRONMENTAL EFFECTS ON PATE RNAL CARE AND MATE CHOICE..............43 Introduction................................................................................................................... ..........43 Methods........................................................................................................................ ..........46 Collection and Transportation.........................................................................................46 Acclimation and Experiments.........................................................................................46 Analyses....................................................................................................................... ...48 Results........................................................................................................................ .............50 Salinity, Native Habitat Type, and Population................................................................50 Sexual Selection on Male Behavior................................................................................51 Behavioral Responses to Sexual Selection......................................................................52 Association between Pre-Parent al and Parental Behavior...............................................53 Discussion..................................................................................................................... ..........54 Female Mating Preferences.............................................................................................54 Male Behavior.................................................................................................................5 7 Summary........................................................................................................................ ..58

PAGE 7

7 4 PRE-SPAWNING BEHAVIOR AND NEST-FANNING PREDICT CLUTCH SUCCESS........................................................................................................................ .......69 Introduction................................................................................................................... ..........69 Methods........................................................................................................................ ..........73 Collection, Transport, and Acclimation..........................................................................73 Experimental Design and Protocol..................................................................................73 Analyses....................................................................................................................... ...76 Results........................................................................................................................ .............78 Clutch Size.................................................................................................................... ...78 Hatching Success.............................................................................................................79 Activity scored as ‘nest-tending’..............................................................................79 Activity scored as ‘approaching the female’............................................................80 Fanning and salinity, alone.......................................................................................80 Hatchling Mass................................................................................................................8 1 Discussion..................................................................................................................... ..........81 Context-Dependent Association betw een Male Traits and Fitness.................................82 Context-Dependent Mating Preferen ces and Mate Choice Benefits...............................84 Summary........................................................................................................................ ..87 5 NATIVE ENVIRONMENT, BUT NOT SA LINITY, INFLUENCES METABOLIC RATE........................................................................................................................... ...........94 Introduction................................................................................................................... ..........94 Methods........................................................................................................................ ..........99 Field Survey................................................................................................................... ..99 Laboratory Study of Metabolism...................................................................................100 Analyses....................................................................................................................... .102 Results........................................................................................................................ ...........104 Field Survey................................................................................................................... 104 Laboratory Study of Metabolism...................................................................................105 Discussion..................................................................................................................... ........106 6 SYNTHESIS.................................................................................................................... .....118 Context-Dependent Mating Preferen ces and Mate Choice Benefits....................................118 Adjustment of Parental Behavi or to Selection Pressures.....................................................123 Future Directions.............................................................................................................. ....125 APPENDIX FIELD SURVEY OF WATER CHEMIS TRY AND SPECIES COMPOSITION AT FOUR STUDY SITES..........................................................................................................128 Introduction................................................................................................................... ........128 Methods........................................................................................................................ ........128 Results and Discussion......................................................................................................... 129

PAGE 8

8 LIST OF REFERENCES............................................................................................................. 149 BIOGRAPHICAL SKETCH.......................................................................................................161

PAGE 9

9 LIST OF TABLES Table page 3-1 Effects of pre-parental behavior and parental beha vior on reproductive success.................60 3-2 The effects of salinity, native habitat t ype (inland versus coastal), and population (nested within habita t type) on pre-parental behavi or and parental behavior.......................61 3-3 The relationships between pre-pare ntal and parental male activity......................................62 4-1 Analysis of deviance for fit of progre ssively simpler logistic regression models evaluating ln(odds of hatching) usin g a Likelihood Ratio (LR) test....................................89 5-1 Sample size and mean (SE) wet mass and metabolic rate for each population by salinity treatment group......................................................................................................1 12 A-1 Water chemistry data from two years of bi monthly sampling at four Florida sites. ........134 A-2 Mean density (No./m2) of fishes and occurrence of non-fish vertebrates and invertebrates from eight box trap samp les at four sites over two years..............................134 A-3 Results of preliminary sampling at sites within Merritt Island National Wildlife Refuge and St. Marks National Wildlife Refuge in 2003...............................................................144 A-4 Observations of nest attendance and spawning in Jordanella floridae ..............................145

PAGE 10

10 LIST OF FIGURES Figure page 1-1 Location of Jordanella floridae study sites..........................................................................18 2-1 Present, future, and total reproductive su ccess as a function of the amount of care a parent provides................................................................................................................ ......38 2-2 The effect of brood size, n , on the optimal amount of parental care....................................39 2-3 The effect of offspring survival without care, l0, on the optimal amount of parental care under Model 2.................................................................................................................. .....40 2-4 The effect of offspring survival without care, l0, on the optimal amount of parental care under Model 3.................................................................................................................. .....41 2-5 Optimal allocation of parental effort to ch ick 1 as a function of th e total amount of care provided....................................................................................................................... .........42 3-1 Spawning success as a function of pre-parent al behavior in fresh (A) and brackish (B) water.......................................................................................................................... ............63 3-2 Spawning success as a function of parental be havior in fresh (A, C, E) and brackish (B, D, F) water.................................................................................................................... ........64 3-3 The mean number of eggs a male received (± SE) as a function of whether or not a male approached the female during the parental phase........................................................65 3-4 Pre-parental behavior of males (i.e., observed on Day 1).....................................................66 3-5 Parental behavior of males (i.e., observed on the first day of eggs).....................................67 3-6 The percent of males tending their nest s (A) and spending time at their nests (B) increased with the number of eggs in the nest......................................................................68 4-1 Clutch size as a function of salinity of th e natal aquarium and A) whether or not the male tended the nest and B) approach ed the female prior to spawning................................91 4-2 The proportion of embryos that hatched as a function of salinity and A) male nesttending prior to spawning, B) male approaching prior to spawning, and C) fanning...........92 4-3 Hatchling mass as a function of salinity and A) male nest-ten ding prior to spawning, B) male approaching prior to spawning, and C) fanning......................................................93 5-1 Three possible processes by which metabolic rate of inland indivi duals could be lower than that of coastal individuals in fresh water....................................................................113

PAGE 11

11 5-2 Salinity during each month at the four sites for the 2003-2004 and 2005-2006 sampling years.......................................................................................................................... ..........114 5-3 Metabolic rate as a function of mass for each of the four populations...............................115 5-4 Metabolic rate as a functi on of time at the experimental salinity for inland and coastal populations.................................................................................................................... ......116 5-5 Metabolic rate as a f unction of salinity treatment and native habitat type.........................117 A-1 Mean salinity, temperature, and pH for each site and each survey year.............................146 A-2 Mean density of flagfish females, males, and juveniles at the four sites over the two sampling years................................................................................................................. ...147 A-3 Mean density of common fish species , averaged across two years of sampling................148

PAGE 12

12 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CONTEXT-DEPENDENT NATURAL AND SEXUAL SELECTION ON MALE NESTING ACTIVITY IN FLAGFISH ( Jordanella floridae ) By Rebecca Elizabeth Hale December 2006 Chair: Colette M. St. Mary Cochair: Craig W. Osenberg Major Department: Zoology Parental care increases offspring fitness and can increase attractiveness to mates. These two selection pressures should covary across environm ents if the benefit of choosing a mate is to select a good care-giver. However, the influe nce of such covariance on care has not been demonstrated, even though it may accelerat e the evolution of parental care. I examined how offspring fitness benefits a nd mating preferences infl uence paternal care across salinities in flagfish ( Jordanella floridae ). Using mathematical modeling, I demonstrated that careÂ’s benefits for offspring should be gr eater in fresh water (FW), where previous shows offspring do worse without care, than in brackish wa ter (BW). Therefore, if females select mates based on the care they provide, then preferences should be stronger in FW. Further, levels of care should be higher in FW than in BW. I tested these predictions in three laboratory experiments. I quantified spawning success as a function of parental care act ivity and found that female pref erences for care varied with salinity, but were not always str onger where offspring benefits we re predicted to be greater. I then quantified the effects of care on hatching success and hatc hling mass and found that some care activities were more beneficial in FW, as predicted by my model, but others were equally

PAGE 13

13 beneficial in FW and BW, possibl y explaining why female preferen ces were not always stronger in FW. Because the offspring and mate choice benefits of care varied with salinity, I expected male behavior to vary. However, a comparison of behavior in FW and BW showed that it did not. I then measured metabolic costs of osmore gulation across salinities to determine whether variable costs might balance the benefits of care, but I found no effect of salinity on metabolic. These data indicate that 1) the benefits of care for offspring can be predicted by how well offspring do without care, 2) mati ng preferences sometimes covary with the benefits of care for offspring, and suggest that 3) male behavior ma y remain constant across environments despite variation in selection on care if the costs of providing care are moderate compared to the magnitude of benefits.

PAGE 14

14 CHAPTER 1 INTRODUCTION The investment of resources into offspring through parental care allows parents to increase their immediate reproductive success at the expense of their future reproduction (Clutton-Brock 1991; Williams 1966). In this c ontext, offspring risk (e.g., Hale et al. 2003; Listøen et al. 2000), parental co sts (e.g., Kaitala et al. 2000; We imerskirch et al. 2000), and mating preferences based on pare ntal care (e.g., Jones and Reynolds 1999; Petersen et al. 2005) can each influence parental care, but the relatio nship among these factors and how they together influence the evolution of parent al care has not been thoroughl y examined. In this work, I evaluate how natural selection and sexual selection together in fluence parental care in the euryhaline fish Jordanella floridae (flagfish). In particular, I describe variation in selection across a salinity gradient in order to explain observed beha vioral plasticity. Species distributed across a broad ecologica l range often experien ce different selection pressures in different parts of thei r range and variation in local sel ection may give rise to distinct patterns of reproductive investme nt. I argue that local ecologi cal conditions can influence the risks offspring face and the costs of investi ng into young, and thereby influence natural and sexual selection on parental care. Specifically, variation in the risks facing offspring will influence both natural and sexual selection because the direct bene fits of providing care and of choosing to mate with good parents are determ ined by the extent to which care increases offspring fitness, which itself may be a function of the offspring risk. By examining variation in behavior across an ecological range expected to influence offspring mortality risk, I explore the processes that connect ecology with behavior. Jordanella floridae belongs to the family Cypri nodontidae and is distributed across peninsular Florida and regions of th e Florida panhandle (Figure 1-1). J. floridae is found

PAGE 15

15 predominantly in inland, freshwat er ditches and ponds but its pe rsistence in coastal, brackish habitats has been noted in South Florida, par ticularly in Everglades National Park and Big Cypress Swamp (Gunter and Hall 1965; Kil by 1955; Loftus and Kushlan 1987). Of J. floridae Â’s closer relatives (Echelle and Echelle 1993), Garmanella pulchra also exhibits a freshwater distribution, but most other Cypr inodontids are found in higher sa linities, including the coastal salt marsh species Cyprinodon variegatus and Floridichthys carpio (Nordlie 1987; Nordlie and Walsh 1989) found in Florida, and the desert pupfishes ( Cyprinodon spp.) found in salt springs of the desert southwest (Soltz and Naiman 1978). Of the Cyprinodontids, only J. floridae is known to provide care for its offspring. J. floridae males establish a nesting terr itory before attracting mates. Spawning occurs on a small area of substrate within this territory as eggs are individua lly released by the female and fertilized. An individual spaw ning bout typically involves the rele ase of 5 to 20 eggs and, in the laboratory, the number of eggs spawned in the sa me nest over the course of a day can reach 80 eggs (Chapters 3 and 4). Males defend and cl ean the nest area until young have hatched, whereas females provide no care for offspring. St. Mary et al. (2004) found that sa linity increases the survival of J. floridae embryos reared without care. This suggested that se lection on paternal care may depend on salinity. However, how parental care should vary with salinity is not entirely clear. Indeed, predictions in the literature for how care should vary as a function of offspring mortality risk, or vulnerability, are contradictory, with some models predicti ng greater care and some less care for more vulnerable young. The difference among these models lies in the assumed relationship between offspring vulnerability and careÂ’s benefits. In Chapter 2, I use mathematical modeling to resolve these conflicting models. I argue that the pr ediction that care will decrease with increasing

PAGE 16

16 offspring vulnerability is faulty because it is based on the inappropriate argument that care should increase with increasing reproductive value of the brood. I show that if one separates a broodÂ’s reproductive value into its components, pare ntal investment should increase with some of these components and decrease with others. I then demonstrate under what circumstances one should expect care to increas e and to decrease with incr easing offspring vulnerability. I then use the model as a guide in experiment al studies of natural and sexual selection on parental care in J. floridae . If salinity influences both natu ral and sexual selection on parental care, then behavior within populations may be plastic in response to salinity and populations from different native salinity environments may exhibit different patterns of plasticity. I focus my experimental work on four J. floridae populations (Figure 1-1), two of which are coastal (St. Marks and Merritt Island) and tw o of which are inland (Otter Creek and Miccosukee). These four sites are distributed across the state and are paired in northern and so uthern replicates to prevent confounding of coastal/inland with latitude-associated factors such as temperature. St. Marks and Otter Creek are north and west of Merritt Island and Mi ccosukee. Data from a two year survey of water chemistry and species co mposition at these sites are presented in the appendix. In Chapter 3, I examine the effect of salin ity on male parental behavior and female preferences for male behavior in a laboratory study of behavioral plasticity. I found that female mating preferences, measured by a pairÂ’s reprodu ctive success, differed between salinities. For example, males enjoyed higher reproductive success if they performed certain nest-tending activities when their nests were empty (prior to spawning), but only when in fresh water. Although the sexual selection benefit of performi ng these activities was large, males did not adjust their activity level with sa linity as one might predict. Inst ead, male behavior was constant

PAGE 17

17 across salinities. This raised two questions. First, is the benefit that females gain from preferring tending males greater in fresh than in brackish water? Second, why do males not tend their nests more in fresh water? I address the first question in Chapter 4. I measure hatching success of embryos fathered by tending and non-tending males in a single coastal popu lation, Merritt Isla nd. If tending males have higher offspring survival th an non-tending males in fresh wa ter, then it would appear that females exhibit preferences for these males because of the direct benefits of male behavior for their offspring. I also measured the effect of artificial egg fanning on su rvivorship in fresh and brackish water as a means of measuring the bene fit of this component of parental care for offspring. In Chapter 5, I address the second qu estion by evaluating whether the costs of care are too high in fresh water for males to exhibit more tending there. A comparison of metabolic rates across salinities for each of the four populations indicated that the costs are similar across salinities. In the final chapter, I relate the results of Chapters 4 and 5 to the patterns of male behavior and female mating prefer ences described in Chapter 3. In addition, I draw conclusions about how natural and sexual select ion simultaneously influence the ev olution of parental care in light of my theoretical and experimental work.

PAGE 18

18 Figure 1-1. Location of Jordanella floridae study sites. The speciesÂ’ range extends from southern, mainland Florida north and west to St. Marks National Wildlife Refuge in Wakulla County. Miccosukee Merritt Island Otter Creek St. Marks Miccosukee Merritt Island Otter Creek St. Marks Miccosukee Merritt Island Otter Creek St. Marks

PAGE 19

19 CHAPTER 2 A RE-EXAMINATION OF THE INFLUENCE OF OFFSPRING REPRODUCTIVE VALUE ON PARENTAL EFFORT Introduction Parents should care for young if the benef it of care to current reproductive success outweighs the cost of reduced future repr oduction (e.g., Andersson et al. 1980; Trivers 1972; Williams 1966). The amount of care parents prov ide depends, in part, on the amount by which care will increase the fitness of their young. Desp ite their importance to parental investment decisions, these marginal fitness ga ins are often overlooked in theore tical and empirical studies. Overlooking marginal fitness gains is part icularly problematic when examining how parental effort should differ be tween broods of different repro ductive value. Fisher (1930) defined reproductive value as the number of offspr ing an individual is exp ected to produce from now until its death, scaled for population growth and relative to that of a newborn. The reproductive value of a brood can then be defined as the sum of the reproductive values of all individuals in the brood and is pr oportional to the brood’s survival to maturity. Many authors have argued that parental effort should increa se with a brood’s reproductive value (e.g., Amat 1996; Carlisle 1985; Koskela et al. 2000; Rytkone n 2002). In this argument, the reproductive value of interest is presumably the brood’s repr oductive value prior to th e parent’s investment decision, which I will refer to as current reproduc tive value. For care to increase with a brood’s current reproductive value, broods of high current value must offer a greater return on the parent’s investment than broods of low current va lue. However, not all broods of relatively high current value will offer greater returns – whethe r such a brood will offer greater returns may depend on why it has high reproductive value. A brood’s reproductive value is influenced by its size and its expected survival to maturity, which, in turn, may be a function of its age and vulnerabi lity. Therefore, if parental

PAGE 20

20 effort should increase with a brood’s reproduc tive value, it must vary with brood size, vulnerability, and age similarly. However, theoretic al studies of optimal parental effort differ in how they predict parental care to increase with each of these factor s. For example, larger broods have higher current reproductive value than sma ller broods because the number of individuals expected to reach maturity is greater for the larger broods, assuming the same survivorship schedule for individuals in both sizes of broods. But optimality models examining the effect of brood size on parental effort do not always predic t effort to increase w ith size. Sargent and Gross (1993) created a model to de scribe parental care patterns in teleost fishes and predicted parental care would increase with brood size. In contrast, Tammaru and Hõrak (1999) predicted that, in birds, larger clutches may actually receive less care if the care must be divided among offspring. In large broods, the amount of care goi ng to individual offspring may be so small that it is not enough for them to survive. A similar conflict among models was illustra ted by Dale et al . (1996), but involves predictions for how effort should change with a brood’s vulnerability. Dale and colleagues argued that offspring that are more vulnerable, eith er because they are in higher risk habitats or because they are in poorer condition, should receiv e more care. Their prediction was consistent with earlier verbal models that considered vulne rability to indicate the need of offspring for parental care (Andersson et al. 1980; Montgomerie and Weatherhead 1988; and subsequently Webb et al. 2002). Supporting this prediction, Listøen et al. (2000) found that pied flycatchers ( Ficedula hypoleuca ) returned to the nest sooner to feed young when young were in poorer condition. Similarly, Tveraa et al. (1 998) found that Antarctic petrels ( Thalassoica antarctica ) delivered more food when rearing a small chick than when rearing a large one of the same age. However, Dale et al. (1996) point ed out that their prediction was c ontrary to predictions based on

PAGE 21

21 a brood’s reproductive value because broods with greater vulnerability are, by definition, of lower reproductive value and, t hus, are predicted to receive less parental care by reproductive value-based hypotheses. These alternative predictions highlight an a pparent paradox between whether care should increase with brood reproductive value or with th e young’s need for care. I suggest that this paradox is artificial beca use both predictions are based on char acteristics of the brood before the parent’s investment decision is made and not on th e fitness gain that results from this decision. Indeed, different factors influe ncing a brood’s current reproductiv e value – including brood size, age, and vulnerability – may influence the benef its of care differently. This is because the change in offspring reproductive value that resu lts from care, not current reproductive value, determines parental benefits. The idea that the change in fitness resulti ng from investment decisions determines selective advantage is not new and has been empha sized in the context of parental investment (Lessells 2002; Maynard Smith 1980; Winkler a nd Wallin 1987), kin selection (Charlesworth and Charnov 1981; Hamilton 1964), optimal foragi ng (Charnov 1976), and seed size (Temme 1986). I suggest that its emphasis in the parental investment literature has been limited because the mechanisms by which offspring fitness is infl uenced by care are not well defined. Here, I develop a set of similar optimality models to ev aluate how the change in a brood’s reproductive value resulting from care is influenced by differe nt components of a brood’s current reproductive value and, in turn, how this change in value infl uences parental effort. I focus on allocation of care among broods, though my conclusions easily apply to the allocat ion of care among broodmates.

PAGE 22

22 Using these models, I will examine how br ood size, vulnerabilit y, and age influence optimal levels of parental care and will demons trate that the apparent paradox regarding the effects of brood reproductive value on care can be resolved. In doing so, I will make three main points. First, I will show that variation in these three factors does not have similar effects on optimal parental effort and, as a result, that a broodÂ’s curr ent reproductive value is a poor predictor of parental effort. S econd, I will demonstrate that the e ffect of offspring vulnerability depends on the assumptions regarding how vulnera ble offspring benefit from parental care. Third, I will return to FisherÂ’s (1930) mathematical formulation of reproductive value to evaluate how a broodÂ’s reproductive value changes with age. The Models The models evaluate how much care a pare nt should currently provide young. They examine a single investment decision and assume that the amount that the parent is able to currently invest is independent of how much it has invested in the past. For example, when evaluating the investment decision of the mother , the amount of resources available for her to currently invest into hatchlings is not traded off against the am ount she invested into the same young as eggs. Let the parentÂ’s pres ent reproductive success, P , be the fitness benefit the parent gains from investing in the current brood. P is a function of brood size ( n ) and the mean fitness gain from each individual offspring ( S ), where S is a function of the amount of care, C : P n S ( C ). (2-1) I assume that parental effort is non-d epreciable (Clutton-Brock 1991), such that S ( C ) does not decrease with brood size, and that all offspring benefit equally from care. Let the parentÂ’s future reproductive success, F , be considered across its lifetime, including the remainder of the current reproductive season and all future seasons. Investing

PAGE 23

23 resources into current young reduces the potential investment into future reproductive efforts. This tradeoff is incorporated by defining future reproductive success as a declining function of the amount of care provided to current young: ), ( C f F (2-2) where f ( C ) is independent of brood size. To determine optimal care effort, C* , I solve for the amount of care that maximizes the sum of present and futu re reproductive success ( R = P + F ), e.g., by setting dR / dC = dP / dC + dF / dC = 0. This optimum arises when dP / dC = dF / dC , (2-3) that is, when the two component s of reproductive success have slopes of the same magnitude but opposite sign. I assume that P increases at a decelera ting rate with increasing C and that F decreases at an acce lerating rate with C (Figure 2-1). Therefore, d2R/dC2 will be negative and the value of C that satisfies equation 2-3 w ill indicate the value at which R is maximal. Using FisherÂ’s (1930) definition of reproductive value, the vulnerability of the brood is characterized by the survivorship schedule of young. In equation (2 -1), vulnerability is captured in the expected fitness of young if the parent decides not to provide care, P (C=0). Broods or individuals of greater vu lnerability have lower P (C=0). Vulnerability is deliberately defined independently of an individualÂ’s response to parent al effort because they are separate variables that may be positively correlated, negatively correlated, or not correlated at all (discussed in Hale et al. 2003). I consider three models that differ in th e shape of the relati onship between care and offspring fitness, defined by S ( C ), and thus the shape of P . In Model 1, S ( C ) = l0 + g ( C ), where l0 denotes the fitness of offspr ing in the absence of care and g ( C ) denotes the increase in fitness

PAGE 24

24 resulting from care and is a pos itive, diminishing function of C . In Model 2, S ( C ) = l0 + g ( C )·(1 – l0), reflecting that the benefits of care ar e negatively proportional to the fitness of offspring in the absence of care. This structur e represents the underlying assumption of Dale et al. (1996) and similar models. In Model 3, S ( C ) = l0 + ( l0 · g ( C )). This model structure follows that of Sargent and Gross (1993) and reflects that the benefits of care are positively proportional to the fitness of offspring in the absence of car e. This represents th e underlying assumption of the reproductive value-based models of optimal parental care. I will evaluate how C * varies as a function of brood size, offspring fitnes s in the absence of care, and brood age in the cont ext of these models. I will show that whether or not parental care should increase as a func tion of a brood’s reproductive va lue depends on which component of reproductive value is variable (brood size, fitness in the abse nce of care, or brood age) and on the assumptions regarding how fitness in the ab sence of care influences care’s benefits for offspring (i.e., Model 1, 2, or 3). As a result, a brood’s repr oductive value is not a general indicator of the benefits to be received from care. Instead, whether care should increase with reproductive value will depend on which assumptions are appropriate to the parental care system under examination. Results Brood Size The effects of brood size are similar in the th ree models; therefore, I will illustrate its effect in the simplest model. When S ( C ) = l0 + g ( C ), care is optimized when n · g ’( C* ) = f ( C* ), (2-4) substituting S ( C ) in equation 2-3, above. As n increases, the slope of both S ( C ) and f ( C ) at C * become steeper, which corresponds to a higher value of C * (Figure 2-2A,B). Therefore, as brood size increases, optimal parental care increases (Figure 2-2C).

PAGE 25

25 Offspring Vulnerability Vulnerability is negatively correlated with l0 and may influence P in a number of ways if variation in l0 results in variation in the response of individua l young to parental effort, S ’( C ). In Model 1, S ’( C ) is independent of l0 (equation 2-4). Therefore, cha nges in offspring vulnerability have no influence on optimal care. Such independence applies when, for example, a fixed amount of provisioning has the same effect on broods in habitats of high and low predator density. The broods in the high-density habitat ar e more vulnerable, but this vulnerability does not influence the effect of provisioning on the fitness of individual offspring. In Model 2, S ( C ) = l0 + g ( C )·(1 l0) and P approaches n as C approaches 1.0. Parental effort is optimal when ( n · g ’( C* )) + ( n · l0 · g ’( C* )) = f ’( C* ). (2-5) Increasing l0 decreases the steepness of S ( C ) and f ( C ) at C *, causing a shift in C * toward lower levels of care (Figure 2-3). Br oods with higher fitness in the ab sence of care shou ld receive less care. A model structure like that of Model 2 is as sumed by the models of Dale et al. (1996), Andersson (1980), and Montgomerie and Weatherhead (1988) that argue pa rental effort should increase with a brood’s vulnerability. Model 2 may describe systems in which parental care primarily affects offspring survival to independen ce and not fecundity or survival after maturity. Under such circumstances, the effect of care on a brood’s collective fitn ess is expected to saturate as survival approaches 100%. Theref ore, broods that have re latively high expected survival in the absence of care have relatively little to gain from receiving care and should receive less. In Model 3, S ( C ) = l0 + ( l0 · g ( C )) and care is optimized when n · l0 · g ’( C* ) = f ’( C* ). (2-6)

PAGE 26

26 Here, the effect of l0 is the opposite from its effect in M odel 2. Just as increasing brood size increased optimal care, so does increasing the fi tness of offspring when they are not provided care. Increasing l0 increases the steepness of S ( C ) and f ( C ) at C *, resulting in an increase in C *. Therefore, under this model, care should decr ease with increasing brood vulnerability (Figure 2-4) because vulnerability ne gatively correlates with the benefit of care for offspring. The structure of this third model is often assu med in models predicting that parental care should increase with the reproductive value of a brood. For example, Sargent and Gross (1993) evaluated a model with similar assumptions with the specific goal to evaluate the effect of clutch age on parental care in br ood-cycling fish such as th e three-spine stickleback ( Gasterosteus aculeatus ). They assumed that embryos would not surv ive without parental ca re, but if parental care were provided, older embryos would have higher expected fitness than younger embryos because they will reach maturity sooner. Indeed, older embryos do have higher reproductive value. Because all embryos have zero fitness in the absence of care, providing care results in greater fitness benefits for th e older embryos than the younger ones. This argument makes a number of assumptions regarding the state of the clutch as it ages and these assumptions may not be valid in most systems. In the final section of the Results, I explore the effect of age on a broodÂ’s reproductive value. Examining the effect of offspring vulnera bility on optimal care in each model demonstrates that offspring vulnerability will only influence optimal effort if differences in vulnerability are associated with differences in the slope of P . When the slope varies with vulnerability, more vulnerable young are predicted to receive either more or less care, depending on the assumed relationship between vulnerability and the benefits individual offspring receive

PAGE 27

27 from care. This illustrates th e limited generality of th e prediction that care should increase with a brood’s current reproductive value. Brood Age Although the models do not explicitly inco rporate brood age, one can explore how P might change with age. As defined in equation 2-1, P is influenced by brood size, n , and the relationship between care and offspring fitness, S ( C ). Both may change with offspring age. First, factors such as predation, disease, and f ilial cannibalism, will cause broods to be smaller on average as they approach independence than th ey were at fertilization (Ackerman and Eadie 2003; Klug and St Mary 2005; St. Mary et al. 2004). As the broods d ecline in size, the benefit of caring declines, as illust rated in Figure 2-2. Second, the shape of S ( C ) may change as the brood ages as the result of three processes. In the first process, g ’( C ) changes with age because the ability to survive without care changes – i.e, their vulnerability changes. Consider a system in which parental care increases the probability that offspring survive to the next day. The benefit of care would decline with age if, for example, care increases daily survivorship by 20% at hatching, but only by 10% once the young approach fledging. Parental effort would decline with br ood age under this circumstance, consistent with dynamic optimization models in wh ich age is explicitly incorporated (Sargent 1990; Webb et al. 2002). However, daily surviv orship could also decline with age, causing optimal effort to increase. For example, older offspring may be more mobile and more likely to be detected by predators. In the second process, g ’( C ) changes with age because young and old offspring respond differently to care, possibly as the result of physiological differences. For example, the efficiency with which food is converted into body mass may change with age. In the third process, older offspring have highe r expected survival to maturity solely by virtue of their having

