THE EFFECTS OF DISPERSAL, INBREEDING, AND MUTUALISMS ON THE FITNESS 
OF FRAGMENTED POPULATIONS 

By 

Isabela Lima Borges 

A DISSERTATION 

Submitted to 
Michigan State University 
in partial fulfillment of the requirements 
for the degree of 

Integrative Biology — Doctor of Philosophy 
Ecology, Evolutionary Biology and Behavior — Dual Major 

2024

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ABSTRACT 

Anthropogenic global change has led to the decline and fragmentation of countless 

natural populations throughout the tree of life and across the globe. Understanding how these 

fragmented populations evolve and interact with their environments is a fundamental question in 

ecology and evolution with potential applications for biodiversity conservation. In this 

dissertation, I researched how dispersal, inbreeding, and mutualisms affect fitness in two 

systems: Poecilia reticulata (the Trinidadian guppy) and Chamaecrista fasciculata (the partridge 

pea). Using observations and experiments, I aimed to answer three main questions: (Chapter 2) 

What are the effects of dispersal on fitness in the Trinidadian guppy?; (Chapter 3) What are the 

individual and population-level drivers of dispersal in the Trinidadian guppy?; (Chapter 4) How 

do inbreeding and mutualisms interact to affect fitness in Chamaecrista fasciculata? I found that 

dispersal behavior in guppies is driven by complex interactions between population, individual, 

and seasonal conditions; that dispersal and inbreeding are important drivers of fitness in guppies 

and C. fasciculata; that mutualisms are affected by inbreeding in C. fasciculata; and that the 

expression of inbreeding depression in C. fasciculata is altered by the loss of mutualistic 

interactions.

Dedicada à minha família, em especial a Carlos Cândido Borges

iii 

 
 
ACKNOWLEDGEMENTS 

None of this research, nor the enjoyment I got out of it, would have been possible without 

the support and guidance of my mentors, friends, and family. I am, first and foremost, grateful to 

Sarah Fitzpatrick for being my Ph.D. advisor. Sarah has been patient, generous, excited, and 

endlessly encouraging throughout my time in her lab, pushing me to become a better scientist 

while supporting me to be a good and happy person. Sarah’s passion for incorporating rigorous 

research into applied conservation has inspired me to continue on a scientific path and always be 

mindful of the impacts of my work. I could not have completed this degree in the midst of a 

global pandemic, away from my family and friends, without her kindness and the warm 

environment she fosters in the Fitzpatrick lab. It has been a difficult but incredible adventure, 

and I could not have chosen a better person to guide me through it. 

I would also like to thank my committee for their scientific guidance and support 

throughout my degree. Nick Haddad, Emily Josephs, and Phoebe Zarnetske provided invaluable 

feedback and lively intellectual discussion throughout our many meetings over the years, and I 

am grateful to Lauren Sullivan for graciously stepping in as a committee member for my defense 

on very short notice. My dissertation is better for their participation in it. Additionally, none of 

my research at KBS would have been possible without the tireless work of Andy Fogiel, Mark 

Hammond, Holly Vander Stel, and Daniel Gorentz, who are vital to keeping this field station 

productive, safe, and welcoming to new researchers. 

Being a member of the Fitzpatrick lab has been an absolute joy. Kyle Jaynes is the best 

co-“first pancake” I could have dreamed of, and it has been magical to see his family grow 

alongside our lab, research, and friendship. I want to thank Meaghan Clark for always bringing 

her fierceness, craftiness, and humor to the lab, keeping everyone on track whether or not that 

iv 

 
 
track happens to be work. Tyler Linderoth, Cinnamon Mittan, and Alex Lewanski are three of 

the strangest and loveliest people I have ever met, and I enjoyed every lab meeting, turtling 

expedition, movie night, and caffeinated lab afternoon in their company. Our Archbold trips, 

where we all got to spend time with the unstoppable Jessica Judson, Caitlin Weid, and their 

infinite fish, were highlights of my Ph.D. I cannot imagine a more fun, supportive, and 

interesting group of people to work with, and I will miss them terribly. 

Beyond the lab, the whole KBS community has been a steady presence throughout the 

years, despite all the turnover that comes from being at a field station. I am thankful for every 

house party, backyard bonfire, and Windmill Island reception. Immediately befriending Ravi 

Ranjan was one of my luckiest decisions, and absolutely necessary for settling into life in the 

U.S. while maintaining my sanity. I cherish every long conversation, homecooked meal, and trip 

we shared, and look forward to many more as we settle into our new (still U.S.) lives. I also want 

to thank Christopher Warneke, Robert Logan, and Alice Puchalsky for every phone call, hike, 

and dinner in their company. Their love for the natural world, and willingness to share it with 

me, made life in Hickory Corners more meaningful. 

Outside of KBS, I remain surprised by the community I found in Kalamazoo. Blaire 

Bohlen and Jamie Smith welcomed me into their homes, always ready to provide a warm meal, a 

soft hug, a weird dog, and expert life advice. They built a family and a life for me here. There are 

so many wonderful people I got to meet through Blaire and Jamie whose friendship made my 

time in Michigan more fun, including Grant Grames, Alan Streeter, Gina Krzebiot, Jerod 

Kauffman, Jared Smith, and David Zakharov. Meeting Victor DeBrito and Patrícia Dias was 

another very lucky strike, and their warmth, generosity, and general carioquisse offered me a 

much needed slice of home in the cold times. All of them made Kalamazoo a happy place to be. 

 v 

Before the Michigan times, there were the Brazil and Toronto friends. I would not be on 

this biology path were it not for my time at UFRJ and the community I found within Caldeirão 

da Dandara. They are a phenomenally talented group of people, travellers and adventurers of all 

sorts who have found their callings within and outside the field that first brought us together. In 

Canada, Leila Forsyth showed me how to be a scientist (and an adult), and her and Penny were 

my home in the exciting and confusing times of late undergrad and early grad school. All these 

people inspire me to care deeply about my work, to build community, to strive for a better world. 

In particular, Rafael Martins, Juliana Silva, and Julia Rodrigues have been my companions in all 

of life’s adventures for the past decade and a half, cheering for me even through the decisions 

that increased the geographic distance between us. My friendships with them ground me and 

build me up. They are my place in the world. 

Finally, my most sincere and eternal gratitude goes to my family. My parents, Heloisa 

and Luiz, invested all of their love and time into making sure my sister and I can be whoever we 

want to be. They taught us to be curious about the world, and confident of our places in it. My 

sister Sofia is the coolest person I know, and her approval is still the only one I care about. 

Though our dreams have taken us far, they know how to always stay close. I miss them every 

day, but am comforted by their unwavering excitement for all my achievements, big and small. 

And as if all of that was not wonderful enough, I now get to build my own family 

alongside Jordan Incorvaia. Of all the beautiful things this Ph.D. gave me, you are my favorite. I 

cannot wait for our next adventure together, and the one after that and the one after that and the 

one after that… 

 vi 

 
TABLE OF CONTENTS 

CHAPTER ONE: Introduction ....................................................................................................... 1 

CHAPTER TWO: Reproductive benefits associated with dispersal in headwater populations of 
Trinidadian guppies (Poecilia reticulata) ....................................................................................... 6 
Abstract ....................................................................................................................................... 6 
Introduction ................................................................................................................................. 6 
Methods....................................................................................................................................... 9 
Results ....................................................................................................................................... 15 
Discussion ................................................................................................................................. 20 
Acknowledgements ................................................................................................................... 25 

CHAPTER THREE: Rollin’ on the river: Local census size, sex ratio, and relatedness drive 
dispersal behavior in the Trinidadian guppy (Poecilia reticulata) ............................................... 26 
Abstract ..................................................................................................................................... 26 
Introduction ............................................................................................................................... 26 
Methods..................................................................................................................................... 31 
Results ....................................................................................................................................... 34 
Discussion ................................................................................................................................. 41 
Acknowledgements ................................................................................................................... 47 

CHAPTER FOUR: Mutualism benefits are lost at high inbreeding levels .................................. 48 
Abstract ..................................................................................................................................... 48 
Introduction ............................................................................................................................... 48 
Methods..................................................................................................................................... 53 
Results ....................................................................................................................................... 58 
Discussion ................................................................................................................................. 66 
Acknowledgements ................................................................................................................... 71 

REFERENCES ............................................................................................................................. 75 

APPENDIX A: Supporting information: Reproductive benefits associated with dispersal in 
headwater populations of Trinidadian guppies (Poecilia reticulata) ........................................... 93 

APPENDIX B: Supporting information: Rollin’ on the river: Local census size, sex ratio, and 
relatedness drive dispersal behavior in the Trinidadian guppy (Poecilia reticulata) ................. 114 

APPENDIX C: Supporting information: Mutualism benefits are lost at high inbreeding……….120 

 vii 

 
 
 
 
 
 
 
CHAPTER ONE: 

Introduction 

Habitat fragmentation is one of the defining human impacts of the Anthropocene 

(Haddad et al., 2015). Fragmentation affects biodiversity at all levels, from continents (Riitters et 

al. 2000; Crooks et al. 2011) to microbes (Mony et al. 2022), whole ecosystems (Mitchell et al. 

2015) to single genes (McCartney-Melstad et al. 2018), ecology to evolution. As an example 

from a single biome, as of 2015, approximately 70% of all remaining forests on Earth lie within 

one kilometer of an edge (Haddad et al., 2015). Understanding the consequences of such 

extensive restructuring of natural landscapes requires research that connects across scales, 

considering the knock-on effects that one scale may have on another.    

At the individual level, habitat fragmentation can decrease organisms’ ability to disperse 

(e.g. Brooker et al., 1999; Coulon et al., 2010; Schtickzelle et al., 2006; Soons et al., 2005). 

When suitable patches are surrounded by a hostile matrix of human-modified environments, 

individual dispersal can become constrained. Decreased patch sizes, increased distance between 

patches, and increased patch isolation all contribute to disrupt dispersal (Niebuhr et al. 2015; 

Cote et al. 2017; Fletcher et al. 2018). Dispersal, however, is an important part of the life-history 

of many species––allowing for the establishment of new territories, release from competition, 

and inbreeding avoidance (Van Dyck & Baguette 2005; Clobert et al. 2012; Baguette et al. 

2013). When individual dispersal is limited by fragmentation, ecological and evolutionary 

processes like range expansion, recolonization, mate choice, and gene flow also cease to function 

as they previously did. 

As a result, population level processes can be deeply changed by habitat fragmentation. 

Ecologically, lower dispersal decreases colonization of new patches and prevents the rescue of 

 1 

declining populations, potentially leading to local extinctions (MacArthur & Wilson 1963; 

Hanski 1998; Mittelbach & Schemske 2015). Evolutionarily, decreased dispersal leads to lower 

rates of gene flow, without which populations become more differentiated from one another, and 

lose genetic diversity through increased drift and inbreeding (Slatkin 1987; Ronce 2007; Clobert 

et al. 2012). Increased inbreeding is linked to loss of fitness (“inbreeding depression”) in many 

systems, and especially in small and declining populations (Brook et al. 2002; Armbruster & 

Reed 2005a; Kardos et al. 2016; Doekes et al. 2021). Inbreeding depression occurs due to 

genome-wide increases in homozygosity, which lead to the expression of previously masked 

recessive deleterious variants and loss of heterozygote advantage (“overdominance”) 

(Charlesworth & Charlesworth, 1987; Charlesworth & Willis, 2009; Keller & Waller, 2002). 

Similarly, the loss of genetic diversity through increased drift can lead to the fixation of 

deleterious alleles, decreasing fitness and eroding populations’ ability to adapt to novel 

conditions over time (Wright 1966; Lande 1976; Willi et al. 2006; Bijlsma & Loeschcke 2012; 

Hoffmann et al. 2017; Spigler et al. 2017). Together, these factors can speed up declines and 

send populations into an extinction vortex (Gilpin & Soulé, 1986). 

As populations and species are lost, whole ecological communities change. Local 

extinctions not only decrease biodiversity as a whole, but they create gaps in biotic interaction 

networks that can lead to further losses (Colwell et al. 2012; Bascompte et al. 2019). Tightly 

coevolved interactions, like many mutualisms and parasitisms, are increasingly at risk of 

coextinctions––once one of the partners is lost, the other is also likely to disappear (Dunn et al. 

2009a). Obligate ant-following birds, for example, depend on patchily-distributed colonies of 

army ants, so as ants disappear from small forest fragments due to the lack of food resources, so 

do the birds (Harper 1989). Even in the absence of local extinctions, fragmentation can still 

 2 

disrupt biotic interactions. For instance, when differential migration ability through a human-

modified landscape creates asynchrony in the interactions between plants and their seed 

dispersers (Rodríguez-Cabal et al. 2007), or when harmful pathogens are introduced to new 

locations through human movement (James et al. 2015). The reshuffling of biotic interactions 

that comes with fragmentation is often a concurrent stressor that species need to contend with 

alongside disrupted dispersal and lower gene flow. 

The negative effects of fragmentation not only propagate up, as described above, from 

individuals to populations to communities, but the consequences of fragmentation at larger scales 

can also feed back down to lower levels of organization. For example, novel pathogens at the 

community level can be more successful at establishing in more genetically depauperate, inbred 

populations (Gibson & Nguyen 2020); and changes in population structure that increase overall 

relatedness can alter individual dispersal behavior (Daniels & Walters 2000). As a result, 

understanding how fragmentation negatively impacts biodiversity can be incredibly complex, 

requiring observation and experimentation across several levels of organization.  

For my dissertation, I investigated how dispersal, inbreeding, and biotic interactions 

impact fitness in two species: Chamaecrista fasciculata, the partridge pea, and Poecilia 

reticulata, the Trinidadian guppy. P. reticulata is a model system in evolutionary ecology, 

naturally occurring in streams in northern South America (de Bragança et al. 2020). In Trinidad, 

these fish are a textbook example of rapid adaptation, evolving different life histories, behavior, 

and morphology in response to selective pressures related to their habitats, especially the 

predator community (Endler 1980, 1987; Reznick 1989; Magurran & Seghers 1991; Magurran et 

al. 1993; Endler & Houde 1995; Reznick et al. 2001). In higher elevation headwater streams, 

guppies coexist with few predators and are naturally fragmented by the distribution of streams 

 3 

across the landscape, waterfall barriers between stream reaches, and riffles that form between the 

pools where guppies spend their lives. As such, these small headwater populations can suffer 

from inbreeding depression (Van Oosterhout et al. 2003; Johnson et al. 2010; Smallbone et al. 

2016), blinking in and out of existence as population sizes fluctuate and extinctions and 

recolonizations occur. 

The first two chapters of my dissertation aim to better understand the causes and 

consequences of dispersal in small populations of P. reticulata. Chapter 2 looks at the fitness 

consequences of individual dispersal, describing the movement patterns of these fish and how 

they correlate with reproductive success. For Chapter 3, I connect population-level factors, 

specifically local pool census size, sex ratio, and relatedness, to individual dispersal behavior. In 

Chapter 2, I find that most fish move little or not at all throughout their lifetimes, but the ones 

that do disperse accrue significant fitness benefits. Males, in particular, are more likely to 

produce offspring, have a higher number of offspring, and have more mates the more they 

disperse. In terms of dispersal drivers (Chapter 3), I find that census size is an important 

predictor of dispersal probability in both streams, and that guppies’ response to census size vary 

with season and fish weight. Sex ratio and relatedness also predict dispersal probabilities, but did 

so differently in the two study streams. Overall, I find that population-level characteristics in 

these streams are driving individual guppy behavior, with consequences to their fitness. These 

chapters  made use of an extensive, multi-year dataset and pedigree collected by Fitzpatrick et al. 

(2015, 2016, 2020) in two natural guppy populations, initially collected to test hypotheses about 

genetic rescue of headwater fish using divergent lower elevation populations.  

Chamaecrista fasciculata is an annual legume native to eastern North America that 

occurs in disturbed prairie habitats in Michigan (Fenster 1991a; Fenster et al. 2003; Bueno et al. 

 4 

2019). As such, its populations are naturally fragmented across the landscape, appearing in 

disturbed and early-successional sites and disappearing as prairies develop. That natural 

fragmentation, however, is exacerbated by the loss of prairie habitat in the U.S. Midwest 

(Samson & Knopf 1994). As an outcrossing but self-compatible plant, C. fasciculata suffers 

from inbreeding depression when occurring in small and isolated populations (Fenster 1991a; 

Fenster & Galloway 2000; Mannouris & Byers 2013). As a legume, C. fasciculata participates in 

a mutualism with nitrogen-fixing soil bacteria (rhizobia), as well as a mutualism with bodyguard 

ants.  

In Chapter 4, I researched whether inbreeding in C. fasciculata affects the strength of its 

mutualistic interactions with ants and bacteria, and whether the presence of rhizobia alters the 

expression of inbreeding depression. Following two generations of controlled crosses in the 

greenhouse and an outdoor common garden experiment with over 2500 individuals, I found that 

inbreeding decreases the plants’ investment in both mutualisms. Importantly, I find that 

inbreeding depression is steeper for plants that interact with rhizobia, as the fitness benefits 

provided by the mutualism are lost at high levels of inbreeding.     

My dissertation aims to fill gaps in our understanding of the effects of fragmentation in 

natural systems, focusing on how dispersal, increased inbreeding, and mutualism loss affect 

fitness. The effects we find of population-level dynamics on individual dispersal behavior 

(Chapter 3), of dispersal on fitness (Chapter 2), and of inbreeding and mutualisms (Chapter 4) 

highlight the need to integrate across scales when considering how organisms will respond to the 

pervasive and ongoing fragmentation of their habitats. 

 5 

CHAPTER TWO: 

Reproductive benefits associated with dispersal in headwater populations of Trinidadian guppies 
(Poecilia reticulata) 

Abstract 

Theory suggests that the evolution of dispersal is balanced by its fitness costs and 

benefits, yet empirical evidence is sparse due to the difficulties of measuring dispersal and 

fitness in natural populations. Here, we use spatially-explicit data from a multi-generational 

capture-mark-recapture study of two populations of Trinidadian guppies (Poecilia reticulata) 

along with pedigrees to test whether there are fitness benefits correlated with dispersal. 

Combining these ecological and molecular datasets allows us to directly measure the relationship 

between movement and reproduction. Individual dispersal was measured as the total distance 

moved by a fish during its lifetime. We analyzed the effects of dispersal propensity and distance 

on a variety of reproductive metrics. We found that number of mates and number of offspring 

produced were positively correlated to dispersal, especially for males. Our results also reveal 

individual and environmental variation in dispersal, with sex, size, season, and stream acting as 

determining factors. 

Introduction 

Dispersal is one of the major processes shaping diversity in the natural world. Broadly 

defined, dispersal is a departure from a local environment, followed by movement and settlement 

in a new location (Clobert et al. 2012). Dispersal followed by successful breeding leads to gene 

flow, which alters the distribution of genotypes across space (Ronce 2007; Clobert et al. 2012). 

Dispersal, then, plays a crucial role in determining population structure, patterns of adaptive 

differentiation, and population dynamics (Kawecki & Holt 2002; Sexton et al. 2009; Kubisch et 

al. 2014; Bonte & Dahirel 2017). The effects of dispersal on these evolutionary and ecological 

 6 

processes crucially depends on the fitness of dispersing individuals, both in terms of their 

survival and reproduction. Yet, dispersal and fitness are two of the most challenging quantities to 

empirically measure in wild populations, and so our understanding of their relationship is 

tenuous (Doligez & Pärt 2008).  

Theory suggests that the probability and extent of dispersal within a population are 

shaped by a balance between fitness costs and benefits incurred by dispersing individuals (Holt 

2003). Dispersal costs may manifest though energetic costs of physically moving, and/or 

increased mortality associated with moving through unsuitable or dangerous environments 

(Bonte et al. 2012). Benefits may arise through a variety of mechanisms, including a reduction in 

competition among relatives (Hamilton 1977; Taylor 1988; Perrin & Mazalov 2000), escape 

from unfavorable natal conditions (Venable & Brown 1988; Meylan et al. 2002), resource 

acquisition (Clobert et al. 2012), and inbreeding avoidance (Pusey & Wolf 1996; Perrin & 

Mazalov 1999). Dispersal should evolve when dispersal benefits outweigh costs, yet we expect 

those benefits to differ between individuals and environments. Given observed variation in 

dispersal propensity, it is important to understand how fitness benefits of dispersal are distributed 

across individuals and environments to understand dispersal evolution. The extent to which 

fitness effects of dispersal depend on individual characteristics such as sex, size, and total 

distance dispersed, as well as environmental factors, remains an open question in most natural 

systems. Here, we integrate whole-population mark-recapture and pedigree analyses in two 

natural populations of Trinidadian guppies (Poecilia reticulata) to quantify the relationship 

between dispersal and reproductive fitness at fine spatial and temporal scales. 