PAGE 28

28 to survive less time until maturity. This increas e in the expected survival to maturity as individuals age could cause g Â’( C ) to increase (e.g., Sargent an d Gross 1993), which would cause the benefit of care to increase as offspring age. The relative influences of d eclining brood size and changing g Â’( C ) can be examined in the context of FisherÂ’s (1930) mathematical fo rmulation of reproductive value (J. Nichols, personal communication). In a population with non-overlapping generations, the reproductive value of an individual in stage a ( Va) is Va 1 la lxmx x a death (2-7) where lx is the probability of survival from birth to age class x and mx is the expected fecundity of an individual of age class x (Fisher 1930). This equation can be simplified for individuals that have not yet matured because the fecundity before maturity = 0. Therefore, Va 1 la lxmx x maturity death Rmla (2-8) where Rm is the expected reproductive success from ma turity to death. The probability of surviving from birth to a given age decreases with age. Therefore, la decreases with age. However, equation 7 defines the reproductive va lue of only one individual in the brood. Reproductive value of the brood is the number of individu als in the brood at time a , na, multiplied by their average reproductive value: Vbrood = naVa. As the brood ages, the number of offspring in the brood declines due to mortality. The proportion of the brood that survives from birth to stage a is la, so the equation for the value of the brood becomes Vbrood , a n0la R mla (2-9)

PAGE 29

29 where n0 is the initial number of offspring. The la values cancel one another such that the reproductive value of the brood is independent of a ; thus, reproductive value does not change as the brood ages. Equation 2-9 suggests that neither changes in brood size nor in the tim e to maturity (the third process, above) as a brood ages influence how much care should be provided to broods of different ages and, instead, that any changes in care with offspring age are due to changes in vulnerability (the first process) or in offspring response (e.g., physiological) to care (the second process). Discussion Although numerous models have explicitly disc ussed how optimal parental effort is determined by the changes in fitness that result from parental effort (e.g., Lessells 2002; Lloyd 1987; Temme 1986), empirical studies are frequen tly motivated by predictions for how a brood’s reproductive value at a single time should influen ce parental investment decisions (e.g., Amat 1996; Dale et al. 1996; Koskela et al. 2000). By focusing on the brood’s current reproductive value, a paradox has arisen regarding the influe nce of reproductive value on optimal care. I aimed to resolve this paradox by emphasizing that the optimal amount of care for a parent to provide depends on the change in brood reproductive value that results from care rather than the brood’s current value. Specifically, I demonstr ated that different components of a brood’s current reproductive value – namely brood size, vul nerability, and age – can be either positively or negatively correlated with the change in offs pring fitness resulting fr om care, indicating that the relationship between current reproductive value and fitness benefits is not invariable. Brood Size The model predicts parental effort to increa se with increasing brood size, consistent with previous models (e.g., Sargent and Gross 1993). Alternatively, parental care may decrease with

PAGE 30

30 brood size when young require a minimum amount of car e to benefit and when that care must be divided among broodmates ('depreciable care', Clutton-Brock 1991). Although a model with depreciable care was not evaluated here, such m odels have shown that individuals in large broods are unable to receive enough parental care to survive and will re ceive less care (Kacelnik 1988; Tammaru and Hõrak 1999). Many studies have shown parental effort to increase with brood size (e.g., Amat 1996; Koskela et al. 2000; St. Mary et al. 2001; Wiklund 1990), but this pattern is by no means ubiquitous. Chick provisioning in birds has been found to both incr ease with (Olsen and Tucker 2003; Sanz and Tinbergen 1999) and be una ffected by (Tolonen and Korpimaki 1996) manipulated brood size. Furthe r, nest fanning in fishes has been shown to both increase (Coleman and Fischer 1991) and remain cons tant (Lindström and Wennström 1994) with increasing brood size. Variation among species in th e effect of brood size on parental effort may be attributable to the effects of brood size on ca re’s efficacy. For example, a given amount of aeration may benefit small and la rge egg masses differently if more flow is necessary to penetrate the center of a large egg mass than a small ma ss (Strathmann and Strathmann 1995). Offspring Vulnerability Young may differ in vulnerability due to differences in pare ntal effort during an earlier life stage, as in the difference in incubation ve rsus provisioning between altricial and precocial young (Nice 1962; Starck and Ricklefs 1998), or due to differences in the risk imposed by their environment (e.g., St. Mary et al. 2004). Young in habitats with high pred ation risk or low food availability are likely to be of greater vulnerability than young in low risk habitats (e.g., Listøen et al. 2000; Weimerskirch et al. 2001). How parental care is predicted to change with offspring vulnerability varies among published models, largely due to how vulnerabil ity is incorporated mathematically. Here,

PAGE 31

31 vulnerability is defined as offspr ing fitness in the absence of cu rrent parental effort. Model 2 assumed that vulnerable offspring will benefit more from parental care. This negative relationship between vulnerability and the change in offspring fitness was also assumed by Dale et al. (1996), who first posed the paradox regard ing predictions based on offspring reproductive value, and by Webb et al. (2002). Other models have assumed the opposite – that less vulnerable offspring receive a greater benefit from care (Kacelnik 1988; Sargent and Gross 1993). For example, Kacelnik (1988) incorporates offspring condition as the response to care, itself. Offspring are in poor condition if they exhibit li ttle response to care, whereas offspring are in good condition if they exhibit a la rge response. Because optimal care depends on the shape of the offspring fitness function, differences in pr edictions among models can primarily result from differences in how vulnerability is defined and in corporated. The models I present here predict that parents should care more for offspring or broods that exhib it a greater response to care, whether those broods are defined as more or less vulnerable. Ideally, vulnerability should be defined independently of the response to care, such as by the fitness of young without care. Brood Age I have demonstrated that the reproductive value of a brood is not a good predictor of optimal parental effort because high reproductive value is not necessarily associated with high benefits of care. Yet a brood’s increasing repr oductive value is a common explanation for why parental effort often increas es as a brood ages (CluttonBrock 1991). Although the model presented above does not explicitly incorporate brood age, its resu lts highlight the influence of some factors that change as a brood ages. I illustrated how the changing benefits of care with brood age are influenced by brood size and individual reproductive va lue and that simultaneous change s in these two factors could negate the effect of each, individually. This argument is supported by a study of nest desertion

PAGE 32

32 in mallards ( Anas platyrynchos ). Ackerman and Eadie (2003) used published data on mean daily egg survival in mallards to determine how clutch size changes over time due to partial clutch predation. By simulating partia l clutch predation, they compar ed seven-day old broods with 17day old broods of identical brood reproductive value and found that females deserted these clutches at equal rates, demons trating that the increase in re productive value of individual young does not cause investment to increase when partial brood mortality is incorporated. In contrast, parental care doe s increase with offspring age in many systems (e.g., Amat 1996; Ridgway 1988; van Iersel 1953), suggesting that the combined influence of brood size, expected survival to maturity, vulnerability, and care efficacy results in th e benefits of providing care to increase as offspring age. For exampl e, if brood size does not decline over time as expected based on mean survivorship, care might in crease over time. For example, in Ackerman and EadieÂ’s (2003) study, deserti on rates did not change with br ood age when the size of older broods was reduced to reflect identical reproduc tive value as younger broods. However, rates increased with the level of brood reduction among broods of the same age. This suggests that if the size of later-stage broods had been reduced by a lesse r amount, causing broods to be larger than expected from mean survivorship rates, th en females might be less likely to desert. Alternatively, parental care could increase w ith offspring age due to the correlated effect of time on the parentÂ’s reproducti ve opportunities. If reproductiv e opportunities for the parent become limited as the end of the reproductive season approaches, the costs of care may decrease as offspring age due to the correlated dec line in both missed mating opportunities and future mating opportunities in the current season. Simila rly, if the probability of survival to the subsequent reproductive season de clines as the investment in the current brood increases, then

PAGE 33

33 the costs of care may decrease as offspring age due to the correlated decline in the cost of care to future reproduction. The effect of future mating opportunities on care is most apparent in dynamic statevariable models of parental care in which future reproductiv e opportunities are explicitly incorporated and vary over time. For exampl e, although the amount of care was not allowed to vary in their models, Sargent (1990), McNama ra et al. (2000), and Webb et al. (2002) all predicted care to be more likely later in the s eason when future mating opportunities are low. In addition, Sargent (1990) and Webb et al. (2002) found that the like lihood of care decreases as the probability of offspring survival without care increases, which occurs as offspring age. Empirical evidence for this effect is mixed and th e difficulty of demonstrating such an effect due to the confounded changes in future mating oppor tunities, size-related f ecundity, and senescence is discussed by Clutton-Brock (1984). Howeve r two recent studies in which future mating opportunities were experimentally manipulated suggest that individua ls do reduce their reproductive effort as future mating opportunitie s decline (Bonneaud et al. 2004; Javois and Tammaru 2004). Recent empirical work examining the eff ects of brood size manipulation on parental behavior indicates a shift away from using a brood’s current re productive value to predict the amount of care parents should provide young. Br ood manipulation experiments used to test the prediction that larger broods should receive more care have produced mixed results (e.g., Maigret and Murphy 1997; Sanz and Tinbergen 1999; Tolonen and Korpimaki 1996; Wiklund 1990). A negative relationship between brood size and parental investment may occur if the reproductive value of individuals in enlarged broods is lower than that of those in unmanipulated or reduced broods. Tammaru and Hõ rak (1999) modeled how, when care is divided

PAGE 34

34 among young, as with provisioning of chicks, young in artificially enlarged broods may receive insufficient care for survival, rende ring the collective value of a larg e brood less than that of an intermediate-sized brood. This is consistent w ith the optimal brood size theory put forth by Lack (1947), in which intermediate br ood sizes yield the greatest life time reproductive success for the parent. This approach emphasizes brood reproductiv e value measured at a fixed point in time, as other models have done, but focu ses on reproductive value of broods after receiving care – e.g., using recruitment data (Hõrak 2003) – instead of prior to care. For species in which the brood’s current reproductive value equals zero, value of a brood that has received care is equivalent to the change in brood value that re sults from care. Therefore, using recruitment data to predict which broods should receive more care may be appropriate for species in which young are completely dependent on care for survival (but see Wolf and Wade 2001), including all birds and mammals and some reptiles, amphibians, and fish. A limitation of using recruitment data to pred ict which broods should receive more care is that enlarged broods may have received a diffe rent amount of care in the past than reduced broods. Recruitment rates indicate the br ood’s reproductive value following an unknown amount of care, and may not correlate with the increase in reproductive value per unit parental effort. Some combination of these data with m easures of parental investment, measured as a reduction in parental survival or future reproduction, may yield measures of fitness benefits per unit investment and would better indicate which broods offer the gr eatest returns per unit care. For species in which care is facultative – i.e., young do not require post-fe rtilization investment for survival – measuring brood reproductive value vi a recruitment rates is insufficient. Instead, the benefits of care in these species, which incl ude most invertebrates, fishes, amphibians, and reptiles, must be measured by comparing offspring fitness with and without care.

PAGE 35

35 Finally, I have assumed that all offspring in a brood receive an equal amount of care. However, parents often differentially allocate resources among broodmates (reviewed in CluttonBrock 1991) and the conclusions of the above mode ls can offer insight in to how care should be allocated among young and, more extremely, the circum stances that give rise to brood reduction. Broods often contain individuals th at differ in age, in paternity , or in vulnerability. These differences can create variation in offspring vul nerability or in the vulnerability-independent response to parental care, g Â’( C ), making it beneficial for parents to differentially allocate resources among young. Which offspring receiv e the most food depends on each individualÂ’s probability of surviv al without care ( S (0)) and response to care ( S Â’( C )). Differential allocation can be examined by m odifying equation (2-1) to incorporate two individual offspring. Assume a brood of two chicks and let the subscripts 1 and 2 denote chicks 1 and 2, respectively. Then, the parentÂ’s pres ent reproductive success is a function of care in the following manner: P S1( C , p ) S2( C ,1 p ) , (2-10) where p is the proportion of care going to chick 1. For a given amount of care, C , the proportion going to chick 1, p , can be optimized. Care should be di vided equally between the chicks unless they differ in their response to care. Such a diffe rence may arise as a result of differences in age, for example, due to different hatching dates, or due to competition among young for resources during an earlier stage, for example, via competit ion for placental resource s during gestation. In either case, the shape of S ( C ) may differ between young due to diffe rences in past investment. Assume that chick 1 and chick 2 differ such that l0,chick1 > l0,chick2 (e.g., Figure 2-3A). Because vulnerability is i nversely related to l0, chick 2 is more vulnerable than chick 1. As in the optimization of care to different broods, the more vulnerable chick, chick 2, should receive more

PAGE 36

36 care ( p* < 0.5; Figure 2-5). But as the total amo unt of care provided increases, the marginal benefit of caring for the more vulnerable chick w ill decrease until it equals the marginal benefit of caring for the less vulnerable chick, until even tually each subsequent increment of care is divided equally between chicks (as in Fretwell and Lucas 1970). Differential allocation should occur whenever broodmates differ in their marginal fitness gains from care; however, the added constraint of resource limitation can give rise to the extreme case of brood reduction. Figure 25 indicates that chick 2 should receive no care when the total allocation of parental resources to care is low, below a threshold value of approximately 0.05. This threshold will vary as a function of the difference between S1’( C ) and S2’( C ); the greater the difference, the higher the threshold value of to tal allocation to care below which one chick is excluded. This means that both resource lim itation and differences among broodmates in the response to care will influence whether or not th e parent withholds care from one offspring. Given this, parents might produce more young than can survive given typical resource availability if resource abundance is occasiona lly high enough to favor allocation to multiple offspring – an explanation for brood reduc tion initially proposed by Lack (1947) and subsequently formalized in a number of mathem atical models (e.g., Bonabeau et al. 1998; Pijanowski 1992; Temme and Charnov 1987). When re sources are abundant, the parent is able to provide enough care to be shared among all it s young, whereas when resources are scarce, the parent provides less care, which is pr ovided exclusively to one individual. Summary Offspring reproductive value is critical in dete rmining the fitness benefits parents receive when providing care. However, there is no sing le prediction for how care should change with offspring reproductive value. Instead, compone nts of offspring reproductive value, including brood size and vulnerability, influenc e optimal care differently. In addition, the effects of

PAGE 37

37 offspring age on optimal care remain unclear as ag e has yet to be defined independently of the parentÂ’s future mating opportunities. Studies of parental investment should shift toward examining how the benefits of care, measured as an increase in the reproductive value of young once care is provided, are influenced by brood ch aracteristics, instead of focusing solely on reproductive value either before or after receiving care.

PAGE 38

38 Figure 2-1. Present, future, and total reproductiv e success as a function of the amount of care a parent provides. 0.00.20.40.60.81.0 CareReproductive Success Present Total Future 0.00.20.40.60.81.0 CareReproductive Success Present Total Future

PAGE 39

39 Figure 2-2. The effect of brood size, n , on the optimal amount of pa rental care. A) Brood size affects the shape of function describing pr esent reproductive success. B) Optimal parental care occurs when d P /d C = -d F /d C ; therefore, C) optimal care increases with brood size. 0.01.0 01234 CareReproductive Success P( n =3) P( n =1) F 0.01.0 024681012 Cared/d C P( n =3) P( n =1) -F 246810 0.200.30 0.40 0.50 Brood size, nC * C * A B C 0.01.0 01234 CareReproductive Success P( n =3) P( n =1) F 0.01.0 024681012 Cared/d C P( n =3) P( n =1) -F 246810 0.200.30 0.40 0.50 Brood size, nC * C * A B C

PAGE 40

40 Figure 2-3. The effect of offs pring survival without care, l0, on the optimal amount of parental care under Model 2. A) Survival without car e affects the shape of function describing present reproductive success. B) Optim al parental care occurs when d P /d C = -d F /d C ; therefore, C) optimal care decreases as offspring survival without care increases. 0.01.0 01234 CareReproductive Success P( l0=0.5) P( l0=0) F 0.01.0 0246810 12 Cared/d C P( l0=0) P( l0=0.5) -F 0.00.20.40.60.81.0 0.00.20.40.6 0.81.0 Survival without care, l0C * C * A B C 0.01.0 01234 CareReproductive Success P( l0=0.5) P( l0=0) F 0.01.0 0246810 12 Cared/d C P( l0=0) P( l0=0.5) -F 0.00.20.40.60.81.0 0.00.20.40.6 0.81.0 Survival without care, l0C * C * A B C

PAGE 41

41 Figure 2-4. The effect of offs pring survival without care, l0, on the optimal amount of parental care under Model 3. A) Survival without car e affects the shape of function describing present reproductive success. B) Optim al parental care occurs when d P /d C = -d F /d C ; therefore, C) optimal care increases w ith offspring survival without care. 0.01.0 01234 CareReproductive Success P( l0=0.5) P( l0=0.2) F 0.01.0 02468 Cared/d C P( l0=0) P( l0=0.5) -F 0.00.20.40.60.81.0 0.00.20.40.6 0.81.0 Survival without care, l 0 C * C * A C B 0.01.0 01234 CareReproductive Success P( l0=0.5) P( l0=0.2) F 0.01.0 02468 Cared/d C P( l0=0) P( l0=0.5) -F 0.00.20.40.60.81.0 0.00.20.40.6 0.81.0 Survival without care, l 0 C * C * A C B

PAGE 42

42 Figure 2-5. Optimal allocation of parental effort to chick 1 as a function of the total amount of care provided. In this model, chick 2 is less vulnerable than chick 1 ( l0,chick1 > l0,chick2). Chick 2 is always expected to receive more can than chick 1, indicated by p * always being less than 0.5 0.00.20.40.60.81.0 0.00.20.40.60.81.0 CareProportion care to chick 1, p* 0.00.20.40.60.81.0 0.00.20.40.60.81.0 CareProportion care to chick 1, p*

PAGE 43

43 CHAPTER 3 NEST TENDING INCREASES REPRODUCTIVE SUCCESS, SOMETIMES – ENVIRONMENTAL EFFECTS ON PATE RNAL CARE AND MATE CHOICE Introduction The optimal amount of care to provide one’s young reflects a balance between the benefits of care to young and the costs to the parent’s residual reproductiv e value (e.g., Sargent and Gross 1993; Webb et al. 2002; Williams 1966). Consistent with this model, pa rents often reduce the care they provide when the costs of doing so ar e high (e.g., Brommer et al. 2000; Weimerskirch et al. 2001) and increase care when the benefits are high (Dale et al. 1996 ; Listøen et al. 2000). These natural selection pressures are not the on ly factors influencing parental investment decisions, as an increasing body of work demonstrat es that female mating preferences can select for male activity that is likely to improve offspring fitness (Mø ller and Thornhill 1998; Pampoulie et al. 2004; Tallamy 2000). For example, mating success can be associated with the quality of a potential mate’s nest (Reynolds and Jones 1999), of the care he will provide young (Forsgren 1997; Lindström et al. 2006; Östlund and Ahnesjö 1998), and whet her he is caring for a current brood (Petersen et al . 2005). Thus, natural and sexu al selection can simultaneously influence optimal care. Natural and sexual selection may favor differe nt amounts of care in species in which the choosy sex (e.g., females) is not the care-givi ng sex (e.g., males) (reviewed in Clutton-Brock 1991), as is the case for many invertebrates, fi shes, and birds (review s in Andersson 1994; Tallamy 2000). Specifically, imposing sexual selecti on on paternal care pred icts an increase in care above the natural selection optima (H oelzer 1989; Iwasa and Pomiankowski 1999; Kirkpatrick 1985). When sexual se lection is acting, there may al so be a conflic t between the interests of males and females (reviewed in Arnqvist and Rowe 2005). Females should favor males that provide the greatest fitness benefits to offspring, whereas males should balance

PAGE 44

44 benefits to offspring against the costs of providi ng care (e.g., Trivers 1972; Westneat and Sargent 1996). For species distributed across a range of e nvironmental conditions, the strength of this conflict can vary across environments. For ex ample, environmental conditions can influence both the costs of care to parent s (Weimerskirch et al. 2001) and benefits of care to offspring (Dale et al. 1996) such that the environments in which care is most beneficial are those in which care is also most costly (Hale et al. 2003). If the stre ngth of female preferences for parental males is correlated with the expected benef it of care to young, then preferences should be stronger in the environments in which care is most beneficial. Vari ation in care across environments, then, should reflect the changes in pa rental costs, benefits to offspring, as well as changes in the strength of mating preferences. In this study, I examine varia tion in parental care both w ithin and among populations of flagfish ( Jordanella floridae ). Care in flagfish is provided entirely by males, who defend nesting territories and guard, clean, and fan eggs from multiple females. Male parental behavior is variable both within (Hale et al. 2003; St. Mary et al. 2001) and among (C. M. St. Mary, unpublished data) populations and a com ponent of this variation can be attributed to variation in salinity (St. Mary et al. 2001). The benefits of parental care u nder various salinities are currently unknown; however, previous work suggests that care may be more beneficial in fresh water. Specifically, an increased rate of fungal infectio n in fresh water appears to reduce survival of unattended embryos (St. Mary et al. 2004). Consequently, egg cleaning may be more beneficial in fresh than in brackish water. Further, if egg cleaning is more beneficial to offspring in fresh water, then the strength of mating preferences for nest tending males should be stronger in fresh water than in brackish water.

PAGE 45

45 I will examine the effects of salinity on male activity prior to and after spawning in four populations of flagfish. Flagfish are native to both fresh and brackis h habitats. Therefore, I will examine the effect of native habitat on behavior al responses to salinity by observing males from both coastal and inland habitats. Behavior may differ among populations native to different habitat types (inland vs. coasta l) due to genetic drift resulti ng from reproductive isolation. Alternatively, behavior may differ consisten tly among habitat types, suggesting adaptation to local conditions. In flagfish, the amount of gene flow across the salinity gradient is unknown. However, the proximity of freshwater habitats in Florida to coastal salt marsh may facilitate gene flow across the salinity gradient within drainages. An effect of native habitat type on behavior would indicate that gene flow is restricted and that selection regimes differ betw een salinities. I will also examine female mating preferences by determining whether a maleÂ’s preand post-spawning activity is associated with his reproductive success. Male flagfish perform nesttending activities, such as fanning and nest-clean ing, prior to spawning (Bonnevier et al. 2003) and these activities may se rve as signals to potential mates of the quality of the nest and of the care a male will provide young (Tallamy 2000). In a ddition, females often mate repeatedly with the same male and a maleÂ’s behavior once he has eggs in his nest may influence whether a female will mate with him again (Tallamy 2000). Therefore, male behavior both prior to and after spawning will be examined w ith respect to his initial and subsequent reproductive success. An effect of male behavior on reproductive succe ss would suggest that ma le activity indicates a maleÂ’s interest in mating and/or that female s dynamically adjust their spawning activity in response to male behavior. I assume that female preferences that are ba sed on male activity are free to vary in strength and direction across salinity treatme nts and populations, whereas I assume that a maleÂ’s interest in mating should be similarly correlated w ith his activity across all

PAGE 46

46 treatments in much the same way that I expect courtship to be similar in all treatments. Indeed, field observations indicat e that courtship t-circling (Mertz and Barlow 1966) precedes spawning in both inland and coastal populations (personal ob servation). As a result, I assume variation across salinities and popul ations in the association between male behavior and reproductive success to indicate variation in female mating preferences. Methods Collection and Transportation Fish were collected from four sites in Florida between May and July of 2003 under Florida Fish and Wildlife Conservation Commission Scie ntific CollectorÂ’s Permit number FNC-03-015, U.S. Fish and Wildlife Service Special Use Permit numbers 58875 and 03008 for St. Marks National Wildlife Refuge and number 03 SUP 59 fo r Merritt Island National Wildlife Refuge. Seine nets, minnow traps, and dip nets were used to collect animals. Otter Creek (OC, Levy County) and Miccosukee (MC, Miami-Dade County) are inland and freshwat er. St. Marks (SM, St. Marks National Wildlife Refuge, Wakulla C ounty) and Merritt Island (MI, Merritt Island National Wildlife Refuge, Brevard County) are coasta l, with freshwater ar eas in close proximity to brackish areas. Animals were transported to the Florida State University (FSU) campus in Tallahassee, where the experiments were conducted, in insulated coolers. They were transferred to 1 m diameter wading pools at the FSU Mission Road Greenhouse, where they experienced the natural daylight cycle. All animals were return ed to their native sites within four months of collection. Acclimation and Experiments Responses to salinity in each of the four popul ations were examined in a factorial design with two native habitat types (coa stal and inland) crossed with two salinity treatments (fresh and brackish). Two populations were nested within each native habitat type. Males and females

PAGE 47

47 from each population were acclimated to either 0.2 ppt (hereafter referred to as 0 ppt) or 15 ppt salinity from their native salinity at a rate of 5 ppt every other day such that all animals reached their target salinity treatment on the same day. Fi sh were then maintained at these salinities for two weeks prior to experimentation. Brackish water (15 ppt) was made by mixing Instant Ocean Aquarium Salt (Aquarium Systems brand) with we ll water, whereas the freshwater treatment (0 ppt) consisted of unaltered well water. The experiment was conducted indoors at 28 ± 1 ° C with 14 h light – 10 h dark. Each male was placed in a 37.5 l aquarium with two ar tificial plants, a carpeted spawning mat, and a filter. One female from the same population and acclimated to the same salinity as the male was placed in a transparent plastic box within this aquarium and water was shared between the box and the aquarium. The female was maintained in the box for 48 h and then released into the aquarium. Twelve OC, nine MC, and 11 SM males were observed in each salinity treatment; 10 and 11 MI males were observed in the fresh and brackish treatments, respectively. Twenty-four h following the release of the female (Day 1), each pair was filmed for 20 min and then each spawning mat was inspected for eggs. Daily from Day 2 to Day 14, each spawning mat was removed and all eggs counted. Pa irs were filmed for 20 min the first day eggs were observed. All filming took place between 0800 and 1200 h. Male behavior was analyzed using Observer Pro 5.0 (Noldus Information Technology 2003). Each observation was divided among time spent at or away from the nest, sw imming, fanning, following the female, courtship (t-circling), or spawning. In a ddition, the frequencies of bites at and away from the nest, chases, and spawning events were recorded. See Hale et al. (2003) for definitions of these activities. A male was following the female, as opposed to ch asing her, if he was swimming no more than

PAGE 48

48 approximately three body lengths behind the female at approximately her same speed and did not make contact with her. Reproductive success was measured as whether or not a pair spawned (spawning success) and as the number of eggs received. Analyses Three composite responses were analyzed: number of approaches toward the female (chases plus follows), number of nest-tending activities (bites at the nest plus fanning events), and proportion time at the nest. I consider male activity to reflect a deci sion either to perform the activity at a level appropriate to the environmen t or to not perform the activity at all. Under this assumption, males that did not perform a gi ven activity did so either because they were categorically inactive or because they assessed the environment and decided that ‘no activity’ was optimal. Ideally, I would like to evaluate onl y those males whose activity level was adjusted to the environment, but this is not possible. As an alternative, each composite behavioral response was analyzed in two ways. First, it wa s treated as a binomial response variable with males scored as either exhibiting or not exhibitin g the response. The effects of salinity, native habitat type, and population on whether or not male s exhibited the response were analyzed using logistic regression. Second, the responses were treated as interval data, with the effects of the treatments on the frequency of the responses anal yzed using log-linear (Poi sson) regression. In these analyses, males that did not perform the activ ity were excluded. Each of these analyses is valuable in the interpretation of male behavior. The logistic regression evaluates activity as a binomial response and considers whether or not a male performed an activity, regardless of whether he was categorically inactive or he deci ded ‘no activity’ was optim al. In contrast, the log-linear regression evaluates the magnitude of activity and considers the male’s energetic investment into the activity.