Poecilia reticulata are live-bearing fish inhabiting streams and rivers throughout Trinidad 

and Tobago. Decades of research on the guppy mating system have shown that female choice is 

 7 

important (Haskins & Haskins, 1950; Haskins & Haskins, 1949; Houde, 1987), often on the basis 

of phenotypic novelty (Brooks & Endler 2001; Hughes et al. 2013a). Given the genetic basis of 

P. reticulata coloration (Haskins & Haskins 1951; Houde 1992), this preference might be a 

mechanism for females to avoid mating with kin (Hughes et al. 1999; Daniel & Rodd 2016). P. 

reticulata populations in headwater streams are also strongly regulated by density dependence, 

with high population density increasing mortality and limiting reproductive success (Reznick et 

al. 2012). Thus, it is likely that guppy dispersal plays an important role in minimizing inbreeding 

and allowing fish to escape high density pools––potentially providing dispersing individuals with 

substantial fitness benefits. 

To date, it is known that male guppies are more likely to disperse than females (Croft et 

al. 2003) and that density-dependent dispersal varies with guppy life stage (De Bona et al. 2019). 

Throughout Trinidad, guppies show strong patterns of population structure associated with 

drainage (Fraser et al. 2015), distance, and barriers to gene flow (Crispo et al. 2006), suggesting 

some dispersal limitation at large spatial scales. P. reticulata is also a prolific global invader, and 

its ability to establish new populations following dispersal or artificial introduction greatly 

impacts freshwater ecosystems worldwide (Deacon et al. 2011). However, no previous study has 

related dispersal to reproduction in the P. reticulata system, or quantified how reproductive 

benefits of dispersal vary among individuals or environments. As such, the crucial link between 

dispersal and fitness––and its consequences for dispersal evolution, gene flow, and population 

structure––remain unknown. Here, we aim to fill this gap with a long-term, multi-generational 

study to ask the overarching question: Is dispersal correlated with reproductive success in P. 

reticulata?  

 8 

Our hypothesis, based on the natural history of guppy populations in headwater 

environments, is that dispersal confers a reproductive benefit at this scale, and that these benefits 

are greater for males than females. That is, guppies that move at all and those that disperse 

farther experience greater reproductive success than philopatric individuals, and reproductive 

benefits are higher for males as they compete for access to females. However, we predict that 

movement is only beneficial at small spatial scales within the low predation environment, as 

long-distance dispersal into downstream high predation reaches results in high mortality (Weese 

et al. 2011). These costs associated with local adaptation to heterogenous predation regimes may 

lead to the evolution of short dispersal distances over evolutionary time. We also analyzed how 

dispersal behavior vary with male body size, and across seasons. We hypothesize that smaller 

fish move more to escape conditions where they are competitively inferior, and all fish move 

more and in the downstream direction during the wet season due to floods and high stream flow 

which facilitate both passive and active movement. Our dataset allowed us to test for correlations 

between fine-scale dispersal and fitness for two natural populations, and contributes to the 

understanding of individual variation in dispersal for a model system in ecology and evolution. 

Methods 

Field capture-mark-recapture 

We studied low predation populations of P. reticulata in neighboring streams, Taylor and 

Caigual, in the Guanapo drainage on the south slope of the Northern Range of Trinidad. Data 

used for this project were collected in a spatially explicit, monthly capture-mark-recapture study 

that spanned June 2009−July 2011. In April 2009, as part of a separate study (see Travis et al. 

2014), guppies from downstream site within the same drainage were translocated upstream of 

our two focal sites. Translocated individuals eventually reached and bred with the focal resident 

 9 

populations (Fitzpatrick et al. 2016, 2020). However, our goal for this study was to focus on 

dispersal behavior of the resident population. Therefore, individuals reported on in this study 

were fish captured in the first 13 months of the mark-recapture study (June 2009−July 2010), 

when immigration of translocated individuals into the focal sites was minimal. To account for 

the presence of few immigrant and hybrids in our dataset, we included a hybrid index covariate 

in all statistical models. This hybrid index varied from 0 (pure resident) to 1 (pure immigrant) 

and was calculated using genetic data, as described below (see Fitzpatrick et al. 2020). Only 5% 

of fish in our dataset were classified as pure immigrants, suggesting that the vast majority of our 

dataset represents the resident populations of Taylor and Caigual prior to the onset of gene flow 

(Fitzpatrick et al. 2016). 

Detailed capture-mark-recapture methods are described in Fitzpatrick et al. (2016). 

Briefly, we selected a portion of each stream that was the upstream-most extent of wild guppy 

populations uninterrupted by waterfall barriers. The sampled reach in Taylor was 240 m in 

length, and 80 m in length in Caigual. Each distinct pool or riffle within focal reaches was 

uniquely labelled and sampled monthly using a combination of hand nets and mesh traps. All 

guppies greater than 14 mm were captured, transferred to the lab, and placed in aerated tanks, 

separated by pool location and sex. Only mature individuals were included in the analyses, as 

determined by gonopodium morphology for males, and the presence of melanophores at the 

cloaca for females. These methods are commonly used in capture-mark-recapture studies of 

Trinidadian guppies, and have been shown to produce low lab mortality and high capture 

probabilities (Reznick et al. 1996). During lab processing, individuals were anesthetized with a 

dilute solution of MS-222, and new recruits were given a unique subcutaneous elastomer mark 

(Northwest Marine Technologies, Inc., Shaw Island, WA, USA). Recruits had three scales 

 10 

collected and dried for DNA extraction, and all individuals were weighed and photographed each 

month. All fish were returned to their exact capture location one to two days after processing. 

During their release, fish were acclimated to stream water and released into the lowest flow 

region of their capture location to minimize accidental passive downstream movement.   

Quantifying dispersal variation 

The exact pool locations of initial capture and subsequent recaptures were recorded every 

sampling event for all individuals. Although sizes and locations of pools and riffles change based 

on seasonal dynamics, they are always noted by the number of meters from the upstream extent 

of the reach. P. reticulata were considered philopatric (non-dispersing) if they were consistently 

captured less than 10 m from their initial capture location, the approximate length of the 

maximum pool size. Individuals were considered dispersers if at any point in the study they were 

captured 10 m or more from the pool where they were first captured, regardless of how many 

sampling occasions it took for that movement to occur.  

We use the threshold of 10 m because that is the length of the largest pool in our study, 

such that any movement beyond it reflects that an individual departed its initial pool and settled 

at a new location for a period of time. Given that most guppy daily activity takes place at the 

within-pool scale and movement between pools places an individual in a new, non-local 

environment, this threshold matches our definition of dispersal as a “a departure from a local 

environment, followed by movement and eventual settlement in a new location.” A potential 

source of error in our estimate of distance could stem from individuals moving and backtracking 

between sampling occasions, such that total movement would be underestimated. Thus, we are 

only able to estimate minimum distance moved.  

 11 

In addition to the categorical classification of philopatric or dispersing, we quantified the 

total distance travelled for all individuals as the cumulative distance the moved during our study 

(Figure 2.1). This estimate considers upstream and downstream movement equally. For example, 

a fish that moved upstream from 0 m to 10 m to 20 m would have the same dispersal distance (10 

+ 10 = 20 m) as a fish that moved upstream from 0 m to 10 m, then back downstream from 10 m 

to 0 m (10 – 0 + 10 = 20 m). Finally, we calculated the range of all dispersing fish, defined as the 

minimum non-cumulative distance spanning all of its locations across the study––in the example 

above, the first fish has a range of 20 m, but the second fish has a range of 10 m.  

Figure 2.1. How dispersal distances were calculated. Arrows represent boundaries between 
pools and riffles along the stream. Dispersal distances for all fish were measures based on the 
distance between the pools where they were captured. Distances between pools were measured 
as the minimum distance an individual would have to move to change pools. For example, the 
yellow fish was initially found in a pool that extended from 0-5 m, then captured in a pool that 
extended from 17-25 m, and so was recorded to have moved 12 m (17 minus 5). The green fish, 
having moved from a pool extending from 7-14 m to the same 17-25 m pool, was considered to 
have moved only 3 m (17 minus 14). Our metric of total distance, then, reflects a conservative 
minimum distance that each fish moved during the study. Fish that were always captured in the 
same pool, like the blue fish represented in the figure, were not considered dispersers.  

 12 

 
 
 
 
Analyses of dispersal patterns 

Total estimated dispersal distances showed an excess of zero values, as 47.5% of fish did 

not disperse at all. As such, we modelled dispersal distance using zero-inflated negative binomial 

distributions, which accommodates separate analyses of zeros and total counts. We found this to 

be the most biologically appropriate model to understand both the probability of moving (zero 

vs. non-zero) as well as total distance moved. For all models described below, we used AIC 

(Akaike Information Criterion) for model selection (Johnson & Omland 2004) to choose between 

models including different independent variables and distributional assumptions, in addition to 

using the R package DHARMa (Hartig 2021) to assess model fit. Covariates in all models 

included stream, sex, hybrid index, and longevity. Longevity was calculated as the total number 

of months a fish was captured in our study, which likely underestimates the true number of 

months that individuals live due to imperfect detection probability and because it does not 

account for the time when fish are too small to be captured (less than 14 mm).  

To test whether P. reticulata move more in the downstream direction and in the wet 

season, we first investigated whether movement differed in distance and direction between 

seasons and streams. We used generalized linear mixed models to test for the effect of season on 

the movement of fish for each capture occasion, with individual and month as random effects. 

Seasonal analyses were done separately for each stream. 

To test the hypothesis that smaller males disperse more often and farther, we tested for 

relationships between standard length and dispersal status and total dispersal distance. Females 

size was not tested because females have indeterminate growth, making size and longevity 

strongly colinear. All models included stream, hybrid index and longevity as covariates, and the 

interaction between hybrid index and stream. 

 13 

Effects of dispersal on fitness 

Lifetime reproductive success was determined for each individual by reconstructing 

pedigrees for Caigual and Taylor populations using individual genotypes from 12 polymorphic 

microsatellite markers (see Fitzpatrick et al. 2016 for genetic data collection methods and 

Fitzpatrick et al. 2020 for details on pedigree reconstruction). Final pedigrees consisted of 1,106 

individuals in Caigual (458 maternal links, 655 paternal links) and 1,725 individuals in Taylor 

(975 maternal links, 994 paternal links) spanning 4-6 overlapping generations. We used the 

pedigrees to estimate the number of offspring that reached a minimum size of 14 mm for each 

individual, as well as the total number of mates for each individual that produced at least one 

offspring that survived to 14 mm. 

We tested for a relationship between dispersal and lifetime reproductive success using 

zero-inflated negative binomial regression because 55% of fish were assigned zero offspring in 

the pedigree analysis. As with distance, we were interested in how our variables affected the 

probability of zero vs. non-zero values of offspring as well as total counts. Number of mates was 

also modeled with a zero-inflated negative binomial regression for similar biological reasons. 

Dispersal status and total distance were used as predictor variables in separate models, with 

stream, hybrid index and longevity as covariates. We included the interaction between hybrid 

index and stream in these models, given higher immigration into the Taylor during the study 

period, and overall higher fitness of hybrid individuals (Fitzpatrick et al. 2020). We analyzed 

males and females separately when modelling reproductive success, because the offspring from 

each fish were not independent of the other parent. In models with males, standard length was 

also included as a covariate, while female size is accounted for by incorporating longevity in 

models of reproduction. 

 14 

Finally, we explored whether guppies with a larger range have an increased number of 

mates, where a “mate” is defined as an individual that shared at least one surviving offspring 

with the focal individual. Range was used in this analysis because it directly reflects the extent to 

which fish explored the stream beyond their natal pools––total distance, on the other hand, 

includes backtracking and reflects overall movement (Figure 2.1). We again used a negative 

binomial zero-inflated model with stream, hybrid index, and longevity as covariates. All analyses 

were conducted in R 3.5.3 (R Development Core Team. 2019).  

Results 

Patterns of dispersal 

A total of 1,357 fish were caught and recaptured at least once in our 13 sampling 

occasions from June 2009 to July 2010. On average, fish in our study were recaptured 3.6 times 

during that time period. A total of 525 (38.7%, Table S1) individuals in our dataset were 

classified as dispersers (i.e., moved ≥10 m during the study). Our data consisted overwhelmingly 

of resident genotypes, with 954 out of 1,357 fish having a hybrid index lower than 0.5, and 726 

of those falling between zero and 0.1 (Figure S2.1). The distribution of dispersal distances was 

highly skewed, with 50% of all dispersing fish moving between 10-26 m, a mean dispersal 

distance of 41.5 m, and a maximum of 248 m (Figure 2.2, Table 2.1). The proportion of fish that 

dispersed was higher in Caigual (48.6%) than in Taylor (28.9%, Table S2.1).  

We found that the likelihood of fish becoming a disperser increased with longevity 

(N=1357, z=7.825, p<0.001) and was higher for guppies of resident descent (N=1357, z=-2.848, 

p=0.004).  Longer-lived fish also moved greater total distances (N=1357, z=9.667, p<0.001; 

Figure S2.2A), as did fish from Taylor (N=1357, z=11.709, p<0.001). We did not find evidence 

that fish in older age classes are more likely to move. That is, we found that most dispersal 

 15 

 
happens between a fish’s first and fifth captures, and there does not seem to be an age threshold 

after which fish begin to disperse (Figure S2.3, Table S2.2).We also uncovered an interaction 

between stream and hybrid index, where fish moved higher total distances if they were from 

Taylor and had low hybrid indexes (N=1357, z=-4.247, p<0.001). Thus, we kept longevity, 

stream, hybrid index, and the interaction between stream and hybrid index as covariates in all 

subsequent models.  

Figure 2.2. Total movement of fish considered dispersers (total distance moved ≥ 10 m) grouped 
by sex and stream. Red lines indicate means for each group. 

 16 

 
 
 
Individual and environmental correlates of dispersal 

Supporting our hypothesis about male-biased dispersal, we found that males were 12% 

more likely to disperse at all than females (N=1357, z=4.425, p<0.001; Table S2.1). Males also 

moved farther total distances (Figure 2.2; N=1357, z=2.451, p=0.014). Contrary to our 

hypothesis, we found positive correlations between male body size and likelihood of dispersing 

(N=604, z=4.142, p<0.001) as well as total distance moved (N=604, z=5.429, p<0.001; Figure 

S2.4). 

The probability of movement was higher in the wet season for the Taylor population 

(N=469, z=6.499, p<0.001; Figure S2.5A), but was consistent across seasons in Caigual (N=651, 

z=-0.41, p=0.685; Figure S2.5B). The Caigual population also showed no seasonality in the 

extent of upstream (N=651, z=-0.698, p=0.485) versus downstream (N=651, z=-0.278, p=0.781) 

dispersal, indicating fish did not have a directional bias throughout the year of the study. 

However, as expected, fish in the Taylor population moved more in both directions during the 

wet season, with a stronger effect of the wet season on increasing downstream dispersal 

distances (N=469, z=11.36, p<0.001), compared to upstream movement (N=469, z=5.036, 

p<0.001). 

Fitness consequences of dispersal 

We observed different effects of dispersal status and distance on reproductive success 

between the sexes. In all models, longevity (Figure S2.2B), male standard length, higher 

immigrant ancestry, and being from Taylor had positive effects on reproductive success, so these 

covariates were always included. Male dispersers had a higher chance of having at least one 

surviving offspring than philopatric males (N=604, z=-2.877, p=0.004; Table 2.1), as well as a 

non-significant trend towards more total offspring (N=604, z=1.693, p=0.090; Table 2.1). The 

 17 

 
chance of males having a non-zero number of offspring also increased with total distance moved 

(N=604, z=-2.243, p=0.025, Figure 2.3A), though there was no effect of total distance on the 

number of offspring for males (N=604, z=1.002, p=0.316). Unlike males, a female’s chance of 

having at least one surviving offspring was not influenced by dispersal status (N=753, z=-1.527, 

p=0.127), or dispersal distance (N=753, z=-1.204, p=0.229). Being classified as a disperser did 

have marginally-significant positive effect on the number of offspring in females (N=753, 

z=1.864, p=0.062), as did total dispersal distance (N=753, z=1.845, p=0.065, Figure 2.3A). 

Given the tight biological links between longevity and reproductive success, we re-ran our 

analyses on subsets of individuals of the same age, and found qualitatively similar results across 

longevities (Figures S2.6 and S2.7). We also conducted fitness analyses using absolute distance 

(distance between first and last capture locations) as a predictor, which gave us qualitatively 

similar results to our metric of total distance, and these analyses are described in Supplemental 

Materials Appendix 2.I. 

We also analyzed the effects of range on number of mates (Figure 2.3B). In these models, 

longevity, hybrid index, and the interaction between lower hybrid index and being from Taylor 

all significantly increased number of mates. Range size had a positive effect on number of mates 

for males (N=604, z=2.754, p=0.006, Table 2.2), and a marginally-significant effect on number 

of mates for females (N=753, z=1.809, p=0.070, Table 2.2). For both sexes, we found no effect 

of range size on the chance of having at least one mate (N=753, z=1.224, p=0.221 for females, 

N=604, z=-0.970, p=0.332 for males). Tables 1 and 2 describe summary statistics for dispersal 

distance, range, reproductive success, and number of mates. Detailed model outputs are provided 

 18 

in Tables S2.3-6.

Figure 2.3. Response of fitness metrics to dispersal status, distance, and range. (A) Males that 
disperse are more likely to have at least one surviving offspring, and show a trend towards 
having more offspring; (B) Female dispersers do not show trends toward increased reproductive 
success compared to females that do not disperse. (C) Males have higher chance of having 
offspring with longer distance moved and females show a trend towards more offspring with 
higher distance moved; (D) Males with larger ranges have an increased number of mates, and 
that effect was not statistically significant for females. Lines show negative binomial regression 
and shading as 95% confidence interval. 

 19 

 
 
Discussion 

This study provides novel insights into the ecological drivers and evolutionary 

consequences of dispersal in two populations of the Trinidadian guppy. Our results reveal 

variation in dispersal propensity and distances within and between our study populations, with 

size, genetic ancestry, and seasonality influencing dispersal. We found strong evidence for sex-

specific reproductive benefits of dispersal consistent with observed patterns of male-biased 

dispersal behavior, thus exemplifying how dispersal evolution can occur in natural populations. 

Characterization of dispersal behavior  

Our study directly measured fine-scale dispersal distances for all adult individuals in two 

populations of P. reticulata for over one year, adding to what is known about dispersal behavior 

in natural settings and the specific characteristics of movement in this system. Overall, we 

observed some variation in dispersal between seasons and streams, but mostly consistent 

patterns: large male guppies were most likely to disperse, and the majority of fish moved short 

distances or not at all. Not surprisingly, the overall proportion of dispersers we observed over the 

course of a year (39% across both sexes and streams, Figure 2,2) was significantly higher than 

what was observed for a single recapture event (Reznick et al. 1996; Croft et al. 2003), but 

similar to estimates from a 5 year period (De Bona et al. 2019), suggesting that variation in 

dispersal can be adequately captured over the course of a year.  

We also detected seasonal variation in dispersal distance in Taylor, where guppies were 

more likely to disperse downstream in the wet season, but not in Caigual (Figure S2.3). This was 

expected given that Taylor has a higher probability of flooding in the wet season, as it is a 

steeper and more channelized stream (Fitzpatrick et al. 2014). Floods, which primarily occur in 

the wet season, may therefore increase the prevalence of long distance dispersal in the wet 

 20 

season in Taylor. We cannot determine whether increased wet season dispersal by P. reticulata is 

primarily active, due to floods and high water levels reducing dispersal barriers, or passive, due 

to fish being carried by stronger stream flows. We also note that while upstream dispersal can be 

interpreted as active, downstream movement could be either active or passive. There were no 

other strong biases in up- vs. downstream dispersal, consistent with previous studies of P. 

reticulata dispersal (Croft et al. 2003; Crispo et al. 2006) and findings of positive rheotaxis in 

this system (Blondel et al. 2020). As a whole, these data suggest that guppy dispersal is often 

active, raising questions about the environmental and individual triggers of this behavior.  