PAGE 49

49 All analyses were conducted using PROC GENMOD in SAS version 8 (SAS Institute 2000) using logistic regression for binomial data and log-linear re gression for frequency data. For log-linear regressions, a ne gative binomial error distributi on was specified to reduce model deviance. Interaction terms (see below) were removed from full sta tistical models using backward elimination if P > 0.10. Main effects of salinity, native habitat type, and population (nested within habitat type) are included in al l analyses, regardless of the significance of their effect, in order to consistently removed variance explained by these variables from all analyses. Summary statistics provided in the text and figures are mean ± standard error. I analyzed the effects of salin ity, native habitat type, and population (nested within native habitat type) on male behavior dur ing a pre-parental phase and a parental phase. I limited my testing of interaction effects to that between sa linity and native habitat type. The pre-parental phase consisted of Day 1 observations of all ma les that had not spawne d in the 24 h since the female was released (73 of 82 males). The pare ntal phase began when eggs were first observed in a nest and parental behavior was measured from the observation of each male on the first day he had eggs (57 of 82 males). I also examined whether a pair’s reproductive success – either the probability of spawning or the number of eggs received – was influenced by male behavior (either pre-parental phase or parental phase), salinity, native habitat type, population (neste d within native habitat type), and all possible twoand three-way interactions betw een male behavior, salinity, and native habitat type. First, I examined the effect of pre-pa rental-phase male behavior on spawning success and number of eggs received measured across the entire 14 day trial. Second, I examined the effect of parental-phase male behavior on spawning su ccess and number of eggs received in the two

PAGE 50

50 days immediately following the observation, as fe males may continue to spawn with the same male once he has eggs in his nest. Analyses of whether or not males spawned or performed each activity include all males for which there were behavioral observations. Anal yses of the number of eggs spawned and the frequency of activity include only those males that spawned and performed the relevant activity, respectively. As a result, sample sizes of the latt er analyses are smaller than those of the former analyses (Tables 3-1 and 3-2). Results Salinity, Native Habita t Type, and Population The effects of salinity, native habitat type, and population (nested wi thin habitat type) on reproductive success were evaluated in models that also included male behavior as independent variables. Behavior during the pre-parental a nd parental phases were considered in separate models (Table 3-1). Only population (nested w ithin habitat type) had a consistent effect on initial spawning success across all three models (pre-parental behavior models, Table 3-1). Initial spawning success for Otter Creek was 54 %, for Miccosukee was 71 %, for Merritt Island was 50 %, and for St. Marks was 91 %. None of these variables had consistent effects on whether the pair spawned again af ter the initial spawning event (p arental behavior models, Table 3-1). Among males that spawned (57 of 82 males), salinity, native habitat type, and population (nested within habitat type) di d not influence the number of eggs received over 14 days (49.1 ± 8.9 eggs), the latency to spawn (4.2 ± 0.6 days), or the mean clutch size (11.0 ± 1.5 eggs), estimated as the total number of new e ggs observed divided by th e number of days on which new eggs were observed for a given male. In addition to the 57 pairs that were observed during the pre-parental phase and then subsequen tly spawned, nine pairs spawned on the day the

PAGE 51

51 female was released before a pre-parental obser vation could be made. These males necessarily had a shorter latency to spawn than the other males used in the analyses and received, on average, nearly twice as many eggs (80.4 ± 34.8 eggs), but had similar mean clutch size (13.8 ± 4.7 eggs). Four of these males were fr om the Miccosukee site, three from Merritt Island, and two from St. Marks. Sexual Selection on Male Behavior A male’s behavior influenced his reproductive success, thou gh it did so differently in fresh and brackish water. In fresh water, pre-pa rental behavior was more important to spawning success, whereas in brackish water, pa rental behavior was more important. In fresh water, males that tended their nest s during the pre-parental phase were more likely to spawn than males that did not (tending x salinity interaction for pre-parental behavior Table 3-1, Figure 3-1). If the direct benefits of nest-directed activity are greater in fresh than in brackish water, as I suggest they are (St. Mary et al. 2004), then females appear to increase their preference for male activity where the benefit of doi ng so is greater, particul arly if a male’s nesttending during the pre-parental phase indi cates his tending once he has eggs. In brackish water, pre-parental behavior was not related to spawning success. However, two activities performed during the parental phase were impor tant to subsequent reproductive success. Males who approached the female during the parental phase were more likely to spawn than males who did not perform these activities (Table 3-1, Figure 3-2) and this effect was stronger in inland populations. However, the males that approached the female actually received fewer eggs on average than males who did not ap proach (Table 3-1, Figure 3-3). Males who spent time at the nest were also more likely to spawn, and this effect was stronger in coastal populations.

PAGE 52

52 In both fresh and brackish water, females pref erred males that tende d the nest during the parental phase, but only if the pair was from a coastal population (Table 3-1, Figure 3-2). I expected that preferences for activities that increase offspring fitness should be stronger in fresh water than in brackish water. Assu ming spawning success indicates, in part, female mating preferences (i.e., single male choice test ; Shackleton et al. 2005 ; Wagner 1998) and that nest-associated activities (nest tending and presence at the nest ) increase offspr ing fitness (Klug and St. Mary 2005), then the data offer mixed suppor t for this prediction. Both pre-parental and parental phase nest-associated act ivities were important to spawning success, but their influences were not always greater in fresh than in brackis h water; males who spent time at the nest during the parental phase were more lik ely to spawn, but only in brackish water – the opposite of what I expected. Further, in both fresh and brackis h water, parental-phase nest tending increased spawning success. Behavioral Responses to Sexual Selection Based on the effects of male behavior on spawning success described above, I expected males to be more likely to perform certain activities in the salinities in which those activities increase spawning success. Specifically, I expected nest tending during the pre-parental phase to be more common in fresh water, for approaches and presence at the ne st during the parental phase to be more common in brackish water, and for nest-tending to be more common among coastal males. Nearly without exception, these expectations were not met. Pre-parental nest tending was not more common in fresh water (Table 3-2, Figure 3-4C,D), nor were parentalphase approaches more common in brackish water (Table 3-2, Figure 3-5A,B). Further, males were not more likely to spend time at the nest during the parental phase when in fresh than in brackish water (Table 3-2, Figure 3-5E). Finally, coastal males were not more likely to tend the

PAGE 53

53 nest during the parental phase than inland males, but among males that tended, coastal males tended more. Two activities that did not influence repr oductive success were affected by salinity. Among coastal males, pre-parental nest tending was more frequent in brackish water. In addition, coastal males were more likely than inland males to approach the female during the pre-parental phase and all males approached her more often in brackish water (Table 3-2, Figure 3-4A, B). In addition to these treatment effects, ma le behavior during the parental phase was influenced by clutch size. The more eggs a male had in his nest during the parental phase, the more likely he was to tend and to spend time at his nest (Table 3-2, Figure 3-6). Association between Pre-Paren tal and Parental Behavior Deciding whether or not to spawn with a male based on his nest tending prior to spawning may offer a female direct benefits either if such nest tending directly increases the success of offspring subsequently sp awned and reared in the nest or if the maleÂ’s activity toward an empty nest indicates his care of a future brood . I used log-linear and logistic regressions to determine whether the frequency or probability, resp ectively, of behavior prior to spawning (i.e., during the pre-parental phase) is a good indicator of a ma leÂ’s behavior once a nest contains eggs (i.e., during the parental phase). In the analyses of the probab ility of performing a particular activity, I included all males that spawned and examined the effects of all twoand three-way interactions between salinity, native habitat type, and Day 1 male behavior (pre-parental behavior) on male behavior on the first day with e ggs (parental behavior). In the analyses of the frequency of activities, I include d only males that performed the activity during the pre-parental phase and then subsequently spawned. This gr eatly reduced sample sizes precluding the testing

PAGE 54

54 of all possible interactio ns. Therefore, I pooled all males rega rdless of salinity treatment, native habitat type, or population. In general, male activity prior to spawning di d not predict post-spawn ing behavior (Table 3-3). Males that spent time at th e nest during the pre-parental pha se were equally likely to do so once they had eggs as males who did not spend tim e at the nest during the pre-parental phase. Further, whether or not a male approached th e female during the pre-parental phase did not predict whether he did so once he had eggs. Sim ilarly, whether a male tended the nest during the pre-parental phase did no t predict whether he did so once he had eggs. The number of times the male approached the female during the pre-parent al phase did not predict his approaches once he had eggs, nor did the frequency of pre-parental nest tending predic t that once he had eggs, or did the pre-parental time spent at the nest pr edict the time spent once he had eggs. Discussion Female Mating Preferences I found no evidence that the tenden cy to spawn is plastic in re sponse to salinity, as there was no main effect of salinity, native habitat type , or their interaction on reproductive success. However, I found evidence that se xual selection on male behavior varies across salinities and that selection on pre-parental beha vior differs from selection on pa rental behavior, indicating that females adjust their mating activity in response to male behavior. In fresh water, where the benefits of care ar e expected to be greater (St. Mary et al. 2004), males who tended nests during the pre-pare ntal phase had higher spawning success than males who did not tend. In brackish water, activity during the parent al phase influenced reproductive success; males that approached the female and spent time at the nest were more likely to spawn. In both salinities, coastal males were more likely to spawn if they tended the nest during the parental phase . If spawning success reflec ts female mating preferences

PAGE 55

55 (Shackleton et al. 2005; Wagner 1998) , as I suggest it does, then these results indicate that preferences are plastic in response to salinity and vary geographically. Females often choose mates based on the qual ity of their nest s ites (Kodric-Brown 1983; Reynolds and Jones 1999) or of the care they will provide yo ung (e.g., Forsgren 1997; Östlund and Ahnesjö 1998) and preferences for males based on these direct benefits of mate choice may change as factors defining quality change. Fo r example, Reynolds and Jones (1999) found that, in gobies, female preference for males with small, more cryptic nest en trances disappeared under low oxygen conditions, where the importance of water flow and egg fanning to offspring survival may outweigh the benefit of reduced nest pr edation. Nest tending in flagfish appears to reduce predation (Klug et al. 2005 ) and may reduce fungal infection (via the removal of infected egg or detritus from the nest). I found that female s are more likely to spawn with males that nest tend before spawning, but only in fresh water. The relative importance of removing detritus may be low in brackish water, where the rates of f ungal infection are lower (St. Mary et al. 2004), such that pre-parental tending is a less reli able indicator of offspring survival in this environment. If pre-parental nest tending is an indicator to females of direct benefits for offspring, then I might expect pre-parental nest tending to be correlated with nest tending after spawning, but this was not the case. An alternative adaptive ex planation is that pre-pa rental nest tending has immediate effects on the nest itself that influence survival of eggs once they are laid such that pre-parental nest tending indicate s hatching success. Indeed, rem oval of detritus from the nest prior to spawning may reduce fungal infection throughout embryo development (Cote and Gross 1993), as the spores of at least some oomycetes (e.g., Saprolegnia ) are attracted to dead material (Smith et al. 1985).

PAGE 56

56 Females may also choose mates based on the amount they are courted. For example, female green swordtails were more likely to re spond to males exhibiting courtship displays than to males performing other activities (Rosenthal et al. 1996), and similar pr eferences for courtship displays have been demonstrated in insects and birds (reviewed in Andersson 1994). In flagfish, males that approached the female in brackish water were more likely to spawn than males who did not. Because the rates of fungal infection ar e low in brackish water, females may shift their choice criterion from male activit y that may improve offspring fitn ess to activity that indicates a maleÂ’s eagerness to spawn, such as whether or not a male approaches. I have argued that females may be more like ly to spawn with males that tend the nest prior to spawning because of the direct benefits such tending offers. An alternative explanation is that females are not choosing mates based on th e direct benefits of parental care but on the indirect, genetic benefits to her offspring (re viewed in Andersson 1994). Females can improve the fitness of their offspring by selecting males of higher genetic quality (e.g., Parker 2003) and male behavior may be an indicator of male quality. As the energetic costs of activity vary with salinity, so may the reliability of behavior as an indicator. In fresh wa ter, where the energetic costs of nest tending are expect ed to be high (Evans 1993), wh ether or not males are tending their nests prior to spawning may be a better indi cator of male quality than in brackish water, where all males, regardless of condition or genetic quality, may be able to tend nests. However, while this explanation holds for the pattern in fresh water, it cannot explain the advantage of post-spawning nest tending in brackish water. Regardless of the type of benefit females ga in from their decisions , these data suggest that female reproductive decisions are plastic not only in the magnitude, and possibly the direction, of preference but also in the traits used to select mates. In addition, female preferences

PAGE 57

57 changed after spawning. These changes may reflect that females make different decisions when assessing males with versus without eggs, when assessing a male for the first time versus reassessing him after already spaw ning with him, or when assessing a male after a period of not spawning versus after having recently spawned. These three possibilities are confounded in this experiment but potentially reflect different mate choice decisions. Male Behavior The results clearly indicate a sexual selecti on advantage of pre-parental nest tending in fresh water, whereas this advantage was not pr esent in brackish water. Given this, and the understanding of how salinity influences embryo hatching success and adult metabolism, I would expect male nest tending to vary with salinity. However, males were not more likely to tend in fresh water and coastal males actuall y tended more in brackish water. Coastal males may have tended less in fresh than in brackish water as a result of negative effects of a novel environment, yet this is not entirely consistent with the results. If a novel salinity environment were to elicit such a respons e, I would expect to s ee effects of salinity on female fecundity, a measure I expect to be tigh tly linked with energy e xpenditure and metabolic rate. However, there were no effects of salinity or native habitat type on clutch size, total eggs laid, or latency to spawn. The results also indicate an advantage of approaching the female during the parental phase in brackish water. However, males were not more likely to approach and did not approach more in brackish water, sugges ting that relatively st rong sexual selection in brackish water does not result in plasticity in approach activity. Similarly, both coastal and inland males were favored if they tended the nest in brackish water, but they we re not more likely to nest tend in brackish water, despite the mating advantage.

PAGE 58

58 There are at least two possible explanations for why plasticity in male behavior did not mirror plasticity in female mating preferences: 1) responses to salinity are the direct consequence of metabolic effects or 2) male behavior reflects a balance of the fitness consequences across environments. I reject the first hypothesis becaus e there are no consistent effects of salinity on behavior either before or afte r spawning; although pr e-parental males approached the female more in brackish water, no other male activit ies exhibited the same pattern. The data are consistent with the second hypothesis. This hypothesis requires that th e increased offspring fitness and sexual selection benefits in fresh wa ter be balanced by the increased energetic or opportunity costs of care in fresh water. As a re sult, the optimum level of care for the male to provide remains constant across environments. Support for this hypothesis requires that the costs and benefits be explicitly quantif ied in both salinity environments. I expect sexual selection benefits to covary w ith the direct benefits of care to offspring, whereas the costs of care may vary independently. In this system, it is possible that the parental costs, direct benefits, and the sexual selection be nefits of care all positively covary such that environments with relatively high benefits also have relatively high costs. A consequence of this covariance would be that the stre ngth of sexual conflict over pate rnal care is variable across salinities, being stronger in fresh water, wher e costs and benefits ar e both high. To my knowledge, variation in the streng th of sexual conflict across populations has not been measured (Arnqvist and Rowe 2005), but I su ggest that a system in which th e benefits of care vary across environments offers a good opportunity to describe such variation because the direct benefits of choosing a good parent as a mate should also vary. Summary In summary, I found that male behavior in fluences reproductive success differently in different environments, suggesting that females use different traits to sele ct mates in different

PAGE 59

59 environments. However, plasticity in male beha vior does not reflect a re sponse to plasticity in sexual selection alone. Males appe ar to adjust their behavior in response to covarying selection pressures. These results reflect the conflict between th e interests of the c hoosy parent and those of the care-giving parent, as se xual selection favors providing more care than may be optimal to provide. In addition, they suggest that the resolution of sexual conflict is not static, but might vary across environments, as the strength of se xual selection and possibl y the costs of providing care vary.

PAGE 60

60 Table 3-1. Effects of pre-pa rental behavior and parental behavior on reproductive success. Spawning success Eggs spawned Model Predictor df 2 P N 2 P N Pre-parental phase 1 Approaches 1 0.07 0.79 73 0.18 0.67 30 Salinity 1 1.07 0.30 1.47 0.22 Habitat type 1 0.65 0.42 0.13 0.72 Population(Habitat) 2 8.32 0.016 0.48 0.79 2 Nest tending 1 1.40 0.24 73 1.10 0.29 30 Salinity 1 7.43 0.006 1.59 0.21 Habitat type 1 0.50 0.480 0.18 0.67 Population(Habitat) 2 9.33 0.009 0.67 0.72 Tending•Salinity 1 10.83 0.001 3 At nest 1 0.04 0.85 73 2.07 0.15 35 Salinity 1 1.37 0.24 4.96 0.026 Habitat type 1 0.65 0.42 0.09 0.76 Population(Habitat) 2 8.57 0.014 1.09 0.58 Parental phase 1 Approaches 1 1.93 0.16 57 0.19 0.67 42 Salinity 1 6.60 0.010 3.93 0.047 Habitat type 1 4.41 0.036 0.09 0.77 Population(Habitat) 2 2.62 0.27 2.77 0.25 Approaches•Salinity 1 4.45 0.035 4.08 0.043 Approaches•Habitat 1 4.03 0.045 2 Nest tending 1 8.84 0.003 57 0.97 0.32 37 Salinity 1 1.88 0.17 0.20 0.65 Habitat type 1 0.53 0.47 0.09 0.77 Population(Habitat) 2 1.51 0.47 0.32 0.85 Tending•Habitat 1 4.83 0.028 3 At nest 1 4.47 0.035 57 1.42 0.23 37 Salinity 1 3.84 0.050 0.01 0.94 Habitat type 1 1.65 0.20 0.0 0.96 Population(Habitat) 2 1.82 0.40 0.85 0.65 At nest•Salinity 1 3.21 0.07 Population is nested within habi tat type. Logistic and linear re gressions were used to examine effects on spawning success and the number of e ggs received, respectivel y, as a function of behavior during the pre-parental and parental phases. Intera ction terms with P > 0.2 were removed from the statistical models.

PAGE 61

61 Table 3-2. The effects of salinity, native habi tat type (inland versus coastal), and population (nested within habitat type ) on pre-parental behavior and parental behavior. Performed (yes/no) Frequency performed Variable Predictor df 2 P N 2 P N Pre-parental phase Approaches Salinity 19.50< 0.01736.69 0.0130 Habitat type 10.490.480.09 0.76 Population(Habitat) 26.150.051.08 0.58 Habitat•Salinity 15.670.02 Nest tending Salinity 11.070.30735.48 0.0227 activities Habitat type 10.170.680.01 0.92 Population(Habitat) 27.990.021.80 0.41 Habitat•Salinity 116.21 < 0.01 Time at nest Salinity 10.330.56732.20 0.1435 Habitat type 10.710.400.08 0.78 Population(Habitat) 22.460.290.89 0.64 Parental phase Approaches Salinity 10.380.54570.22 0.6442 Habitat type 10.010.910.09 0.77 Population(Habitat) 29.010.011.12 0.58 Clutch size 11.070.300.01 0.94 Nest tending Salinity 10.00.95571.59 0.2137 activities Habitat type 10.300.585.20 0.02 Population(Habitat) 21.310.521.18 0.55 Clutch size 111.19< 0.013.36 0.07 Time at nest Salinity 10.440.51570.18 0.6741 Habitat type 11.590.212.39 0.12 Population(Habitat) 20.150.930.30 0.86 Habitat•Salinity 15.180.02 Clutch size 18.83< 0.01 3.10 0.08 Population is nested within habitat type. Logi stic regression was used to examine the proportion of males that performed the activity. Log-linea r regression was used to examine the number of activities performed. Tests of interaction effect s were limited to the interaction between habitat type and salinity. Interaction terms with P > 0. 2 were removed from the statistical models.

PAGE 62

62 Table 3-3. The relationships between pr e-parental and parental male activity. Performed (yes/no) Frequency performed Predictor df 2 P N 2 P N Approaches 1 0.61 0.43 47 0.22 0.96 21 Salinity 1 2.95 0.09 Habitat type 1 0.07 0.79 Population(Hab) 2 0.05 0.09 Nest tending 1 2.29 0.13 47 0.04 0.84 20 Salinity 1 0.70 0.40 Habitat type 1 0.28 0.60 Population(Hab) 2 0.08 0.96 At nest 1 3.07 0.08 47 0.17 0.68 23 Salinity 1 0.04 0.83 Habitat type 1 63.00 0.43 Population(Hab) 2 0.13 0.94 Population is nested within habitat type. Logi stic regression was used to examine the proportion of males that performed the activity. Loglinea r regression was used to examine the number of activities performed, with males pooled across treatments.

PAGE 63

63 Figure 3-1. Spawning success as a function of pre-parental beha vior in fresh (A) and brackish (B) water. (A) Males that tended the ne st were significantly more likely to spawn than non-tenders in fresh water (Logistic regression, fresh water only: N = 38, tending: 2 1 = 10.88, P = 0.001; habitat type: 2 1 = 0.05, P = 0.82; population(habitat): 2 1 = 10.57, P = 0.005). (B) Tending and non-tending males were equally likely to spawn in brackish water (Logistic regression, brackish water only: N = 35, tending: 2 1 = 2.29, P = 0.13; habitat type: 2 1 = 1.54, P = 0.21; population(habitat): 2 1 = 1.37, P = 0.51). 0 20 40 60 80 100 InlandCoastal Native habitat type B 0 20 40 60 80 100 InlandCoastalPercent spawned Tended Did not tend A 0 20 40 60 80 100 InlandCoastal Native habitat type 0 20 40 60 80 100 InlandCoastal Native habitat type B 0 20 40 60 80 100 InlandCoastalPercent spawned Tended Did not tend A

PAGE 64

64 Figure 3-2. Spawning success as a function of parental behavior in fresh (A, C, E) and brackish (B, D, F) water. Approaching the female (A and B), tending the nest (C and D), and spending time at the nest (E and F) during the parental phase were each important in determining spawning success in brackish wa ter. Specifically, inland males needed to approach the female in brackish water to increase spawning success (B). In contrast, coastal males needed to spend time at and tend the nest in order to increase spawning success (D and F). Stars indicate significant differences. 0 20 40 60 80 100 InlandCoastal Tended Did not tend 0 20 40 60 80 100 Inland Coastal 0 20 40 60 80 100 InlandCoastal 0 20 40 60 80 100 InlandCoastal Approached Did not approach 0 20 40 60 80 100 Inland Coastal At nest Not at nest 0 20 40 60 80 100 InlandCoastalPercent spawned Percent spawnedPercent spawnedD F C E A B Native habitat typeNative habitat type* * * * * * 0 20 40 60 80 100 InlandCoastal Tended Did not tend 0 20 40 60 80 100 Inland Coastal 0 20 40 60 80 100 InlandCoastal 0 20 40 60 80 100 InlandCoastal Approached Did not approach 0 20 40 60 80 100 Inland Coastal At nest Not at nest 0 20 40 60 80 100 InlandCoastalPercent spawned Percent spawnedPercent spawnedD F C E A B 0 20 40 60 80 100 InlandCoastal Tended Did not tend 0 20 40 60 80 100 Inland Coastal 0 20 40 60 80 100 InlandCoastal 0 20 40 60 80 100 InlandCoastal Approached Did not approach 0 20 40 60 80 100 Inland Coastal At nest Not at nest 0 20 40 60 80 100 InlandCoastalPercent spawned Percent spawnedPercent spawned 0 20 40 60 80 100 InlandCoastal 0 20 40 60 80 100 InlandCoastal Tended Did not tend 0 20 40 60 80 100 Inland Coastal Tended Did not tend 0 20 40 60 80 100 Inland Coastal 0 20 40 60 80 100 InlandCoastal 0 20 40 60 80 100 InlandCoastal 0 20 40 60 80 100 InlandCoastal Approached Did not approach 0 20 40 60 80 100 InlandCoastal Approached Did not approach 0 20 40 60 80 100 Inland Coastal At nest Not at nest 0 20 40 60 80 100 Inland Coastal At nest Not at nest At nest Not at nest 0 20 40 60 80 100 InlandCoastal 0 20 40 60 80 100 InlandCoastalPercent spawned Percent spawnedPercent spawnedD F C E A B Native habitat typeNative habitat type* * * * * *

PAGE 65

65 Figure 3-3. The mean number of eggs a male r eceived (± SE) as a function of whether or not a male approached the female during the parental phase. All males are included regardless of whether or not they received eggs after their initial spawning. Inland and coastal populations were pooled. 0 5 10 15 20 25 ApproachedDid not approach Male ActivityEggs received Fresh Brackish 0 5 10 15 20 25 ApproachedDid not approach Male ActivityEggs received Fresh Brackish

PAGE 66

66 Figure 3-4. Pre-parental beha vior of males (i.e., observed on Day 1). (A) Proportion of males who made advances toward the female. (B) Number of approaches toward the female, excluding males that did not appro ach (n = 30). (C) Proportion of males who tended nests. (D) Number of nest tendi ng activities performed by males, excluding those that did not tend (n = 27). (E) Proportion of males who spent time at the nest. (F) Percent time spent at the nest by male s, excluding those who spent no time at the nest (n = 35). Means and SE are calculated from ln-transformed values and backtransformed for the figures. Star s indicate significant differences. 0 5 10 15 20 25 30 35 40 45 InlandCoastal 0 2 4 6 8 10 12 14 InlandCoastal Native Habitat Type D F Native Habitat Type 0 20 40 60 80 100 InlandCoastal C E 0 20 40 60 80 100 InlandCoastal Fresh Brackish APercent males Percent malesPercent males 0 2 4 6 8 10 12 14 16 18 InlandCoastal BNo. approaches Percent time at nest No. tending activities 0 20 40 60 80 100InlandCoastal* * * * 0 5 10 15 20 25 30 35 40 45 InlandCoastal 0 2 4 6 8 10 12 14 InlandCoastal Native Habitat Type D F Native Habitat Type 0 20 40 60 80 100 InlandCoastal C E 0 20 40 60 80 100 InlandCoastal Fresh Brackish APercent males Percent malesPercent males 0 2 4 6 8 10 12 14 16 18 InlandCoastal BNo. approaches Percent time at nest No. tending activities 0 20 40 60 80 100InlandCoastal 0 20 40 60 80 100InlandCoastal* * * *

PAGE 67

67 Figure 3-5. Parental be havior of males (i.e., observed on th e first day of eggs). (A) Proportion of males who made advances toward the fe male. (B) Number of approaches toward the female, excluding males that did not a pproach (n = 42). (C) Proportion of males who tended nests. (D) Number of nest tending activities performed by males, excluding those that did not te nd (n = 37). (E) Proportion of males who spent time at the nest. (F) Percent time spent at the ne st by males, excluding those who spent no time at the nest (n = 41). Means and SE are calculated from ln-transformed values and back-transformed for the figures. Stars indicate signi ficant differences. 0 10 20 30 40 50 60 70 InlandCoastal 0 20 40 60 80 100 InlandCoastal 0 30 60 90 120 150 180 InlandCoastal 0 20 40 60 80 100 InlandCoastal Fresh Brackish 0 20 40 60 80 100 InlandCoastal 0 5 10 15 20 25 InlandCoastal Native Habitat Type B D F Native Habitat Type A CPercent males Percent malesPercent males No. approaches Percent time at nest No. tending activitiesE* *Habitat 0 10 20 30 40 50 60 70 InlandCoastal 0 10 20 30 40 50 60 70 InlandCoastal 0 20 40 60 80 100 InlandCoastal 0 20 40 60 80 100 InlandCoastal 0 30 60 90 120 150 180 InlandCoastal 0 30 60 90 120 150 180 InlandCoastal 0 20 40 60 80 100 InlandCoastal 0 20 40 60 80 100 InlandCoastal Fresh Brackish 0 20 40 60 80 100 InlandCoastal 0 20 40 60 80 100 InlandCoastal 0 5 10 15 20 25 InlandCoastal 0 5 10 15 20 25 InlandCoastal Native Habitat Type B D F Native Habitat Type A CPercent males Percent malesPercent males No. approaches Percent time at nest No. tending activitiesE* *Habitat

PAGE 68

68 Figure 3-6. The percent of male s tending their nests (A) and sp ending time at their nests (B) increased with the number of eggs in the nest . Note that the bin size ranges for egg number are not constant. Nu mbers above bars indicate the number of males in that bin. 0 20 40 60 80 100 1-1011-2021-4041-60Percent males nest tending 0 20 40 60 80 100 1-1011-2021-4041-60 No. eggs in nestPercent males at nestn = 42 n = 7 n = 4 n = 4 n = 42 n = 7 n = 4 n = 4 B A 0 20 40 60 80 100 1-1011-2021-4041-60 0 20 40 60 80 100 1-1011-2021-4041-60Percent males nest tending 0 20 40 60 80 100 1-1011-2021-4041-60 No. eggs in nest 0 20 40 60 80 100 1-1011-2021-4041-60 No. eggs in nestPercent males at nestn = 42 n = 7 n = 4 n = 4 n = 42 n = 7 n = 4 n = 4 B A

PAGE 69

69 CHAPTER 4 PRE-SPAWNING BEHAVIOR AND NEST-FANNING PREDICT CLUTCH SUCCESS Introduction The amount of parental care a parent provides its young may vary among environments if the risks facing offspring vary. A number of models have predicted that offspring that are more vulnerable or face greater risks s hould receive more parental care (A ndersson et al. 1980; Dale et al. 1996; Montgomerie and Weatherhead 1988; and Ch apter 2). Underlying this prediction is the assumption that vulnerable offspring will benefit more from parental care, yet this assumption remains largely untested. In many species with parental care, care is facultative – i.e., offspring are not entirely dependent upon care for survival. These species provide an excel lent opportunity to test the assumption because the survival of offspring w ith and without care can easily be compared between highand low-risk environments. If the assumption is met, then parental care should be expected to vary among environm ents as the models predict. Variation in the benefits of parental care for offspring can have an impact not only on parental effort, but also on mating preferences that are based on parental care activity. Female mate choice appears to play a large role in shap ing male parental care, as preferences for both good fathers and males already caring for young have been demonstrated (e.g., Forsgren 1997; Møller and Thornhill 1998; Pete rsen et al. 2005; Reynolds and Jones 1999; Tallamy 2000) and mating with a male that will prov ide high quality parental care s hould confer direct benefits to offspring (Hoelzer 1989). Recent work examining variation in mati ng preferences has focused on whether the benefits of mate choice vary among environmen ts, causing mating preferences to be contextdependent. These studies have s hown that both the direct (paren tal care, protec tion and food for

PAGE 70

70 mate) and indirect (good genes, sexy sons) be nefits of mate choi ce can vary with the environment. For example, the correlation between sexually select ed traits and fitness has been shown to depend on male condition in stalk-eyed flie s (David et al. 2000), so cial context in sideblotched lizards (Alonzo and Sinervo 2001), and season (Schmoll et al. 2005) and density (Welch 2003) in grey tree frogs. Although variation across environments in the benefits of mate choice should result in varia tion in mating preferences, studie s linking the correlation between male phenotype and offspring fitnes s to female mating preferences are rare (Forsg ren et al. 1996; Tomkins et al. 1999). In Jordanella floridae (flagfish), males provide parent al care for offspring and the parental care traits preferred by females and the strength of female preferences vary with salinity (Chapter 3). Here, I examine whether this variati on reflects variation in be nefits – i.e., variation in the association betwee n male behavioral trai ts and clutch success. Territorial male J. floridae exhibit a set of conspicuous acti vities involved in courtship, mating, and parental care (Hale et al. 2003; Me rtz and Barlow 1966; St. Mary et al. 2001). When caring for embryos, males fan the nest and b ite at the nest substrate, which appears to remove debris. Egg fanning is commonly thought to aerate developing embryos (van Iersel 1953), but the effect of fanning on embryo development and oxygen consumption is not known. Further, males exhibit nest biting and fanning prio r to spawning (e.g., Chapter 3; Bonnevier et al. 2003), suggesting that they may be involved in nest preparati on (Foster 1967) or in signaling male quality to potential mates. J. floridae embryos can survive without parental care, but their hatching success in the absence of care depends on the salinity environment. In a st udy by St. Mary et al. (2004), embryos reared in fresh water (~ 0 ppt) were less likely to hatch than embryos reared in brackish

PAGE 71

71 water (5 – 15 ppt), apparently due to a higher prevalence of fungal infection in fresh than in brackish water. Because embryos reared in fres h water are more vulnerable to infection in the absence of care than those reared in brackish water, a number of models predict that embryos will benefit more from care when in fresh than wh en in brackish water (e.g., Chapter 2; Dale et al. 1996). These models suggest that the strength of natural selection on parental care should be greater in fresh water where the benefits of care for offspring are gr eater. However, sexual selection also act on male behavior, and the patte rn of care across salinit ies that is favored by sexual selection is not necessarily the same pattern favored by natural selection. Males can care for multiple clutches simultaneously, so females may have the opportunity to choose among males that are curren tly caring for young and thos e that are not. In addition, some males may already be caring for th e female’s offspring. Females are able to partition their daily egg complement among multiple males or to the same male upon repeated occasions, spawning between five and 20 eggs in any given spawning event. Therefore, a male currently caring for offspring may be caring for th ose of the female that is assessing him. Female preferences for males at these different nest stages appear to differ. In a previous study in J. floridae (Chapter 3), I evaluated whether a male’s activity could predict his subsequent mating success and whether the asso ciation between activity and mating success was affected by salinity. I assumed that spawning success reflected, in part, female mating preferences. Pairs were housed together for two weeks and, thus , females had the opportunity to assess the male when his nest was empty and, for those pairs that spawned, again when his nest contained a clutch of eggs. In fresh water, a male’s activity while his nest was empty (‘preparental phase’, Chapter 3) pred icted whether or not the pair would spawn or not over the two weeks. Males that tended the nest were more likely to spawn than males that did not tend the

PAGE 72

72 nest, when the nests were empty, and this was c onsistent with the hypothesi s that the strength of female mating preferences are grea ter in the salinity in which the benefits of care for offspring are greater. In contrast, prefer ences for males tending nests that contained clutches (‘parental phase’, Chapter 3) were equally stro ng in fresh and brackish water. The goal of this study is to evaluate the eff ect of male activity on clutch success in two salinity environments for one of the J. floridae populations examined the earlier study. I evaluate the alternative hypotheses that 1) the be nefits of preand postspawning nest tending for offspring are greater in fresh wate r, consistent with predictions of parental care models and 2) the benefits for offspring vary with salinity in a ma nner consistent with the direction of female mating preferences. Based on the results of the pr evious study of female mating preferences, the second hypothesis generates two predictions. First, males that tend their nest prior to spawning should have higher clutch success than non-tending males, and this difference should be greater in fresh than in brackish water. Second, post-spawning tending should increase clutch success equally in both fresh and brackish water. The effects of preand post-spawning ne st tending on hatching success and hatchling mass were evaluated in fresh and brackish water for males from Merritt Island, Florida. I also evaluated the effect of pre-spawning appro aches toward the female because approaches influenced spawning success in some treatments in the previous study. Both hatching success and hatchling mass are know n to be influenced by parental care (e.g., Crespi and Lessig 2004; Eggert et al. 1998) and may indica te benefits of mating decisions that are based on parental care activity.

PAGE 73

73 Methods Collection, Transport, and Acclimation Male and female Jordanella floridae were collected from Merritt Island National Wildlife Refuge (Brevard County), Florida, in May and November 2005 and June 2006 under U.S. Fish and Wildlife Special Use Permit number 03 SUP 59. Salinity at this collection site ranged from 3 to 15 ppt over two years of bimont hly sampling and sites within approximately 4 km2 varied between 2 and 18 ppt when measured on a single day (appendix). Fish were transported in their native water in insulated c oolers to Florida State University, Tallahassee, where experiments were conducted. Water at the collection site measured 3.4 ppt in May 2005, 3.0 ppt in November 2005, and 5.4 ppt in June, 2006. The fish were removed from their native water upon arriving at the laboratory and males and females were separately adjusted to target salinities of either freshwater (~0.2 ppt hereafte r referred to as 0 ppt ) or brackish (15 ppt) at a rate no greater than 5 ppt every two days. These two salinities are referred to as the natal salinities, as they are the salinities at which cl utches were spawned. Saline water was mixed by adding Instant Ocean Aquarium Salt (Aquarium Sy stems brand) to well water. Fresh water consisted of unaltered well water. Animals were fed frozen brine shrimp (San Francisco brand) daily and dried algae (Hikari brand sinking pellet s) on alternate days. They were maintained indoors at 26 °C with an artificial light regime of 14 h light: 10 h dark. Experimental Design and Protocol Rearing salinity, artificial fanning, and a male ’s pre-spawning activity were manipulated in a three factor design. Ar tificial fanning was applied as a surrogate for post-spawning nest tending in order to rear embryos independently of the father. Th e four combinations of rearing salinity (0 or 15 ppt) and fanning tr eatment (fanning present or absent) were applied to clutches after they, and the nest in which they were sp awned, had been removed from their parents and

PAGE 74

74 placed in a separate rearing aquarium. These treatments were blocked across adult spawning pairs. The male’s pre-spawning activity, in cont rast, was measured before a pair spawned and thus, characterized the natal environment, including the nest. Pairs were nested within activity category (performed/did not perform) as determ ined by whether or not the male of the pair performed specific pre-spawning activities. Two pre-spawning male activities were evaluated: nest tending, which included biting at the nest substrate and fanning the nest, and approaching the female, which included chases and follows (Chapter 3). The introduction of pairs to spawning aquaria was identical to that used in Chapter 3. One male and one female were introduced to a 37 .5 l spawning aquarium of the same salinity as their housing tank and were separated by a transparen t, plastic mesh partition. Each side of the partition was equipped with artificial plants and the male’s side also contained a carpeted “nest” (10 cm x 10 cm) and a water filter. Two days following the introduction of pairs to these aquaria, the mesh partition was removed. The following day (Day 1), each pair was filmed for 15 min to quantify whether the male approached th e female or tended the nest. In all cases, the pair had not yet spawned, so these observations indicate prespawning (pre-parental, sensu Chapter 3) behavior. Nests were checked for eggs daily beginning on Day 1 and continuing for three to eight weeks. Pairs set up in the su mmer of 2006 were maintained longer than pairs established in 2005 because I noti ced that spawning ra tes notably increased about four weeks after pairs were introduced. Spawning pairs were maintained until they had produced a clutch for each of the four salinity-by-fanning treatment combinations. However, not all pairs produced four clutches, resulting in an incomplete block design. Individuals that did not spawn over the three to eight week period were rem oved and later placed in a new pair.

PAGE 75

75 Upon finding a clutch in a nest, the nest was removed from the spawning pair and placed alone in a 20 l rearing aquarium (20 cm x 50 cm x 20 cm deep) at a salinity of either 0 or 15 ppt and a small water pump directed for horizontal flow approximately 1 cm above the glass bottom. This aquarium was not aerated aside from any aeration caused by flow from the water pump. The pump was placed near the substrate and did not create any visible disturbance of the water surface. Clutch sizes ranged from seven to 84 (mean = 36, SD = 23) embryos. The blocked design required that approximately half of the cl utches were reared at a different salinity than their natal salinity. An earlier study in which embryos were spawned in fresh water and then reared at salinities rang ing from 0 to 15 ppt suggests that the effect of such movement is small and should not obscure the effects of treatments in the current study (St. Mary et al. 2004). Clutches placed in the “fanning” treatment we re exposed to water flow for approximately 2 s every 20 s, corresponding to the mean fanning bout duration and inter-bout interval for fanning by males identified in an earlier study (Hal e et al. 2003). The water flow rate of the pump was set to 17 gallon/h for ~30 cm verti cal flow (Little Giant 40 GPH Statuary Fountain Pump). Flow cycle was regulated using Ho meSeer Home Control Software (HomeSeer Technologies 2005). The software was programmed to send a signal every 20 s to turn on a unit (X10 appliance module) attached to each pump. A second signal was sent 2 s later telling the unit to turn off. Nests were inspected daily and all hatchl ings and live eggs were counted. Upon hatching, fry were removed to a separate 1.2 l container. Two days following hatching, the embryos were sacrificed in MS-222 and preserve d in 10 % formalin. The fry were later freezedried and weighed. Average hatchling mass was obtained by simultaneously weighing 10-20 hatchlings from a clutch and di viding by the number weighed.

PAGE 76

76 Analyses The effects of two types of parental care (tending of th e nest prior to spawning and simulated fanning of the clutch) and one non-parental care activity (approaching the female prior to spawning) on hatching success and hatchling mass were evaluated. Tending and fanning are assumed to be parental care because they are di rected toward the nest and are likely to have direct effects on offspring fitness (Trivers 1972). In contrast, approaching the female when the nest is empty is not assumed to be a parental care because the male’s activity is not directed toward the nest nor can it serve to reduce egg predation if the nest is empty. The effects of natal salinity, rearing salinity, fanning treatment, male activity (tending or approaching), and clutch size on hatching success and ln(hatchling mass) were evaluated using logistic and log-linear regressi on, respectively, using generali zed linear mixed models in the lme4 package (Bates and Sarkar 2006) of the programming language R (R Development Core Team 2006), which uses restricted maximum lik elihood to estimate model fit and parameter values. The effects of the two pre-spawning male act ivities were evaluated in separate models because they were not considered independent – ma les that tended the nest were also likely to approach the female (see Results). In addi tion, evaluating the effects of these activities separately parallels the analysis of the eff ects of activity on spawning success conducted in Chapter 3, to which I intended to comp are the results of the current study. Main effects and all possible interactions be tween rearing salinity, fanning treatment, and male activity, as well as the main effects of na tal salinity, clutch size (a covariate) and the random effect of pair, were included in each mode l. The significance of each term was evaluated by comparing the deviances of progressively simp ler models. Deviance is defined as negative twice the difference in likelihood between the model of interest and a saturated model (i.e., a

PAGE 77

77 model with a separate parameter describing each data point). Beginning with the model containing terms for all the main and interaction effects, I sequentially removed the term that would result in the smallest change in model likelihood. This allowed me to compare the likelihood of a series of nested models, each with one fewer parameters than the previous model. The deviance of each model was compared with that of the model with one fewer parameters using analysis of deviance, which employs a li kelihood ratio test to determine whether removing a term results in a significantly poorer fit of the model to the data (Agresti 1996). Therefore, the goal was to identify the simplest model with a deviance not significantly different from more complex models. The likelihood ratio test statistic is approximately 2-distributed with degrees of freedom equal to the difference in the number of parameters of the models being compared (Agresti 1996) – in this ca se, degrees of freedom = 1. Additional analyses were necessary in orde r to present the mean and standard error estimates for clutch size, hatching success, and ha tchling mass while controlling for the effects of fixed and random factors. These values were plo tted as the grand mean of clutch size, hatching success, or hatchling mass plus residuals from re gression models including fixed effect terms for those factors for which I wanted to control and the random effect term for pair. However, the lmer function of R does not calcula te residuals for logistic and log-linear regression models. Therefore, new regression models with Gaussian error distributions were evaluated in which clutch size and hatchling mass were ln-transform ed and the probability of hatching was arcsinesquare root-transformed (Sokal and Rohlf 1995). Residuals from these models were then added to the appropriately-transformed grand mean value. These adjusted values were used to calculate treatment means and standa rd errors before back-transforming to the appropriate scale. The terms used in each model are provided in the figure legends.

PAGE 78

78 Results A total of 55 clutches were obtained from 19 pa irs. Nine of the pairs spawned four times, yielding clutches for each of the four salinity-by-fanning treatments (i.e., nine complete blocks), with the remaining pairs spawni ng between one and three clutches each. Pre-spawning behavior was quantified for 15 of the pairs (including eight complete blocks ) and 45 clutches. Five of these pairs included a male that tended the nest and eight included a male that approached the female. All of the males that tended th e nest also approached the female. Clutch Size Clutch size ranged between seven and 84 eggs and was marginally affected by the maleÂ’s activity prior to spawning (Figure 4-1). Clutch size was not affected by the salinity of the natal aquarium. The effects of whether or not the ma le tended the nest or approached the female on clutch size were evaluated in separate models , each of which included terms for natal salinity, the interaction of natal salinity with male activity, and the random effect of pair. Males that tended the nest had marginally la rger clutches than males that did not tend the nest (Likelihood ratio test, tending: LR = 3.12, df = 1, P = 0.08; natal salinity: LR = 0.25, df = 1, P = 0.61; interaction: LR = 2.09, df = 1, P = 0.15). Simila rly, males that approached the female also had marginally larger clutches (Likelihood ratio test, approaching: LR = 2.94, df = 1, P = 0.09; natal salinity: LR = 0.63, df = 1, P = 0. 43; interaction: LR = 2.46, df = 1, P = 0.12). This pattern may indicate that either females spawned more e ggs with active males or active males consumed fewer eggs in the period between spawning and co llection of the clutch. Because clutch size varied with treatment, clutch size was included as a covariate in the analyses examining hatching success and hatchling mass. However, there was no effect of clutch si ze on hatching success or hatchling mass when male activity wa s taken into account (see below).

PAGE 79

79 Hatching Success Hatching success was evaluated as a function of the fixed eff ects of natal salinity, rearing salinity, fanning treatment, male activity, and clutch size, and the random effect of pair. In the first analysis, male activity was scored accordi ng to whether or not the male approached the female prior to spawning. In a second analysis , male activity was scored according to whether or not the male tended the nest prio r to spawning. In a third anal ysis, the effects of fanning were evaluated without male activity. This third analysis was cond ucted because male activity was scored for only a subset of the pairs (45 clutches, 15 pairs, 8 comp lete pair blocks); therefore, by removing male activity from the model, the sample size in the regression could be increased to 55 clutches and 9 complete pair blocks and a pr esumably better estimate of the effects of fanning and salinity could be obtained. Activity scored as ‘nest-tending’ Hatching success was influenced by the interaction between rearing salinity and whether or not a male tended his nest prio r to spawning (tending by rearing sa linity interaction, Table 41). Qualitatively, tending increased hatching success when clutches were reared in fresh water and decreased hatching success when clutches were r eared in brackish water (Figure 4-2), but these effects of nest tending were not significant when fresh and brackish treatments were evaluated separately (fresh water only: te nding, LR = 0.68, df = 1, P = 0.41, tending x fanning interaction, LR = 0.26, df = 1, P = 0.61; brackish water only: tending, LR = 0.002, df = 1, P = 0.97, tending x fanning interaction, LR = 2.82, df = 1, P = 0.09). For some of the pairs that spawned clutch es for multiple treatments, it was possible calculate the within-pair effect of salinity on hatching success. Hatching success was 8.9 % (± 8.5 % SE, N = 3) higher in fresh water th an in brackish water among nest tending males and only 1.6 % (± 5.6 % SE, N = 8) higher among non-tending males.

PAGE 80

80 Activity scored as ‘approaching the female’ Approaching the female prior to spawning had a similar effect on hatching success as tending the nest did; there was a significant eff ect of the interaction between approaching the female and salinity (Table 4-1). Approachi ng the female increased hatching success when clutches were reared in fresh water and decrease d hatching success when clutches were reared in brackish water (Figure 4-2), though these effects of nest tending were not significant when fresh and brackish treatments were evaluated separately (fresh water only: approaching, LR = 0.004, df = 1, P = 0.95, approaching x fanning interactio n, LR = 3.10, df = 1, P = 0.078; brackish water only: tending, LR = 1.10, df = 1, P = 0.30, tending x fanning interaction, LR = 0.006, df = 1, P = 0.94). Evaluating the effect of salinity on hatchi ng success among pairs that spawned multiple clutches, hatching success was 7.1 % (± 4.4 % SE, N = 6) higher in fresh water than in brackish water among males that approached the female a nd was 0.6 % (± 5.6 % SE, N = 8) less in fresh water among non-approaching males. Clutch size and natal salinity did not influen ce hatching success in either the nest tending or approaching models. Fanning and salinity, alone Fanning increased hatching success (Figure 4-2), but there was no effect of salinity or the fanning by salinity interaction on hatching success (Table 4-1). Among pairs for which multiple clutches were evaluated, fanning increased hatc hing success more for clutches reared in fresh water (8.5 % ± 3.2 SE, N = 11) than for those rear ed in brackish water (6.6 % ± 7.9 SE, N = 12). Together, the results of these three anal yses indicate that tending the nest and approaching the female prior to spawning, and fanning the nest once it contains a clutch, each increase hatching success more in fresh than in brackish water (Figure 4-1), even though these

PAGE 81

81 effects were not always significant when evalua ted separately for each salinity. Further, the effect of salinity was similarly influenced by nest tending, approaching, and fanning. Hatching success was higher in brackish than in fresh wa ter in the no tending, no approaching, and no fanning treatments, but was higher in fresh than in brackish water for the tending, approaching, and fanning treatments. Hatchling Mass Hatchling mass was not significantly influen ced by rearing salinity, fanning treatment, male activity, clutch size, or natal salinity (Tab le 4-2, Figure 4-3). St atistical models were reduced to include only the main effects of re aring salinity, fanning treat ment, and male activity, but the effects of these terms were not significant (Table 4-4). Discussion The goal of this study was to identify whether the association between clutch success and male behavior, both observed (tending and appr oaching) and simulated (fanning), differs between fresh and brackish water. There are two alternative sets of predictions for how these components of male behavior shou ld influence clutch success. The first set of predictions is based on the effect of salinity on hatching success in J. floridae , evaluated in light of current models of optimal parental investment. J. floridae embryos are more vulnerable in fresh water than they are in brackish water. Numerous ve rbal models of optimal parental effort have predicted that offspring that are highly vulnera ble should receive more care than less vulnerable offspring because vulnerable offspring will benefi t more from care (e.g., Andersson et al. 1980; Dale et al. 1996; Montgomeri e and Weatherhead 1988). In Ch apter 2, I formalized this prediction in a mathematical model and predicted th at as the survival of a clutch that is not provided care declines, the benefit of receiving ca re should increase (Chapter 2, Model 2). As a

PAGE 82

82 result, the model predicts that the benefits of pare ntal care – in this study, the effects of male nest tending and artificial fanning – should be greater in fresh th an in brackish water. Alternatively, female preferences for nest-asso ciated activity vary with salinity, but they are not always stronger where th e benefits of parental care fo r offspring are assumed to be greater (Chapter 3). For exampl e, female preferences for males that are caring for clutches are equally strong in fresh and brackis h water, suggesting th at the benefits of parental care for offspring may be the same in the two salinities, contrary to the predictio ns of current models. The goal of this work was to quantify the eff ects of parental care on developing clutches to simultaneously test predictions of current parental care models and to explore the possible direct benefits to offspring of female mating preferences. Context-Dependent Association between Male Traits and Fitness These data provide evidence that the direct be nefit of parental care – specifically, tending the nest prior to spawning – varies among salinity environments. If male activity also indicates heritable components of male quality , then parental care activity in J. floridae may serve the dual roles of increasing clutch success directly and signaling to potential ma tes not only the direct benefits of mating with a given male, but also the indirect , genetic benefits. The data are consistent with the predictions of the mathematical model presented in Chapter 2. I found that tending the nest prior to spawning increased hatching success to a greater extent in fresh than in brackish water. Therefore, J. floridae , and species with similar parental care systems, may be good species in which to te st these models’ predictions for variation of parental care as a function of offspring vulnerability – namely, that offspring of greater vulnerability should receive more care. This predic tion is evaluated in light of the data presented in Chapter 3 in the general Discussion (Chapter 6).

PAGE 83

83 I evaluated the effects of two parental care activities (preand post-spawning nest tending) and one non-care activity (pre-spawni ng approaches toward the female) and found a consistent trend in the effect of activity on ha tching success. In fresh water, clutches that received parental care or were fathered by males that approached the female had higher hatching success than those that were not. In brackish water, these positive effects were either reduced or reversed. In addition, I found the effect of sali nity on hatching success to be affected similarly by parental care and non-parental care activity. Brackish water in creased hatching success when clutches did not receive care and were not fathered by males that approached, but this effect of salinity was reduced or reversed for clutches that received care or had fathers that approached the female. These data provide evidence that the amount of variance in clutch success that can be explained by variance in the fatherÂ’s traits can va ry with the environment. A similar effect of environment has been demonstrated in species la cking parental care. For example, Sheldon et al.(2003) showed that male coloration in moor frogs ( Rana arvalis ) predicted whether offspring would survive the threat of predation from larger predators, but not small predators, indicating that male color is a better indicator of offspr ing fitness in some predator environments than others. Similarly, Welch (2003) found that grey tree frog ( Hyla versicolor ) males with long calls father tadpoles that are larger at metamorphosis when the tadpoles are reared at high density, but that their size was equal to those of shor t call-males when reared at low density. In these two species, an asso ciation between male traits and offspring fitness is presumably due to the heritability of fitness-associ ated traits that covary with the male trait. However, in species that provide parental care for young, the associ ation may be due to the direct effects of parental activity. In J. floridae , the association may be both direct and indirect. The

PAGE 84

84 effect of artificial fanning on hatching success indicates that nest fa nning provides direct, nongenetic benefits for offspring. Fanning was appl ied to clutches at random, regardless of male pre-spawning behavior. For fanning to have in creased hatching success, the effect of fanning must have modified the rearing environment in such a way that embryos had higher survival. Tending the nest before it contains a clutch also may provide direct benefits to offspring. Prespawning nest-tending, which consists of biti ng at the nest substrat e and fanning, may be parental care in the form of nest construc tion and maintenance (Clu tton-Brock 1991; Foster 1967). If so, pre-spawning nest-tending may be a signal of a nestÂ’s qual ity. For example, a previous study of embryo survival in J. floridae suggested that embryo mortality may have been due to fungal infection and that in fection had a greater effect on surv ival in fresh than in brackish water (St. Mary et al. 2004). If the prevalence of fungal spores and hyphae is lower in brackish than in fresh water, then male activity that ma intains a clean nest may have a greater impact on infection rates in fresh than in brackish water (Cote and Gross 1993). Alternatively, approaches toward the female and tending the nest prior to spawning may indicate indirect fitness benefits such as the quality-related genes males will pass to offspring. For example, male activity level may be an i ndicator of health and resistance to infection (Hamilton and Zuk 1982; Holmes and Zohar 1990), which, if passed to offspring, may help offspring resist parasites or fungal infection in the nest environment. Context-Dependent Mating Preferences and Mate Choice Benefits The traditional view that mating preferences ar e static with respect to the direction of preferences and the traits that are preferred ha s been refuted by repeated evidence that mating preferences are dependent upon th eir context (reviewed in Jenn ions and Petrie 1997). For example, mating preferences vary with predator density in Brachyrhaphis episcope (Simcox et al. 2005), with predator dens ity and social context in Poecilia reticulata (Dugatkin and Sargent

PAGE 85

85 1994; Endler and Houde 1995), with time of season in Ficedula albicollis (Qvarnström et al. 2000), with water turbidity in ci chlids (Seehausen et al. 1997), and with dissolved oxygen in Pomatoschistus microps (Reynolds and Jones 1999). Cont ext-dependent mating preferences may result from variation in the ability to asse ss mates, as when turbid water or low light conditions limit the ability to assess sexual orna ments, from the costs of assessment (Jennions and Petrie 1997), or from variation in th e benefits of mate choice (Qvarnström 2001). Qvarnström (2001) argued that female prefer ences for male traits should be dependent upon the association between the male trait and th e success of his offspring. Using stalk-eyed flies as an example, she argued that female pref erences for eye-span shou ld be greater when food conditions are poor because a greater amount of vari ance in male fitness is attributed to variance in male eye-span under poor conditions th an under good conditions (D avid et al. 2000). The same argument can be made when the benef its of mate choice are direct, such as when resources are provided by the male to the female or her offspring. Females should favor males with high quality territories or nest sites and thos e that will provide higher quality parental care. For example, females of many species prefer to mate with males that already have young in their nests (Forsgren et al. 1996; Pete rsen et al. 2005; Tallamy 2000), s uggesting that they favor males that have demonstrated they will care for offspr ing. Females also prefer males that provide better care for young (Forsgren 1997; Östlund a nd Ahnesjö 1998). However, evidence that female preferences are stronger where a male’s territory, nest, or pare ntal behavior better predicts direct benefits is l acking. While the examples above provide evidence that preferred males offer higher direct benefits, I know of no studies demonstrating that context-dependent female mating preferences for male traits are asso ciated with a context-dependent direct benefits of mate choice.

PAGE 86

86 The results presented here provide this evidence for J. floridae . In Chapter 3, I found that female mating preferences were sa linity-dependent. I evaluated the effects of male activity on whether pairs spawned at all and whether they sp awned more than once. In fresh water, pairs were more likely to spawn if the male tended his nest than if the male did not tend, when the nests were empty (i.e., during the ‘pre-parental ph ase’, Chapter 3). In both fresh and brackish water, pairs were more likely to spawn a s econd time if the male tended the nest once it contained embryos (‘parental phase’, Chapter 3) . I argued that spawni ng rates indicated the female’s willingness to spawn and, thus, reflec ted female mating preferences. Therefore, I concluded that female mating pr eferences were based on pre-pare ntal activity when in fresh water, but were based on parental activity when in brackish water. A goal of the current study was to determ ine whether female preferences differed between salinities because the male behavioral traits that predict clutch success were different. I predicted that 1) tending of an empty nest would predict cl utch success in fresh water, whereas 2) tending of a nest containing embryos would pred ict clutch success in fres h and brackish water. To test the first prediction, I evaluated whether males that tended the ne st prior to spawning had different hatching success than t hose that did not tend, and whethe r the effect of tending differed between salinities. To test the second pred iction, I evaluated whet her tending (i.e., fanning artificially) a nest that contai ned a clutch influenced hatching success differently in fresh and brackish water. As predicted, there was a significant interac tion between the effect s of salinity tending the nest on hatching success; tending increased hatching success in fresh water but actually decreased it in brackish water, though these effects were not si gnificant. In addition, fanning increased hatching success in both fresh and brackis h water. These data suggest that females

PAGE 87

87 adjust the strength of their mating preferences in accordance with the magnitude of the direct benefits they receive from choosing good mates. In contrast, one aspect of the results run contrary to my predictions. Males that approached the female prior to spawning ha d considerably higher hatching success in both environments than males that did not, yet females showed no preference for males who approached her prior to spawning in the earlier study. Further, I would have expected males to approach females equally often in both salinities, but they approached more in brackish water in the earlier study. Costs of mate assessment may explain why females showed no preference for approaching males, even though approaching males had higher hatching success in both environments. Female preferences may be w eakened if the costs of assessment are high, preventing them from gathering enough informa tion about potential mates to choose the best male (Jennions and Petrie 1997). Being approached by males may be costly, as females occasionally have their fins torn and become bruised when housed with highly active males (personal observation). This cost may favor fema les either mating quickly or leaving the maleÂ’s territory without mating, either of which could limit the femalesÂ’ ab ility to assess male quality. Summary I have shown that females use some, though not all, information about potential mates that is presented in the mate Â’s behavior when making mating decisions. Further, this study demonstrates that context-de pendent mating preferences in J. floridae are associated with context-dependent direct benefits of mate choice and indicate th at the strength and direction of sexual selection on parental care is variable. Variation in the stre ngth and/or direction of sexual selection on parental care that arises from variation in ma ting preferences may provide the opportunity for parental care activity to become an exaggerated signal of male quality and to be

PAGE 88

88 expressed outside the parenting c ontext (Chapter 3 and Bonnevier et al. 2003) and for parental activity to evolve separately in different enviro nments. It remains to be shown whether this context-dependent sexual selection can resu lt in divergence in pa rental activity among environments.