In this study we cannot draw conclusions about the causes of dispersal, but are instead 

interested in characterizing dispersal patterns within these streams and understanding their 

fitness consequences. Similar to previous studies of P. reticulata (Croft et al. 2003; De Bona et 

al. 2019), we found that dispersal was male-biased: 42% of males and 36% of females in our 

study were classified as dispersers. This corroborates studies on a wide range of other species, 

supporting the hypothesis that polygynous mating systems tend to have male-biased dispersal 

(Trochet et al. 2016). The “resource competition hypothesis” for the evolution of sex-biased 

dispersal (Greenwood 1980) proposes that competition for local resources and mates leads to 

increased dispersal of the sex that most strongly competes for mates, while extensions of this 

hypothesis suggest that systems in which female choice is prevalent should evolve male-biased 

dispersal (Dobson 1982). Though abundant evidence supports these hypotheses in mammals and 

birds, data on other taxa remain scarce (Trochet et al. 2016), making P. reticulata an important 

example of how these ideas may apply to other species. Finally, contrary to our expectations, but 

in line with previous results (Croft et al. 2003), we found that larger males dispersed more often 

and farther––suggesting that larger fish are more likely to leave local pools or survive dispersal. 

 21 

This may be because larger fish are less susceptible to predation by gape-limited predators 

(Gosline & Rodd 2008), better at acquiring resources (Potter et al. 2018), or stronger swimmers 

(more able to resist passive downstream movement and/or complete active upstream movement; 

Kieffer, 2000; Radinger & Wolter, 2014). 

Fitness correlates of dispersal 

Quantifying the fitness of dispersing individuals is a fundamental step towards 

understanding how dispersal shapes the ecology and evolution of natural populations. Successful 

gene flow between populations, colonization of new environments, and maintenance of range 

boundaries all crucially depend on the ability of dispersing individuals to reproduce. We found 

strong effects of dispersal on reproductive success in our study: 60.25% of non-dispersers had 

zero offspring, compared to only 37.79% of dispersers. We also saw a trend towards increased 

number of offspring for dispersers, where dispersing males had 35% more offspring than non-

dispersing males when holding all other variables constant. Given that most fish do not leave 

behind any successful offspring, this metric reflects the most important reproductive 

achievement in these populations, namely, having at least one offspring survive to the minimum 

markable size of 14 mm (Figure S2.1A). Thus, our results demonstrate that dispersal is a major 

factor in male reproductive success. 

Several aspects of guppy natural history and mating biology may contribute to the 

increased reproductive output of dispersing males. Theoretical models for the evolution of sex-

biased dispersal suggest that the strength selection for male dispersal in polygynous systems 

depends on female choice, kin recognition, and the strength of inbreeding depression in the 

system (Lehmann & Perrin 2003). In P. reticulata, female choice plays a major role in 

determining male reproduction, such that increasing encounters with females is in itself not 

 22 

sufficient to increase number of successful matings (Pitcher et al. 2008; Zajitschek & Brooks 

2008; Daniel & Rodd 2016). Rather, females prefer unrelated males and those with novel 

phenotypes, and are able to differentiate between kin and non-kin individuals (Hughes et al. 

1999, 2013a; Hampton et al. 2009; Daniel & Rodd 2016, 2020). Female preference may be 

important for inbreeding avoidance in this system, given ample evidence that guppy populations 

in small headwater streams such as Caigual and Taylor can suffer from significant inbreeding 

depression (Van Oosterhout et al. 2003; Crispo et al. 2006; Fitzpatrick et al. 2014, 2016). Male 

color, a key phenotypic marker used in female choice, is highly heritable (Endler 1980; Endler & 

Houde 1995; Kemp et al. 2009), such that dispersal between pools may offer males a chance to 

exploit unrelated females’ preference for novelty. Overall, the correlation between fitness and 

dispersal we observed for males reflects a selective pressure for increased dispersal resulting 

from the combination of female choice and inbreeding avoidance in this system, providing an 

empirical example for how these factors interact to shape dispersal evolution. 

Unlike males, females’ chances of having at least one surviving offspring was not 

affected by any dispersal metric. If female choice is adaptive (e.g., inbreeding avoidance or 

“good genes”; Anderson, 1982; Kodric-Brown & Brown, 1984; Perrin & Mazalov, 1999; Pusey 

& Wolf, 1996; Zahavi, 1977), we would expect females that dispersed farther to benefit from a 

larger pool of available mates, potentially leading to the observed trend in higher number of 

offspring for dispersing females. However, female reproduction in polygynous systems is more 

directly limited by resources than mates (Greenwood 1980; Perrin & Mazalov 2000; Magurran 

2005). For example, a study on density-dependent dispersal in P. reticulata found that females 

were likely to disperse out of high density pools (De Bona et al. 2019). Again, these findings are 

in line with hypotheses for the evolution of sex-biased dispersal in polygynous species: increased 

 23 

philopatry in resource-limited females and increased dispersal in mate-limited males 

(Greenwood 1980; Perrin & Mazalov 2000; Lehmann & Perrin 2003). 

Given the observed reproductive correlates of dispersal, males in the studied populations 

are under strong selection to disperse. Yet, dispersal was generally limited and occurred over 

short distances, even for males (Figure 2.2). One potential explanation for these contrasting 

results is the relationship between local adaptation, habitat heterogeneity, and dispersal in this 

system. P. reticulata populations adapted to low predation levels are likely to incur severe 

survival costs if they disperse into stream sections with higher predation, as will any of their 

offspring that retain low-predation phenotypes such as brighter coloration and weaker 

antipredator behaviors (Haskins et al. 1961; Endler 1980, 1987; Magurran & Seghers 1991; 

Reznick et al. 1996, 2001). As a consequence, this local adaptation to a gradient in predation 

regime should limit dispersal propensities through selective deaths of maladapted dispersers 

and/or adaptive dispersal decisions by individuals (McPeek & Holt 1992; Bolnick & Otto 2013; 

Berdahl et al. 2015). In other words, if dispersal is beneficial at the within-reach, between-pool 

scale––where we observed reproductive benefits––but costly between steep environmental 

gradients, most individuals should disperse only small distances. 

In natural systems, it is notoriously difficult to directly determine causality between a 

behavior, such as dispersal, and reproductive success. In an observational study such as this, it is 

always possible that unaccounted for variables may lead to the observed correlation between 

reproductive success and dispersal. We dealt with this issue by including two important drivers 

of fitness and dispersal, longevity and size, as covariates in all our models of reproductive 

success. Though longevity and size are significant predictors of both dispersal distance and 

lifetime reproductive success in our system (Figures S2.2), accounting for them in our models 

 24 

 
allows us to statistically separate their effects on fitness from those of dispersal status and 

distance as best as possible. Within the constraints of an observational study of natural 

populations, our approach strongly suggests a causative link between dispersal and reproduction, 

yet experiments that directly manipulate dispersal and longevity would be needed to fully 

disentangle these factors. 

Conclusions 

We provide empirical evidence that fine-scale dispersal correlates to reproductive benefits in 

Poecilia reticulata. Our study highlights the value of multi-generational capture-mark-recapture 

data for understanding individual consequences of dispersal, especially when paired with fitness 

estimates from molecular data. Our characterization of dispersal and reproductive success sheds 

light on how the benefits of dispersal behavior vary between individuals and environments. 

Given drastic reductions in population connectivity caused by anthropogenic changes to the 

environment across the globe (Haddad et al. 2015b), understanding patterns of dispersal and its 

consequences for fitness in natural populations will be vital to understanding population 

persistence in the future (Kokko & López-Sepulcre 2006; Crispo et al. 2011) 

Acknowledgements 

We thank the many field assistants who helped catch, process, and release guppies in Trinidad 

and whose names are listed in Supplementary Appendix 2.II. The Fitzpatrick Lab, KBS Writing 

Group, and three anonymous reviewers provided valuable feedback on previous versions of this 

manuscript. We thank D.N. Reznick for intellectual contributions to this work. This project was 

funded by National Science Foundation grants to W.C.F. and L.M.A. (DEB-0846175) and 

S.W.F. (DEB-2016569). This is KBS contribution 2300. 

 25 

 
CHAPTER THREE: 

Rollin’ on the river: Local census size, sex ratio, and relatedness drive dispersal behavior in the 
Trinidadian guppy (Poecilia reticulata) 

Abstract 

Dispersal is a major determinant of individual fitness, population structure, and range 

dynamics in natural populations. In organisms that disperse actively, the decision to move from 

one place to another is the fundamental building block of all dispersal dynamics that follow. 

Understanding what drives individual movement behavior, then, is fundamental to understanding 

dispersal. For this study, we asked how seasonal, local, and individual characteristics affects 

dispersal probabilities in two populations of Trinidadian guppies (Poecilia reticulata). We were 

especially interested in how the local composition of pools affected dispersal, focusing on the 

effects of local census size, sex ratio, and average relatedness. We found that guppies were more 

likely to disperse out of pools with fewer individuals, and pools with lower average relatedness. 

In one of the study streams, fish were also more likely to disperse out of male-biased pools, 

especially females. Overall, we find that intrinsic and extrinsic factors interact to drive dispersal 

behavior, explaining some of the dispersal variation we observe between individuals in the wild. 

Introduction 

Dispersal is an important ecological and evolutionary force that structures populations 

and influences individual fitness (Bowler & Benton 2005; Clobert et al. 2009, 2012). While 

some organisms disperse randomly, many show non-random dispersal behaviors driven by both 

intrinsic individual characteristics and extrinsic environmental conditions (Clobert et al. 2009). 

These small-scale individual movement decisions, multiplied across individuals, space, and time, 

scale up to define population- and species-level patterns of dispersal and gene flow across a 

landscape. From the perspective of individual organisms, multiple factors can potentially drive 

 26 

dispersal behavior. For example, poor habitat conditions can lead individuals to search for higher 

quality patches, high competition can push out less competitive individuals, and searching for 

available and/or unrelated mates can motivate reproductive-age individuals to search for new 

locations and opportunities. These different drivers will have consequences for where dispersers 

establish, how their fitness is affected by dispersal, and how populations are structured across a 

landscape (Morales et al. 2010; Kubisch et al. 2014; Spiegel et al. 2017). 

It is often difficult to determine the relative importance of different drivers of dispersal in 

natural populations. Local context (e.g. conspecific abundance, sex ratio, predator presence, 

etc.), individual characteristics (e.g. sex, age, body condition), and seasonality lead to within-

population variation in dispersal decisions and distances, with important consequences to 

ecological and evolutionary processes (Peniston et al. 2023). For instance, increased temporal 

variation in dispersal within a metapopulation framework can decrease population growth rates 

(Watson et al. 2012), while in an evolutionary context that same increased variation is expected 

to facilitate local adaptation (Peniston et al. 2019)  

Studies that explore drivers of dispersal in natural populations often look at how 

individuals with differences in traits vary in their dispersal responses to local conditions. For 

example, small milkweed beetle (Tetraopes tetrophthalmus) males are more likely to leave male-

biased patches than larger males, suggesting that avoidance of male-male competition is a driver 

of dispersal in this system (Lawrence 1987); while female red deer (Cervus elaphus) disperse 

more frequently when populations sex ratios become highly female-biased and male dispersal 

decreases (Pérez-González & Carranza 2009), suggesting an important role of inbreeding 

avoidance in this species. Competition and inbreeding avoidance may affect dispersal decisions 

differently for males compared to females, and for individuals of different sizes, depending on 

 27 

the mating system and resource availability of different populations. As such, variation in 

dispersal may be explained by temporal changes in intensity of competition or mating behavior. 

By analyzing the effects of individual, local, and seasonal variation on dispersal behaviors, we 

can better understand which drivers are most important in different scenarios. 

Here, we make use of a multi-year capture-mark-recapture study of Trinidadian guppies 

(Poecilia reticulata) to ask: What are the extrinsic and intrinsic factors influencing dispersal 

decisions, and how do they change between seasons? We do this within two established 

headwater guppy populations, looking at factors affecting movement between patches (“pools”) 

within the same population. As these animals are highly philopatric and rarely disperse long 

distances (Croft et al. 2003; De Bona et al. 2019; Borges et al. 2022), this within-population 

movement reflects the scale at which guppies are interacting with one another: competing, 

mating, and reproducing. We define dispersal as any movement that can lead to gene flow 

(Ronce 2007), and given our lack of information about where and when matings take place in 

our dataset, we use the terms “dispersal” and “movement” interchangeably. In this study, we do 

not look at long-distance, between-population dispersal, range expansion, or habitat colonization 

dynamics. Instead, we asked how the local composition of pools, alongside individual size 

and sex, affected monthly dispersal decisions for over a thousand unique individuals across 

two streams (Figure 3.1). 

Our dataset allows us to assess the effects of three variables related to local guppy 

composition: local fish abundance (“census size”), sex ratio, and relatedness. In a previous study 

of the same two populations, we described dispersal variation between sexes, sizes, and across 

seasons, finding that dispersal has fitness benefits in these streams, especially for males (Borges 

et al. 2022). Here, we shift our focus from the consequences of dispersal to its causes. Though, to 

 28 

our knowledge, only two other studies have looked at dispersal in the Trinidadian guppy (Croft 

et al. 2003; De Bona et al. 2019), decades of research in this important model system allow us to 

make predictions about how dispersal decisions will be affected by local and individual factors. 

Understanding the relative importance of different drivers of dispersal in guppies not only offers 

an example of how dispersal occurs in natural systems, but also fills an important gap in our 

understanding of this model species, connecting existing knowledge on competition, mate 

choice, and seasonality to their effects on the ecological and evolutionary force that is dispersal. 

Figure 3.1. Conceptual figure of dispersal drivers analysed in this study. Both extrinsic and 
intrinsic factors affect individuals’ dispersal decisions, in isolation and in interaction with one 
another. In this example, the colorful male does disperse (1), while the female does not (0). 

For one, density-dependence has a major influence on guppy life-history: higher guppy 

density leads to decrease offspring size, delayed maturation, and higher mortality in low-

predation headwater environments (Reznick et al. 2012; Bassar et al. 2013; Travis et al. 2023). 

In the context of dispersal, DeBona et. al (2019) showed that in the process of colonizing new 

sections of a stream, guppies respond differently to density based on life stage, sex, and the 

 29 

extrinsic factorsseasonalitylocal pool compositioncensus sizesex ratiorelatednessintrinsic factorssexweightdispersal decision= 1= 0 
progression of colonization. Importantly, guppy densities vary predictably with season, 

increasing during the dry season (January–May) and decreasing during the wet season (June–

December) when resources are scarce, competition is stronger, and mortality is higher (Reznick 

1989; Fitzpatrick et al. 2014, 2016). We expect, then, that increased conspecific density will 

increase individual dispersal probabilities, as guppies avoid competitors, and that this effect will 

be stronger in the wet season. 

In addition to density, sex ratio and relatedness may also be important drivers of dispersal 

in P. reticulata. In the guppy mating system, males actively search and compete for mates, while 

females tend to choose their mates based on male phenotypes such as color (Houde 1987; Endler 

& Houde 1995; Zajitschek & Brooks 2008). This female choice is thought to be a mechanism for 

inbreeding avoidance (Johnson et al. 2010), as females have the ability to recognize kin, male 

color patterns are heritable, and females show preference for novelty and rarity (Zajitschek & 

Brooks 2008; Hughes et al. 2013a; Daniel & Rodd 2016). Given this mating system, we expect 

that males disperse most often out of male-biased pools in the search for mates, potentially 

explaining males’ observed higher rates of dispersal (Reznick et al. 1996; Croft et al. 2003; 

Borges et al. 2022). In contrast, we expect that females respond more strongly to high 

relatedness within pools, dispersing in search of less-related mates. Ultimately, differences in 

how the sexes respond to sex ratio and relatedness may also explain why males accrue more 

fitness benefits than females from movement within streams (Borges et al. 2022). 

Understanding local drivers of dispersal behavior in natural environments is inherently 

difficult because it requires fine scale, individual, multi-site and multi-year datasets. The 

Trinidadian guppy system offers just such an opportunity, supported by a thorough 

understanding of other ecological and evolutionary processes in this organism. Disentangling the 

 30 

 
relative importance of local, individual, and temporal factors in driving dispersal in this system 

will shed light not only on how these fish move across their environments, but on how factors 

such as inbreeding avoidance and competition influence dispersal in natural settings.  

Methods 

Field capture-mark-recapture 

This study was conducted on two populations of P. reticulata in neighboring headwater 

streams, Taylor and Caigual, in the Guanapo drainage in northern Trinidad. The data were 

collected in a monthly capture-mark-recapture study between July 2009−October 2010, with the 

full mark-recapture methods described in Fitzpatrick et al. (2016). To summarize, we selected 

reaches in each stream that were the upstream-most extent of wild guppy populations 

uninterrupted by waterfall barriers (240 m in length for Taylor, and 80 m in length in Caigual). 

Every pool within these reaches was uniquely labeled based on its distance to the upstream limit 

of the reach, such that labels remained consistent across season as pools changed in size. For 

every month of the study, each pool was sampled using hand nets and mesh traps. Every guppy 

larger than 14 mm was captured, temporarily moved to a field laboratory, marked, and measured. 

Here we only include sexually mature individuals in our analyses, with maturity being 

determined by gonopodium morphology for males, and the presence of melanophores at the 

cloaca for females. These mark-recapture methods are commonly used in studies of wild 

Trinidadian guppies, showing low lab mortality and high capture probabilities (Reznick et al. 

1996). For each sampled fish, we also collected tissue for DNA extraction, and all individuals 

were weighed and photographed each month. All fish were returned to their exact capture 

location one to two days after processing.  

 31 

 
 
Quantifying dispersal probability 

The pool location of each capture was recorded for every individual at every sampling 

occasion. For this study, every movement between two distinct pools was considered as a 

dispersal event, regardless of the distance between pools. Given that sampling occasions were 

separated by one month, it is possible that individual fish moved away then backtracked between 

sampling dates, and we are not able to observe such movement in our data, such that we are 

quantifying the minimum dispersal between pools for these individuals. 

For each timepoint in which an individual fish was captured, we included a variable in 

the dataset recording whether it was captured at a different pool in the following month (i.e. 

whether or not it dispersed from that timepoint to the next), regardless of the distance between 

pools. This is in contrast with our analyses of the same dataset in Borges et al. (2022), where we 

were interested in the distances traveled between pools and only considered movement above 10 

m to be indicative of dispersal. Here, we were not interested in dispersal distances and therefore 

considered any movement between distinct pools to be reflective of a dispersal decision made by 

an individual in response to the local pool composition. We then used this variable as a binomial 

response variable in our “probability of dispersal” models to determine which individual and 

environmental characteristics affected dispersal probability. We found that in Caigual, 31% of 

individuals never moved between pools, while that number was 70% for Taylor (Figure 3.2). The 

maximum number of dispersal events within a fish’s lifetime within our study was 7, achieved 

by 12 individuals in Caigual. 

Modeling dispersal response to pool characteristics 

We were interested in how guppy dispersal probabilities respond to the local, biotic, 

intraspecific composition of the pools, specifically: number of guppy individuals in that pool 

 32 

(“census size”), sex ratio, and average relatedness from the focal individual. Sex ratio was 

calculated as the number of males divided by the number of females in a given pool at a given 

capture occasion. Both census size and sex ratio values were the same for all fish in a given pool 

in a given month. In contrast, average relatedness was calculated for each individual fish at each 

pool in each timepoint as the average relatedness value for that focal fish when compared to all 

other fish present in that pool at that time. Relatedness values were extracted from a relatedness 

matrix based on 12 microsatellite loci genotyped for each individual fish in the study. 

Relatedness between two individuals was calculated as the Jaccard similarity between alleles at 

all microsatellite loci, and a matrix was populated with all pairwise relatedness values between 

individuals. Average within-pool relatedness was calculated as the mean pairwise relatedness of 

the focal individual to all other individuals present in a pool at a specific time. Methods for 

obtaining the microsatellites is described in detail in Fitzpatrick et al. 2016. Relatedness and sex 

ratio were only tested as drivers of dispersal for pools that contained more than one individual, 

since in those pools the sex ratio variable is equal to the sex of the that individual, and its 

relatedness to itself is equal to one. 

To test for individual characteristics driving dispersal probabilities, we included fish sex 

and weight as dependent variables in all models. Every model also included fish ID as a random 

effect to account for individual behavioral consistency in dispersal from one timepoint to the 

next. We also tested how dispersal probabilities change between the wet and dry seasons by 

including season as a dependent variable.  

All our models include dispersal probability as our response variable, modeled with a 

binomial distribution. We modeled each stream separately, given extreme differences in 

population sizes between the two locations, which would not only create an imbalance in terms 

 33 

 
 
of sample size within a joint model, but also may lead to different biological interpretations of 

dispersal behavior in response to the local factors we investigated. For each potential local driver 

of dispersal (e.g. census size), we also build separate models for males and females to assess the 

effects of each driver in isolation, as well as its interactions with season and fish size. We 

decided to separate the sexes due to the correlations between fish size and sex (females guppies 

are larger and grow indeterminately, while males stop growing after they reach maturity; 

Constanz 1989) and the different biological interpretations of some of our parameters of interest. 

For example, females dispersing out of male-biased pools suggest that they may be avoiding 

male harassment, while males dispersing out of male-biased pools might be avoiding 

competition with other males. We attempted to build a full model containing all potential drivers 

of dispersal, yet that model did not converge for either stream. Therefore, we present results for 

models that tested each local driver of dispersal separately (six total models, three for each 

stream). For example, the census size model included as variables: census size, fish sex, weight, 

and season, but not sex ratio or relatedness (Table 3.1). All statistical analyses were conducted in 

R 1.3.1093 (R Development Core Team. 2019). We used AIC (Akaike Information Criterion) for 

model selection (Johnson & Omland 2004) and the R package DHARMa (Hartig 2021) to assess 

the fit of candidate models. 

Results 

Individual and temporal drivers of dispersal 

In agreement with a different analyses of this same dataset (Borges et al. 2022), we found 

that male guppies are more likely to disperse, and that fish weight is also positively correlated 

with dispersal probability. Similarly, we found that fish were more likely to disperse during the 

 34 

 
wet season, when water volumes in these streams are higher and resource abundance is lower, 

potentially leading to increased incidental and intentional dispersal, respectively. 

We found that a significant portion of the variation in dispersal probabilities in both 

Caigual and Taylor was due to the two variables we modeled as random effects: month and 

individual. In Taylor, random effects accounted for between 68-89% of the proportion of the 

variance explained by our models, while in Caigual that range was 49-84% of the R2 values in 

our models (Table 3.1). This suggests that variation between individuals, and consistency within 

individuals, regardless of sex and weight, are important components of dispersal behavior 

variability. Similarly, variation between months, even when accounting for seasons, was a driver 

of dispersal variability.   

Local drivers of dispersal 

We were primarily interested in the effects of local pool composition on the probabilities 

of dispersal. The measured variables were census size, sex ratio, and relatedness, which changed 

between each pool and time point. In Caigual, pool census sizes varied more widely and were up 

to an order of magnitude larger than in Taylor. Sex ratio also differed between the streams, with 

pools in Caigual being more female-biased. Average relatedness within pools was higher in 

Caigual than Taylor (Figure 3.2). 

Census size. We found negative effects of census size on dispersal probabilities in both 

Caigual (N=2367, z=-4.765, p<0.001) and Taylor (N=460, z=-2.725, p=0.006; Figure 3.3C). 

Guppies in a pool by themselves in Caigual had a 91% probability of dispersal in the next time 

step, compared to 47% for guppies in a pool of average Caigual census size, and 4% at the 

maximum census size. The full census model explained 40% of the variance in dispersal 

probabilities in Caigual (marginal R2 = 0.404, conditional R2=0.170), but only 5.4% of the 

 35 

 
 
variation in Taylor (conditional R2 = 0.210). In Taylor, guppies alone in a pool have a 13% 

probability of dispersal in the next time step, compared to 3% for guppies in a pool of average 

Taylor census size, and 0.3% at the maximum. In both streams, the negative effect of census size 

was also significant in the separate male (Taylor: N=244, z=-2.196, p=0.028; Caigual: N=669, 

z=-2.182, p=0.029) and female models (Taylor N=216, z=-1.942, p=0.052; Caigual: N=1698, 

z=-4.205, p<0.001). 

Sex ratio. We also found a significant positive effect of sex ratio on dispersal probability  

in Caigual (N=2336, z=7.478, p<0.001; Figure 3.4A), with the full sex ratio model explaining 

about 42% of variance in dispersal (marginal R2 = 0. 419, conditional R2=0.110). Guppies in 

female-only pools disperse 41% of the time, and that probability grows to 88% in even sex-ratio 

pools, and 99% in male-only pools. That effect was significant for the full and separate sex 

models in Caigual, suggesting both males (N= 659, z=4.413, p<0.001) and females (N=1677, 

z=6.364, p<0.001) are more likely to move out of male-biased pools. In Taylor, sex ratio did not 

significantly affect dispersal probabilities (N=335, z=0.380, p=0.704; Figure 3.4B).  

Relatedness. In Caigual, we found a significant negative effect (N= 2313, z=-3.736, 

p<0.001) of relatedness on movement probability (Figure 3.5A). This model had the lowest 

marginal R2 of the three Caigual models, with fixed effects explaining only 7.6% of the variance 

(though the conditional R2 was still high at 48.7%). In pools with zero average relatedness, 

dispersal probabilities were 81%, increasing to 68% at an average relatedness of 0.17, and 

decreasing to 9% when relatedness equals 1. We see similar results in Taylor, with a significant 

negative effect of dispersal on relatedness (N= 288, z=-2.993, p=0.003; Figure 3.5D). In the 

separate sex models, there is no significant effect of relatedness itself on dispersal for males 

(Caigual: N=603, z=1.679, p=0.093, Figure 3.5B; Taylor: N=138, z=1.265, p=0.206, Figure 

 36 

3.5E), and in the female-only models the negative effect remains for Caigual (N=1610, z=-3.418, 

p<0.001, Figure 3.5C) but not for Taylor (N=150, z=-1.688, p=0.091, Figure 3.5F).  

Interactions between local, individual, and temporal drivers of dispersal 

We were also interested in how individual characteristics mediate guppies’ dispersal 

behavior in response to local conditions. We tested for this by looking for significant interactions 

between local, temporal, and individual predictors of dispersal probability, and again found 

varying results depending on local drivers and stream.  

Caigual. We found a significant positive interaction between census size and season in 

explaining dispersal probabilities (N=2367, z=2.571, p=0.010), suggesting that fish are 

responding more strongly to census during the dry season, when guppy populations in general 

are more abundant in response to higher resource availability (Figure 3.3A). This interaction was 

significant in the full model, almost significant in the female model (N=1698, z=1.803, p=0.071), 

and not significant for males. We also found a significant interaction between census size and 

fish weight (N=2367, z=3.507, p<0.001). This interaction is significant in the full model, and for 

males (N=669, z=2.252, p=0.024) and females (N=1698, z=5.120, p<0.001) separately. As a 

positive interaction, it suggest that larger guppies’ dispersal probabilities are less responsive to 

local pool census (Figure 3.3B). 

When looking at sex ratio, we found a significant interaction between pool sex ratio and 

individual sex in Caigual (N=2336, z=-2.109, p=0.035). As a negative interaction, it suggests 

that males’ dispersal probabilities are less responsive to pool sex ratio than females (Figure 

3.4A). Finally, we found a significant interaction between relatedness and season for Caigual 

males (N=603, z=-2.194, p=0.028, Figure 3.5B), but not in the full relatedness model or the 

female-only model. 

 37 

 
 
 
Taylor. We found no significant interactions between census size or sex ratio and season 

or individual characteristics in Taylor. For relatedness, we again saw a significant interaction 

with season in the male-only model (N=138, z=-2.328, p=0.019, Figure 3.5E). 

 38 

 
 
 
A 

C 

E 

G 

B 

D 

F 

H 

Figure 3.2. Distribution of individual movement and local pool composition. Red lines show 
average value of each distribution. Left (black) plots represent Caigual data, right (gray) plots 
represent Taylor. A and B) How many times individuals moved, reflecting that most fish move 
little or not at all. C and D) Caigual census sizes (and overall population size) are much larger 
and more widely distributed than in Taylor census sizes. E and F) Sex ratio in Caigual is more 
female biased (lower than 0.5, which represents a 1:1 ratio), while in Taylor it is more even. G 
and F) Average relatedness of a focal individual to the rest of its pool (excluding pools with a 
single individual) is more widely distributed in Taylor, with a higher average in Caigual. 

 39 

 
 
Figure 3.3. Probability of dispersal out of a pool in response to local census size. A) In Caigual, 
the response to census size differs between seasons, and is stronger in the dry season. B) 
Representing the interaction between census size and fish weight, different color lines represent 
small, average, and large fish. The dispersal response to census size is more extreme for smaller 
fish. C) In Taylor, we see that increased census size decreases dispersal probabilities, and does so 
with no interactions with season, weight, or sex. 

Figure 3.4. Probability of moving out of a pool in response to local sex ratio. A) In Caigual, all 
fish disperse more out of male-biased pools, and the response from females to sex ratio is steeper 
than males’. B) Local sex ratio did not affect dispersal probabilities in Taylor. 

 40 

0%25%50%75%100%050100150200Census sizeProbability of movementSeasonDWCaigualA0%25%50%75%100%050100150200Census sizeProbability of movementWeight0.090.250.41CaigualB0%25%50%75%100%481216Census sizeProbability of movementTaylorC0%25%50%75%100%0.000.250.500.751.00Sex ratioProbability of movementSexFMCaigualA0%25%50%75%100%0.000.250.500.751.00Sex ratioProbability of movementTaylorB 
 
Figure 3.5. Probability of moving out or in of a pool in response to local average relatedness 
relative to the focus individual. A) In Caigual, fish are most likely to move out of pools with 
lower average relatedness. B) For males, there was an interaction between relatedness and 
season, with the negative pattern flipped during the dry season, making makes more likely to 
disperse out of pools with the most relatives. C) When modeling Caigual females alone, the 
negative effect was still significant, though it did not interact with season. D) Similarly to 
Caigual, guppies in Taylor were more likely to disperse out of lower average relatedness pools. 
E) We also observed an interaction between seasonality and relatedness in Taylor males, but 
unlike in Caigual, the relationship was not flipped in the dry season, only weaker. F) In Taylor 
females, we observe a similar trend of decreased dispersal with increased relatedness, though 
that effect was not statistically significant. 

Discussion 

We found a consistent effect of local census size and relatedness on dispersal 

probabilities in Trinidadian guppies, where individuals were more likely to move out of emptier 

pools and pools with lower average relatedness in both of our study streams. In Caigual, we 

found that fish were most likely to move out of male-biased pools. We also find evidence that 

dispersal responses to census size and sex ratio in Caigual vary based on season, focal fish 

weight, and focal fish sex. Overall, these results suggest that guppies are making dispersal 

 41 

0%25%50%75%100%0.00.10.20.30.4Probability of movementCaigualA0%25%50%75%100%0.00.10.20.30.4SeasonDWCaigual malesB0%25%50%75%100%0.00.10.20.30.4Caigual femalesC0%25%50%75%100%0.00.20.40.60.8Probability of movementTaylorD0%25%50%75%100%0.00.20.40.60.8Within pool relatednessSeasonDWTaylor malesE0%25%50%75%100%0.00.20.40.60.8Taylor femalesF 
decisions based on local stream conditions, and that these decisions can differ systematically 

between the sexes, seasons, fish size, and streams. 

Our census size results show that dispersal behavior is another life-history trait affected 

by local population sizes. We were not able to measure density directly, as we lacked depth 

information for pools at our two study sites, but we assume that our metric of guppy census size 

and density are correlated. The negative effects of census size on dispersal probability across 

both streams suggest that guppies disperse most out of low-density pools, a surprising result 

given the known negative effects of competition on fitness at high densities in these low-

predation populations (Reznick et al. 2012). We also find that the smallest fish are the ones most 

likely to move out of emptier pools, and are not dispersing out of crowded pools with a higher 

probability than larger (and assumed to be more competitive) individuals, as would be predicted 

if competition was a strong driver of movement behavior.  

We propose some potential explanations for these non-intuitive results. First, this study 

looked only at dispersal in sexually mature guppies, where resource competition may not be as 

strong as it is for juveniles, who are gape-limited to fewer potential food sources (Dial et al. 

2017) and more susceptible to cannibalism by adults (Loekle et al. 1982; Magurran & Seghers 

1990). A previous study looking at dispersal dynamics during colonization of low-predation 

habitat by guppies from high-predation habitats found that juveniles disperse most out of more 

crowded pools (positive density-dependence), while adults changed density-dependent dispersal 

based on the stage of colonization (De Bona et al. 2019). In that study, adult guppies show 

negative density-dependence at the beginning of colonization of a new habitat, which is 

attributed in part to schooling behavior in high-predation guppies in a novel environment 

(Magurran & Seghers 1991; Magurran et al. 1993; De Bona et al. 2019). That, however, is not 

 42 

expected to be a factor in our study of headwater fish populations. Another potential mechanism 

behind the negative-density dependent dispersal we observed is census size being an honest 

indicator of habitat quality. That is, the pools with more individuals are the ones with the most 

resources, lower mortality, and higher reproductive output, effectively attracting, producing, and 

retaining more individuals. When moving between pools, conspecific attraction as a means to 

assess habitat quality can be an effective strategy when animals are unable to directly determine 

quality (Fletcher 2006).  

It is also possible that the increased mating opportunities in pools with more individuals 

outweigh the costs of higher resource competition for adult guppies. Dispersing P. reticulata, 

especially those who move longer distances, show increased reproductive output (Borges et al. 

2022). For males, dispersal significantly increases the likelihood of having any offspring at all 

(Borges et al. 2022), suggesting that finding mates is an important consequence––and therefore 

potential driver––of dispersal for these animals. Alongside our results on the effects of sex ratio 

and relatedness, discussed below, the increased dispersal of guppies from empty pools and the 

increased reproductive output of dispersers points to the importance of reproduction over 

competition in driving individual dispersal decisions. 

Comparing differences in response to census size between the seasons , we found some 

evidence of competition influencing dispersal. As predicted, the probability of dispersal from 

crowded pools is higher in the wet season, when resources are more scarce (Reznick 1989; 

Fitzpatrick et al. 2014, 2016). This points to temporal variation in the role of competition as a 

driver of dispersal, in line with what is known about the effects of seasonality on other aspects of 

guppy life history. 

 43 

 
These mechanisms––decreased competition in adults when compared to juveniles, high 

census sizes as an honest signal of habitat quality, heightened importance of finding mates, and 

seasonal fluctuations in competitive pressures––are not mutually exclusive, and may all 

contribute to the negative effect of census on dispersal probability in the studied streams. Our 

results add nuance to what is known about the importance of competition driving dispersal in this 

model system. Future studies linking census size and density, and directly assessing the role of 

competition on dispersal could further our knowledge of density-dependent dispersal in guppies. 

In contrast to density, less is known about the effects of sex ratio and relatedness on 

guppy dispersal behavior. The mating biology of P. reticulata in low-predation habitats is 

defined by vibrantly colored males pursuing females, who then choose their partners based 

primarily on color pattern, showing strong preferences for novelty (Houde 1987; Endler & 

Houde 1995; Hughes et al. 2013b). Males compete to attract mates, indirectly by displaying their 

colors for females to choose and directly by interfering with other males’ displays (Liley 1966; 

Houde 1987; Jirotkul 1999; Auld et al. 2015). When sex ratios are more male-biased, 

interference behaviors between males are more common and individual males copulate less often 

(Jirotkul 1999). Females can also incur negative fitness consequences from excessive mating 

attempts from males, foregoing foraging to escape courtship displays and sneaky mating 

attempts, especially in more male-biased pools (Magurran & Seghers 1994). Given this natural 

history, we expected both sexes to avoid male-biased pools: males to avoid competition for 

mates, and females to avoid harassment. In Caigual, we observed that exact pattern, with female 

avoidance of male-biased pools being stronger than males’. This suggests that male harassment 

of females in these streams may be more detrimental to female fitness than male-male 

competition is to males.  

 44 

We did not observe the same effects of sex ratio on dispersal in Taylor, which could be 

due to lower sample sizes decreasing the power of our analyses. Interestingly, average sex ratios 

between Taylor (0.55) and Caigual (0.38) were different––almost even (1:1) in Taylor and more 

female-biased in Caigual (Figure 3.2). Population-level sex ratios are known to vary between 

populations and over time in P. reticulata adults, with potential consequences for sexual 

selection (Pettersson et al., 2004; Rodd & Reznick, 1997; Seghers, 1983). Given that juvenile sex 

ratios tend to be even in this species (Pettersson et al. 2004), this suggests increased mortality of 

males in Caigual, or it may be a consequence of their higher overall dispersal abilities (moving 

out of the studied reach entirely; Borges et al. 2022). Male guppies are known to increase 

investments in courtship in female-biased environments (Jirotkul 1999), which may increase 

both male-male competition and male harassment of females, driving the increased sex-biased 

dispersal we see in Caigual compared to Taylor. Future studies linking sex ratio variation with 

mating behaviors and sex-biased dispersal could help explain the differences we observed 

between streams. 

We also find that relatedness affected dispersal probabilities in both Caigual and Taylor. 

Our results go against expectations based on inbreeding avoidance in the guppy mating system 

(Pitcher et al. 2008; Zajitschek & Brooks 2008; Johnson et al. 2010), with fish being more likely 

to disperse out of pools with lower average relatedness (Figure 3.5). De Bona et al. (2019) found 

similar results in their study of dispersal during colonization, suggesting a potential benefit to 

guppies sharing a pool with kin. Habitat quality may also be at play, where individuals that 

encounter high quality pools have more offspring and tend to stay in those pools, leading to a 

pattern of increased relatedness to their neighbours alongside a decrease in dispersal probabilities 

(De Bona et al. 2019). In both that study and ours, however, the effects of relatedness on 

 45 

dispersal probabilities were substantially weaker than the effects of census size, which appears to 

be the main driver of dispersal dynamics in these streams.  

Interestingly, we found an interaction between seasonality and relatedness driving male 

dispersal in both Caigual and Taylor (Figure 3.5). This interaction shows that the negative effects 

of relatedness on dispersal are only present for males during the wet season, while for females 

that effect remains constant throughout the year. For males in the dry season, when resources are 

more abundant, the direction of this effect is reversed in Caigual and becomes flat in Taylor. If 

males avoid pools with more relatives in search of females to whom they appear more novel, and 

therefore more attractive, this pattern may be driven by variation in reproductive behaviors 

between seasons. Specifically, if females are more receptive during the dry season, males should 

disperse out of pools with more relatives then. Future studies linking reproductive behavior and 

seasonality in Poecilia reticulata will increase our understanding of how relatedness drives 

dispersal patterns in these populations. 

Finally, across all of our models, we consistently found that individual and monthly 

variation explained a lot of the variance in dispersal probabilities. This was true even when 

accounting for between-individual differences in sex and weight (marginal vs. conditional R2 

values, Table 3.1). This suggests that there may be individual consistency in behavior that is 

driven by factors we did not include in our models. The study of guppy “personalities,” 

individual consistency in behavior across time and context (Réale et al. 2010; Carter et al. 2013; 

Kaiser & Müller 2021), has shown that these animals can vary in boldness, exploratory behavior, 

and sociability (Budaev 1997; Harris et al. 2010; Kniel & Godin 2019). These personality traits 

are expected to influence how individuals respond to competition and mates, as well as their 

dispersal behavior (Cote et al. 2010b, 2010a; Spiegel et al. 2017). Though guppy personalities 

 46 

are partially shaped by their exposure to predation, as well as their sex and body size (Harris et 

al. 2010), our study suggest that these factors may not be enough to explain individual 

consistency in dispersal behavior in natural low-predation settings. This points to a gap in our 

understanding of the links between individual behavior variation in guppies and their effects on 

movement and dispersal (Spiegel et al. 2017).  

As a whole, our results highlight the complexity of dispersal decision-making in natural 

habitats. The local composition of habitat patches, reflected in metrics like census size, sex ratio, 

and relatedness vary dynamically across space and time, and different individuals will respond 

differently to them based on their own traits, shedding light on the selective pressures driving 

dispersal in this system. Overall, we see that finding mates may be more important than resource 

competition for determining local dispersal behavior in Poecilia reticulata, with variation across 

season and streams. Knowing the drivers of animals’ movement decisions at this fine scale is 

necessary for a thorough understanding of the selective forces behind patterns of dispersal at the 

population level.  

Acknowledgements 

We thank the many field assistants who helped catch, process, and release guppies in Trinidad. 

The Fitzpatrick Lab, KBS Writing Group provided valuable feedback on previous versions of 

this manuscript. We thank D.N. Reznick for intellectual contributions to this work. This project 

was funded by a National Science Foundation grant to S.W.F. (DEB-2016569). This is KBS 

contribution X. 