PAGE 89

89 Table 4-1. Analysis of deviance for fit of pr ogressively simpler logistic regression models evaluating ln(odds of hatching) us ing a Likelihood Ratio (LR) test. Terms removed from model* LR† df P Tend the nest clutch size 0 1 1.00 tend•fan•rearing salinity 0.32 1 0.57 fan•rearing salinity 0.32 1 0.57 natal salinity 2.38 1 0.12 tend•fan 3.26 1 0.07 tend•rearing salinity 9.23 1 < 0.01 tend 0.32 1 0.57 Approach the female clutch size 0 1 1.00 approach•fan•rearing salinity 0.26 1 0.61 approach•fan 0.44 1 0.51 fan•rearing salinity 1.16 1 0.28 natal salinity 1.60 1 0.21 approach•rearing salinity 15.36 1 < 0.01 approach 0.92 1 0.34 Fanning clutch size 0 1 1.00 fan•rearing salinity 0.38 1 0.54 natal salinity 1.0 1 0.32 rearing salinity 2.64 1 0.10 fan 14.73 1 < 0.01 *Three sets of nested models were evaluated. In the first, male activity was scored as whether or not the male tended the nest. In the second, male activity was scored as whether or not the male approached the female. Terms were removed by backward elimination in the order presented from a model that included terms for the main effects of male activity, rearing salinity, fanning treatment, clutch size, and natal salinity; the effects of all twoand three-way interactions between male activity, rearing salin ity, and fanning treatment; and the random effect of pair. In the third analysis, male activity was not included as an explanatory variable such that the most complex model included fanning, rearing salinity, thei r interaction, clutch si ze, natal salinity, and pair. Terms were removed in the order that produced the smallest change in model likelihood between progressive steps. †LR test statistic is calculated for the comparison of the model including the term with a simplified model lacking the term and is approximately 2-distributed with df = the difference in number of parameters between the models being compared. Therefore, a P < 0.05 indicates that the te rm should not be removed from the model.

PAGE 90

90 Table 4-2. Analysis of deviance for fit of pr ogressively simpler log-li near regression models evaluating ln(hatchling mass) usi ng a Likelihood Ratio (LR) test. Terms removed from model* LR df P† Tend the nest natal salinity 0.38 1 0.54 clutch size 0.59 1 0.44 tend•fan•salinity 0.88 1 0.34 fan•rearing salinity 0.33 1 0.56 tend•fan 0.88 1 0.35 tend•rearing salinity 0.48 1 0.49 tend 0.22 1 0.64 Approach the female clutch size 0.21 1 0.64 natal salinity 0.26 1 0.61 approach•fan•rearing salinity 2.72 1 0.10 fan•rearing salinity 0.18 1 0.67 approach•fan 0.41 1 0.52 approach•rearing salinity 0.43 1 0.51 approach 0.67 1 0.41 Fanning fan•rearing salinity 0.002 1 0.97 natal salinity 0.34 1 0.56 clutch size 0.62 1 0.43 salinity 1.00 1 0.32 fan 1.90 1 0.17 *An explanation of the models is provided in Table 4-1. †In these analyses, removing terms did not significantly change the models’ deviances, suggesting poor fit of the original models.

PAGE 91

91 Figure 4-1. Clutch size as a func tion of salinity of the natal aqua rium and A) whether or not the male tended the nest and B) approached the female prior to spawning. Means ± 1 SE are calculated as the back-transformed mean ln(clutch size) plus residuals from the model ln(clutch size) = nata l salinity + pair, where pa ir was modeled as a random effect. Natal salinity did not have a si gnificant effect on clutch size (see text). NoYes 0 1020 3040 50 Tended the nestClutch size Fresh Brackish 010 2030 4050 Approached the femaleClutch size NoYes A B NoYes 0 1020 3040 50 Tended the nestClutch size Fresh Brackish 010 2030 4050 Approached the femaleClutch size NoYes NoYes 0 1020 3040 50 Tended the nestClutch size Fresh Brackish 010 2030 4050 Approached the femaleClutch size NoYes A B

PAGE 92

92 Figure 4-2. The proportion of embryos that hatched as a function of salinity and A) ma le nest-tending prior to spawning, B) ma le approaching prior to spawning, and C) fanning. Means ± 1 SE are calculated as the back-tra nsformed mean arcsine-square root-transformed proportion hatched pl us residuals from the model arcine (proportion hatched) = fanning + pair in A and B, and arcine (proportion hatched) = natal salinity + pair in C. In both models, pa ir was modeled as a random effect. Natal salinity did not have a si gnificant effect on th e proportion of embryos hatched, but was included in the second model because residuals could not be obtained fr om a model with a random effect, alone. Note scale ranges from 0.6 to 1.0. A B C 0.60.7 0.80.91.0 Tended the nestProportion hatched Fresh Brackish 0.60.70.8 0.91.0 Approached the female 0.6 0.70.80.91.0 Fanning NoYes NoYes NoYes 0.60.7 0.80.91.0 Tended the nestProportion hatched Fresh Brackish 0.60.70.8 0.91.0 Approached the female 0.6 0.70.80.91.0 Fanning NoYes NoYes NoYes

PAGE 93

93 Figure 4-3. Hatchling mass as a function of salinity and A) male nest-tending prior to spawning, B) male approaching prior to spawning, and C) fanning. Means ± 1 SE are calculated as th e back-transformed mean arcsine-square root-transformed proportion hatched plus residua ls from the model arcine (proportion hatched) = fanning + pair in A and B, and arcine (proportion hatched) = natal salinity + pair in C. In both models, pair was modeled as a random effect. Natal salinity did not have a significant effect on the proportion of embryos hatched, but was included in the second model because residuals could not be obtained from a model with a random effect, alone . Note scale ranges from 0.08 to 0.1 NoYes 0.08 0.09 0.100.110.12 Tended the nestHatchling mass (mg) Fresh Brackish 0.080.09 0.100.110.12 Approached the female 0.080.09 0.100.110.12 Fanning NoYes NoYes NoYes 0.08 0.09 0.100.110.12 Tended the nestHatchling mass (mg) Fresh Brackish 0.080.09 0.100.110.12 Approached the female 0.080.09 0.100.110.12 Fanning NoYes NoYesA B C

PAGE 94

94 CHAPTER 5 NATIVE ENVIRONMENT, BUT NOT SALI NITY, INFLUENCES METABOLIC RATE Introduction Phenotypic plasticity is expected to evolve if the selection imposed on a trait varies across environments. A speciesÂ’ pattern of plas ticity (i.e., reaction norm ) then depends on the range of environments to which it is exposed, its frequency of exposure, and the strength and nature of selection in each e nvironment (de Jong 1995; de Jong 2005; Via et al. 1995; Via and Lande 1985). Populations that are reproductively isolated may exhi bit different reaction norms if they experience different subsets of the speci esÂ’ range of environments or experience the environments at different frequencies. These differences in experience may result in not only differences in the shape of the reaction nor m among populations but also in each populationÂ’s mean phenotype across environments. Metaboli c rates associated with osmoregulation in euryhaline fishes (those tolerant of a broad sa linity range) are a useful system in which to evaluate the influence of envir onmental range on phenotypic plastici ty because 1) metabolic rate shows a high degree of plasticity across salinities, 2) sp ecies that differ in their salinity ranges typically exhibit different reaction norms, and 3) within species, populations may experience different ranges in salinity, allowing the comparis ons between interand intra-specific effects of range on reaction norm. The effects of salinity on metabolic rate have been examined in a variety of fishes and plastic responses tend to vary w ith ecological distribution and with the osmotic concentration of blood plasma. The osmotic concentration of plasma of most fishes is intermediate between the concentrations of freshwater (< 0.5 ppt) and seawater (~ 34 ppt) (Evans 1993). This means that the primary regulatory demand in seawater is to minimize water loss and ion gain, whereas the primary regulatory demand in fresh water is to retain ions while excreting excess water.

PAGE 95

95 Therefore, while maintaining homeostasis of pl asma osmotic concentration, an individualÂ’s osmoregulatory demands will increase with the gradient between environmental and plasma osmotic concentrations (Evans 1993). Because species vary in osmoregulatory efficiency (defined here as the metabolic cost of osmoregula tion) as well as plasma osmotic concentration, the effect of this gradient on metabolic rate and, thus, of salinity on metabolic rate, varies widely among species. For example, metabolic rates ca n increase with (e.g., Moser and Hettler 1989), decrease with (e.g., Claireaux and Lagardere 1999) , be quadratic functions of (e.g., Brocksen and Cole 1972; Febry and Lutz 1987; Haney and Nord lie 1997; Parvatheswarar ao 1965), or remain unaffected by (e.g., Chittenden 1971; Muir and Niimi 1972) salinity. Despite considerable varia tion in response to salinity among species, Nordlie (1978) showed that, among euryhaline species (i.e., thos e with broad salinity to lerances), metabolic reaction norms varied with a speci esÂ’ typical salinity range. No rdlie found four general patterns of metabolic responses among the st udies he reviewed. Metabolic rate was (Type 1) independent of salinity, (Type 2) concave and minimal at th e salinity at which plasma and the environment were iso-osmotic; (Type 3) concave and minimal at a salinity presumed to be optimal with respect to the typical sa linity range, though not the iso-osmoti c salinity, or (Type 4) convex and maximal at some presumed optimal salinity and reflecting a limited ability to maintain prolonged plasma homeostasis across salinities . He argued that a response of the first type represents an absence of metabolic costs associated with movi ng across salinities. Further, he showed that response Types 1 and 2 are more common among speci es that live in a broa d range of salinities, whereas Type 3 and 4 responses are more co mmon among species living within narrower ranges (Nordlie 1978). Among the studie s he reviewed, no species exhibi ted either a continual increase or decrease in metabolic rate with salinity.

PAGE 96

96 NordlieÂ’s review indicates that species can exhibit different metabolic reaction norms if they experience different salinity ranges, even th ough they share a toleran ce for a wide range of salinities. Further, his results suggest that a similar effect of salinity range on the metabolic response to salinity may be apparent among populat ions of the same, eur yhaline species if those populations typically experience diffe rent salinity ranges. If so , then populations experiencing the broadest range of salinity may exhibit a meta bolic response toward the Type 1 end of the spectrum, whereas those populati ons experiencing the narrowest range may exhibit a response toward the Type 4 end of the spectrum. A difference among populations in the res ponse to salinity may occur in some Cyprinodontiform fishes (pupfish, killifish, livebearers, and ot hers), a large group of salinitytolerant species thought to have evolved from a euryhaline ancestor (Hedgpeth 1957). Extant species of this order show consid erable diversification in natural salinity range; some species are predominantly freshwater (e.g., Jordanella floridae , Cyprinodon variegatus hubbsi , Fundulus crysotus ), whereas others are estuarine (e.g., Cyprinodon v. variegatus , Fundulus heteroclitus , and Rivulus marmoratus ), marine (e.g., Floridichthys carpio ), or even hypersaline (e.g., desert pupfishes, Cyprinodon spp. ) (Kaill 1967; Nordlie 2000; Nordlie and Walsh 1989; Page and Burr 1991; Soltz and Naiman 1978). Although Cyprinodontif orm fishes differ in their typical salinity environment, they retain tolerances for a wide range of salini ties. For example, Cyprinodon variegatus hubbsi , Jordanella floridae , C. v. variegatus , and Floridichthys carpio , all in the family Cyprinodontidae, inhabit prog ressively more saline habitats yet exhibit similar abilities to osmoregulate, as exemplified by their nearly co nstant plasma osmotic concentrations across salinities from just above fresh water to just above seawater (Jordan et al. 1993; Nordlie and Walsh 1989).

PAGE 97

97 In fresh water and in salinities above that of seawater, however, osmoregulatory responses of predominantly freshwater Cyprinodon tids diverge from those of their more marine relatives. The two predominantly freshwater species, Jordanella floridae and C. v. hubbsi , have plasma osmotic concentrations lower than more marine species in fresh water and higher than more marine species in water with salinity above approximately 40 pp t (Jordan et al. 1993; Nordlie and Walsh 1989). The lower plasma osmotic concentrations in fr esh water may reflect that these species are unable to meet the hi gher osmoregulatory demands of the freshwater environment. Alternatively, th e lower concentrations may reflect physiological adaptations that allow them to maintain plasma osmotic concentra tions in fresh water that are lower than their relativesÂ’ without negatively infl uencing other physiological proce sses. In other words, if the benefit of maintaining homeostasis when moving acr oss salinities is that cellular processes are buffered from such changes, then perhaps the cell ular processes of these two freshwater species are less affected by low ion levels than those of their more marine relatives. Because regulation of plasma osmotic c oncentrations differs between typicallyfreshwater and typically-coastal/marine species, then costs of osmoregulation, measured as metabolic rates, may also differ between these two groups. Jordanella floridae (flagfish) is a euryhaline pupfish that is found in coastal habitats up to salinities of approximately 20 ppt, but is predominantly found in inland streams and ponds. Because J. floridae has a narrower salinity range than C. v. variegatus , I predict J. floridae to exhibit a Type 2 or 3 metabolic response, with minimum metabolic rate at or near the iso-osmo tic level of approximately 10 15 ppt (Nordlie and Walsh 1989), in contrast to th e Type 1 response exhibited by C. v. variegatus (Haney and Nordlie 1997). Given this hypothesis, I assume that selection favors a reduction in metabolic costs at all salinities, but that se lection is stronger in salinities in which metabolic rates are higher

PAGE 98

98 – i.e., in fresh water and full seawater. If in land populations typically ex perience fresh water and coastal populations typically experience brackish wa ter, then selection to reduce metabolic rates should be stronger for inland than for coastal in dividuals because inland and coastal individuals experience the stronger, fresh water selection envi ronment at different fr equencies (e.g., de Jong 2005; Via and Lande 1985). As a result, I pred ict that the metabolic response should differ between inland and coastal popula tions. Specifically, I predic t the reaction norm of inland populations to differ such that the metabolic rate in fresh wa ter is reduced relative to the freshwater rate of coastal populations. This e ffect could result from inland populations having a lower mean metabolic rate, a shallower (less negative) slope of the reaction norm between fresh water and the iso-osmotic level (15 ppt), or mini mum metabolic rate at a lower salinity (i.e., a shift from a Type 2 toward a T ype 3 response) (Figure 5-1). In this study, I combine a laboratory examination of the effect of salinity on metabolic rate in each of four populations with a two-year field survey of abiotic conditions at four population locales designed, in part, to describe the natural salinity range of each population. Two of the populations were inland a nd two were coastal. If the pa ttern of salinity variation in a habitat imposes selection on metabolic responses, then responses in the two coastal populations should differ from responses in the two inland populations. Alternatively, differences among populations may be due to genetic drift. If dr ift played a dominant role in the evolution of metabolic responses, then geographically distan t populations should be more divergent than closer populations. Hence any observed differen ces should be more related to individual population differences and geographi c distance than they are to si milar salinity conditions. Unlike an earlier study designed to determ ine whether a specie s’ natural salinity distribution influences osmoregulatory ability (Nordlie and Walsh 1989), I assume that inland

PAGE 99

99 and coastal J. floridae will each be able to osmoregulate acr oss all salinities. Instead, I am asking whether selection on the efficiency of su ch regulation (i.e., its cost as measured by metabolic rate) differs among salinity environments. Methods Two coastal, St. Marks (Wakulla County) and Merritt Island (Breva rd County), and two inland, Otter Creek (Levy County) and Micco sukee (Miami-Dade C ounty), populations of Jordanella floridae (flagfish) in Florida were sampled (Figure 1-1). Both coastal regions contain a wide range of salinity conditions, with salinity varying primarily as a result of water flow management within the two refuges. Within these coastal regions, the collection sites were located within leveed pools and were not influenced by tidal cycle. The two inland sites were sufficiently distant from the coast to prevent the influx of saline water: Otter Creek was approximately 10 km from the Gulf of Mexico and Miccosukee was approximately 27 km from the Atlantic Ocean. Habitat type s were replicated latitudinally, with St. Marks (coastal) and Otter Creek (inland) in the northern half of th e state and Merritt Island (coastal) and Miccosukee (inland) more southern. Field work and fish collections were conducted under Florida Fish and Wildlife Conservation Commission Scientific CollectorÂ’s Permit numbers FNC-03-015 and FNC-05-012, U.S. Fish and Wildlife Service Special Use Permit numbers 58875, 03008, and JH06-2005 for St. Marks National Wildlife Refuge, and U.S. Fish and Wildlife Service Special Use Permit number 03 SUP 59 for Merritt Island National Wildlife Refuge. Field Survey Each of the four collection sites was sample d bimonthly from July 2003 to July 2004 and again from May 2005 to May 2006. Water temperat ure, pH, and salinity were measured during each sampling visit using a YSI Model 45 or Model 75 probe. Fish assemblages were sampled at each site using a box trap. Those results are avai lable in the appendix and will not be discussed

PAGE 100

100 further. In addition, locations within Merrit t Island and St. Marks National Wildlife Refuges other than the collection sites we re sampled to describe the amount of spatial salinity variation and to identify the range of salinities in which J. floridae are found in each refuge. Laboratory Study of Metabolism Fish were collected from Me rritt Island, Otter Creek, and Miccosukee in March 2004, from St. Marks in October 2004, and again from Miccosukee in July 2005 using a seine net, minnow traps, and dip nets. Water temperatures at collection were between 24.7 and 29.9 ° C. They were transported in their native water in insulated coolers to Florida State University in Tallahassee, Florida, where they were housed under a 14h light:10h da rk light cycle. Fish were maintained in the water from their collection site after arriving in the laboratory until the water had reached the temperature of the room. Theref ore, the fish were slowly acclimated to the experimental temperatures before being tr ansferred to water of other salinity. Standard metabolic rates for fish from each of the four populations were quantified at each of five salinities – 0.2, 7.5, 15, 22.5, and 30 ppt – for a total of 20 treatment combinations. Males and females from each population were di vided among the five salinity treatments to create approximately equal sex ratios across treat ments. Fish were adjusted to their target salinity at a rate no greater than 5 ppt per da y by adding Instant Ocean Aquarium Salt (Aquarium Systems brand) to well water. Fresh water (0.2 ppt) consisted of unaltere d well water. The fish were adjusted in groups such that there were tw o replicate groups for each of the 20 salinity-bypopulation treatments, but their metabolic rates were measured individually. The fish were fed frozen brine shrimp (San Francisco Bay Brand, San Francisco, CA) ad libitum daily and were acclimated for a minimum of four months prior to experimentation. Metabolic rates of 13, 17, 20, 17, and 22 individuals (total 89) were measured in 0.2, 7.5, 15, 22.5, and 30 ppt, respectively, when summed across populations.

PAGE 101

101 Standard metabolic rates were measured in a recirculating metabolic chamber modeled after Julian et al. (2003). Prio r to and between trials (see belo w), the chamber was maintained “open”: water cycled from the chamber through a water bath containing water aerated to oxygen saturation. During a trial, the chamber remain ed “closed”: water bypassed the water bath and was not aerated. Food was withheld from the expe rimental fish for 36 h prior to a trial and the fish was isolated in approximately 2 l of wa ter from the metabolic chamber at approximately 1500 h the day before a trial. All trials bega n between 0900 and 1100 h. A fish was placed in the housing portion of the chamber by hand and a llowed to acclimate with the system open and with the flow adjusted to 0.05 l/min. Barometric pressure, relative humidity, and temperature of the chamber water bath (25 to 27 °C across trials) were recorded at the start of each trial. Ten minutes following the introduction of the experimental fish, the system was closed to prevent the introduction of aerated water. The dissolved oxygen concentration was allowed to fall to approximately 40 % saturation before the system was flushed with oxygen-saturated water. After ten minutes of re-saturati on, the system was again closed and a second rate measurement was taken. Following the second measurement, the individual’s wet mass was measured. Dissolved oxygen concentration was measured c ontinuously beginning with the introduction of the fish to the chamber with an Ocean Optics FOXY Fiber Optic Oxygen Sensor System (Ocean Optics Inc.) placed immediately ups tream of the chamber housing the experimental fish. Data were recorded directly to a computer using Ocean Optics Sensors software. Control trials were conducted following the same procedure as trials with fish but without the fish present. Control trials were allowed to run an average of 3 h and, in most control trials, dissolved oxygen declined gradually over time. Oxygen consumption values for control trials were averaged within each salinity treatment.

PAGE 102

102 Metabolic rate (VO 2) was calculated from the time it t ook oxygen concentrations to fall from 100 % to 70 % saturation, af ter adjusting for the change in oxygen measured during control trials. For both control and experi mental trials, the change in mg O2 h-1 was calculated from percent saturation by adjusting for salinity, temp erature, and chamber volume using Truesdale et al. (1955) and Davis (1975) and dividing by the elapsed time. Th e two measurements taken from each individual were averaged before adding th e change in oxygen in the appropriate control measurement. All analyses were conducting using the control-adjusted VO 2 measurements. Analyses Treatment effects were evalua ted with maximum likelihood in a generalized linear mixed model with Gaussian error distribution using the lmer function of the lme4 package (Bates and Sarkar 2006) of the programming language R (R Core Development Team 2006). Because only one of the 89 individuals could be measured per da y, there was considerable variation in the time for which individuals were housed at test salinities prior to experimentation (range: 123 – 352 days), which was included as a covariate in the analyses. During this time, mortality reduced sample sizes and entirely eliminated individuals from the freshwater treatment for Merritt Island. Treatment effects on ln(VO 2) were evaluated in a model that included the fixed effects of sex, all possible interactions among ln(mass), salin ity treatment, native habitat type (inland or coastal), and time at the experimental salinity before measurement, and the random effect of population. Mass and the time at the experimental salinity before measurement were modeled as continuous variables, whereas sali nity, native habitat type, populat ion, and sex were categorical. The significance of each fixed effect term was evaluated by comparing the deviances of progressively simpler models. Deviance is define d as negative twice th e difference in likelihood between the model of interest and a saturated model (i.e., a model with a separate parameter

PAGE 103

103 describing each data point). Beginning with the model containing terms for all main and interaction effects, I sequentially removed the term that would result in the smallest change in model likelihood. This allowed me to compare the likelihood of a series of nested models, each with one fewer parameters than the previous m odel. The deviance of each model was compared with that of the model with one fewer paramete rs using analysis of deviance, which employs a likelihood ratio test to determine whether removing a term results in a signif icantly poorer fit of the model to the data (Agresti 1996). Therefore, the goal was to identify th e simplest model with a deviance not significantly diffe rent from more complex mode ls. The likelihood ratio test statistic is approximately 2-distributed with degrees of free dom equal to the difference in the number of parameters of the models being comp ared (Agresti 1996) – in this case, degrees of freedom = 1. The term for the random effect of population was reta ined in all models. Additional analyses were necessary in order to graphically present the mean and standard error estimates for VO 2 while controlling for the effects of fixed and random factors. These values were plotted as the grand mean VO 2 plus residuals from regre ssion models including fixed effect terms for those factors for which I want ed to control and the random effect term for population. The fixed effect term s included in these models were ln(mass), salinity, and/or time at the experimental salinity, and the random e ffect of pair, even though not all these terms explained a significant amount of the total variance in VO 2 (i.e., even if they were not included in the best fit model as identified with analysis of deviance). For example, to plot the effect of salinity and native habitat type on VO 2, I evaluated the model ln(VO 2) = ln(mass) + time + + population and added th e residuals from this model to the lntransformed grand mean VO 2. These residual-adjusted values were used to calculate treatment

PAGE 104

104 means and standard errors before back-transformi ng to the appropriate scale. The terms used in order to obtain each set of residuals are provided in the figure legends. If J. floridae exhibits a Type 2 response to salinity , then metabolic rates should be lowest at 15 ppt. I predicted that metabolic rates in fresh water would be lower for inland populations than coastal ones. I suggested that lower rate s in fresh water could re sult from one of three patterns (Figure 5-1): 1) metabolic rates of inla nd individuals are lower than those of coastal individuals in 0.2 ppt, but simila r in the other four salinities; 2) inland rates are lower than coastal rates in all salinities; or 3) inland popula tions have their reaction norm shifted relative to coastal populations such that their minimum ra te occurs at a salinity lower than 15 ppt. Alternatively, if habitat type does not explain differences in me tabolic rate, r eaction norms may differ solely due to genetic drift, in which cas e there should be a strong effect of population on the response to salinity. Results Field Survey Water chemistry varied among the four collect ion sites and across months (Figure 5-2). The two inland collection sites (Otter Creek and Miccosukee) consistently had low salinities, ranging between 0.02 and 0.8 ppt across both years. Salinity at Merritt Island ranged between 3.0 and 15.1 ppt, whereas salinity at the other co astal site, St. Marks, was unexpectedly low, ranging from 0.0 to 1.4 ppt. When additional site s at Merritt Island and St. Marks were sampled, J. floridae were found at salinities ranging from 1.9 to 17.7 ppt at Merritt Island and from 0.1 to 22.2 ppt at St. Marks (Table A-3), suggesting th at fish at the two coastal sites experience considerable salinity variation as they move among adjacent pools in the refuges. Neither pH nor temperature varied considerably among collec tion sites, though temperature varied over time

PAGE 105

105 and the southern sites, Merritt Island and Mi ccosukee were slightly warmer than the two northern sites (appendix). Laboratory Study of Metabolism Metabolic rate was influenced by ln(mass), na tive habitat type, and the amount of time at the experimental salinity prior to measurement according to the equation ln(VO 2) = 1.23 + 1.08·ln(mass) – 0.14·hab itat type – 0.0017·time + population, in which habitat type was coded as a dummy vari able with coastal = 0 and inland = 1 (LR test, ln(mass): 95 % CI = 0.93 – 1.24, 2 = 97.24, df = 1, P < 0.001; habitat: 95 % CI = 0.27 – 0.017, 2 = 4.24, df = 1, P = 0.040; time: 95 % CI = 0.0027 – 0.00076, 2 = 8.44, df = 1, P = 0.004, N = 89; Table 5-1). There also were marginal effects of the interaction between habitat type and time (estimate = 0.0017, 95 % CI = -0.0001 – 0.0035; LR test, 2 = 3.44, df = 1, P = 0.061) and of the interaction between habitat type and ln(mass) (estimate = 0.27, 95 % CI = -0.57 – 0.035; LR test, 2 = 3.07, df = 1, P = 0.079). These interaction terms were removed before estimating the coefficients in the mode l above. Differences am ong populations explained very little of the total vari ation in metabolic rate in the best fit model; variance among populations was 4.36 x 10-11. Metabolic rate increased with body mass (Figure 5-3) and decreased with time at the experimental salinity. The effect of time wa s greater in coastal than in inland populations (Figure 5-4). Using residuals from a model estimating the effects of ln(mass), salinity, and population and then evaluating the effect of time on the residual-adjusted ln(VO 2) separately for inland and coastal individuals, ln(VO 2) decreased marginally with time among coastal individuals (estimate = 0.0013, 95 % CI = 0.0027 – 0. 00014), whereas it did not decrease with time among inland individuals (estimate = 0.0004, 95 % CI = 0.0015 – 0.00058).