 47 

 
 
CHAPTER FOUR: 

Mutualism benefits are lost at high inbreeding levels 

Abstract 

Inbreeding depression is a major concern for the persistence of small, isolated 

populations. In plants, the severity of inbreeding depression often depends on the environment in 

which fitness is assessed. Here, we explore how inbreeding interacts with mutualisms to affect 

plant fitness in an outcrossing legume, Chamaecrista fasciculata. We crossed plants for two 

generations to produce seeds that differ in their extent of inbreeding, and conducted a common 

garden experiment where plants were exposed to mutualistic nitrogen-fixing bacteria (rhizobia) 

and bodyguard ants. We found that plant inbreeding decreased plants’ investment in the 

mutualisms with rhizobia and with ants. Plant fitness also decreased with inbreeding, yet the 

slope of that decline was different for plants that did or did not receive rhizobia. The benefits of 

rhizobia for biomass and total seed set were lost at high inbreeding values, such that inbreeding 

depression was steeper in the presence of this mutualist. Overall, we found that mutualisms are 

affected by inbreeding, and that rhizobia can change the severity of inbreeding depression in C. 

fasciculata. 

Introduction 

Understanding the fitness consequences of inbreeding is a longstanding goal in ecology 

and evolutionary biology, with deleterious effects of inbreeding having been documented since 

the inception of evolutionary research (Darwin 1876; Charlesworth & Charlesworth 1987). Here 

we define inbreeding as mating with relatives more often than expected by chance (Keller & 

Waller 2002). As anthropogenic changes drive population declines and habitat fragmentation, 

increased rates of inbreeding in small populations are now a major threat to the conservation of 

biodiversity (O’Grady et al. 2006). Studies that have tested for inbreeding depression, the 

 48 

reduction in fitness associated with inbreeding, have found it to be common in small, isolated 

populations (e.g. Spielman et al. 2004; Armbruster & Reed 2005a; Honnay & Jacquemyn 2007). 

However, the severity of fitness declines associated with inbreeding vary widely depending on 

the environmental context in which fitness is assessed (Cheptou & Donohue 2011; Fox & Reed 

2011; Reed et al. 2012).  

Greenhouse and laboratory studies generally detect less inbreeding depression than 

studies of the same populations in natural environments, where multiple environmental stresses 

are present (Fenster & Galloway 2000; Armbruster & Reed 2005a; Reed et al. 2012; Springer et 

al. 2020). This points to the context-dependency of fitness declines, where the effects of 

inbreeding can be greatly modified by the presence of abiotic stress (Keller et al. 2002; Reed et 

al. 2003, 2012; Armbruster & Reed 2005a), competition (Schmitt & Ehrhardt 1990; Cheptou et 

al. 2000; Meagher et al. 2000; Richardson & Smiseth 2017), herbivory (Carr & Eubanks 2002; 

Campbell et al. 2013; Kalske et al. 2014), disease (Haag et al. 2003; Calleri et al. 2006; Ross-

Gillespie et al. 2007), or predation (Auld & Relyea 2010). These environmental factors and 

biotic interactions have the potential to relieve or exacerbate inbreeding depression in natural 

settings. For example, there is ample evidence that parasitism and disease prevalence are higher 

in populations with lower genetic diversity and higher inbreeding (Smallbone et al. 2016; Gibson 

& Nguyen 2020). As a consequence, the presence of parasites can exacerbate fitness loss in 

inbred populations (Smith et al. 2009). 

Mutualisms are ubiquitous biotic interactions that positively affect the fitness of both 

participating organisms (Kiers et al. 2010). Though the genetic bases of various mutualisms are 

well-established (Gorton et al. 2012; Boutin et al. 2014; Marcionetti et al. 2019), there are no 

general hypotheses for the effects of inbreeding on mutualisms––in contrast with established 

 49 

predictions for these dynamics in other biotic interactions. In the context of plants, many of the 

physiological pathways involved in the defense against pathogens and herbivores are similar to 

those that mediate the recruitment and maintenance of mutualisms, such as root exudates (Bais et 

al. 2006; Narula et al. 2009; Huang et al. 2014), volatile organic compounds (Kost & Heil 2008; 

Dudareva et al. 2013), and immune loci (Wood et al. 2018). Many of the mechanisms through 

which inbreeding affects other biotic interactions can also alter mutualisms, although these 

dynamics remain largely unexplored. Two studies have explicitly looked at the effects of plant 

inbreeding (defined by one generation of selfing vs. outcrossing) on resource mutualisms, 

finding that it can affect plants’ interaction with mycorrhizal fungi (Botham et al. 2009) and with 

nitrogen-fixing bacteria (Rehling et al. 2019). Here, we build on this work by controlling 

inbreeding values along a continuum, and testing effects of inbreeding depression when plants 

are or are not exposed to mutualists. 

Beyond the basic gap in understanding of the interactions between mutualism and 

inbreeding, there is also an applied urgency to this question. The loss of mutualistic interactions 

often results from many of the same global change factors that lead to increased inbreeding in 

natural populations: fragmentation, habitat deterioration, and climate change (Dunn et al. 2009b; 

Hegland et al. 2009; Jorge & Howe 2009; Aslan et al. 2013; Renner & Zohner 2018). When 

predicting extinctions, accounting for the disruption of biotic interactions significantly increases 

predicted biodiversity loss (Colwell et al. 2012; Bascompte et al. 2019). Notable examples 

include increased asynchrony in the phenology of plants and pollinators due to differential 

responses to warming (Hegland et al. 2009; Forister et al. 2010), and coral bleaching due to 

temperature stress in warming oceans (Knowlton 2001). The loss and reshuffling of these 

mutualisms directly affects fitness of participant species and may interact with inbreeding in 

 50 

these same organisms, potentially speeding up biodiversity declines when these stressors occur 

concurrently. 

In this study, we answer two related questions: (1) What are the effects of host 

inbreeding on the presence and strength of mutualisms?; and (2) How does the presence or 

absence of a mutualist affect the expression of host inbreeding depression (i.e. the loss of 

fitness that comes from inbreeding)? We conducted a common garden experiment to test how 

inbreeding in the legume Chamaecrista fasciculata altered two of its mutualisms: a resource 

mutualism with nitrogen-fixing soil bacteria (rhizobia), and a defense mutualism with nectar-

feeding ants (Figure 4.1A). The mutualism between legumes and rhizobia is a widespread, 

economically important interaction known to affect plant fitness (Kiers et al. 2010; Masson-

Boivin & Sachs 2018) and vary with plant and bacterial genotype (Heath et al. 2010; Gorton et 

al. 2012; Burghardt et al. 2017). This mutualism can be inferred by the presence of specialized 

nodules on plant roots, where rhizobia provide plants with nitrogen in exchange for 

carbohydrates. Chamaecrista fasciculata also participates in a defense mutualism with ants, 

where plants provide nectar through extra-floral nectaries on their leaves in exchange for ant-

mediated defense against herbivores (Bentley 1977; do Nascimento & Del-Claro 2010).  

Previous studies have shown that fragmentation of natural habitats and increased rates of 

selfing cause inbreeding depression in C. fasciculata, which is primarily outcrossing (Fenster 

1991b, 1991a; Sork & Schemske 1992; Mannouris & Byers 2013). However, the extent to which 

plant inbreeding affects plants’ mutualism with rhizobia and ants, and the consequences of these 

effects on plant fitness, have not yet been explored in C. fasciculata or other systems. For 

question (1), we tested the effect of plant inbreeding on the number of mutualism structures (root 

nodules and extra-floral nectaries), with two alternative predictions: (1A) inbred plants increase 

 51 

investment in mutualism structures to attract mutualists and potentially compensate for the 

physiological stress of inbreeding; (1B) inbred plants are less able to invest in mutualism 

structures, such that they produce fewer nodules and EFNs (Figure 4.1B). For question (2), we 

tested whether the presence of mutualistic rhizobia affects the expression of inbreeding 

depression, with three potential predictions: (2A) resources provided by the presence of a 

rhizobia buffer plant fitness declines, such that plants can at least partially compensate for other 

physiological costs of inbreeding and maintain high fitness; (2B) the mutualism breaks down 

with inbreeding, as plants lose the ability to recruit, control, or provide for rhizobia, causing the 

benefits of the interaction to decrease as inbreeding increases; or (2C) the mutualisms are not 

affected by inbreeding, such that inbred and outbred plants invest similarly in the interactions 

and accrue similar benefits from them (Figure 4.1C).  

Figure 4.1. Description of system and predictions of results. A) Illustration of C. fasciculata 
showing the two mutualisms studied here: the aboveground ant defense mutualism and the 
belowground nitrogen-fixing rhizobia mutualism. B) Predicted effects of inbreeding on 
mutualism investment. The number of nodules is represented by orange solid lines, reflecting 
plant investment in the rhizobia mutualist. The number of extra-floral nectaries (EFNs) is 
represented by green dashed lines, reflecting plant investment in the ant mutualist. Darker lines 
with a positive slope show hypothesis (1A), with investment in mutualisms increasing with 
inbreeding. Lighter lines with a negative slope depict hypothesis (1B), where more inbred plants 
invest less in the mutualism. C) Predicted effects of inbreeding on plant fitness with and without 
rhizobia. The black dashed line represents an expected decline in fitness with inbreeding. 

 52 

plant inbreedingmutualism investmentinbreeding depressionwith mutualistwithout mutualistmutualismA)C)(2A)(2B)(2C)B)plant inbreedingplant fitness(1A)(1B)EFN numbernodule number 
Figure 4.1 (cont’d) The solid lines in different colors represent different predictions for how the 
addition of a mutualist may affect the expression of inbreeding depression (i.e. the slope of the 
line). It is assumed that the presence of mutualists increases plant fitness, shown as the higher 
intercept of the solid lines. In blue (2A) the rhizobia partially rescue plants from inbreeding 
depression by maintaining similar fitness across different levels of inbreeding; in red (2B), the 
mutualism benefit is lost as inbreeding increases; and in gray (2C) the mutualism does not affect 
the slope of inbreeding depression, offering similar benefits to plants across the whole gradient 
of inbreeding. 

Methods 

This study was conducted with the annual legume Chamaecrista fasciculata, and its 

Bradyrhizobium sp. and ant mutualists. Using an outdoor common garden experiment, we first 

tested whether plant inbreeding affected the presence and number of mutualistic structures 

involved in the interactions with rhizobia and ants (root nodules and extra-floral nectaries, 

respectively). Then, we asked whether the presence of rhizobia modified the expression of 

inbreeding depression in C. fasciculata.  

To obtain seeds for this experiment, we manipulated two generations of plant crosses for 

C. fasciculata individuals collected from two natural populations in the US Midwest (Westland, 

MI, and Sandusky, OH). Four maternal families from each population were used to generate all 

seeds used in the experiment, and crosses were conducted manually in a greenhouse at Kellogg 

Biological Station in Hickory Corners, MI (Figure 4.2). See Supplemental Materials for details. 

After two generations, we had a total of 3423 F2 seeds varying from twice-selfed to outbred, 

which we used in the common garden experiment described below.  

All C. fasciculata seeds were geminated in the greenhouse, then planted into 2.5 x 10 

inch “cone-tainers” (Stuewe & Sons, Inc., Tangent OR, USA) filled with Premier Pro-Mix FPX 

2.8cf soil from BFG Supply Co. (Grand Rapids MI, USA). Parental and F1 generations were 

kept in the greenhouse for the duration of their lifecycle, while the experimental F2 plants 

 53 

 
remained in the greenhouse only until rhizobia inoculations were completed, then were moved to 

an outdoor common garden. 

For the common garden experiment, we attempted to geminate 3209 seeds, resulting in 

2571 live seedlings. Out of those experimental plants, 1715 C. fasciculata individuals of all 

crossing backgrounds were inoculated with Bradyrhizobium sp., while 856 plants did not receive 

any rhizobia. We used two rhizobia strains that were obtained from previous collections of 

naturally occurring C. fasciculata at the Kellogg Biological Station, which did not differ in their 

ability to nodulate with experimental plants or the number of nodules they produced (Figure 

S4.1). The strains were collected in 2017 from nodules in C. fasciculata grown at the Kellogg 

Biological Station, such that both experimental plant populations were naïve to the strains used 

in the common garden. We added 2mL of liquid culture to each plant, with control plants 

receiving the same volume of sterile media (see Supplemental Materials for full culturing 

methods). 

Figure 4.2. Outline of crossing design and inoculations. A) For crosses, two populations were 
used. Each box depicts a maternal line from one population, with different sizes representing the 
two different populations. From left to right, cross types were: between populations, within 
population between maternal lines, within maternal lines, and selfed. All cross types were 
applied to all maternal lines in both populations, for two generations. B) Individuals resulting 
from these crosses were then divided into two inoculation treatments: one with rhizobia (full 
gradient line) and one without rhizobia (dashed black line). 

 54 

+ rhizobia-rhizobiax 2 generations 
 
 
Common garden: 

All individuals were then moved to a fenced outdoor common garden at the Kellogg 

Biological Station. The common garden allowed for natural levels of insect herbivory and 

pollination, while preventing mammalian herbivory. We scored the day of first flower for each 

individual, and height at flowering on that date.  

To study the plant-ant mutualism, we selected 58 individuals of C. fasciculata that 

spanned our inbreeding gradient. For the month after plants were placed outside, we conducted 

two surveys of the number of extra-floral nectaries (EFN) on each of those 58 individuals and 

quantified insect herbivory once as the percentage of plant leaves with any herbivore damage. 

We conducted seven ant surveys on those plants, where we counted the number of ants on each 

plant, then calculated an average number of ants per survey per plant. Once plants began to set 

seed (late August), we manually collected seed pods daily. Plants started being harvested in 

October, once they senesced and no longer had any green leaves or flowers. At time of harvest, 

we separated above and belowground biomass. Plant fitness was quantified as the total number 

of viable seeds produced throughout the growing season, as well as above and belowground 

biomass at senescence. We also estimated the strength of plants’ interactions with rhizobia by 

counting the number of nitrogen-fixing nodules on the roots of each plant at the end of the 

experiment.  

Data analyses: 

We used statistical modeling to test how inbreeding and mutualisms affected plant traits. 

For all models described below, we used the R package DHARMa (Hartig 2021) to assess model 

fit. For comparing candidate models that varied in explanatory variables and distributional 

 55 

 
assumptions, we selected best models based on AIC (Akaike Information Criterion, Johnson & 

Omland 2004). 

For all plant fitness and trait responses, our main explanatory variable of interest was 

plant inbreeding, which we calculated using the pedigree obtained from two generations of 

crosses. We used the R package pedtools (Vigeland 2021) to build the pedigree from our cross 

data, and the R package ribd (Vigeland 2021) to calculate inbreeding coefficients from the 

constructed pedigree. To do so, we assumed no initial relatedness between the populations, no 

differences in baseline levels of inbreeding, and treated all maternal lines within populations as 

being equally related to one another. To account for between-population differences that may 

affect traits and fitness, we also calculated ancestry proportion by counting how many of the four 

grandparents of each focal F2 individual originated from a parental population, divided by four. 

We chose the large population from Ohio as our reference, such that an ancestry proportion equal 

to 1 reflects an individual with all grandparents originating from the Ohio population and 0 

reflects an individual with all grandparents from the Michigan population. 

- Effects of plant inbreeding on mutualisms: We first tested for the effects of plant 

inbreeding on the mutualism traits and responses: number of root nodules (indicator for the 

legume-rhizobia mutualism), number of EFNs (indicator of plant investment in ant mutualist 

recruitment), and average number of ants on plants across seven ant surveys (direct measure of 

mutualist recruitment). Number of nodules and ants were modeled with a negative binomial 

distribution. Probabilities of contamination (plants that were not inoculated but produced 

nodules) and nodulation failure (plants that were inoculated but did not produce nodules) were 

modeled with a binomial distribution, and number of EFNs was approximately normal. 

 56 

Herbivory was measured once during the season as the percentage of plants’ leaves that had any 

herbivory damage, and was modeled with a zero-inflated beta distribution. 

The independent variables of interest tested in these models were pedigree inbreeding 

(both linear and quadratic effects) and inoculation status (i.e. whether or not it received rhizobia). 

We also included proportion of ancestry from each population in all models. For herbivory, we 

included average number of ants as a predictor. All models included planting date as a random 

effect, which was removed from final models if it did not improve model fit or increased 

conditional R2. For all nodulation variables, we also included harvesting date as a random effect, 

as it could affect nodule senescence and counts.  

- Effects of inbreeding on plant traits: We then tested the extent to which inbreeding 

affected the different plant traits and fitness. To do so, we modelled: total seed set, aboveground 

biomass, belowground biomass, plant height, and individual seed mass. The independent 

variables tested in these models were pedigree inbreeding (both linear and quadratic effects), 

rhizobia inoculation, ancestry proportion, and interactions between these variables. All potential 

two-way interactions between independent were tested, and dropped from the model if they were 

not significant. 

- Effect of mutualisms on inbreeding depression: Due to small sample sizes in the ant 

portion of the study, we could only examine the effect of rhizobia on the slope of inbreeding 

depression (Figure 4.1B). This was done using the same models described above, where a 

significant interaction between inoculation status (i.e. rhizobia presence) and inbreeding was 

interpreted as an effect of this mutualism on the expression of inbreeding depression. 

 57 

 
 
 
Results 

Effects of plant inbreeding on mutualisms 

We first examined the effect of plant inbreeding on plant interactions with mutualistic 

ants and rhizobia. For rhizobia, we used root nodule number as a proxy for the investment from 

the plants in this mutualism. For ants, we used the number of extra-floral nectaries as proxy of 

plant investment, average number of ants across all surveys as a direct measurement of the 

strength of the interaction, and measured herbivory as a proxy for the effectiveness of bodyguard 

ant mutualists. 

We found a significant negative effect of inbreeding on EFN number (N=50, z=-2.900, 

p=0.004; model conditional R2=0.845; marginal R2=0.577; Figure 4.3A), even when accounting 

for the positive effect of plant size on ENFs (aboveground biomass; z=5.577, p<0.001; Figure 

4.3B). EFN number, in turn, had a significant positive effect on the average number of ants that 

visited plants across seven ant surveys (N=58, z=3.542, p<0.001; model R2=0.311; Figure 4.3C). 

For herbivory (model conditional R2=0.509; marginal R2=0. 472), we detected no significant 

effect of average ant number (N=58, z=0.075, p=0.94043), but did find significant positive 

effects of inbreeding (linear: z=2.022, p=0.043; and quadratic: z=3.056,  p=0.002), as well as a 

significant negative interaction between quadratic inbreeding and rhizobia (z=-2.072, p=0.038; 

Figure 4.3D). As a whole, an increase in pedigree inbreeding from the minimum to the maximum 

value in our dataset is predicted to decrease EFN number by approximately 20%, decrease ant 

numbers by 2%, and lead to a 48% increase in herbivory.  

We then examined the effects of inbreeding on the plant-rhizobia mutualism. For nodule 

number, our best supported model (conditional R2=0.269; marginal R2=0.071) found nearly 

significant negative effects of plant inbreeding on nodule counts (N=1432, z=-1.821, p=0.068) 

 58 

and a significant positive effect of inbreeding on the likelihood of plants producing zero nodules 

(z=2.788, p=0.005; Figure 4.4A). Supporting the effectiveness of our rhizobia treatments, we 

found that the strongest effect in these models was inoculation status––a positive effect on 

nodule counts and (z=6.901 p<0.001) a negative effect on the likelihood of producing zero 

nodules (z=-17.354, p<0.001). An increase in inbreeding from minimum to maximum values in 

our dataset is predicted to decrease nodule number by approximately 17%. 

We observed some plants that were not inoculated yet did form nodules (i.e. 

contamination, 176 out of 856 non-inoculated plants, 20%), as well as a lower number of plants 

that did not form nodules despite being inoculated (i.e. failure to nodulate, 40 out of 1715 

inoculated plants, 2%). We modeled the probability of contamination and failure to nodulate and 

found a nearly significant negative effect of inbreeding on the probability of contamination 

(N=1546, z=-1.726, p=0.084), and a significant positive effect of inbreeding on the probability of 

nodulation failure (N=1512, z=2.162, p=0.031). That is, more inbred plants were less likely to 

recruit rhizobia from the environment when they were not inoculated (Figure 4.4B), and plants 

that failed to nodulate despite being inoculated tended to be more inbred (Figure 4.4C). Ancestry 

proportions were not significant predictors of either contamination or failure to nodulate, 

suggesting that both populations used in the study were similarly naïve to and compatible with 

the used rhizobia strains. 

 59 

 
Figure 4.3. Predicted values based on best models. A) EFN number decreases with increasing 
inbreeding, even when accounting for biomass. B) EFN number peaks at intermediate values of 
aboveground plant biomass. C) The average number of ants across three survey occasions during 
the growing season increases with increased number of EFNs. D) The percent of leaves with 
herbivory increases with increased plant inbreeding, and does so more steeply for plants that are 
not inoculated with rhizobia (red). 

 60 

010203040−2−10123Pedigree inbreedingEFN numberA0102030400.00.51.01.52.02.5Aboveground biomass (g)EFN numberB01234010203040ENF countAverage number of antsC−0.20.00.20.40.6−2−10123Pedigree inbreedingPercent herbivoryrhizobiaNYD 
 
 
Figure 4.4. Predicted values based on best models. A) Number of rhizobia root nodules 
decreases with increased plant inbreeding, plants inoculated with rhizobia produced more 
nodules than those that did not, but both rhizobia treatments are shown due to contamination. B) 
Probability of contamination (producing nodules despite not being inoculated) decreases with 
increased inbreeding, though this effect is not significant. C) Probability of nodulation failure 
(not producing nodules despite inoculation) increases with increased inbreeding. 

Effects of inbreeding on plant traits 

As expected, we found that inbreeding in Chamaecrista fasciculata generally led to 

decreases in plant fitness. Effects differed by trait, but most showed some response to 

inbreeding, and those responses were generally negative (i.e. more inbred plants having lower 

values). Some of these effects (e.g. seed set) were also present in the first generation of crosses 

(F1, Borges personal obs.), but here we only show and discuss the effects for the experimental 

generation, F2. 

The best fit model for total seed set explained over 80% of the variation in this metric 

(marginal R2=0.813), and showed a significant decrease in seed counts with increased inbreeding 

(N=2566, z=-6.78, p<0.001). We also found a significant negative quadratic effect of pedigree 

on seed number (N=2566, z=-5.18, p<0.001), suggesting that plants with intermediate inbreeding 

values produce the highest number of seeds. Inoculation with rhizobia (N=2566, z=27.08, 

p<0.001), average number of ants (N=57, z=2.464, p=0.014), and ancestry proportion (N=2566, 

z=5.07, p<0.001) all had significant positive effects on seed counts. The interaction between 

 61 

050100150200−2−10123Pedigree inbreedingNumber of nodulesrhizobiaNYA0%25%50%75%100%−10123Pedigree inbreedingProbability of contaminationB0%25%50%75%100%−10123Pedigree inbreedingProbability of nodulation failureC 
 
inbreeding and rhizobia was also significantly positive (N=2566, z=3.12, p=0.002), suggesting 

that plants that receive rhizobia show steeper declines in seed set with increased inbreeding 

(Figure 4.5A). 

For aboveground biomass, our best fit model explained approximately a third of the 

variation (marginal R2= 0.182, conditional R2= 0.301). Inbreeding had a significantly negative 

effects on biomass (N=1619, linear: z=-3.428, p<0.001; and quadratic: z=-3.357 p<0.001). 

Ancestry proportion also had significantly negative effects on biomass (z=-4.943, p<0.001), 

while rhizobia had strong positive effects on it (z=17.284, p<0.001). We found a significant 

interaction between inbreeding and rhizobia presence (z=-1.930, p=0.054) in this model, 

suggesting that the biomass benefits that rhizobia provide to plants decrease with increased 

inbreeding (Figure 4.5C). 

 62 

 
Figure 4.5. Predicted values based on best models. A) Total seed number (fitness) decreases 
with increased plant inbreeding, and does so more steeply for plants that were inoculated with 
rhizobia. B) Ancestry proportions affect number of seeds produced, with values closer to 0 
(higher ancestry from smaller Michigan population) producing fewer seeds. Left panel shows 
trends for plants that were not inoculated with rhizobia, right panel shows trends for plants that 
did get inoculated. C) Similarly to the trend we see for seed set, aboveground biomass decreases 

 63 

0100200300−2−10123Pedigree inbreedingTotal seed numberrhizobiaNYANY−2−10123−2−101230100200300Pedigree inbreedingTotal seed numberprop.E0.000.250.500.751.00B01234−2−10123Pedigree inbreedingAboveground biomass (g)CNY−2−10123−2−1012301234Pedigree inbreedingAboveground biomass (g)D−2−1012−2−10123Pedigree inbreedingIndividual seed mass (g)ENY−2−10123−2−10123−2−1012Pedigree inbreedingIndividual seed mass (g)FFigure 4.5 (cont’d) with increased plant inbreeding, and the decrease is steeper for plants that 
received rhizobia. D) Ancestry proportions also affected aboveground biomass, but in the 
opposite direction of seed set. Plants with more Michigan ancestry (lower values) were larger. E) 
The mass of individual seeds (averaged per envelope) decreases with increased inbreeding, and 
in contrast to total seed set and aboveground biomass, the decrease is less steep for plants that 
were inoculated with rhizobia. F) Similarly to biomass, plants with increased Michigan ancestry 
produce on average larger seeds. 

For belowground biomass, a generalized linear model with a gamma distribution best fit 

the data, though it had a low explanatory power (marginal R2=0.006). We found significant 

positive effects of rhizobia on belowground biomass (N=1436, z=-5.106, p<0.001), and 

significant negative effects of inbreeding (z=3.080, p=0.002), but no interaction between the two 

(Figure 4.6A). 

We found significant negative effects of inbreeding (N=1973, z=--5.695, p<0.001) on 

plant height at flowering, with our best model explaining almost half of the variation in height 

(marginal R2= 0.265, conditional R2= 0.492). We also see that ancestry (z=-16.193, p<0.001) 

affects height,  with plants that have more ancestors from the Ohio population being smaller. We 

also included age at flowering (i.e. days from planting until first flower) as a variable in this 

model and found significant positive effect of it on height (z=18.102, p<0.001; Figure 4.S2). 

There was also no significant effect of rhizobia on height at flowering (Figure 4.6B).  

We also modeled individual seed mass, which we obtained by dividing the total weight of 

seeds from seed pods collected from the same plant each day by the number of seeds in those 

pods (one value per plant per day where seeds where collected). Due to this methodology, this 

response variable had low variation, and our best fit linear model had low R2 

(conditional=0.040). This model found significant negative effects of inbreeding (N=1951, z=-

6.469, p<0.001) and ancestry proportions (z=-3.876, p<0.001). We also found a significant 

interaction between inbreeding and rhizobia (z=2.344, p<0.001), suggesting that the loss in seed 

 64 

mass that comes with increased inbreeding is less severe when plants interact with rhizobia 

(Figure 4.5E). There was no significant effect of rhizobia alone in this model. 

Figure 4.6. Predicted values based on best models. A) Belowground biomass decreases with 
increased inbreeding, and though plants that received rhizobia have higher belowground 
biomass, there is no statically significant interaction between inbreeding and inoculation with 
rhizobia. B) There was no effect of rhizobia inoculation on plant height at flowering, but height 
did decrease with increased plant inbreeding. 

Effects of mutualisms on inbreeding depression  

For each trait, we tested for interactions between inbreeding (both linear and quadratic) 

and rhizobia to determine whether this mutualism altered the slope of inbreeding depression. For 

total seed set and aboveground biomass, we found interactions between inbreeding and rhizobia 

where plants with rhizobia had a steeper decline in these traits compared to plants that did not 

(Figures 4.5A and 4.5C). In contrast, the interactions we observe for individual seed mass 

(Figure 4.5E) and herbivory (Figure 4.3D), show that plants that received rhizobia had a 

shallower slope for these traits, and so did not lose seed mass or increase herbivory as much as 

their non-inoculated counterparts. 

 65 

0.02.55.07.510.0−2−10123Pedigree inbreedingBelowground biomass (g)A204060−2−10123Pedigree inbreedingHeight at flowering (cm)B 
 
 
 
Discussion 

We found that inbreeding decreased Chamaecrista fasciculata’s investment in the 

mutualism structures involved in the interactions with both nitrogen-fixing rhizobia and ant 

bodyguards. The presence of rhizobia, in turn, affected the expression of inbreeding depression 

in the plants, but did so differently for different plant traits. Interactions between rhizobia and 

plant inbreeding for the most important fitness metrics, total seed counts and aboveground 

biomass, suggest that the benefits of the rhizobia mutualism are lost at high levels of inbreeding.  

In accordance with previous studies of inbreeding depression in C. fasciculata, we 

generally found that inbred plants have smaller stature and lower fecundity (Fenster 1991b; 

Fenster & Galloway 2000; Erickson & Fenster 2006). There was substantial variation in the 

strength of inbreeding depression for different traits, with total seed set, height, and aboveground 

biomass being strongly negatively affected by inbreeding. This suggests that inbred individuals 

are not as capable of acquiring or maintaining resources as their outbred counterparts (Paige 

2010; Kariyat & Stephenson 2019)––potentially increasing plants’ need for defense and resource 

mutualisms. For example, drought, another stress that generally decreases plant biomass and 

resource availability, has been shown to increase the effectiveness and importance of ant and 

rhizobia mutualisms in legumes (Staudinger et al. 2016; Leal & Peixoto 2017). However, we 

found that inbreeding decreased the investment of plants’ in the two mutualisms we studied. That 

is, more inbred plants generally produced fewer extra-floral nectaries, attracted fewer ants, 

produced fewer root nodules, and were more likely to fail to nodulate when inoculated with 

rhizobia. 

Our study adds to the small body of evidence that plant inbreeding affects belowground 

mutualisms (e.g. arbuscular mycorrhizal fungi in Botham et al. 2009; rhizobia in Rehling et al. 

 66 

2019). The potential mechanisms behind the decrease in mutualism structures and benefits are 

multifold. Broadly, it is possible that inbred plants’ decreased access to resources lead to a 

decreased ability to invest in these interactions. Providing rhizobia with carbon is an energy-

intensive process (between five and ten grams of plant carbon input to one gram rhizobia 

nitrogen output; Xu & Wang 2023), and though the costs of extra-floral nectaries and nectar are 

less well-defined (Bentley 1977; Rutter & Rausher 2004; Marazzi et al. 2013; Pyke & Ren 

2023), both interactions can be assumed to impose metabolic costs on plants. In root nodules, 

rhizobia are dependent on the carbon provided by plants, and are not able to multiply and fix 

nitrogen in the absence of these resources (Xu & Wang 2023). Ants, in turn, are active partners 

in the ant-plant mutualism, showing preference for nectar with specific sugar and amino-acid 

compositions, and visiting plants and leaves with higher sugar concentrations (Ness 2003). An 

inbred plant that cannot provide adequate rewards to its mutualists may lose those interactions 

and their benefits as a result. 

A second suite of potential mechanisms for the decrease in mutualism structures with 

inbreeding pertains to the biochemical and physiological complexity involved in attracting ants 

and rhizobia. Inbreeding is known to affect plant chemical exudates related to pollination and 

herbivore defense (Ferrari et al. 2006; Kariyat et al. 2012; Campbell et al. 2013), and plant-

rhizobia and plant-ant interactions are mediated through similarly specific chemical signals (Heil 

2008; Kost & Heil 2008; Dudareva et al. 2013; Cao et al. 2017; Xu & Wang 2023). The legume-

rhizobia interaction, in particular, is incredibly complex, involving genotype-by-genotype 

interactions between the two partners (Sachs & Simms 2006; Heath & Tiffin 2009; Gorton et al. 

2012; Burghardt et al. 2017; Nobarinezhad & Wallace 2020). Root nodules are plant organs that 

get formed in response to these specific lock-and-key signals between plants and rhizobia, 

 67 

relying on several chemical cues in order to produce an environment where nitrogen fixation can 

occur (Xu & Wang 2023). Similarly to how host-pathogen interactions can be modified by 

inbreeding when immunity is disrupted with increased homozygosity (e.g. Reid et al. 2003, 

2007; Stevens et al. 1997), it is possible that some of these mutualism molecular signals get 

disrupted by the intrinsic genomic changes that come from inbreeding. Our finding that more 

inbred plants are more likely to fail at nodulation when inoculated with rhizobia provides some 

support for this idea. 

Another important facet of the ant-plant and plant-rhizobia mutualisms is their 

dependence on the local environments experienced by plants. Plants’ interaction with rhizobia 

will be favored in nitrogen-poor environments where plants cannot acquire sufficient nitrogen 

from root foraging alone (Akçay & Simms 2011; Keller & Lau 2018). In our common gardens, 

we used nitrogen-poor soil and did not make any nitrogen additions, so that plants were more 

likely to nodulate when inoculated with rhizobia. This low nitrogen availability can be assumed 

to be a main driver of the fitness benefits we observed with nodulation, and it likely modulated 

the fitness declines we observed when inbred plants produced fewer nodules. Similarly with ants, 

the number of extra-floral nectaries and the amount of nectar produced by plants can be induced 

by herbivore damage (Mondor & Addicott 2003; Koricheva & Romero 2012), and the benefits of 

ant defense will only manifest in the presence of herbivores that can be deterred by ants (Bentley 

1976; Kelly 1986).  

In our common garden, mammalian herbivory was fully absent and insect herbivory was 

low (on average, each plant had evidence of herbivore damage on only 7% of its leaves), such 

that the fitness loss that followed from decreased investment in extra-floral nectaries was likely 

weaker than in environments with higher herbivore pressure. Ultimately, the interactions we 

 68 

found between inbreeding depression and mutualisms will be dependent on the environments 

that plants experience. The decrease in mutualism benefit we see at high inbreeding levels is 

more likely to be observed, and to be relevant to plant population dynamics, in stressful (i.e., low 

nitrogen and high herbivory) environments. Importantly, the anthropogenic changes that lead to 

increased inbreeding and mutualism loss in natural populations often also increase other kinds of 

stress, like disrupted soil nutrient cycles (Mosier 1998; Sardans & Peñuelas 2012; Smith et al. 

2016) and increased insect herbivory at habitat fragment edges (Urbas et al. 2007; Guimarães et 

al. 2014; Murphy et al. 2016). As such, the relevance of the interactions we observe between 

mutualism loss and inbreeding depression may become increasingly relevant as global change 

progresses.  

Beyond the environmental dependency of these interactions, they are also likely to vary 

across plant populations with different evolutionary histories. Local adaptation to mutualists and 

herbivores can change the population-level investment of plants in the ant and rhizobia 

interactions (Bronstein 1998; Vittecoq et al. 2012; Pringle & Gordon 2013; Keller & Lau 2018; 

Magnoli & Lau 2020), and a previous history of inbreeding may change the expression of 

inbreeding depression in plants (e.g. purging; Byers & Waller 1999; Fowler & Whitlock 1999; 

Glémin 2003; Keller et al. 2002). In this study, we used two populations assumed to be naïve to 

experimental rhizobia strains, not likely to be in contact via gene flow, and with very different 

estimated population sizes (small Michigan roadside population vs. large Ohio preserve 

population). Interestingly, many of our models revealed significant negative quadratic effects of 

inbreeding on plant traits, suggesting some outbreeding depression between these two 

populations, as the lowest pedigree inbreeding values could only result from between-population 

crosses. We also used a metric of ancestry proportion in all of our models to account for any 

 69 

 
effects of population, and found significant ancestry effects on plant traits but not on nodulation. 

We then tested for interactions between ancestry and inbreeding for all traits, as they would 

suggest differences in the expression of inbreeding depression between the two populations––

potentially indicative of purging in the small Michigan population. However, those interactions 

were never significant and were dropped from final traits models. These results suggest that our 

two experimental populations suffered similarly from inbreeding depression and interacted 

similarly with rhizobia, despite population-specific trait differences. 

Broadly, our results highlight the ecological dependencies and consequences of 

inbreeding depression. Inbreeding is increasingly recognized and managed as a stressor in small 

and fragmented population populations, but we still know little about how it affects and is 

affected by biotic interactions. Even in a system like Chamaecrista fasciculata, where inbreeding 

depression has been previously characterized (Fenster & Galloway 2000; Erickson & Fenster 

2006), considering the ant and rhizobia mutualisms changes our understanding of how 

inbreeding manifests in this species. Incorporating realistic ecological interactions into 

inbreeding studies could provide better predictions for how inbreeding depression plays out in 

natural environments, where interactions are often being disrupted by the same anthropogenic 

pressures that cause inbreeding (Liao & Reed 2009). Our results suggest that the presence of a 

mutualist does not alleviate inbreeding depression, making mutualism loss and inbreeding 

compounding stressors that need to be considered jointly for predicting the fate of populations 

(Rudgers et al. 2020; Kiesewetter & Afkhami 2021). Our study also points to the potential 

consequences of inbreeding in one species scaling to whole communities: two generations of 

inbreeding in our focal plant impacted ant presence, nodulation (and consequentially soil 

nitrogen), and herbivory. Considering the full spectrum of ecological consequences of inbreeding 

 70 

has the potential to improve our predictions about the fate of declining and fragmented 

populations and communities in the face of global change.  

Acknowledgements 

Susan Magnoli provided Chamaecrista seeds collected across the Midwest to kick off this 

project. Mark Hammond and Jeff Conner managed the KBS greenhouses and Plant Ecology 

Field Lab, offering much needed advice at every stage of the common garden experiment. Sam 

Pendrick’s help in the field, alongside the whole Fitzpatrick Lab, made seed collections possible. 

The Fitzpatrick Lab and KBS Writing Group provided valuable feedback on previous versions of 

this manuscript. This project was funded by a J. S. Karling Award from the Botanical Society of 

America and a Student Research Award from the American Society of Naturalists awarded to 

ILB. This is KBS contribution X. 

 71 

 
 
TABLES: 

Table 1.1. Average and maximum values for dispersal distance of both sexes in both streams, 
and for number of offspring of dispersing (moved ≥ 10 m) and non-dispersing fish. 

Stream  Sex 

Average 

Maximum 

dispersal 

Dispersal 

distance 

Distance 

(m) 

(m) 

Average Number of 

Offspring per 

Individuals 

Maximum Number of 

Offspring 

Non-

Non-

Dispersers 

Dispersers 

Dispersers 

Dispersers 

Taylor 

Caigual 

F 

M 

F 

M 

17.57 

19.12 

13.88 

16.94 

248 

170 

157 

216 

1.74 

1.42 

0.75 

1.31 

4.03 

4.66 

1.02 

3.38 

24 

26 

11 

43 

51 

22 

20 

55 

 72 

 
 
 
 
 
 
 
 
 
 
 
 
 
Table 1.2. Average and maximum values for range of both sexes in both streams, and for 
number of mates of dispersing (moved ≥ 10 m) and non-dispersing fish.  

Average 

Maximum 

Stream  Sex 

Range 

Range 

(m) 

(m) 

Average Number of 

Maximum Number of 

Mates per Individuals  

Mates 

Non-

Non-

Dispersers 

Dispersers 

Dispersers 

Dispersers 

Taylor 

Caigual 

F 

M 

F 

M 

14.38 

17.19 

9.65 

11.69 

173 

158 

56 

61 

0.64 

0.52 

0.42 

0.45 

1.17 

1.54 

0.48 

1.32 

7 

7 

4 

8 

9 

7 

12 

18 

 73 

 
 
 
 
 
Table 3.1. Model summaries for best models of probability of dispersal in Caigual and Taylor. 

 74 

CaigualTaylorPredictorsOdds RatiosCIpOdds RatiosCIpCensus 0.150.07 – 0.32<0.0010.660.49 – 0.890.006Sex (M)0.540.40 – 0.72<0.0011.350.76 – 2.380.308Season (W)3.510.58 – 21.240.1722.080.69 – 6.320.196Weight0.080.03 – 0.19<0.0010.530.04 – 7.040.63Census * Season2.831.28 – 6.260.01–––Sex ratio99.7529.86 – 333.28<0.0011.430.23 – 8.910.704Sex (M)0.640.34 – 1.180.1521.290.28 – 5.930.74Season (W)1.560.27 – 9.150.6231.830.45 – 7.390.394Weight0.030.01 – 0.07<0.0014.930.19 – 129.810.339Sex ratio * Sex0.170.03 – 0.880.0351.490.13 – 16.420.746Relatedness8404.663.56 – 1111356.20<0.001 0.690.00 – 1201.260.922Quadratic relatedness00.00 – 0.00<0.001 0.000.00 – 328.020.22Sex (M)0.490.35 – 0.70<0.001 1.180.52 – 2.710.691Season (W)1.960.24 – 16.190.531 3.190.53 – 19.300.207Weight0.020.01 – 0.04<0.001 3.230.10 – 103.220.507Conditional R-squared = 0.474Marginal R-squared = 0.112Conditional R-squared = 0.410Marginal R-squared = 0.110Conditional R-squared = 0.419Marginal R-squared = 0.027Conditional R-squared = 0.247Marginal R-squared = 0.075Response: binary variable describing whether individuals moved in the next time stepMarginal R-squared =  0.170Conditional R-squared = 0.404Marginal R-squared = 0.054Conditional R-squared = 0.211 
 
 
 
 
 
 
 
 
 
 
REFERENCES 

Akçay, E. & Simms, E.L. (2011) Negotiation, sanctions, and context dependency in the legume-
rhizobium mutualism. American Naturalist, 178, 1–14. 