PAGE 106

106 Metabolic rates were higher for coastal individuals than for inland individuals (Figure 5-5). Using residuals from a model estimating the effects of ln(mass), salinity, time, and population and then evaluating th e mean residual-adjusted ln(VO 2) separately for inland and coastal individuals, mean VO 2 of coastal individuals was 0.38 ± 0.002 SE mg O2/h/ind, whereas mean VO 2 of inland individuals was 0.34 ± 0.002 SE mg O2/h/ind. Contrary to my predictions, metabolic rate was not influenced by salinity and metabolic rate in fresh water did not notably differ betw een inland and coastal individuals (Figure 5-5), despite an overall higher metabolic rate in coastal than in inland individuals. Discussion The data are consistent with the hypothesis that an individual’s natural salinity range influences metabolic rate. Inland fish, which experience a lower and na rrower salinity range than coastal fish, had lower overall metabolic rates th an coastal fish. This is also consistent with the mechanism presented in Figure 5-1B. In add ition, I predicted that metabolic rate in fresh water, but not necessarily in other salinities, w ould be lower for inland fish than coastal ones. Despite the overall difference between inland and coastal individuals, their metabolic rates were more similar in fresh water than at other salini ties. As a result, these data do not support my hypothesis regarding the effect of ha bitat type on metabolic rate at the lower end of the salinity range. Lower metabolic rates were measured for individuals from habitats with generally lower salinities. However, these lower metabolic rates cannot be explained as either an evolutionary or developmental response by inland individuals to re duce what was predicted to be relatively high metabolic costs in fresh water, because in neith er inland nor coastal populations were metabolic costs higher in fresh than in brackish water. Instead, these data suggest that J. floridae exhibits a

PAGE 107

107 Type 1 response to salinity (Nor dlie 1978) and experiences sim ilar osmoregulatory costs across all treatment salinities. Habitat-associated metabolic rates could be e xplained by differences in abiotic conditions other than salinity or in biotic conditions (reviewed in McNab 2002). For example, temperature can have a large effect on metabolic rate (e .g., Claireaux et al. 2000; Lardies et al. 2004; Schurmann and Steffensen 1997) and differences in temperature between inland and coastal sites could explain the difference in metabolic rate. Si milarly, exposure to adult or embryo predators may influence overall energy expenditure if di fferences in predator communities between habitats affect activity levels (e.g., Brett 1964; Goolish 1991). For example, the presence of either adult or embryo predators may influen ce energy expenditure by in creasing vigilance and nest defense activity. It is po ssible that temperature or pred ator community could have longlasting developmental effects on metabolic rates or that they could have resulted in evolved differences in metabolic rate. However, a comp arison of conditions in th e four sites indicates little difference in water chemistry or species composition between inland and coastal sites. Of the four populations, Merritt Island had the highest mean pH and ch loride content, whereas St. Marks had the lowest, with the two inland popula tions exhibiting intermediate means for both variables (appendix). Further, th e only adult predators found with regularity at the sites were juvenile Centrarchids (sunfish ) (appendix), and they were found with consistency only at St. Marks (coastal) and Otter Creek (inland); thus , they cannot explain th e habitat-associated differences in metabolic rate. Similarly, densitie s of fish that are likely to be important embryo predators, such as Gambusia holbrooki , Poecilia latipinna , Fundulus spp . , and Lucania spp . , are similar across sites (appendix). Therefore, dens ities of embryo predators also cannot explain the differences in metabolic rate between inland and coastal individuals.

PAGE 108

108 An alternative explanation is that metabolic rates differ between coastal and inland individuals because they respond differently to prolonged salinity exposure. I found that metabolic rate of coastal indivi duals, but not inland individual s, decreased with the amount of time an individual spent at the e xperimental salinity prior to measurement. Coastal individuals likely experience both temporal a nd spatial salinity variation if they move among adjacent areas of their habitat (Figure 1-1; Tabl e A-3). It is possible that pl asma homeostasis during movement among habitats is facilitated by temporarily elev ating metabolic rate upon initial exposure to a new salinity, and that this elev ation may be more pronounced in coastal populations that more frequently experience salinity changes. Under this hypothesis, plasma ion concentrations of coastal individuals should vary less with salinity than those of inland individuals. While such a comparison has not be made in J. floridae , data from C. v. variegatus and C. v. hubbsi support the hypothesis; plasma ion concentrations are more stable across a salinity range from 0 to 70 ppt in the more saline-distributed C. v. variegates (Jordan et al. 1993). In ad dition, this hypothesis is consistent with the finding of Potts (1954) that the energetic requirements of osmoregulation, itself, may be quite low and instead that costs of salinity tolerance for euryhaline fish are felt during the acclimation period (Styczy ska-Jurewicz 1970). Under this hypothesis, differences in metabolic rate among treatment groups should gra dually decrease over time as these immediate acclimation costs decline. Both inland and coastal J. floridae exhibited a Type 1 metabo lic response to salinity and metabolic rates of the four popula tions were most similar in fr esh water. Rejection of my hypothesis, that metabolic rates of inland and co astal individuals should differ most in fresh water, is complicated by the absence of any measurements for Merritt Island individuals in fresh water. The Merritt Island colle ction site experienced the great est amount of temporal salinity

PAGE 109

109 variation. Therefore, it may be expected th at Merritt Island individuals , in particular, should have considerably higher metabolic rates in fr esh water than individuals from the other three populations. The absence of this treatment group makes it difficult to evaluate whether typical salinity range influences the shap e of the reaction norm. Theref ore, the conclusion that inland and coastal individuals have similar metabolic re sponses to salinity must be made contingent on the assumption that fresh water metabolic rate s for the two coastal popul ations are similar. It is not surprising for a species that can expe rience temporal and spatial salinity variation to exhibit a Type 1 response to salinity (Nordlie 1978). For example, a Type 1 response has been measured in Cyprinodon variegatus (Haney and Nordlie 1997), Morone saxatilis (Chittenden 1971) and Kuhlia sandvicensis (Muir and Niimi 1972) and indicates low energetic costs of moving across salinities (Nordlie 1978). This homeostasic ability may be maintained regardless of native habitat type – for example, Hedgpe th (1957) argued that euryhalinity is highly phylogenetically conserved and Nord lie et al. (1992) suggested that tolerances for a broad range of salinities should be retained if there are even occasional influxes of more saline water to inland sites, such as due to storm surge, or if reproductive isolation from coastal populations is less than complete. Other studies of metabolic responses to salinity in Cyprinodontids have found mixed evidence for Type 1 responses. For example, C. v. variegatus is reported to exhibit a Type 1 response within the fresh to seawater range, with metabolic rates declin ing at salinities above approximately 40 ppt, but both C. v. variegatus and its relative, C. v. hubbsi , exhibit some variation in metabolic rate at salinities below 40 ppt. (Haney a nd Nordlie 1997; Jordan et al. 1993; Nordlie et al. 1991). Another euryhaline species, Cyprinodon salinus , seasonally moves across a salinity gradient and maintains a patt ern of plasma osmotic concentration across

PAGE 110

110 salinities similar to those of C. v. hubbsi and J. floridae , but its metabolic rate increases with salinity (Stuenkel a nd Hillyard 1981). Of these three species, C. v. hubbsi is ecologically most similar to J. floridae . Both species have salinity ranges lower than other Cy prinodontids and they exhi bit similar patterns of plasma homeostasis across salini ties – they have lower plasma osmotic concentrations than C. v. variegatus in fresh water and highe r concentrations than C. v. variegatus at salinities above seawater (Jordan et al. 1993; Nordlie and Walsh 1989). Beca use of their ecological and physiological similarities, I woul d expect metabolic rates of C. v. hubbsi to be lower than those of C. v. variegatus , just as metabolic rates of inland J. floridae were lower than those of coastal J. floridae . This is not the case, howe ver, as Jordan et al. (1993) showed no difference in mean metabolic rates between C. v. hubbsi and C. v. variegatus .. Mean metabolic rates of the two C. variegatus subspecies may not have had tim e to diverge considerably if C. v. hubbsi represents the early stages of a range expansion or if C. v. hubbsi ’s relatively small popul ation size (Gilbert 1992) has limited the ability of selection to shap e metabolic response. Another explanation for why the J. floridae populations differ in metabolic rate, whereas the C. variegatus subspecies do not, may be that J. floridae ’s more distant relation to an estuarine/marine ancestor is associated with greater variation in osmoregul ation. To explore these alternatives, and whether or not lower metabolic rates are commonly associated with a freshwater distributi on in Cyprinodontids, a broader comparison of metabolic rates and salin ity distribution across th is group of euryhaline species would be necessary to determine the relative influences of phylogeny and salinity distribution on metabolic rates. Finally, these results have implications for variation in reproduc tive behavior observed across salinities in J. floridae (Chapter 3). My laboratory stu dy of salinity effects on paternal

PAGE 111

111 care and female mating preferences in J. floridae found that females were more likely to spawn with a male if he performed nest-tending activities prior to spawning, but that females only exhibited this preference in fresh water. This led to the prediction th at males should perform these nest-tending activities more often in fresh water than in brackish water, but my results showed that they did not: there was no eff ect of salinity treatment on male activity. One proposed explanation for this patt ern was that the energetic costs of increased tending in fresh water outweighed its benefit. Cost s were predicted to be higher in fresh water than in brackish water under the assumptions that J. floridae exhibit similar metabolic responses to salinity as C. variegatus and that higher metabolic demands in fresh water required more time to be spent foraging and less time spent in nest preparati on. Spending time tending the nest, therefore, imposes a greater cost due to lost foraging time in fresh than in brackish water. However, the absence of an effect of salinity on metabolic de mands suggests that the time-allocation cost is equivalent across salinities and indicates that th ere must be an alternat ive explanation for the nest-tending pattern of males.

PAGE 112

112 Table 5-1. Sample size and mean (SE) wet mass and metabolic rate for each population by salinity treatment group. Habitat Population Salinity (ppt) N Mass (SE) VO 2 (SE)* Inland Otter Creek 022.65 (0.20)0.26 (0.002) 7.522.98 (0.61)0.36 (0.044) 1562.56 (0.32)0.34 (0.019) 22.542.73 (0.27)0.36 (0.013) 3042.64 (0.28)0.40 (0.025) Miccosukee 061.53 (0.15)0.34 (0.021) 7.542.18 (0.37)0.32 (0.016) 1532.43 (0.22)0.28 (0.008) 22.531.36 (0.12)0.28 (0.018) 3061.64 (0.16)0.33 (0.015) Coastal St. Marks 051.73 (0.12)0.35 (0.018) 7.571.52 (0.11)0.36 (0.019) 1561.71 (0.05)0.39 (0.009) 22.561.75 (0.12)0.40 (0.026) 3081.57 (0.07)0.45 (0.016) Merritt Island 00 7.542.20 (0.06)0.40 (0.017) 1552.10 (0.58)0.32 (0.026) 22.542.69 (0.29)0.34 (0.028) 3041.93 (0.09)0.45 (0.014) * VO 2 is presented as the back-transformed grand mean ln(VO 2) plus residuals from the model ln(VO 2) = ln(mass) + time.

PAGE 113

113 Figure 5-1. Three possible processes by which me tabolic rate of inland individuals could be lower than that of coastal individuals in fresh water. A) The shape of the reaction norm could be shallower, such that metabolic rate increases more slowly as salinity deviates away from the iso-osmotic level. B) The shape of the reaction norm could be the same, but the mean across all populati ons could be lower for inland individuals than coastal individuals. C) The salinity at which metabolic rate is minimal could be lower for inland than for coastal individuals. Arrows indicate the change from coastal (solid curve) to inland (d ashed curve) reaction norms. 051015202530 VO 2 051015202530 Salinity (ppt) 051015202530 A B C 051015202530 VO 2 051015202530 Salinity (ppt) 051015202530 A B C

PAGE 114

114 Figure 5-2. Salinity during each month at the four sites for the 2003-2004 and 2005-2006 sampling years. Closed symbols, coas tal populations; filled symbols, inland populations. Salinity (ppt) 0 2 4 6 8 10 12 14 16 Merritt Island St. Marks Otter Creek Miccosukee 2003-2004 0 2 4 6 8 10 12 14 16 MayJulSepNovJanMarMay Month 2005-2006Salinity (ppt) 0 2 4 6 8 10 12 14 16 Merritt Island St. Marks Otter Creek Miccosukee 2003-2004 0 2 4 6 8 10 12 14 16 Merritt Island St. Marks Otter Creek Miccosukee 2003-2004 0 2 4 6 8 10 12 14 16 MayJulSepNovJanMarMay Month 2005-2006 0 2 4 6 8 10 12 14 16 MayJulSepNovJanMarMay Month 2005-2006A B

PAGE 115

115 Figure 5-3. Metabolic rate as a function of mass for each of the four populations. Residualadjusted metabolic rates (VRes) are presented as the back-transformed grand mean ln(VO 2) plus residuals from the model ln(VO 2) = habitat type + time + salinity + population. Closed symbols, coastal populat ions; open symbols, inland populations. The effect of mass did not differ am ong habitat types or treatments ( b = 1.08).

PAGE 116

116 Figure 5-4. Metabolic rate as a function of time at the experi mental salinity for inland and coastal populations. Residual-adjusted metabolic rates (VRes) are presented as the back-transformed grand mean ln(VO 2) plus residuals from the model ln(VO 2) = ln(mass) + salinity + population. Closed sy mbols, coastal populations; open symbols, inland populations. Dotted line, coastal popula tions; dashed line, inland populations.

PAGE 117

117 Figure 5-5. Metabolic rate as a function of salinity treatment a nd native habitat type. Treatment means (± 1 SE) are presented for inland (dashed lines) and coastal (dotted lines) individuals. Rates (VRes) are presented as the back-transformed grand mean ln(VO 2) plus residuals from the model ln(VO 2) = ln(mass) + time + population, such that populations were pooled within native habitat type after removing the effect of population. Closed symbols, coastal populat ions; open symbols, inland populations. X-axis jitter is added to facilitate visualization.

PAGE 118

118 CHAPTER 6 SYNTHESIS I have argued that if environmental conditions influence offspring fitn ess and the benefits of parental care for offspring, then environmental conditions can indirectly influence sexual selection on parental behavior a nd parental behavior, itself. I have examined the influence of salinity environment on metabolic rate, paternal care activity, mating success, and clutch success in flagfish ( Jordanella floridae ) in an effort to describe how natural and sexual selection shape patterns of care. In this final discussion, I br idge the results of the three experimental projects and the modeling work to draw conclusions about how selective pressures shape parental care across an environmental gradient. Specifically, I will addr ess three topics before discussing the larger implications of this work. First, I will briefly examine whether the patterns of mating success revealed in Chapter 3 correspond to parallel patterns in the benef its of mate choice described in Chapter 4. A more lengthy discussion of female mating preferences and the benefits of mate choice is presented in Chapter 4. Second, I will discuss the effects of parental care and salinity environment on hatching success described in Ch apter 4 and whether they support a fundamental prediction of the optimality model presented in Ch apter 2. Third, I will examine why variation in female mating preferences and male parental activity did not entirely conform to my predictions. Finally, I will discuss the potential di rections that a continuation of this work could take and what the results of this work may reveal about ge neral patterns in the evolution of parental care. Context-Dependent Mating Preferences and Mate Choice Benefits Female preferences for good fathers should be st ronger in environments in which parental care will have a greater impact on offspring fitness. Numerous st udies have demonstrated that

PAGE 119

119 females select mates based on the quality of parent al care that males will provide her offspring, (e.g., Forsgren 1997; Östlund and Ahnesjö 1998; Petersen et al. 2005; Tallamy 2000); however, evidence that female preferences are stronger where a male’s territo ry, nest, or parental behavior better predicts direct benefits is lacking. The results presented in Chapters 3 and 4 provide this evidence for J. floridae . In Chapter 3, I found that female mating preference s were salinity-dependent. I evaluated the effects of male activity on whethe r pairs spawned at all, and wh ether they spawned more than once. In fresh water, pairs were more likely to spawn if the male tended his nest when the nests were empty (during the ‘pre-parental’ phase). In brackish water, wh ether or not the pair spawned at all over the observati on period did not depend on male activity. Instead, whether or not the pair spawned more than once depended on the male’s activity while he was caring for young (during the ‘parental’ phase). I argued that spawning rates i ndicated the female’s willingness to spawn and, thus, reflected female mating preferences. Therefore, I concluded that female mating preferences were based on pre-pare ntal activity when in fresh water, but were based on parental activity when in brackish water. The goal of the work presented in Chapter 4 was to determine whether female preferences differed between salinities because the male behavioral traits that predict offspring success were different. I predicte d that tending of an empty nest would predict offspring success in fresh water, whereas tending of a nest contai ning a clutch would predict offspring success in brackish water. I quantified whether clutches from males that tended their empty nests prior to spawning had higher hatching success than those from non-tending males and whether the difference in success differed between fresh and brackish water. I also provided some clut ches with artificial

PAGE 120

120 fanning to determine whether tending during the parental phase increased clutch success. I found that both having a nest-tending father and being fanned during development increased hatching success, but each only increase d hatching success in fresh water. These data were partially consistent with my predictions, providing evidence that female mating preferences vary with the benefits of mate choice, but also raising questions as to whether there are additional benefits of mate choice in J. floridae . The data indicate that female preferences during the pre-parent al phase are adaptive with resp ect to the benefits of mate choice. Females preferred males that tended thei r nests in fresh water (Chapter 3), where nesttending increased hatching success (Chapter 4), but not in brack ish water, where male tending had minimal effect on hatching success. In cont rast, female preferences during the parental phase could not be explained by hatching success. Fanning, which is one component of the parental-phase nest-tending that females prefe rred in brackish water, did not increase hatching success in brackish water, where females exhi bited preference. In stead, fanning increased hatching success in fresh water, where female s exhibited no preferen ce for parental-phase activity. The pattern of preferences for and mate ch oice benefits of nest-tending during the preparental phase provide conclusi ve evidence that context-depende nt association between male behavior and clutch success can explain cont ext-dependent mating preferences. They also indicate that the strength of natural and sexual selection on pa rental activity can positively covary across environments. A more extensive di scussion of the implications of these results is presented in Chapter 4. Together, variation among environments in natural and sexual se lection pressures on parental care could give rise to considerable intraand inter-sp ecific variation in parental care

PAGE 121

121 patterns (Jennions and Petrie 1997), particularly if divergence in parent al care pattern is associated with ecological divergence. For exam ple, as a species spreads to new habitats, the demands of offspring for parental care may ch ange and new patterns of parental care may emerge. Further, natural and sexual selection to gether acting on parental care should result in more rapid evolution of care th an under natural selection alone. Therefore, sexual selection on parental care may play an important role in the e volution of parental care patterns, particularly in taxa with highly variable patterns of care, such as fishes and birds, in which sexual selection likely plays a large role in the transitions betw een no care, uniparental care, and biparental care (Burley and Johnson 2002; Gittleman 1981; Mank et al. 2005; McKitrick 1 992; Tullberg et al. 2002), and in anurans, in which parental care has arisen as ma ny as 40 times (Reynolds et al. 2002). Offspring Vulnerability and Parental Care Benefits The optimality model presented in Chapter 2 makes a series of assumptions about the relationship between offspring vulnerability, pa rental care effort, and present reproductive success in order to resolve a conflict between two existing verbal models of parental effort. While the model demonstrates how the conflic t can be resolved, em pirical support for the assumptions of the model were lacking. The data presented in Chapter 4 provide support for an important prediction that is a direct consequence of these assumptions. In the basic model, I assumed that as the am ount of parental care provided to a clutch increases, clutch fitness increases, but approaches an asymptotic value. This was presented as PDec, which described present reproductive success as a decelerating function of parental care, C . I defined the vulnerability of a clutch as the clut chÂ’s expected survival if care is not provided ( l0) and I assumed, via the structure of B ( C ), that greater vulnerability is associated with lower

PAGE 122

122 fitness if no care is provided ( P ( C = 0)), but that vulnerability does not affect the asymptotic fitness value (Figure 2-2A). An asymptotic value of clutch fitne ss is applicable if parental care functions to increase offspring su rvival to independence, as is likely the case in fishes such as Jordanella floridae that provide parental care for developing embryos . Parental care may increase clutch survivorship to the level at which all embryos hatch, but parental careÂ’s influence likely ends there. Under these assumptions the model predicted that the benefits of parental care for vulnerable clutches will be greater than those for less vulnerable clutches. In J. floridae , clutch vulnerability is influenced by the salinity enviro nment because, in the absence of parental care, embryos reared in fresh water have lower ha tching success than embryos reared in brackish water (St. Mary et al. 2001), ma king embryos in fresh water more vulnerable. The model then predicts that clutches reared in fresh water should benefit more from parental care than those reared in brackish water. The experiment presented in Chapter 4 provides a test of this prediction. I examined the effects of two types of pare ntal activity, pre-spawning nest -tending and post-spawning nest fanning, on hatching success. The results were cons istent with the modelÂ’s prediction: clutches reared in fresh water benefited more from pre-spawning nest-tending and from post-spawning nest-fanning than did clutches reared in brackish water. A review of the forms of care common among sp ecies suggests that this pattern may be widespread and not a special case of parental care in fishes. In fact, the prediction that benefits increase with vulnerability may be appropriate wh enever care benefits are asymptotic. Embryo and juvenile defense and attendance (including vi viparity) is the most common form of care in fishes (Blumer 1979), amphibians (Crump 1995), a nd reptiles (de Fraipo nt et al. 1996) and

PAGE 123

123 should result in an asymptotic benefit curve. Ev en in birds, in which the young of most species rely on provisioning (Cockburn 2006), an asymptotic benefit curve may describe parental care that affects offspring recruitment rate, but not the offspringÂ’s mating success or fecundity. In capital-breeding species, for example, the re productive success of a first-year breeder will depend on its resource intake during the prev ious winter and spring (Stearns 1992), and presumably not on its parentsÂ’ pr ovisioning rate, suggesting that the benefits of care in some bird species may be asymptotic and maximal when all o ffspring are recruited to the adult population. The same may apply in capital-breeding mammal s (Boyd 2000). If so, then an asymptotic benefit curve, and the predictions for parental ca re benefits that such curve suggests, may be appropriate for the majority of sp ecies that exhibit parental care. Finally, the data presented in Chapter 4 support earlier verbal models of optimal parental effort that predict parents should provide more care for vulnerable offspring because those are the offspring that will benefit most from car e (Andersson et al. 1980; Dale et al. 1996; Montgomerie and Weatherhead 1988). In contra st, they do not support arguments for parental effort that are based on the offspringÂ’s reproductiv e value. Recall that reproductive value-based predictions rely on the argument that the offspringÂ’s reproduc tive value prior to receiving parental care determines care benefits. However, J. floridae clutches reared in brackish water are of higher reproductive value prior to receivin g care, but benefit less from nest-tending and fanning, suggesting that clutches of higher repr oductive value should actu ally receive less care than those of lower value. In the next secti on, I will explore whether the amount of care males male provide conforms to predictions of the model presented in Chapter 2. Adjustment of Parental Behavior to Selection Pressures The goal of the model presented in Chapter 2 was to predict how parental care should vary as a function of clutch characteristics and pa rental costs. The model predicted that parental

PAGE 124

124 care should increase as the survival of offspring that are not provided care declines. In Chapter 4, I showed that, in the absence of parental care, clut ches that are reared in fresh water have lower hatching success than those reared in brackish water. I also showed that the benefit of nest-tending for hatching success was greater in fre sh than in brackish water. In Chapter 3, I showed that the sexual selection be nefit of nest-tending is also greater in fresh water, when the nest-tending is directed toward an empty nest. Together these data indica te that both natural and sexual selection favor more nest-tending in fresh than in brackish water when that tending is directed toward empty nests. Given this, males observed in Chapter 3 should have been more likely to tend their empty nests in fresh than in brackish water, but they did not. I experimentally explored one explanation for the apparently maladaptive pattern in male behavior by quantifying metabolic rates of flagfish across a range of salini ties (Chapter 5). If parental care is more costly to provide in fresh water than in brackish water, then the benefits of tending in fresh water may be negated by careÂ’s ener getic costs. Contrary to this explanation, metabolic rates were very similar in fresh and br ackish water, indicating that the energetic costs of care are not so high in fresh water as to preven t males from increasing th eir nest-tending rates. Therefore, the question of why males do not adjust their pre-parental activity with salinity remains largely unanswered. In contrast, the absence of variation in male activity during the parental phase was initially surprising, but can be explained by co mparing the results presented in Chapter 3 and Chapter 4. Tending the nest on ce it contained eggs increased th e probability that males would spawn again, but only when in brackish wate r (Chapter 3). However, fanning the clutch increased hatching success only in fre sh water (Chapter 4). Therefore, it seems that the natural

PAGE 125

125 and sexual selection benefits of te nding a nest that contains eggs r un contrary to one another and, by doing so, may make the net benefit of tendi ng the same in fresh and brackish water. Future Directions Two questions remain regarding female mating preferences and male parental activity in flagfish. First, why do females prefer activitie s that do not increase o ffspring fitness, while showing no preference for activitie s that do? Second, why do male s not adjust their tending of empty nests in response to sa linity? The answers to both of these questions may lie in unmeasured costs and benefits of mate choice and parental activity. Mate choice has been shown to be costly in a number of species (reviewed in Jennions and Petrie 1997) and it is possible that the costs of assessing males for J. floridae females may prevent them from choosing better fathers. For example, males that approached the female when caring for an empty nest had higher hatching success in fresh than in brackish water in Chapter 4, but female preferences for approaching males did not differ between salinities in Chapter 3. In Chapter 4, I suggested that receptivity to appro aching males may be invari able with respect to salinity if being approached by males is costly. As a result, it may be advantageous for a female to make a quick decision as to whether or not she will mate with an approaching male, and the advantage of making the decision quickly may outw eigh the benefit of adjusting assessment to the local salinity. To explore whether assessm ent costs could explain the poor match between the effects of male activity on offspring success and female preferences for that activity, the consequences of mating with approaching and nest -tending males should be quantified. Such a study should focus specifically on the survivor ship and future fecundity of females. Similarly, lack of variation in male activity when the benefits of that activity vary across salinities may be explained by costs not measured in these studies. For example, tradeoffs between tending the nest and excluding either eg g predators or males from adjacent territories

PAGE 126

126 may preclude males from increasing tending rates in some salinities. However, in order to explain male tending activity, thes e costs would have to be greater in the salinities in which tending is preferred, or would have to be so high as to outweigh the benefit of adjusting tending rate. A test of this tradeoff hypothesis woul d require conducting similar experiments as presented here, but in a setting that includes bo th adjacent territorial males and egg predators. Finally, the data presented in these chapters indicate that the ab iotic environment can influence how offspring respond to parental care and that variation in offspring response can have cascading effects on female mating preferences and possibly parental effort. This effect of environment may help to explain an apparent para dox in the evolution of parental care. Namely, the primary benefit of providing pare ntal care is to increase offspr ing fitness. Yet, for parental care to arise in a lineage that pr eviously lacked parental care, offspring that previously survived well without care must suddenly exhibit depende nce on care. In other words, an evolved increase in offspring vulnerability seems to be a necessary precursor to th e evolution of parental care. Environmental effects on offspring dependence may help explain how selection gives rise to parental care. A species that previously lacked parental care may shift its distribution into habitats in which offspring vulnerability slightly increases, creating the selective pressure that favors parental care. For example, Baylis (1981) suggested that a fres hwater distribution may favor territoriality and parental care in fishes beca use of the combined effects of a greater need to lay demersal eggs and the low availability of suit able demersal habitat. An association between distribution shifts and the evolut ion of parental care would pr ovide one solution to the paradox that offspring need to undergo evolutionary chan ges that increase their vulnerability before the

PAGE 127

127 appearance of parental care. Instead, the distribut ion shift would be a prec ursor that creates the selective environment that favors parental care. Why offspring dependence on parental care ar ises and reaches such extremes as the highly altricial young of passerines and mammals is one of the la rge, unanswered questions in the evolution of parental care. An exploration of whether evolutionary transitions in parental care are associated with distribut ional or ecological shifts may re veal whether patterns like that proposed by Baylis (1981) are important and wi despread and whether th e cascading effects of changes in offspring vulnerability may extend to the evolutionary orig ins of parental care.

PAGE 128

128 APPENDIX FIELD SURVEY OF WATER CHEMISTR Y AND SPECIES COMPOSITION AT FOUR STUDY SITES Introduction Jordanella floridae is a Cyprinodontid found thr oughout peninsular Florida and the coastal region of FloridaÂ’ s panhandle in the area of St . MarkÂ’s National Wildlife Refuge (Wakulla County) (Page and Burr 1991) . A few studies (G unter and Hall 1965; Kilby 1955; Loftus and Ku shlan 1987) indicate that J. floridae can be found in brackish habitats, reaching salinities of 39 ppt in isolated, drying pools (Brockmann 1974). However, their distribution among freshwater swamps and wetlands is considerably broader and better described (e.g., Carlson and Duever 1978; Hale 2001; Loftus and Kushlan 1987; Tagatz 1967; Trexler et al. 2005). In order to compare natural and sexual selection on parental care beha vior as a function of salinity, it was necessary that I identify populations persisting in brackish en vironments and compare biotic and abiotic conditions between these and more inland, freshw ater habitats. The re sults of a two year survey of water chemistry and specie s composition at two inland (Otter Creek, Miccosukee) and two coastal (St. Marks, Me rritt Island) sites (see Figure 1-1) are described. Methods Each of the four sites was sampled bi monthly from July 2003 to July 2004 and again from May 2005 to May 2006 under. O tter Creek and Miccosukee were also sampled in May 2003. All work was c onducted under Florida Fish and Wildlife Conservation Commission Scientific Coll ectorÂ’s Permit numbers FNC-03-015 and FNC05-012 and U.S. Fish and Wildlife Servic e Special Use Permit numbers 58875, 03008,

PAGE 129

129 and JH-06-005 for St. Marks National W ildlife Refuge and number 03 SUP 59 for Merritt Island National Wildlife Refuge. Water temperature, pH, specific conductivity, and salinity were measured during each sampling visit using a YSI Model 63 or Model 85 meter (Yellow Springs Instruments, Inc.). Chloride content wa s analyzed using ion chromatography using a Dionex Series 4500i Chromatographer and ELab Chromatography System V.4 software (OMS Tech, 1992), after filtering water sa mples through a 0.20 µm PES Membrane (Corning Brand, Corning, New York). Sp ecies composition and density were sampled using a 0.5 m2 box trap. Once the trap was placed, animals were removed with ten sweeps of a dip net and placed in a bucket to be counted. Density of female, male, and juvenile Jordanella floridae , density of other fish speci es, and occurrence of non-fish vertebrates and invertebrates we re measured in each of eight samples for a total of 4.0 m2 sampled per site visit. The selection of ei ght samples per site was based on preliminary sampling at three sites indicating stabilization of density estimates with approximately six to eight samples. Individual samples were separated by a minimum of 5 m. Corner depth, vegetative cover (measured as the per cent of horizontal trap area having vegetation in the water column), and distance to shore were also measured for each sample. Densities, corner depth, cover, and distance to shore were then averaged across the eight samples per site, whereas the occurrence of invertebrates and non-fish vertebrates was pooled across all eight samples. Results and Discussion Water chemistry varied among the four si tes and among months (Table A-1). The two inland sites (Otter Creek and Miccosukee) consistently had low salinities, ranging between 0.02 and 0.8 ppt across both years (F igure A-1). Salinity at Merritt Island

PAGE 130

130 ranged between 3.0 and 15.1 ppt, whereas salinity at the other coastal site, St. Marks, was unexpectedly low, ranging from 0.0 to 1. 4 ppt. Although salinity at St. Marks was consistently low throughout the study, prel iminary sampling at both Merritt Island and St. Marks indicated considerable spatial variati on in salinity (Table A-3). For example, J. floridae were observed in water exceeding 20 ppt at Tower Pond within St. Marks, though they could not be found there upon return visits. Therefore, if migration among adjacent habitats occurs, it is probable that the two coastal populations each consist of individuals distributed acro ss a broad salinity range, even though only the Merritt Island collection site, itself, experienced considerable temporal salinity va riation. Temperature varied over time and across sites with th e two southern sites, Merritt Island and Miccosukee, typically warmer than the two nor thern sites during any given survey month. Jordanella floridae abundance varied but the patt ern was not consistent across years or sites (Table A-2; Fi gure A-2). Juveniles were more common than adults at all sites except for St. Marks a nd during many sampling periods only juveniles were found. Further, females tended to be more common than males. J. floridae was rare at St. Marks and collection of additional animals from areas adjacent to the collection site (but within 1 km) was necessary to conduct the laboratory studies in Chapters 3, 4, and 5. Temporal variation in J. floridae density at all sites sugge sts either considerable movement of individuals among microhabitats or temporally variable reproductive rates. Both of these processes may explain the observe d variation. All of the collection sites are connected to other aquatic habi tats, including larger wetlands, and thus individual fish are not physically restricted to the sampling site. It is possible that the occasional absence and then reappearance of J. floridae at each site is the result of movement of adults

PAGE 131

131 among neighboring habitats. Indeed, studies of freshwater fish communities in the Florida Everglades ecosystem have repeatedly found that J. floridae colonize wetlands very quickly following dry period s and that their density fall s gradually as other species immigrate (Jordan et al. 1998; Trexler et al. 2005), suggest ing that they readily and quickly move among adjacent habitats. C onsistent with these earlier studies, J. floridae density in the current survey tended to be high following dry months. Water levels at the four sites varied over time and during some months it was not possible to obtain eight box trap samples (Table A-1) because the area inundated with water was too small. These dry months were followed by months with higher than average J. floridae density. For example, in May of 2004, 2005, and 2006, Miccosukee was too dry to obtain eight samples. In July of 2004 and 2005, the two post-dry months for which I have data, J. floridae density was higher than the mean across all months of the sample year at that site (Table A-2). The same occurred at Merritt Island, at which fewer than eight samples were obtained during May of 2004 and 2005 and July of 2006. At the same time, temporal variation in density, particularly that of juveniles, could result from an annual reproductive cycle. I observe d nesting males during March at St. Marks and during July at Merritt Isla nd and Miccosukee (Table A-4), whereas none were located during fall and wi nter. Peaks in juvenile a bundance tend to occur during the summer and fall months, which may reflect a reproductive season that extends from early spring (Loftus and Kushlan 1987) until mid or la te fall. This is a longer reproductive season than suggested by Fost er (1967), who stated that J. floridae likely breed between April and August based on the seasonal progres sion of gonad development. Further, my observation of nesting and spaw ning during March at St. Marks, together with FosterÂ’s

PAGE 132

132 observations of gonad development, suggest th at reproduction at the northern end of the species’ range is not notably delayed relative to that at the southern e nd. In contrast to juveniles, adults were gene rally rare, which suggests eith er that they experience high mortality or that they are more transient th an juveniles, entering the shallow habitats selected for the survey only to nest and in stead spending most of their time in adjacent habitats. The presence of predators of adult and embryonic J. floridae may influence reproductive behavior if the pres ence of predators alters the allocation of time or energy to predator avoidance or egg defense (e.g., Klug et al. 2005). Fish species composition was dominated by Fundulid and Poecilid fishes, with Gambusia holbrooki and Heterandria formosa the most common species at all sites (Figure A-3). Poecilia latipinna was common at Otter Creek, Miccosukee, and Merritt Island, but rare at St. Marks, which instead had nearly as many Elassoma spp. and Etheostoma spp. as Heterandria formosa . Of the species encountered , the potential predators of J. floridae adults were Aphredoderus sp., Esox sp . , the Cichlids, and the Cent rarchids. In general, these species were rare and appeared to be juveniles (< 6 cm Esox sp., < 4 cm Aphredoderus and Centrarchids) . The presence of only juveniles may be due to the shallow depth of the areas sampled (Table A-2). In conclusion, water chemistry varied among the four sites, but not as expected – St. Marks was more similar to the two inland s ites than to Merritt Island, the other coastal site. In contrast, species composition was similar among all sites and was dominated by Gambusia holbrooki and Heterandria formosa . Therefore, neither salinity at the sampling site nor species composition can expl ain differences in behavior and metabolic

PAGE 133

133 rate associated with native habitat type, as defined a priori as inland or coastal. Instead, I argue that fish at all sites are able to m ove among adjacent habitats and, as a result, individuals at both coastal popula tions are likely to experience spatial variation in salinity that inland populations do not. As a result, metabolic similarity be tween the two coastal populations may result from the fact that they each may experience a broad range of salinity as individuals move among adjacent ha bitats, despite vastly different salinity ranges at the survey sites.

PAGE 134

134 Table A-1. Water and box trap sample data from two years of bimonthly sampling. TempSpecificSalinity[Chloride]No. box trapDistance toSample SiteMonth(°C) conductivity ( µ S) (ppt)pH( µ M)samplesshore (m)% coverDepth (cm) 2003-2004 St. MarksJul30.01540.107.405.2281.793226.6 Sep27.21380.106.364.67681.567625.8 Nov12.51440.16.8281.776627.4 Jan13.81580.179.85883.067425.8 Mar19.61220.16.779.07881.936529.4 May27.53340.27.4761.7481.967128.7 Jul29.22410.16.6231.0281.559426.9 Merritt IslandJul32.00192609.908.306821.683.196928.3 Sep28.12401015.16.45208.882.885436.3 Nov17.12067012.47.157219.282.945929.1 Jan18.2132007.68.214967.482.134331.0 Mar27.159107.68.21262282.567430.1 May25.4115806.67.724063.221.75030.8 Jul34.778404.38.752789.2882.457427.8 Otter CreekMay33.101770.8031.472215.6 Jul28.03620.207.3015.39281.494618.7 Sep25.007.6013.66881.236622.6 Nov15.47420.47.8577.181.361517.2 Jan12.811940.67.7210.99281.166117.9 Mar15.75600.37.1252.3881.344819.8 May24.73670.67.7217.47531.533417.1 Jul24.73670.27.0214.83282.258129.5 MiccosukeeMay28.505460.208.6036.07252.052641.6 Jul32.003030.108.0024.12881.764535.1 Sep29.102140.107.1414.76981.668737.7 Nov193500.27.230.16181.258330.2 Jan16.24370.27.239.0382.068333.0 Mar24.24150.27.4537.4781.887331.9 May29.94130.27.4738.0481.605734.8 Jul30.45160.27.735.151.1810022.1 2005-2006 St. MarksMay301280.16.4560.1483881.9610021.0 Jul3310260.57.778076.2382.138424.1 Sep28.78900.46.976222.91381.774627.2 Nov25.127141.47.5112646.0481.002428.5 Jan13.91230.182.951335.8 Mar21.51700.15.71357.891781.416122.7 May27.36100.36.023234.19680.632417.3 Merritt IslandMay27.762303.446623.5641.059023.6 Jul4.86888082.2416629.4 Sep32656038.4550685.6287.258933.6 Nov23.2530037.6545890.1382.889333.7 Jan21119106.87.8120361.283.389132.5 Mar25143308.36.98159091.882.053622.5 May30.897005.47.1810033341.35022.7 Otter CreekMay30.36190.37.26341.248580.91615.7 Jul356.1371.174118.7 Sep24.410520.57.751083.7781.24819.3 Nov20.4481.827.83235.747881.08017.6 Jan12.66520.37.7288.780381.388417.2 Mar25.85980.36.71485.818382.156917.8 May21.61490.15.939.24569881.575924.2 MiccosukeeMay25.15530.3150.980581.604937.1 Jul31.62430.17.8694.704181.257635.0 Sep29.62290.17.21349.005983.288632.8 Nov26.92830.17.78278.34380.987625.5 Jan143860.2452.108981.4610026.3 May237440.46.19967.623760.9110028.0

PAGE 135

135 Table A-2. Mean density (No./m2) of fish es and occurrence of nonfish vertebrates and invertebrates from eight box trap samp les at four sites over two years. Jordanella floridae CyprinodontidaeAphredoderidae SiteMonthfemalemalejuvtotal Cyprinodon variegatusAphredoderusLepomisMicropterus unknown 2003-2004 St. MarksJul0.25000.250.750000 Sep00003.50000 Nov0.25000.2550000 Jan0.50.2500.7510000 Mar00001.750000 May000000000 Jul00.2500.2500000 Merritt IslandJul0.250.251.5200000.8 Sep1.250.253.755.2500000 Nov002.252.2500000 Jan00.252.75300000 Mar0011000.2500.5 May000000.25000.25 Jul0.50.2533.7500000 Otter CreekMay3.31.304.700000 Jul000.250.2500000 Sep20.7579.750000.250 Nov0.250.54.755.500005.25 Jan3.2504.257.500000.7 Mar0.2500.50.7500000 May0.7000.700000 Jul004.54.500000 MiccosukeeMay1.60.802.400004 Jul1.250.2501.50000.251.75 Sep222.56.500001.5 Nov1.251.513.7500001.25 Jan5.752613.7500.5002.75 Mar10.250.51.7500002 May00.52.252.7500000.5 Jul00.41.21.600000.5 2005-2006 St. MarksMay000010000 Jul000000000 Sep000000000 Nov000000000 Jan00000.250000 Mar00002.250000 May000000000 Merritt IslandMay002201001 Jul0.250.255.255.7500000.25 Sep3.2532.258.500.25001.75 Nov0.2500.75100.25000 Jan001.251.2500001 Mar00.2533.2500000 May0000000.2500 Otter CreekMay000000000.6 Jul000000000 Sep0.37503.253.62500000 Nov000000000 Jan0.7500.51.2500000 Mar0.250.250.5100000 May0.50.7501.2500.25000.25 MiccosukeeMay1211.2514.250000.250.25 Jul0.5013.2513.75000.2501 Sep6122.2529.2500000.75 Nov9.2507.2516.500000 Jan0.50.257.58.2500000 May000000.2500.250 Centrarchidae

PAGE 136

136 Table A-2. Continued. CichlidaeCyprinidaeElassomatidaeEsocidaeIctaluridae SiteMonth CichlidaeCyprinidNotropis harperiElassomaEsoxNoturus unknown 2003-2004 St. MarksJul0000000 Sep0000000 Nov0000000 Jan0000000 Mar0000000 May0000000 Jul0000000 Merritt IslandJul0000000 Sep0001000 Nov0001.75000 Jan0003.5000 Mar0.25002000 May0000.5000 Jul00012.5000 Otter CreekMay0000000 Jul0000000 Sep0001.25000 Nov0000000 Jan0000000 Mar0000000 May0000000 Jul0000.25000 MiccosukeeMay00000.2500 Jul000.50000 Sep0000.25000 Nov00.2503.5000 Jan0007.750.2500.5 Mar0001.5000 May0001.50.2500 Jul0002.5000 2005-2006 St. MarksMay0000000 Jul0000000 Sep0000000 Nov0000000 Jan0000000 Mar0000000 May0000000 Merritt IslandMay0.5000.25000 Jul0002.5000 Sep0.25004.25000 Nov0005.5000 Jan0.75008.5000 Mar0000000 May0000.25000 Otter CreekMay0000000 Jul0000000 Sep0000000 Nov0001.25000 Jan0000.5000 Mar0000.5000 May00018.75000 MiccosukeeMay0002.750.2500 Jul0001000 Sep0005.25000 Nov0000000 Jan00000.2500 May00017.2500.250

PAGE 137

137 Table A-2. Continued. Fundulidae SiteMonth Fundulus confluentusF. crysotusF. grandisFundulusLeptoleucania aumataLucania goodeiL. parva 2003-2004 St. MarksJul0000000 Sep2.2500.50000 Nov0002.5000.5 Jan0000.5001 Mar0000002 May0000000 Jul0.5000000 Merritt IslandJul000006.40 Sep00.500.501.50 Nov010009.50 Jan01.7500013.50 Mar00.500060 May00.7500024.50 Jul00.2500090 Otter CreekMay0000000 Jul000002.70 Sep0000010 Nov0000030 Jan0000000 Mar0000000 May00.2500000 Jul00.50000.50 MiccosukeeMay0000000 Jul000006.50 Sep02.50000.50 Nov00.250000.50 Jan0001100 Mar00.7500000 May0100020 Jul05.75001.500 2005-2006 St. MarksMay0000070 Jul1.25000000 Sep2.25000000 Nov1.75000001 Jan200000.50 Mar1.2500001.52.5 May0.5000000 Merritt IslandMay000005.50 Jul00.50101.50 Sep0.250.250003.50 Nov01.2501.506.50 Jan0.751.2500013.50 Mar000000.70 May0000000 Otter CreekMay000000.570 Jul000003.50 Sep00.2500000 Nov00.50001.50 Jan0000000 Mar000000.50 May020001.50 MiccosukeeMay000001.50 Jul0400010 Sep02.5001.500 Nov00000.500 Jan000005.50 May0101020

PAGE 138

138 Table A-2. Continued. PercidaePoecilidae SiteMonth Etheostoma fusiformeGambusia Heterandria formosaPoecilia latipinnaPoecilid juv. 2003-2004 St. MarksJul025.506.7518.5 Sep0552.517.50 Nov054.250.527.50 Jan034.25110.250 Mar0183.54.250 May032000 Jul020003.75 Merritt IslandJul044.824.46.40 Sep042.259.7513.514 Nov052.541.2529.750 Jan02135.75160 Mar012.522.52.750 May048.5345.50 Jul027.520.258.50 Otter CreekMay035.6241.60 Jul010.7121.30 Sep06.758.2513.50.75 Nov145.526.52.750 Jan08.7000 Mar0563.750 May0.53.259.510 Jul04.757.258.250 MiccosukeeMay01.758.75120 Jul13.5100.2502 Sep1616.757.2500 Nov54.5200 Jan20.56.51.7500 Mar315.750.7500 May34.75100 Jul17.253.7500 2005-2006 St. MarksMay0547.500 Jul0182.58.750 Sep027.54.2514.750 Nov03513.7538.50 Jan016.752.510.250 Mar0190.2521.7530.50 May02000 Merritt IslandMay0113.2512.2500 Jul021.52941.250 Sep017.2525.7552.750 Nov028.530.254.253.25 Jan012.25311.750 Mar071500 May032.257.7500 Otter CreekMay09.77.74.60.6 Jul03111.752.750 Sep021.757.55.51.5 Nov0.54.531.250 Jan0.511.751.52.250 Mar01.51.2540 May1413.755.7500 MiccosukeeMay23.256.750.250 Jul17.2530.250 Sep06.25100 Nov00000 Jan090.7500 May21610.2500

PAGE 139

139 Table A-2. Continued. AnuraCaudata SiteMonth Hyla cinereaRana pipiensRana grylioR. sphenocephala unknown SirenNotophthalmus unknown 2003-2004 St. MarksJul00000000 Sep00000000 Nov00000000 Jan00000000 Mar00000000 May00000000 Jul00001000 Merritt IslandJul00000000 Sep00000000 Nov00000000 Jan00001000 Mar00001010 May00000000 Jul00001000 Otter CreekMay00000000 Jul00000000 Sep00001000 Nov00000000 Jan00000000 Mar00000000 May00000000 Jul00000000 MiccosukeeMay00001000 Jul00000000 Sep00001000 Nov01100000 Jan00001010 Mar00000000 May00010000 Jul00001000 2005-2006 St. MarksMay00000000 Jul10001000 Sep00001000 Nov00000000 Jan00000000 Mar00000000 May00000000 Merritt IslandMay00000000 Jul00001000 Sep00000000 Nov00000000 Jan00000000 Mar00000000 May00000000 Otter CreekMay00000000 Jul00000000 Sep00000000 Nov00001000 Jan00000000 Mar00000100 May00011001 MiccosukeeMay00001000 Jul00001000 Sep00000000 Nov00001000 Jan00000000 May00010010

PAGE 140

140 Table A-2. Continued. BivalviaGastropodaAnnelida SiteMonthBivalve CampylomaPhysellaPlanorbellaPomacea unknownHirudineaother Dolomedes Mite 2003-2004 St. MarksJul0000000010 Sep0000000010 Nov0110000010 Jan0000010010 Mar0100010000 May0000000001 Jul1000110010 Merritt IslandJul0000010000 Sep0000000000 Nov0000000000 Jan1000000000 Mar0000000011 May0000000000 Jul0000000010 Otter CreekMay1000110000 Jul1000110110 Sep1000000010 Nov1000010000 Jan0000110000 Mar1000110010 May0000000000 Jul0000100000 MiccosukeeMay0000000010 Jul0000000000 Sep0000001010 Nov0000001010 Jan0000000000 Mar0000011001 May0000010000 Jul0000000000 2005-2006 St. MarksMay1001001010 Jul0000000000 Sep1000000010 Nov0000010010 Jan0000000010 Mar0000010000 May0110000000 Merritt IslandMay0000000001 Jul0000000011 Sep0000000011 Nov0000000011 Jan0000000001 Mar0000000001 May0000000000 Otter CreekMay1000100010 Jul1001100000 Sep1000000010 Nov1000000000 Jan1000110110 Mar1011110000 May1000000010 MiccosukeeMay0001000000 Jul0000000000 Sep0000100000 Nov0000010000 Jan0000000110 May0010000000 Arachnida

PAGE 141

141 Table A-2. Continued. CrustaceaEphemeropteraOdonata naiad SiteMonthAmphipodCrayfishIsopodaShrimpEphemeropteraDamselflyAeshnidGomphidLibelluidaeDragonfly 2003-2004 St. MarksJul1001010010 Sep1001111010 Nov1101100010 Jan1101111010 Mar1111111010 May1001010010 Jul1001110010 Merritt IslandJul0100011000 Sep0001000000 Nov0101010010 Jan0001000010 Mar0001010010 May0001000000 Jul0100000000 Otter CreekMay0001000010 Jul0101000010 Sep0101000000 Nov1101000010 Jan0101000010 Mar0101100110 May0000000000 Jul0101110010 MiccosukeeMay0101000010 Jul0000000000 Sep1101100000 Nov1101100010 Jan1101000010 Mar1001110010 May0101000010 Jul1001010010 2005-2006 St. MarksMay1001111010 Jul1101010010 Sep1001010010 Nov1001000010 Jan1001010111 Mar1001011010 May1001011010 Merritt IslandMay0101010010 Jul1101000000 Sep0101010000 Nov1101000010 Jan0101011010 Mar0101000010 May0001000000 Otter CreekMay0101101000 Jul0101100000 Sep0001010010 Nov0001000110 Jan0101000100 Mar1101010010 May0101000010 MiccosukeeMay0001000010 Jul0101000010 Sep0101010010 Nov1101100010 Jan1101010010 May1001010000

PAGE 142

142 Table A-2. Continued. NeuropteraColeoptera SiteMonthBelostomatidGerridae Lethocerus NepidNotonectidNeuropteranCorixidaeDyticidGyrinidaeHalyplid 2003-2004 St. MarksJul0000001100 Sep0000000100 Nov0100001101 Jan0000001000 Mar0000001111 May1001101101 Jul0000000101 Merritt IslandJul1000001100 Sep0000101000 Nov1000001000 Jan0000001001 Mar1000001101 May0000000101 Jul1000001001 Otter CreekMay0000000101 Jul0000000000 Sep0000000000 Nov0001000101 Jan0000000000 Mar0001000001 May0000000100 Jul0000000111 MiccosukeeMay0000000100 Jul0000000000 Sep0000000000 Nov1100001000 Jan0000000100 Mar0100000100 May1000000000 Jul1000000000 2005-2006 St. MarksMay1011001101 Jul0000000101 Sep0101100100 Nov0001001001 Jan0000001001 Mar1000001000 May1000001101 Merritt IslandMay1000001011 Jul1000001001 Sep1010001101 Nov1010001101 Jan1000000101 Mar0000000000 May0000001101 Otter CreekMay0000000110 Jul0000000101 Sep0000000000 Nov0001000000 Jan0000000101 Mar0000000100 May0000000011 MiccosukeeMay1000000101 Jul0000000000 Sep0000000000 Nov0000000000 Jan0000001100 May0000000101 Hemiptera

PAGE 143

143 Table A-2. Continued. Coleoptera, continuedHeteropteraDiptera SiteMonthHydrophilidunknownNaucoridChironomidCulicidaeunknown 2003-2004 St. MarksJul101100 Sep111000 Nov001000 Jan100100 Mar001001 May101000 Jul101000 Merritt IslandJul111100 Sep101000 Nov100100 Jan000000 Mar001100 May000100 Jul101100 Otter CreekMay000100 Jul101010 Sep100000 Nov000100 Jan000100 Mar000100 May000100 Jul100100 MiccosukeeMay000000 Jul000000 Sep000000 Nov000100 Jan101100 Mar101100 May001000 Jul101000 2005-2006 St. MarksMay111100 Jul000000 Sep011100 Nov001100 Jan011000 Mar001000 May101000 Merritt IslandMay001100 Jul001100 Sep101100 Nov001000 Jan001000 Mar101100 May000100 Otter CreekMay100100 Jul100100 Sep100100 Nov100000 Jan100100 Mar100000 May100000 MiccosukeeMay001000 Jul000100 Sep000100 Nov000000 Jan101100 May101000

PAGE 144

144 Table A-3. Results of preliminary samp ling at sites within Merritt Island National Wildlife Refuge and St. Marks National Wildlife Refuge in 2003. Refuge Site within refuge Date Salinity J. floridae present Merritt Island Freshwater Pond* May 1.9 ppt Yes Pool G May 8.2 Yes Pool I May 17.5 Yes Walking Trail (0.2 mi from tower) May 29.0 No Wildlife Drive #9 May 10.5 Yes Hwy 406, W of Wildlife Drive May 17.7 Yes St. Marks Gambo Bayou† May 0.1 Yes Mounds Pool 2 March 0.2 Yes, nesting Tower Pond April-May 2.0 – 22.2 Yes *The site identified as Merritt Island throughout this work is located in Freshwater Pond. †The site identified as St. Marks in Chap ters 3 and 4 is located at Gambo Bayou.

PAGE 145

145 Table A-4. Observations of nest attendance and spawning in Jordanella floridae . Site Date Number of nests Spawning observed Mean territory diameter Mean nest depth Miccosukee July, 1999 5 Yes 16.5 cm 8.7 cm Loxahatchee* July, 1999 9 Yes 38.6 8.7 St. Marks† March, 2003 5 Yes 26.8 7.5 Merritt Island July, 2003 9 Yes *Loxahatchee is located in Ma rtin County, Florida, and was sampled in 1999 but was not included as part of the fourpopulation experimental design pres ented in this dissertation. †Nests at St. Marks were located in Mounds Pool 2, whereas the field survey was conducted at Gambo Bayou.

PAGE 146

146 Figure A-1. Mean salinity, temperature, and pH for each site and each survey year. Error bars indicate range. St. Marks and Merritt Island are coastal; Otter Creek and Miccosukee are inland. 0 1 2 3 4 5 6 7 8 9 10 St. MarksMerritt IslandOtter CreekMiccosukee SitepH 0 5 10 15 20 25 30 35 40Temperature (°C) 0 2 4 6 8 10 12 14 16Salinity (ppt) 2003-2004 2005-2006 0 1 2 3 4 5 6 7 8 9 10 St. MarksMerritt IslandOtter CreekMiccosukee SitepH 0 1 2 3 4 5 6 7 8 9 10 St. MarksMerritt IslandOtter CreekMiccosukee SitepH 0 5 10 15 20 25 30 35 40Temperature (°C) 0 5 10 15 20 25 30 35 40Temperature (°C) 0 2 4 6 8 10 12 14 16Salinity (ppt) 2003-2004 2005-2006 0 2 4 6 8 10 12 14 16Salinity (ppt) 2003-2004 2005-2006

PAGE 147

147 Figure A-2. Mean density of fl agfish females, males, and juveniles at the four sites over the two sampling years. A and B: inland populations; C and D: coastal populations. Miccosukee was not sampled in March of 2006. Note the difference in scale in panel D. 2003 -20042005 -2006 D. Inland: Miccosukee0 3 6 9 12 15 18 21 24MayJulSepNovJanMarMayJulMayJulSepNovJanMarMay C. Inland: Otter Creek0 1 2 3 4 5 6 7 8 A. Coastal: St. Marks0 1 2 3 4 5 6 7 8 Females Males Juveniles B. Coastal: Merritt Island0 1 2 3 4 5 6 7 8Density (fish/m2) 2003 -20042005 -2006 2003 -20042005 -2006 D. Inland: Miccosukee0 3 6 9 12 15 18 21 24MayJulSepNovJanMarMayJulMayJulSepNovJanMarMay C. Inland: Otter Creek0 1 2 3 4 5 6 7 8 A. Coastal: St. Marks0 1 2 3 4 5 6 7 8 Females Males Juveniles B. Coastal: Merritt Island0 1 2 3 4 5 6 7 8Density (fish/m2)

PAGE 148

148 Figure A-3. Mean density of co mmon fish species, averaged acro ss two years of sampling. St. Marks and Merritt Island are coas tal; Otter Creek and Miccosukee are inland. 0 10 20 30 40 50 60 70 80 90J. floridaeG. holbrookiH. formosaP. latipinnaElassoma sp. Etheostoma sp. F. confluentusF. crysotusL. goodei CentrarchidaeFish speciesDensity (fish/m2) Coastal: St. Marks Coastal: Merritt Island Inland: Otter Creek Inland: Miccosukee 0 10 20 30 40 50 60 70 80 90J. floridaeG. holbrookiH. formosaP. latipinnaElassoma sp. Etheostoma sp. F. confluentusF. crysotusL. goodei CentrarchidaeFish speciesDensity (fish/m2) Coastal: St. Marks Coastal: Merritt Island Inland: Otter Creek Inland: Miccosukee

PAGE 149

149 LIST OF REFERENCES Ackerman, J. T., and J. M. Eadie. 2003. Curre nt versus future reproduction: an experimental test of parental investment decisi ons using nest desertion by mallards ( Anas platyrhynchos ). Behavioral Ecology a nd Sociobiology 54:264-273. Agresti, A. 1996. An introduction to categoric al data analysis. John Wiley & Sons, New York. Alonzo, S. H., and B. Sinervo. 2001. Mate choice games, context-dependent good genes, and genetic cycles in the side-b lotched lizard, Uta stansbur iana. Behavioral Ecology and Sociobiology 49:176-186. Amat, J. A. 1996. Nest defense by chinstrap penguins Pygoscelis antarctica in relation to offspring number and age. Journal of Avian Biology 27:177-179. Andersson, M. 1994. Sexual Selection. Princeton University Press, Princeton. Andersson, M., C. G. Wiklund, a nd H. Rundgren. 1980. Parental defense of offspring: a model and an example. Animal Behavior 28:536-542. Arnqvist, G., and L. Rowe. 2005. Sexual Conflic t. Princeton University Press, Princeton. Bates, D., and D. Sarkar. 2006. Linear mixed-e ffects models using S4 classes. Pp. Fit linear and generalized linear mixed-effects models. R package version 0.9975-0. Baylis, J. R. 1981. The evolution of parental care in fishes, with reference to Darwin's rule of male sexual selection. Environmental Biology of Fishes 6:223-251. Blumer, L. S. 1979. Male parental care in th e bony fishes. Quarterly Review of Biology 54:149161. Bonabeau, E., J. L. Deneubourg, and G. Therau laz. 1998. Within-brood competition and the optimal partitioning of parental investment. American Naturalist 152:419-427. Bonneaud, C., J. Mazuc, O. Chastel, and H. West erdahl. 2004. Terminal investment induced by immune challenge and fitness traits associat ed with major histocompatibility complex in the house sparrow. Evolution 58:2823-2830. Bonnevier, K., K. Lindstrom, and C. M. St. Mary. 2003. Parental care and mate attraction in the Florida flagfish, Jordanella floridae . Behavioral Ecology and Sociobiology 53:358-363. Boyd, I. L. 2000. State-dependent fertility in pinnipeds: contrasti ng capital and income breeders. Functional Ecology 14:623-630. Brett, J. R. 1964. The respiratory meta bolism and swimming performance of young sockeye salmon. Journal of the Fisheries Re search Board of Canada 21:1491-1501. Brockmann, F. W. 1974. Seasonality of fish es in a South Florida brackish canal. Florida Scientist 37:65-70.