Alphei, J. & Scheu, S. (1993) Effects of biocidal treatments on biological and nutritional 
properties of a mull-structured woodland soil. Geoderma, 56, 435–448. 

Anderson, M. (1982) Sexual selection, natural selection and quality advertisement. Biological 
Journal of the Linnean Society, 17, 375–393. 

Armbruster, P. & Reed, D.H. (2005a) Inbreeding depression in benign and stressful 
environments. Heredity, 95, 235–242. 

Armbruster, P. & Reed, D.H. (2005b) Inbreeding depression in benign and stressful 
environments. Heredity, 95, 235–242. 

Aslan, C.E., Zavaleta, E.S., Tershy, B. & Croll, D. (2013) Mutualism Disruption Threatens 
Global Plant Biodiversity: A Systematic Review. PLoS ONE, 8. 

Auld, H.L., Jeswiet, S.B. & Godin, J.G.J. (2015) Do male Trinidadian guppies adjust their 
alternative mating tactics in the presence of a rival male audience? Behavioral Ecology and 
Sociobiology, 69, 1191–1199. 

Auld, J.R. & Relyea, R.A. (2010) Inbreeding depression in adaptive plasticity under predation 
risk in a freshwater snail. Biology Letters, 6, 222–224. 

Baguette, M., Blanchet, S., Legrand, D., Stevens, V.M. & Turlure, C. (2013) Individual 
dispersal, landscape connectivity and ecological networks. Biological Reviews, 88, 310–326. 

Bais, H.P., Weir, T.L., Perry, L.G., Gilroy, S. & Vivanco, J.M. (2006) The role of root exudates 
in rhizosphere interactions with plants and other organisms. Annual Review of Plant Biology, 57, 
233–266. 

Bascompte, J., García, M.B., Ortega, R., Rezende, E.L. & Pironon, S. (2019) Mutualistic 
interactions reshuffle the effects of climate change on plants across the tree of life. Science 
Advances, 5, 1–9. 

Bassar, R.D., Lopez-Sepulcre, A., Reznick, D.N. & Travis, J. (2013) Experimental evidence for 
density-dependent regulation and selection on trinidadian guppy life histories. American 
Naturalist, 181, 25–38. 

Bentley, B.L. (1976) Plants Bearing Extrafloral Nectaries and the Associated Ant Community: 
Interhabitat Differences in the Reduction of Herbivore Damage. Ecology, 57, 815–820. 

Bentley, B.L. (1977) Extrafloral Nectaries and Protection by Pugnacious Bodyguards. Annual 

 75 

Review of Ecology and Systematics, 8, 407–427. 

Berdahl, A., Torney, C.J., Schertzer, E. & Levin, S.A. (2015) On the evolutionary interplay 
between dispersal and local adaptation in heterogeneous environments. Evolution, 69, 1390–
1405. 

Bijlsma, R. & Loeschcke, V. (2012) Genetic erosion impedes adaptive responses to stressful 
environments. Evolutionary Applications, 5, 117–129. 

Blondel, L., Klemet-N’guessan, S., Scott, M.E. & Hendry, A.P. (2020) Asymmetric isolation and 
the evolution of behaviors influencing dispersal: Rheotaxis of guppies above waterfalls. Genes, 
11. 

Bolnick, D.I. & Otto, S.P. (2013) The magnitude of local adaptation under genotype-dependent 
dispersal. Ecology and Evolution, 3, 4722–4735. 

Bona, S. De, Bruneaux, M., Lee, A.E.G., Reznick, D.N., Bentzen, P. & López-Sepulcre, A. 
(2019) Spatio-temporal dynamics of density-dependent dispersal during a population 
colonisation. Ecology Letters, 22, 634–644. 

Bonte, D. & Dahirel, M. (2017) Dispersal: a central and independent trait in life history. Oikos, 
126, 472–479. 

Bonte, D., Dyck, H. Van, Bullock, J.M., Coulon, A., Delgado, M., Gibbs, M., et al. (2012) Costs 
of dispersal. Biological Reviews, 87, 290–312. 

Borges, I.L., Dangerfield, J.C., Angeloni, L.M., Funk, W.C. & Fitzpatrick, S.W. (2022) 
Reproductive benefits associated with dispersal in headwater populations of Trinidadian guppies 
(Poecilia reticulata). Ecology Letters, 25, 344–354. 

Botham, R., Collin, C.L. & Ashman, T.L. (2009) Plant-mycorrhizal fungus interactions affect the 
expression of inbreeding depression in wild strawberry. International Journal of Plant Sciences, 
170, 143–150. 

Boutin, S., Sauvage, C., Bernatchez, L., Audet, C. & Derome, N. (2014) Inter individual 
variations of the fish skin microbiota: Host genetics basis of mutualism? PLoS ONE, 9, 1–17. 

Bowler, D.E. & Benton, T.G. (2005) Causes and consequences of animal dispersal strategies: 
Relating individual behaviour to spatial dynamics. Biological Reviews of the Cambridge 
Philosophical Society, 80, 205–225. 

Bragança, P.H.N. de, Guimarães, E.C., Brito, P.D. de & Uttum, F.P. (2020) On the natural 
occurrence of Poecilia reticulata Peters, 1859 (Cyprinodontiformes: Poeciliidae). Cybium, 44, 
309–316. 

Bronstein, J.L. (1998) The contribution of ant-plant protection studies to our understanding of 

 76 

mutualism. Biotropica, 30, 150–161. 

Brook, B.W., Tonkyn, D.W., O’Grady, J.J. & Frankham, R. (2002) Contribution of inbreeding to 
extinction risk in threatened species. Ecology and Society, 6. 

Brooker, L., Brooker, M. & Cale, P. (1999) Animal Dispersal in Fragmented Habitat: Measuring 
Habitat Connectivity, Corridor Use, and Dispersal Mortality. Conservation Ecology, 3. 

Brooks, R. & Endler, J.A. (2001) Direct and indirect sexual selection and quantitative genetics of 
male traits in guppies (Poecilia reticulata). Evolution, 55, 1002–1015. 

Budaev, S. V. (1997) “Personality” in the Guppy (Poecilia reticulata): A Correlational Study of 
Exploratory Behavior and Social Tendency. Journal of Comparative Psychology, 111, 399–411. 

Bueno, E., Kisha, T., Maki, S.L., Wettberg, E.J.B. Von & Singer, S. (2019) Genetic diversity of 
Chamaecrista fasciculata (Fabaceae) from the USDA germplasm collection. BMC Research 
Notes, 12, 1–7. 

Burghardt, L.T., Guhlin, J., Chun, C.L., Liu, J., Sadowsky, M.J., Stupar, R.M., et al. (2017) 
Transcriptomic basis of genome by genome variation in a legume-rhizobia mutualism. Molecular 
Ecology, 26, 6122–6135. 

Byers, D.L. & Waller, D.M. (1999) Do Plant Populations Purge Their Genetic Load? Effects of 
Population Size and Mating History on Inbreeding Depression. Annual Review of Ecology and 
Systematics, 30, 479–513. 

Calleri, D. V., Reid, E.M.G., Rosengaus, R.B., Vargo, E.L. & Traniello, J.F.A. (2006) 
Inbreeding and disease resistance in a social insect: Effects of heterozygosity on 
immunocompetence in the termite Zootermopsis angusticollis. Proceedings of the Royal Society 
B: Biological Sciences, 273, 2633–2640. 

Campbell, S.A., Thaler, J.S. & Kessler, A. (2013) Plant chemistry underlies herbivore-mediated 
inbreeding depression in nature. Ecology Letters, 16, 252–260. 

Cao, Y., Halane, M.K., Gassmann, W. & Stacey, G. (2017) The Role of Plant Innate Immunity 
in the Legume-Rhizobium Symbiosis. Annual Review of Plant Biology, 68, 535–561. 

Carr, D.E. & Eubanks, M.D. (2002) Inbreeding alters resistance to insect herbivory and host 
plant quality in Mimulus guttatus (Scrophulariaceae). Evolution, 56, 22–30. 

Carter, A.J., Feeney, W.E., Marshall, H.H., Cowlishaw, G. & Heinsohn, R. (2013) Animal 
personality: What are behavioural ecologists measuring? Biological Reviews, 88, 465–475. 

Charlesworth, D. & Charlesworth, B. (1987) Inbreeding Depression and its Evolutionary 
Consequences. Annual Review of Ecology and Systematics, 18, 237–268. 

 77 

Charlesworth, D. & Willis, J.H. (2009) The genetics of inbreeding depression. Nature Reviews 
Genetics, 10, 783–796. 

Cheptou, P.O. & Donohue, K. (2011) Environment-dependent inbreeding depression: Its 
ecological and evolutionary significance. New Phytologist, 189, 395–407. 

Cheptou, P.O., Imbert, E., Lepart, J. & Escarre, J. (2000) Effects of competition on lifetime 
estimates of inbreeding depression in the outcrossing plant Crepis sancta (Asteraceae). Journal 
of Evolutionary Biology, 13, 522–531. 

Clobert, J., Baguette, M., Benton, T.G. & Bullock, J.M. (2012) Dispersal ecology and evolution. 
Oxford University Press, Oxford, UK. 

Clobert, J., Galliard, J.F. Le, Cote, J., Meylan, S. & Massot, M. (2009) Informed dispersal, 
heterogeneity in animal dispersal syndromes and the dynamics of spatially structured 
populations. Ecology Letters, 12, 197–209. 

Colwell, R.K., Dunn, R.R. & Harris, N.C. (2012) Coextinction and persistence of dependent 
species in a changing world. Annual Review of Ecology, Evolution, and Systematics, 43, 183–
203. 

Cote, J., Bestion, E., Jacob, S., Travis, J., Legrand, D. & Baguette, M. (2017) Evolution of 
dispersal strategies and dispersal syndromes in fragmented landscapes. Ecography, 0–2. 

Cote, J., Clobert, J., Brodin, T., Fogarty, S. & Sih, A. (2010a) Personality-dependent dispersal: 
Characterization, ontogeny and consequences for spatially structured populations. Philosophical 
Transactions of the Royal Society B: Biological Sciences, 365, 4065–4076. 

Cote, J., Fogarty, S., Weinersmith, K., Brodin, T. & Sih, A. (2010b) Personality traits and 
dispersal tendency in the invasive mosquitofish (Gambusia affinis). Proceedings of the Royal 
Society B: Biological Sciences, 277, 1571–1579. 

Coulon, A., Fitzpatrick, J.W., Bowman, R. & Lovette, I.J. (2010) Effects of habitat 
fragmentation on effective dispersal of Florida scrub-jays. Conservation Biology, 24, 1080–1088. 

Crispo, E., Bentzen, P., Reznick, D.N., Kinnison, M.T. & Hendry, A.P. (2006) The relative 
influence of natural selection and geography on gene flow in guppies. Molecular Ecology, 15, 
49–62. 

Crispo, E., Moore, J.S., Lee-Yaw, J.A., Gray, S.M. & Haller, B.C. (2011) Broken barriers: 
Human-induced changes to gene flow and introgression in animals: An examination of the ways 
in which humans increase genetic exchange among populations and species and the 
consequences for biodiversity. BioEssays, 33, 508–518. 

Croft, D.P., Albanese, B., Arrowsmith, B.J., Botham, M., Webster, M. & Krause, J. (2003) Sex-
biased movement in the guppy (Poecilia reticulata). Oecologia, 137, 62–68. 

 78 

Crooks, K.R., Burdett, C.L., Theobald, D.M., Rondinini, C. & Boitani, L. (2011) Global patterns 
of fragmentation and connectivity of mammalian carnivore habitat. Philosophical Transactions 
of the Royal Society B: Biological Sciences, 366, 2642–2651. 

Daniel, M.J. & Rodd, F.H. (2016) Female guppies can recognize kin but only avoid incest when 
previously mated. Behavioral Ecology, 27, 55–61. 

Daniel, M.J. & Rodd, F.H. (2020) Kin recognition in guppies uses self-referencing based on 
olfactory cues. American Naturalist, 197. 

Daniels, S.J. & Walters, J.R. (2000) Inbreeding Depression and Its Effects on Natal Dispersal in 
Red-Cockaded Woodpeckers A. The Condor, 102, 482–491. 

Deacon, A.E., Ramnarine, I.W. & Magurran, A.E. (2011) How reproductive ecology contributes 
to the spread of a globally invasive fish. PLoS ONE, 6. 

Dial, T.R., Hernandez, L.P. & Brainerd, E.L. (2017) Morphological and functional maturity of 
the oral jaws covary with offspring size in Trinidadian guppies. Scientific Reports, 7, 1–10. 

Dobson, F.S. (1982) Competition for mater and predominant juvenile male dispersal in 
mammals. Animal Behaviour, 30, 1183–1192. 

Doekes, H.P., Bijma, P. & Windig, J.J. (2021) How depressing is inbreeding? A meta-analysis of 
30 years of research on the effects of inbreeding in livestock. Genes, 12. 

Doligez, B. & Pärt, T. (2008) Estimating fitness consequences of dispersal: A road to “know-
where”? Non-random dispersal and the underestimation of dispersers’ fitness. Journal of Animal 
Ecology, 77, 1199–1211. 

Dudareva, N., Klempien, A., Muhlemann, J.K. & Kaplan, I. (2013) Biosynthesis, function and 
metabolic engineering of plant volatile organic compounds. New Phytologist, 198, 16–32. 

Dunn, R.R., Harris, N.C., Colwell, R.K., Koh, L.P. & Sodhi, N.S. (2009a) The sixth mass 
coextinction: Are most endangered species parasites and mutualists? Proceedings of the Royal 
Society B: Biological Sciences, 276, 3037–3045. 

Dunn, R.R., Harris, N.C., Colwell, R.K., Koh, L.P. & Sodhi, N.S. (2009b) The sixth mass 
coextinction: Are most endangered species parasites and mutualists? Proceedings of the Royal 
Society B: Biological Sciences. 

Dyck, H. Van & Baguette, M. (2005) Dispersal behaviour in fragmented landscapes: Routine or 
special movements? Basic and Applied Ecology, 6, 535–545. 

Endler, J.A. (1980) Natural Selection on Color Patterns in Poecilia reticulata. Evolution, 34, 76–
91. 

 79 

Endler, J.A. (1987) Predation, light intensity and courtship behaviour in Poecilia reticulata 
(Pisces: Poeciliidae). Animal Behaviour, 35, 1376–1385. 

Endler, J.A. & Houde, A.E. (1995) Geographic Variation in Female Preferences for Male Traits 
in Poecilia reticulata. Evolution, 49, 456–468. 

Endlweber, K. & Scheu, S. (2006) Establishing arbuscular mycorrhiza-free soil: A comparison of 
six methods and their effects on nutrient mobilization. Applied Soil Ecology, 34, 276–279. 

Erickson, D.L. & Fenster, C.B. (2006) Intraspecific hybridization and the recovery of fitness in 
the native legume Chamaecrista fasciculata. Evolution, 60, 225–233. 

Fenster, C.B. (1991a) Gene flow in Chamaecrista fasciculata (Leguminosae). II. Gene 
establishment. Evolution, 45, 410–422. 

Fenster, C.B. (1991b) Effect of male pollen donor and female seed parent on allocation of 
resources to developing seeds and fruit in Chamaecrista fasciculata (Leguminosae). American 
Journal of Botany, 78, 13–23. 

Fenster, C.B. & Galloway, L.F. (2000) Inbreeding and outbreeding depression in natural 
populations of Chamaecrista fasciculata (Fabaceae). Conservation Biology, 14, 1406–1412. 

Fenster, C.B., Vekemans, X. & Hardy, O.J. (2003) Quantifying gene flow from spatial genetic 
structure data in a metapopulation of Chamaecrista fasciculata (Leguminosae). Evolution, 57, 
995–1007. 

Ferrari, M.J., Stephenson, A.G., Mescher, M.C. & Moraes, C.M. De. (2006) Inbreeding effects 
on blossom volatiles in Cucurbita pepo subsp. texana (Cucurbitaceae). American Journal of 
Botany, 93, 1768–1774. 

Fitzpatrick, S.W., Bradburd, G.S., Kremer, C.T., Salerno, P.E., Angeloni, L.M. & Funk, W.C. 
(2020) Genomic and Fitness Consequences of Genetic Rescue in Wild Populations. Current 
Biology, 30, 1–6. 

Fitzpatrick, S.W., Gerberich, J.C., Angeloni, L.M., Bailey, L.L., Broder, E.D., Torres-Dowdall, 
J., et al. (2016) Gene flow from an adaptively divergent source causes rescue through genetic 
and demographic factors in two wild populations of Trinidadian guppies. Evolutionary 
Applications, 9, 879–891. 

Fitzpatrick, S.W., Gerberich, J.C., Kronenberger, J.A., Angeloni, L.M. & Funk, W.C. (2015) 
Locally adapted traits maintained in the face of high gene flow. Ecology Letters, 18, 37–47. 

Fitzpatrick, S.W., Torres-Dowdall, J., Reznick, D.N., Ghalambor, C.K. & Funk, W.C. (2014) 
Parallelism Isn’t Perfect: Could Disease and Flooding Drive a Life-History Anomaly in 
Trinidadian Guppies? The American Naturalist, 183, 290–300. 

 80 

Fletcher, R.J. (2006) Emergent properties of conspecific attraction in fragmented landscapes. 
American Naturalist, 168, 207–219. 

Fletcher, R.J., Reichert, B.E. & Holmes, K. (2018) The negative effects of habitat fragmentation 
operate at the scale of dispersal. Ecology, 99, 2176–2186. 

Forister, M.L., McCall, A.C., Sanders, N.J., Fordyce, J.A., Thorne, J.H., O’Brien, J., et al. (2010) 
Compounded effects of climate change and habitat alteration shift patterns of butterfly diversity. 
Proceedings of the National Academy of Sciences of the United States of America, 107, 2088–
2092. 

Fowler, K. & Whitlock, M.C. (1999) The variance in inbreeding depression and the recovery of 
fitness in bottlenecked populations. Proceedings of the Royal Society B: Biological Sciences, 
266, 2061–2066. 

Fox, C.W. & Reed, D.H. (2011) Inbreeding depression increases with environmental stress: An 
experimental study and meta-analysis. Evolution, 65, 246–258. 

Fraser, B.A., Künstner, A., Reznick, D.N., Dreyer, C. & Weigel, D. (2015) Population genomics 
of natural and experimental populations of guppies (Poecilia reticulata). Molecular Ecology, 24, 
389–408. 

Gibson, A.K. & Nguyen, A.E. (2020) Does genetic diversity protect host populations from 
parasites ? A meta-analysis across natural and agricultural systems. Evolution Letters, 1–17. 

Glémin, S. (2003) How are deleterious mutations purged? Drift versus nonrandom mating. 
Evolution, 57, 2678–2687. 

Gorton, A.J., Heath, K.D., Pilet-Nayel, M.L., Baranger, A. & Stinchcombe, J.R. (2012) Mapping 
the genetic basis of symbiotic variation in legume-rhizobium interactions in Medicago 
Truncatula. G3: Genes, Genomes, Genetics, 2, 1291–1303. 

Gosline, A.K. & Rodd, F.H. (2008) Predator-induced plasticity in guppy (Poecilia reticulata) life 
history traits. Aquatic Ecology, 42, 693–699. 

Greenwood, P.J. (1980) Mating systems, philopatry and dispersal in birds and mammals. Animal 
Behaviour, 28, 1140–1162. 

Guimarães, C.D.D.C., Viana, J.P.R. & Cornelissen, T.G. (2014) A Meta-Analysis of the Effects 
of Fragmentation on Herbivorous Insects. Environmental Entomology, 43, 537–545. 

Haag, C.R., Sakwińska, O. & Ebert, D. (2003) Test of synergistic interaction between infection 
and inbreeding in Daphnia magna. Evolution, 57, 777–783. 

Haddad, N.M., Brudvig, L.A., Clobert, J., Davies, K.F., Gonzalez, A., Holt, R.D., et al. (2015a) 
Habitat fragmentation and its lasting impact on Earth’s ecosystems. Science Advances, 1, 1–10. 

 81 

Haddad, N.M., Brudvig, L.A., Clobert, J., Davies, K.F., Gonzalez, A., Holt, R.D., et al. (2015b) 
Applied Ecology: Habitat fragmentation and its lasting impact on Earth’ s ecosystems. Science 
Advances, 1, 1–9. 

Hamilton, W.D. (1977) Dispersal in stable habitats. Nature, 269, 578–581. 

Hampton, K.J., Hughes, K.A. & Houde, A.E. (2009) The allure of the distinctive: Reduced 
sexual responsiveness of female guppies to “redundant” male colour patterns. Ethology, 115, 
475–481. 