PAGE 150

150 Brocksen, R. W., and R. E. Cole. 1972. Physiol ogical responses of three species of fishes to various salinities. Journal of the Fisher ies Research Board of Canada 29:399-&. Brommer, J., H. Kokko, and H. Pietiainen. 2000 . Reproductive effort and reproductive values in periodic environments. Amer ican Naturalist 155:454-472. Burley, N. T., and K. Johnson. 2002. The evol ution of avian parental care. Philosophical Transactions of the Royal Society of L ondon Series B-Biological Sciences 357:241-250. Carlisle, T. R. 1985. Parental response to brood size in a cichlid fish. Animal Behaviour 33:234-238. Carlson, J. E., and M. J. Duever. 1978. Seas onal fish population fluctuation in South Florida swamp. Proceedings of the Annual Conference of the Southeastern Association of Fish and Wildlife Agencies 31:603-611. Charlesworth, B., and E. L. Charnov. 1981. Kin selection in age-struct ured populations. Journal of Theoretical Biology 88:103-119. Charnov, E. L. 1976. Optimal foraging, the marginal value theorem. Theoretical Population Biology 9:129-136. Chittenden, M. E. 1971. Effects of handling and salinity on oxygen requirements of striped bass, Morones axatilis . Journal of the Fisheries Research Board of Canada 28:1823-&. Claireaux, G., and J. P. Lagardere. 1999. In fluence of temperature, oxygen and salinity on the metabolism of the European sea bass. Journal of Sea Research 42:157-168. Claireaux, G., D. M. Webber, J. P. Lagardere, and S. R. Kerr. 2000. Influence of water temperature and oxygenation on the aerobic metabolic scope of Atlantic cod ( Gadus morhua ). Journal of Sea Research 44:257-265. Clutton-Brock, T. H. 1984. Re productive effort and terminal inve stment in iteroparous animals. American Naturalist 123:212-229. Clutton-Brock, T. H. 1991. The Evolution of Parental Care. Princeton University Press, Princeton. Cockburn, A. 2006. Prevalence of different modes of parental care in birds. Proceedings of the Royal Society, Series B 273:1375-1383. Coleman, R. M., and R. U. Fischer. 1991. Brood size, male fanning effort and the energetics of a nonshareable parental invest ment in bluegill sunfish, Lepomis macrochirus (Teleostei, Centrarchidae). Ethology 87:177-188. Cote, I. M., and M. R. Gross. 1993. Reduced di sease in offspring: a be nefit of coloniality in sunfish. Behavioral Ecol ogy and Sociobiology 33:269-274.

PAGE 151

151 Crespi, E. J., and H. Lessig. 2004. Moth ers influence offspring body size through postoviposition maternal effects in the redbacked salamander, Plethodon cinereus . Oecologia 138:306-311. Crump, M. L. 1995. Parental care. Pp. 518-567 in H. Heatwole and B. K. Sullivan, eds. Amphibian Biology. Surrey Beat ty & Sons, Chipping Norton. Dale, S., R. Gustavsen, and T. Slagsvold. 1996. Risk taking during parental care: a test of three hypotheses applied to the pied flycatcher. Be havioral Ecology and Sociobiology 39:31-42. David, P., T. Bjorksten, K. Fowler, and A. Pomiankowski. 2000. Condition-dependent signalling of genetic variation in stalk-eyes flies. Nature 406:186-188. Davis, J. C. 1975. Minimal dissolved oxygen re quirements of aquatic life with emphasis on Canadian species: a review. J ournal of the Fisheries Resear ch Board of Canada 32:22952332. de Fraipont, M., J. Clobert, and R. Barbault. 1996. The evolution of oviparity with eggs guarding and viviparity in li zards and snakes: a phylogenetic analysis. Evolution 50:391400. de Jong, G. 1995. Phenotypic plasticity as a pr oduct of selection in a variable environment. American Naturalist 145:493-512. de Jong, T. J. 2005. Evolution of phenotypic plas ticity: patterns of plasti city and the emergence of ecotypes. New P hytologist 166:101-118. Dugatkin, L. A., and R. C. Sargent. 1994. Ma le-male association pattern s and female proximity in the guppy, Poecilia reticulata . Behavioral Ecology a nd Sociobiology 35:141-145. Echelle, A. A., and A. F. Echelle. 1993. Allo zyme variation and systematics of the New World Cyprinodontines (Teleostei: Cyprinodontidae ). Biochemical Systematics and Ecology 21:583-590. Eggert, A. K., M. Reinking, and J. K. Muller. 1998. Parental care impr oves offspring survival and growth in burying beetles. Animal Behaviour 55:97-107. Endler, J. A., and A. E. Houde. 1995. Geographic variation in female preferences for male traits in Poecilia reticulata . Evolution 49:456-468. Evans, D. H. 1993. Osmotic and ionic regulation. Pp. 315-341 in D. H. Evans, ed. The Physiology of Fishes. CRC Press, Boca Raton. Febry, R., and P. Lutz. 1987. Energy part itioning in fish: the activity-related cost of osmoregulation in a euryhalin e cichlid. Journal of Expe rimental Biology 128:63-85. Fisher, R. A. 1930. The Genetical Theory of Natural Selection. Clarendon Press, Oxford.

PAGE 152

152 Forsgren, E. 1997. Female sand gobies prefer good fathers over dominant males. Proceedings of the Royal Society of London Series B-Biological Sciences 264:1283-1286. Forsgren, E., A. Karlsson, and C. Kvarnemo. 19 96. Female sand gobies gain direct benefits by choosing males with eggs in their nests. Behavioral Ecology and Sociobiology 39:91-96. Foster, N. R. 1967. Comparative studies on th e biology of killifishes (Pisces, Cyprinodontidae). Pp. 369. Cornell University, Ithaca. Gilbert, C. R. 1992. Rare and endangered biot a of Florida, II. Fishes . University Press of Florida, Gainesville. Goolish, E. M. 1991. Anaerobic swimming me tabolism of fish: sit-and-wait versus active forager. Physiological Zoology 64:485-501. Gittleman, J. L. 1981. The phylogeny of parental care in fishes. Animal Behaviour 29:936-941. Gunter, G., and G. E. Hall. 1965. A biological investigation of the Cal oosahatchee Estuary of Florida. Gulf Research Reports 2:1-72. Hale, R. E., C. M. St Mary, and K. Lindstrom. 2003. Parental responses to changes in costs and benefits along an environmental gradient . Environmental Biology of Fishes 67:107-116. Hale, R. E. 2001. Parent al care in flagfish ( Jordanella floridae ): Do fathers alter parental effort in response to offspring mortality risk? Pp. 50. Department of Zoology. University of Florida, Gainesville. Hamilton, W. D. 1964. Genetical evolution of social behaviour . Journal of Theoretical Biology 7:1-52. Hamilton, W. D., and M. Zuk. 1982. Heritable true fitness and bright birds a role for parasites. Science 218:384-387. Haney, D. C., and F. G. Nordlie. 1997. Influe nce of environmental salinity on routine metabolic rate and critical oxygen tension of Cyprinodon variegatus . Physiological Zoology 70:511518. Hedgpeth, J. W. 1957. Es tuaries and lagoons. II. Biol ogical aspects. Pp. 693-729 in J. W. Hedgpeth, ed. Treatise on marine ecology and paleoecology. The Ge ological Society of America, New York. Hoelzer, G. A. 1989. The Good Parent Pro cess of Sexual Selection. Animal Behaviour 38:1067-1078. Holmes, J. C., and S. Zohar. 1990. Pathology and host behaviour. Pp. 34-63 in C. J. Barnard and J. M. Behnke, eds. Parasitism and Host Behaviour. Taylor-Francis, London.

PAGE 153

153 HomeSeer Technologies, LLC. 2005. HomeSeer Home Contro l Software Version 2.0. Bedford, NH. Hõrak, P. 2003. When to pay the cost of reproduction? A brood size manipulation experiment in great tits ( Parus major ). Behavioral Ecology an d Sociobiology 54:105-112. Iwasa, Y., and A. Pomiankowski. 1999. G ood parent and good genes models of handicap evolution. Journal of Theoretical Biology 200:97-109. Javois, J., and T. Tammaru. 2004. Reproducti ve decisions are sensi tive to cues of life expectancy: the case of a moth. Animal Behaviour 68:249-255. Jennions, M. D., and M. Petrie. 1997. Variati on in mate choice and mating preferences: a review of causes and consequen ces. Biological Reviews 72:283-327. Jones, J. C., and J. D. Reynolds. 1999. The influence of oxygen stress on female choice for male nest structure in the common goby. Animal Behaviour 57:189-196. Jordan, F., K. J. Babbitt, and C. C. McIvor. 1998. Seasonal variation in habitat use by marsh fishes. Ecology of Freshwater Fish 7:159-166. Jordan, F., D. C. Haney, and F. G. Nordlie. 1993. Plasma osmotic regulation and routine metabolism in the Eustis pupfish, Cyprinodon variegatus hubbsi (Teleostei, Cyprinodontidae). Copeia:784-789. Julian, D., W. G. R. Crampton, S. E. Wohlgemuth , and J. S. Albert. 2003. Oxygen consumption in weakly electric Neotropical fishes. Oecologia 137:502-511. Kaill, W. M. 1967. Ecology and behavior of the Cyprinodontid fishes Jordanella floridae (Goode and Bean) Floridichthys carpio (Gunther), and Cyprinodon variegatus , L. Pp. 159. Zoology. Cornell University, Ithaca. Kaitala, A., X. Espadaler, and R. Lehtonen. 2000. Ant predation and the cost of egg carrying in the golden egg bug: experiments in the field. Oikos 89:254-258. Kacelnik, A. 1988. Short-term adjustment of parental effort in starlings. Pp. 1843-1856 in H. Ouellet, ed. Acta XIX Congressus Internationa lis Ornithologici. Univeristy of Ottawa Press, Ottawa. Kilby, J. D. 1955. The fishes of two gulf coas tal marsh areas of Florid a. Tulane University, New Orleans. Kirkpatrick, M. 1985. Evolution of female choice and male parental investment in polygynous species the demise of the sexy s on. American Naturalist 125:788-810. Klug, H., A. Chin, and C. M. St. Mary. 2005. Th e net effects of guarding on egg survivorship in the flagfish, Jordanella floridae . Animal Behaviour 69:661-668.

PAGE 154

154 Klug, H., and C. M. St. Mary. 2005. The net effects of guarding on egg survivorship in the flagfish, Jordanella floridae . Animal Behaviour 69:661-668. Kodric-Brown, A. 1983. Determinants of male reproductive success in pupfish ( Cyprinodon pecosensis ). Animal Behavior 31:128-137. Koskela, E., P. Juutistenaho, T. Mappes, and T. A. Oksanen. 2000. Offspring defence in relation to litter size and age: Experiment in the bank vole Clethrionomys glareolus . Evolutionary Ecology 14:99-109. Lack, D. 1947. The significance of clutch size. Ibis 89:302-352. Lardies, M. A., T. P. Catalan, and F. Bozinovic. 2004. Metabolism and life-history correlates in a lowland and highland population of a terr estrial isopod. Canadian Journal of ZoologyRevue Canadienne De Zoologie 82:677-687. Lessells, C. M. 2002. Parentally biased favouri tism: why should parents specialize in caring for different offspring? Philosophical Transacti ons of the Royal Society of London Series BBiological Sciences 357:381-403. Lindström, K., C. M. St. Mary, and C. Pampoulie. 2006. Sexual selection fo r male parental care in the sand goby, Pomatoschistus minutus . Behavioral Ecology and Sociobiology 60:4651. Lindström, K., and C. Wennström. 1994. Exp ected future reproductive success and paternal behavior in the sand goby, Pomatoschistus minutus (Pisces, Gobiidae). Journal of Fish Biology 44:469-477. Listøen, C., R. F. Karlsen, and T. Slagsvold. 200 0. Risk taking during pare ntal care: a test of the harm-to-offspring hypothesis. Behavioral Ecology 11:40-43. Lloyd, D. G. 1987. Selection of offspring si ze at independence and other size-versus-number strategies. American Naturalist 129:800-817. Loftus, W. F., and J. A. Kushlan. 1987. Freshwat er fishes of southern Florida. Bulletin of the Florida State Museum, Biol ogical Sciences 31:147-344. Maigret, J. L., and M. T. Murphy. 1997. Co sts and benefits of pare ntal care in eastern kingbirds. Behavioral Ecology 8:250-259. Mank, J. E., D. E. L. Promislow, and J. C. Avise. 2005. Phylogenetic prespectives in the evolution of parental care in rayfinned fishes. Evolution 59:1570-1578. Maynard Smith, J. 1980. A new theory of sexual investment. Behavioral Ecology and Sociobiology 7:247-251. McNab, B. K. 2002. The physiological ecology of vertebrates: a view from energetics. Cornell University Press, Ithaca.

PAGE 155

155 McKitrick, M. C. 1992. Phylogenetic anal ysis of avian parental care. Auk 109:828-846. McNamara, J. M., T. Székely, J. N. Webb, and A. I. Houston. 2000. A dynamic game-theoretic model of parental care. Journal of Theoretical Biology 205:605-623. Mertz, J. C., and G. W. Barlow. 1966. On the reproductive behavior of Jordanella floridae (Pisces: Cyprinodontidae) with sp ecial reference to a quantitat ive analysis of parental fanning. Zeitschrift fur Ti erpsychologie 23:537:554. Møller, A. P., and R. Thornhill. 1998. Male pa rental care, differential parental investment by females and sexual selection. Animal Behaviour 55:1507-1515. Montgomerie, R. D., and P. J. Weatherhead. 1988. Risks and rewards of nest defense by parent birds. Quarterly Review of Biology 63:167-187. Moser, M. L., and W. F. Hettler. 1989. Routine metabolism of juvenile spot, Leiostomus xanthurus (Lacepede), as a function of temperature, salinity and weight. Journal of Fish Biology 35:703-707. Muir, B. S., and A. J. Niimi. 1972. Oxygen consumption of euryhali ne fish aholehole ( Kuhlia sandvicensis ) with reference to salinity, swimming, and food consumption. Journal of the Fisheries Research Board of Canada 29:67-&. Nice, M. 1962. Development of behavior in precocial birds. Transa ctions of the Linnean Society 8:1-211. Noldus Information Technology. 2003. The Ob server 5.0. Wageningen, The Netherlands. Nordlie, F. G. 1978. Influence of environmen tal salinity on respiratory oxygen demands in euryhaline teleost, Ambassis interrupta Bleeker. Comparative Biochemistry and Physiology a-Physiology 59:271-274. Nordlie, F. G. 1987. Salinity tolerance a nd osmotic regulation in the diamond killifish, Adinia xenica . Environmental Biology of Fishes 20:229-232. Nordlie, F. G. 2000. Patterns of reproduction and development of selected resident teleosts of Florida salt marshes. Hydrobiologia 434:165-182. Nordlie, F. G., D. C. Haney, and S. J. Walsh. 1992. Comparisons of salinity tolerances and osmotic regulatory capabilities in populations of sailfin molly ( Poecilia latipinna ) from brackish and fresh waters. Copeia:741-746. Nordlie, F. G., and S. J. Walsh. 1989. Adaptiv e radiation in osmotic regulatory patterns among 3 species of cyprinodontids (Teleostei, Atherinomorpha). Physiological Zoology 62:12031218.

PAGE 156

156 Nordlie, F. G., S. J. Walsh, D. C. Haney, and T. F. Nordlie. 1991. The influence of ambient salinity on routine metabolism in the teleost Cyprinodon variegatus Lacepede. Journal of Fish Biology 38:115-122. Olsen, J., and A. D. Tucker. 2003. A broodsize manipulation experiment with peregrine falcons, Falco peregrinus , near Canberra. Emu 103:127-132. OMS Tech. 1992. E-Lab Chromatography System Version 4. OMS Tech, Miami, Florida. Östlund, S., and I. Ahnesjö. 1998. Female fi fteen-spined sticklebacks prefer better fathers. Animal Behaviour 56:1177-1183. Page, L. M., and B. M. Burr. 1991. A field guide to freshwater fishes of North America north of Mexico. Houghton-Mifflin Co., Boston. Pampoulie, C., K. Lindstrom, and C. M. St Mary . 2004. Have your cake and eat it too: male sand gobies show more parental care in th e presence of female partners. Behavioral Ecology 15:199-204. Parker, T. H. 2003. Genetic benefits of mate choice separated from differential maternal investment in red junglefowl ( Gallus gallus ). Evolution 57:2157-2165. Parvatheswararao, V. 1965. Influence of diffe rent temperature-salinity combinations on oxygen consumption in the fresh water fish Etroplus Maculatus (Teleostei). Helgolander Wissenschaftliche Meeresuntersuchungen 12:302-&. Pijanowski, B. C. 1992. A revision of Lack's brood reduction hypothesis. American Naturalist 139:1270-1292. Petersen, C. W., C. Mazzoldi, K. A. Zarrella, a nd R. E. Hale. 2005. Fertilization mode, sperm characteristics, mate choice and parental care patterns in Artedius spp. (Cottidae). Journal of Fish Biology 67:239-254. Potts, W. T. W. 1954. The energetics of osmotic regulation in brackish and fresh-water animals. Journal of Experimental Biology 31:618-630. Qvarnström, A. 2001. Context-dependent genetic benefits from mate choice. Trends in Ecology & Evolution 16:5-7. Qvarnström, A., T. Part, and B. C. Sheldon. 2000 . Adaptive plasticity in mate preference linked to differences in reproduc tive effort. Nature 405:344-347. R Development Core Team. 2006. R: A language and environment for statistical computing. R Foundation for Statistical Com puting. Vienna, Austria. Reynolds, J. D., N. B. Goodwin, and R. P. Freck leton. 2002. Evolutionary transitions in parental care and live beari ng in vertebrates. Philosophica l Transactions of the Royal Society of London Series B-Bi ological Sciences 357:269-281.

PAGE 157

157 Reynolds, J. D., and J. C. Jones. 1999. Female preference for preferred males is reversed under low oxygen conditions in the common goby (Pomatoschistus microps). Behavioral Ecology 10:149-154. Ridgway, M. S. 1988. Developmental stage of offspring and brood defense in smallmouth bass ( Micropterus dolomieui ). Canadian Journal of ZoologyRevue Canadienne De Zoologie 66:1722-1728. Rosenthal, G. G., C. S. Evans, and W. L. Mille r. 1996. Female preference for dynamic traits in the green swordtail, Xiphophorus helleri . Animal Behaviour 51:811-820. Rytkonen, S. 2002. Nest defence in great tits Parus major : support for parental investment theory. Behavioral Ecology and Sociobiology 52:379-384. St. Mary, C. M., E. Gordon, and R. E. Hale. 2004. Environmental effects on egg development and hatching success in Jordanella floridae , a species with parental care. Journal of Fish Biology 65:760-768. St. Mary, C. M., C. Noureddine, and K. Linds tröm. 2001. Environmental effects on male reproductive success and parental care in the Florida flagfish ( Jordanella floridae ). Ethology 107:1035-1052. Sanz, J. J., and J. M. Tinbergen. 1999. En ergy expenditure, nestling age, and brood size: an experimental study of parental behavior in the great tit Parus major . Behavioral Ecology 10:598-606. Sargent, R. C. 1990. Behavioral and evolut ionary ecology of fishes conflicting demands during the breeding season. Annale s Zoologici Fennici 27:101-118. Sargent, R. C., and M. R. Gross. 1993. Williams’ principle: an explanation of parental care in teleost fishes. Pp. 275-293 in T. J. Pitcher, ed. The Behavior of Teleost Fishes. Chapman & Hall, London. SAS Institute. 2000. User's Guide: Statistics , Version 8. SAS Institute, Inc., Cary, North Carolina. Schmoll, T., V. Dietrich, W. Winkel, J. T. Eppl en, F. Schurr, and T. Lubjuhn. 2005. Paternal genetic effects on offspring fitness are cont ext dependent within the extrapair mating system of a socially monogamous passerine. Evolution 59:645-657. Schurmann, H., and J. F. Steffensen. 1997. E ffects of temperature, hypoxia and activity on the metabolism of juvenile Atlantic cod. Journal of Fish Biology 50:1166-1180. Seehausen, O., J. J. M. vanAlphen, and F. W itte. 1997. Cichlid fish diversity threatened by eutrophication that curbs sexual selection. Science 277:1808-1811.

PAGE 158

158 Shackleton, M. A., M. D. Jennions, and J. Hunt . 2005. Fighting success and attractiveness as predictors of male mating success in the black field cricket, Teleogryllus commodus : the effectiveness of no-choice tests. Beha vioral Ecology and Sociobiology 58:1-8. Sheldon, B. C., H. Arponen, A. Laurila, P. A. Cr ochet, and J. Merila. 2003. Sire coloration influences offspring survival under predation ri sk in the moorfrog. Journal of Evolutionary Biology 16:1288-1295. Simcox, H., N. Colegrave, A. Heenan, C. Howa rd, and V. A. Braithwaite. 2005. Contextdependent male mating preferences for unfamiliar females. Animal Behaviour 70:14291437. Smith, S. N., R. A. Armstrong, J. Springate, and G. Barker. 1985. Infection and colonization of trout eggs by Saprolegniaceae. Transactions of the British Mycological Society 85:719723. Sokal, R. R., and R. J. Rohlf. 1995. Biomet ry: the principles and pr actice of statistics in biological research. W. H. Fr eeman and Company, New York. Soltz, D. L., and R. J. Naiman. 1978. The natu ral history of native fishes in the Death Valley system. Natural History Museum of Los Angeles County, Science Series 30:1-76. Starck, J. M., and R. E. Ricklefs. 1998. Patterns of development: the altricial-precocial spectrum. Pp. 3-30 in J. M. Starck and R. E. Ricklefs , eds. Avian growth and development: evolution within the altricial-precocial spectrum. Oxford University Press, New York. Stearns, S. C. 1992. The Evolution of Life Histories. Oxford University Press, Oxford. Strathmann, R. R., and M. F. Strathmann. 1995 . Oxygen supply and limits on aggregation of embryos. Journal of the Marine Biological Association of th e United Kingdom 75:413428. Stuenkel, E. L., and S. D. Hillyard. 1981. Th e effects of temperature and salinity acclimation on metabolic rate and osmo regulation in the pupfish Cyprinodon salinus . Copeia 1981:411417. Styczy ska-Jurewicz, E. 1970. Bioenergetics of osmoregulation in aquatic animals. Polish Archives of Hydrobiology 17:295-302. Tagatz, M. E. 1967. Fishes of the St. Johns Ri ver, Florida. Quarterly Journal of the Florida Academy of Sciences 30:25-50. Tallamy, D. W. 2000. Sexual selection and th e evolution of exclus ive paternal care in arthropods. Animal Behaviour 60:559-567. Tammaru, T., and P. Hõrak. 1999. Should one in vest more in larger broods? Not necessarily. Oikos 85:574-581.

PAGE 159

159 Temme, D. H. 1986. Seed size variability a consequence of variable genetic quality among offspring. Evolution 40:414-417. Temme, D. H., and E. L. Charnov. 1987. Brood size adjustment in birds economical tracking in a temporally varying environment. Journal of Theoretical Biology 126:137-147. Tolonen, P., and E. Korpimaki. 1996. Do kestrels adjust their parental effo rt to current or future benefit in a temporally varying en vironment? Ecoscience 3:165-172. Tomkins, J. L., L. W. Simmons, R. J. Knell, a nd K. A. Norris. 1999. Correlates of ball size and rolling speed in the dung beetle Kheper nigroaeneus (Coleoptera: Scarabaeidae). Journal of Zoology 248:483-487. Trexler, J. C., W. F. Loftus, and S. Perry . 2005. Disturbance fr equency and community structure in a twenty-five year in tervention study. O ecologia 145:140-152. Trivers, R. L. 1972. Parental inve stment and sexual selection. Pp. 136-179 in B. Campbell, ed. Sexual Selection and the Descent of Man. Heinemann, London. Truesdale, G. A., A. L. Downing, and G. F. Lowden. 1955. The solubility of oxygen in pure water and sea-water. Journal of Applied Chemistry 5:53-62. Tullberg, B. S., M. Ah-King, and H. Temrin. 20 02. Phylogenetic reconstr uction of parental-care systems in the ancestors of birds. Philosophi cal Transactions of the Royal Society of London Series B-Biological Sciences 357:251-257. Tveraa, T., B. E. Sæther, R. Aanes, and K. E. Erikstad. 1998. Regulation of food provisioning in the Antarctic petrel; the importance of parental body condition and chick body mass. Journal of Animal Ecology 67:699-704. van Iersel, J. J. A. 1953. An analysis of the parental behaviour of the male threespined stickleback ( Gasterosteus aculeatus L.). Behaviour (Supplement) 3:1-159. Via, S., R. Gomulkiewicz, G. De Jong, S. M. Scheiner, C. D. Schlichting, and P. H. Vantienderen. 1995. Adaptive phenotypic plas ticity: consensus and controversy. Trends in Ecology & Evolution 10:212-217. Via, S., and R. Lande. 1985. Genotype-environm ent interaction and the evolution of phenotypic plasticity. Evolution 39:505-522. Wagner, W. E. 1998. Measuring female mati ng preferences. Animal Behaviour 55:1029-1042. Webb, J. N., T. Szekely, A. I. Houston, and J. M. McNamara. 2002. A theoretical analysis of the energetic costs and consequences of parent al care decisions. Philosophical Transactions of the Royal Society of London Series B-Biological Sciences 357:331-340.

PAGE 160

160 Weimerskirch, H., P. A. Prince, and L. Zimme rmann. 2000. Chick provisioning by the yellownosed albatross Diomedea chlororhynchos : response of foraging effort to experimentally increased costs and demands. Ibis 142:103-110. Weimerskirch, H., L. Zimmermann, and P. A. Prince. 2001. Influence of environmental variability on breeding effort in a long-lived seabird, the yellow-nosed albatross. Behavioral Ecology 12:22-30. Welch, A. M. 2003. Genetic benefits of a female mating preference in gray tree frogs are context-dependent. Evolution 57:883-893. Westneat, D. F., and R. C. Sargent. 1996. Sex and parenting: the effect s of sexual conflict and parentage on parental strategies. Tre nds in Ecology & Evolution 11:A87-A91. Wiklund, C. G. 1990. Offspring protection by merlin Falco columbarius females the importance of brood size and expected offspri ng survival for defense of young. Behavioral Ecology and Sociobiology 26:217-223. Williams, G. C. 1966. Natural selection cost s of reproduction and a refinement of Lack's principle. American Naturalist 100:687-690. Winkler, D. W., and K. Wallin. 1987. Offspri ng size and number: a life history model linking effort per offspring and total effo rt. American Naturalist 129:708-720. Wolf, J. B., and M. J. Wade. 2001. On the assign ment of fitness to pare nts and offspring: whose fitness is it and when does it matter? Journal of Evolutionary Biology 14:347-356.

PAGE 161

161 BIOGRAPHICAL SKETCH Rebecca Hale was born April 14, 1974, in Wa shington, D.C., and grew up in Fairfax County, Virginia. She graduated from Falls C hurch High School in Falls Church, Virginia, in 1996 and received her B.A. in biology from Oberlin College in Oberlin, Ohio, in 1996. In 1997, she worked as a member of the AmeriCorps program with the USDA/A gricultural Research Service in Fort Lauderdale, Florida, on projects related to the control and management of exotic invasive plant species. She earned her maste rÂ’s degree at the Univer sity of Florida in 2001 conducting research into the ro le of dissolved oxygen on parental care and offspring development in flagfish ( Jordanella floridae ).