Hanski, I. (1998) Metapopulation dynamics. Nature, 396, 41–49. 

Harper, L.H. (1989) The persistence of ant-following birds in small amazonian forest fragments. 
Acta Amazonica, 19, 249–263. 

Harris, S., Ramnarine, I.W., Smith, H.G. & Pettersson, L.B. (2010) Picking personalities apart: 
Estimating the influence of predation, sex and body size on boldness in the guppy Poecilia 
reticulata. Oikos, 119, 1711–1718. 

Hartig, F. (2021) DHARMa: Residual Diagnostics for Hierarchical (Multi-Level / Mixed) 
Regression Models. R package version 0.4.1. https://CRAN.R-project.org/package=DHARMa. 

Haskins, C.P. & Haskins, E.F. (1949) The Role of Sexual Selection as an Isolating Mechanism in 
Three Species of Poeciliid Fishes. Evolution, 3, 160. 

Haskins, C.P. & Haskins, E.F. (1950) Factors Governing Sexual Selection as an Isolating 
Mechanism in the Poeciliid Fish Lebistes Reticulatus. Proceedings of the National Academy of 
Sciences, 36, 464–476. 

Haskins, C.P. & Haskins, E.F. (1951) The Inheritance of Certain Color Patterns in Wild 
Populations of Lebistes reticulatus in Trinidad. Evolution, 5, 216. 

Haskins, C.P., Haskins, E.F., McLaughlin, J.J.A. & Hewitt, R.E. (1961) Polymorphism and 
population structure in Lebistes reticulatus, an ecological study. In Vertebrate Speciation (ed. by 
Blair, W.F.). University of Texas Press, Austin, pp. 320–395. 

Heath, K.D., Stock, A.J. & Stinchcombe, J.R. (2010) Mutualism variation in the nodulation 
response to nitrate. Journal of Evolutionary Biology, 23, 2494–2500. 

Heath, K.D. & Tiffin, P. (2009) Stabilizing mechanisms in a legume-rhizobium mutualism. 
Evolution, 63, 652–662. 

Hegland, S.J., Nielsen, A., Lázaro, A., Bjerknes, A.L. & Totland, Ø. (2009) How does climate 
warming affect plant-pollinator interactions? Ecology Letters, 12, 184–195. 

Heil, M. (2008) Indirect defence via tritrophic interactions. New Phytologist, 178, 41–61. 

 82 

Hoffmann, A.A., Sgrò, C.M. & Kristensen, T.N. (2017) Revisiting Adaptive Potential, 
Population Size, and Conservation. Trends in Ecology and Evolution, 32, 506–517. 

Holt, R.D. (2003) On the evolutionary ecology of species’ ranges. Evolutionary Ecology 
Research, 5, 159–178. 

Honnay, O. & Jacquemyn, H. (2007) Susceptibility of common and rare plant species to the 
genetic consequences of habitat fragmentation. Conservation Biology, 21, 823–831. 

Houde, A.E. (1987) Mate choice based upon naturally occurring color-pattern variation in a 
guppy population. Evolution, 41, 1–10. 

Houde, A.E. (1992) Sex-linked heritability of a sexually selected character in a natural 
population of poecilia reticulata (Pisces: Poeciliidae) (guppies). Heredity, 69, 229–235. 

Huang, X., Chaparro, J.M., Reardon, K.F., Zhang, R., Shen, Q. & Vivanco, J.M. (2014) 
Rhizosphere interactions: root exudates, microbes, and microbial communities 1, 275, 267–275. 

Hughes, K.A., Du, L. & Reznick, D.N. (1999) Familiarity leads to female mate preference for 
novel males in the guppy, Poecilia reticulata. Animal Behaviour, 58, 907–916. 

Hughes, K.A., Houde, A.E., Price, A.C. & Rodd, F.H. (2013a) Mating advantage for rare males 
in wild guppy populations. Nature, 503, 108–110. 

Hughes, K.A., Houde, A.E., Price, A.C. & Rodd, F.H. (2013b) Mating advantage for rare males 
in wild guppy populations. Nature, 503, 108–110. 

James, T.Y., Toledo, L.F., Rödder, D., Silva Leite, D. da, Belasen, A.M., Betancourt-Román, 
C.M., et al. (2015) Disentangling host, pathogen, and environmental determinants of a recently 
emerged wildlife disease: Lessons from the first 15 years of amphibian chytridiomycosis 
research. Ecology and Evolution, 5, 4079–4097. 

Jirotkul, M. (1999) Operational sex ratio influences female preference and male-male 
competition in guppies. Animal Behaviour, 58, 287–294. 

Johnson, A.M., Chappell, G., Price, A.C., Helen Rodd, F., Olendorf, R. & Hughes, K.A. (2010) 
Inbreeding Depression and Inbreeding Avoidance in a Natural Population of Guppies (Poecilia 
reticulata). Ethology, 116, 448–457. 

Johnson, J.B. & Omland, K.S. (2004) Model selection in ecology and evolution. Trends in 
Ecology and Evolution, 19, 101–108. 

Jorge, M.L.S.P. & Howe, H.F. (2009) Can forest fragmentation disrupt a conditional mutualism? 
A case from central Amazon. Oecologia, 161, 709–718. 

Kaiser, M.I. & Müller, C. (2021) What is an animal personality? Biology and Philosophy, 36, 1–

 83 

25. 

Kalske, A., Mutikainen, P., Muola, A., Scheepens, J.F., Laukkanen, L., Salminen, J.P., et al. 
(2014) Simultaneous inbreeding modifies inbreeding depression in a plant-herbivore interaction. 
Ecology Letters, 17, 229–238. 

Kardos, M., Taylor, H.R., Ellegren, H., Luikart, G. & Allendorf, F.W. (2016) Genomics 
advances the study of inbreeding depression in the wild. Evolutionary Applications, 9, 1205–
1218. 

Kariyat, R.R., Mauck, K.E., Moraes, C.M. De, Stephenson, A.G. & Mescher, M.C. (2012) 
Inbreeding alters volatile signalling phenotypes and influences tri-trophic interactions in 
horsenettle (Solanum carolinense L.). Ecology Letters, 15, 301–309. 

Kariyat, R.R. & Stephenson, A.G. (2019) Inbreeding depression: it’s not just for population 
biologists. American Journal of Botany, 106, 331–333. 

Kawecki, T.J. & Holt, R.D. (2002) Evolutionary Consequences of Asymmetric Dispersal Rates. 
American Naturalist, 160, 333–347. 

Keller, K.R. & Lau, J.A. (2018) When mutualisms matter: Rhizobia effects on plant 
communities depend on host plant population and soil nitrogen availability. Journal of Ecology, 
106, 1046–1056. 

Keller, L.F., Grant, P.R., Grant, B.R. & Petren, K. (2002) Environmental Conditions Affect the 
Magnitude of Inbreeding Depression in Survival of Darwin’s Finches. Evolution, 56, 1229–1239. 

Keller, L.F. & Waller, D.M. (2002) Inbreeding effects in wild populations. Trends in Ecology 
and Evolution, 17, 230–241. 

Kelly, C.A. (1986) Extrafloral nectaries: ants, herbivores and fecundity in Cassia fasciculata. 
Oecologia, 58, 2657–2668. 

Kemp, D.J., Reznick, D.N., Grether, G.F. & Endler, J.A. (2009) Predicting the direction of 
ornament evolution in Trinidadian guppies (Poecilia reticulata). Proceedings of the Royal Society 
B: Biological Sciences, 276, 4335–4343. 

Kieffer, J.D. (2000) Limits to exhaustive exercise in fish. Comparative Biochemistry and 
Physiology - A Molecular and Integrative Physiology, 126, 161–179. 

Kiers, E.T., Palmer, T.M., Ives, A.R., Bruno, J.F. & Bronstein, J.L. (2010) Mutualisms in a 
changing world: An evolutionary perspective. Ecology Letters, 13, 1459–1474. 

Kiesewetter, K.N. & Afkhami, M.E. (2021) Microbiome-mediated effects of habitat 
fragmentation on native plant performance. New Phytologist, 232, 1823–1838. 

 84 

Kniel, N. & Godin, J.G.J. (2019) Characterizing the (co)variance of personality traits in female 
Trinidadian guppies (Poecilia reticulata). Environmental Biology of Fishes, 102, 1351–1363. 

Knowlton, N. (2001) The future of coral reefs. Proceedings of the National Academy of Sciences 
of the United States of America, 98, 5419–5425. 

Kodric-Brown, A. & Brown, J.H. (1984) Truth in Advertising : The Kinds of Traits Favored by 
Sexual Selection. The American Naturalist, 124, 309–323. 

Kokko, H. & López-Sepulcre, A. (2006) From individual dispersal to species ranges: 
Perspectives for a changing world. Science, 313, 789–791. 

Koricheva, J. & Romero, G.Q. (2012) You get what you pay for: Reward-specific trade-offs 
among direct and ant-mediated defences in plants. Biology Letters, 8, 628–630. 

Kost, C. & Heil, M. (2008) The defensive role of volatile emission and extrafloral nectar 
secretion for lima bean in nature. Journal of Chemical Ecology, 34, 2–13. 

Kubisch, A., Holt, R.D., Poethke, H.J. & Fronhofer, E.A. (2014) Where am I and why? 
Synthesizing range biology and the eco-evolutionary dynamics of dispersal. Oikos, 123, 5–22. 

Lande, R. (1976) Natural Selection and Random Genetic Drift in Phenotypic Evolution. 
Evolution, 30, 314. 

Lawrence, W.S. (1987) Dispersal: an alternative mating tactic conditional on sex ratio and body 
size. Behavioral Ecology and Sociobiology, 21, 367–373. 

Leal, L.C. & Peixoto, P.E.C. (2017) Decreasing water availability across the globe improves the 
effectiveness of protective ant–plant mutualisms: a meta-analysis. Biological Reviews, 92, 1785–
1794. 

Lehmann, L. & Perrin, N. (2003) Inbreeding Avoidance through Kin Recognition: Choosy 
Females Boost Male Dispersal. American Naturalist, 162, 638–652. 

Liao, W. & Reed, D.H. (2009) Inbreeding-environment interactions increase extinction risk. 
Animal Conservation, 12, 54–61. 

Liley, N.R. (1966) Ethological Isolating Mechanisms in Four Sympatric Species of Poeciliid 
Fishes Author ( s ): N . R . Liley Source : Behaviour . Supplement , No . 14 , Ethological 
Isolating Mechanisms in Four Sympatric Species of Poeciliid Fishes ( 1965 ), pp . III-VII ,. 
Behaviour. Supplement, 1–197. 

Loekle, D.M., Madison, D.M. & Christian, J.J. (1982) Time dependency and kin recognition of 
cannibalistic behavior among poeciliid fishes. Behavioral and Neural Biology, 35, 315–318. 

MacArthur, R.H. & Wilson, E.O. (1963) An Equilibrium Theory of Insular Zoogeography. 

 85 

Evolution, 17, 373. 

Magnoli, S.M. & Lau, J.A. (2020) Novel plant–microbe interactions: Rapid evolution of a 
legume–rhizobium mutualism in restored prairies. Journal of Ecology, 108, 1241–1249. 

Magurran, A.E. (2005) Evolutionary ecology: the Trinidadian guppy. Oxford Series in Ecology 
and Evolution. Oxford University Press, Oxford, UK. 

Magurran, A.E. & Seghers, B.H. (1990) Population Differences in the Schooling Behaviour of 
Newborn Guppies, Poecilia reticulata. Ethology, 84, 334–342. 

Magurran, A.E. & Seghers, B.H. (1991) Variation in Schooling and Aggression Amongst Guppy 
(Poecilia reticulata) Populations in Trinidad. Behaviour, 118, 214–234. 

Magurran, A.E. & Seghers, B.H. (1994) A cost of sexual harassment in the guppy, Poecilia 
reticulata. Proceedings of the Royal Society B: Biological Sciences, 258, 89–92. 

Magurran, A.E., Seghers, B.H., Carvalho, G.R. & Shaw, P.W. (1993) Evolution of adaptive 
variation in antipredator behaviour. Marine Behaviour and Physiology, 23, 29–44. 

Mannouris, C. & Byers, D.L. (2013) The impact of habitat fragmentation on fitness-related traits 
in a native prairie plant, Chamaecrista fasciculata (Fabaceae). Biological Journal of the Linnean 
Society, 108, 55–67. 

Marazzi, B., Bronstein, J.L. & Koptur, S. (2013) The diversity, ecology and evolution of 
extrafloral nectaries: Current perspectives and future challenges. Annals of Botany, 111, 1243–
1250. 

Marcionetti, A., Rossier, V., Roux, N., Salis, P., Laudet, V. & Salamin, N. (2019) Insights into 
the genomics of clownfish adaptive radiation: Genetic basis of the mutualism with sea anemones. 
Genome Biology and Evolution, 11, 869–882. 

Masson-Boivin, C. & Sachs, J.L. (2018) Symbiotic nitrogen fixation by rhizobia — the roots of a 
success story. Current Opinion in Plant Biology, 44, 7–15. 

McCartney-Melstad, E., Vu, J.K. & Shaffer, H.B. (2018) Genomic data recover previously 
undetectable fragmentation effects in an endangered amphibian. Molecular Ecology, 27, 4430–
4443. 

McPeek, M.A. & Holt, R.D. (1992) The Evolution of Dispersal in Spatially and Temporally 
Varying Environments. The American Naturalist, 140, 1010–1027. 

Meagher, S., Penn, D.J. & Potts, W.K. (2000) Male-male competition magnifies inbreeding 
depression in wild house mice. Proceedings of the National Academy of Sciences of the United 
States of America, 97, 3324–3329. 

 86 

Meylan, S., Belliure, J., Clobert, J. & Fraipont, M. De. (2002) Stress and body condition as 
prenatal and postnatal determinants of dispersal in the common lizard (Lacerta vivipara). 
Hormones and Behavior, 42, 319–326. 

Mitchell, M.G.E., Suarez-Castro, A.F., Martinez-Harms, M., Maron, M., McAlpine, C., Gaston, 
K.J., et al. (2015) Reframing landscape fragmentation’s effects on ecosystem services. Trends in 
Ecology and Evolution, 30, 190–198. 

Mittelbach, G.G. & Schemske, D.W. (2015) Ecological and evolutionary perspectives on 
community assembly. Trends in Ecology and Evolution, 30, 241–247. 

Mondor, E.B. & Addicott, J.F. (2003) Conspicuous extra-floral nectaries are inducible in Vicia 
faba. Ecology Letters, 6, 495–497. 

Mony, C., Uroy, L., Khalfallah, F., Haddad, N. & Vandenkoornhuyse, P. (2022) Landscape 
connectivity for the invisibles. Ecography, 2022, 1–14. 

Morales, J.M., Moorcroft, P.R., Matthiopoulos, J., Frair, J.L., Kie, J.G., Powell, R.A., et al. 
(2010) Building the bridge between animal movement and population dynamics. Philosophical 
Transactions of the Royal Society B: Biological Sciences, 365, 2289–2301. 

Mosier, A.R. (1998) Soil processes and global change. Biology and Fertility of Soils, 27, 221–
229. 

Murphy, S.M., Battocletti, A.H., Tinghitella, R.M., Wimp, G.M. & Ries, L. (2016) Complex 
community and evolutionary responses to habitat fragmentation and habitat edges : what can we 
learn from insect science ? Current Opinion in Insect Science, 14, 61–65. 

Narula, N., Kothe, E. & Behl, R.K. (2009) Role of root exudates in plant-microbe interactions. 
Journal of Applied Botany and Food Quality, 82, 122–130. 

Nascimento, E.A. do & Del-Claro, K. (2010) Ant visitation to extrafloral nectaries decreases 
herbivory and increases fruit set in Chamaecrista debilis (Fabaceae) in a Neotropical savanna. 
Flora: Morphology, Distribution, Functional Ecology of Plants, 205, 754–756. 

Ness, J.H. (2003) Catalpa bignonioides alters extrafloral nectar production after herbivory and 
attracts ant bodyguards. Oecologia, 134, 210–218. 

Niebuhr, B.B.S., Wosniack, M.E., Santos, M.C., Raposo, E.P., Viswanathan, G.M., Luz, M.G.E. 
Da, et al. (2015) Survival in patchy landscapes: The interplay between dispersal, habitat loss and 
fragmentation. Scientific Reports, 5, 1–10. 

Nobarinezhad, M.H. & Wallace, L.E. (2020) Fine-scale patterns of genetic structure in the host 
plant Chamaecrista Fasciculata (Fabaceae) and its nodulating rhizobia symbionts. Plants, 9, 1–
20. 

 87 

O’Grady, J.J., Brook, B.W., Reed, D.H., Ballou, J.D., Tonkyn, D.W. & Frankham, R. (2006) 
Realistic levels of inbreeding depression strongly affect extinction risk in wild populations. 
Biological Conservation, 133, 42–51. 

Oosterhout, C. Van, Trigg, R.E., Carvalho, G.R., Magurran, A.E., Hauser, L. & Shaw, P.W. 
(2003) Inbreeding depression and genetic load of sexually selected traits: How the guppy lost its 
spots. Journal of Evolutionary Biology, 16, 273–281. 

Paige, K.N. (2010) The functional genomics of inbreeding depression: A new approach to an old 
problem. BioScience, 60, 267–277. 

Peniston, J.H., Backus, G.A., Baskett, M.L., Fletcher, R.J. & Holt, R.D. (2023) Ecological and 
evolutionary consequences of temporal variation in dispersal. Ecography, 1–20. 

Peniston, J.H., Barfield, M. & Holt, R.D. (2019) Pulsed Immigration Events Can Facilitate 
Adaptation to Harsh Sink Environments. The American Naturalist, 194, 000–000. 

Pérez-González, J. & Carranza, J. (2009) Female-biased dispersal under conditions of low male 
mating competition in a polygynous mammal. Molecular Ecology, 18, 4617–4630. 

Perrin, N. & Mazalov, V. (1999) Dispersal and Inbreeding Avoidance. The American Naturalist, 
154, 282–292. 

Perrin, N. & Mazalov, V. (2000) Local competition, inbreeding, and the evolution of sex-biased 
dispersal. American Naturalist, 155, 116–127. 

Pettersson, L.B., Ramnarine, I.W., Becher, S.A., Mahabir, R. & Magurran, A.E. (2004) Sex ratio 
dynamics and fluctuating selection pressures in natural populations of the Trinidadian guppy, 
Poecilia reticulata. Behavioral Ecology and Sociobiology, 55, 461–468. 

Pitcher, T.E., Rodd, F.H. & Rowe, L. (2008) Female choice and the relatedness of mates in the 
guppy (Poecilia reticulata): Mate choice and inbreeding depression. Genetica, 134, 137–146. 

Potter, T., King, L., Travis, J. & Bassar, R.D. (2018) Competitive asymmetry and local 
adaptation in Trinidadian guppies. Journal of Animal Ecology, 0–3. 

Pringle, E.G. & Gordon, D.M. (2013) Protection mutualisms and the community: Geographic 
variation in an ant-plant symbiosis and the consequences for herbivores. Sociobiology, 60, 242–
251. 

Pusey, A. & Wolf, M. (1996) Inbreeding avoidance in animals. Trends in Ecology and Evolution, 
11, 201–206. 

Pyke, G.H. & Ren, Z.X. (2023) Floral nectar production: what cost to a plant? Biological 
Reviews, 2. 

 88 

R Development Core Team. (n.d.) R: A language and environment for statistical computing. R 
Foundation for Statistical Computing, Vienna, Austria. 2019. 

Radinger, J. & Wolter, C. (2014) Patterns and predictors of fish dispersal in rivers. Fish and 
Fisheries, 15, 456–473. 

Réale, D., Dingemanse, N.J., Kazem, A.J.N. & Wright, J. (2010) Evolutionary and ecological 
approaches to the study of personality. Philosophical Transactions of the Royal Society B: 
Biological Sciences, 365, 3937–3946. 

Reed, D.H., Fox, C.W., Enders, L.S. & Kristensen, T.N. (2012) Inbreeding-stress interactions: 
Evolutionary and conservation consequences. Annals of the New York Academy of Sciences, 
1256, 33–48. 

Reed, D.H., Lowe, E.H., Briscoe, D.A. & Frankham, R. (2003) Fitness and adaptation in a novel 
environment: Effect of inbreeding, prior environment, and lineage. Evolution, 57, 1822–1828. 

Rehling, F., Matthies, D. & Sandner, T.M. (2019) Responses of a legume to inbreeding and the 
intensity of novel and familiar stresses. Ecology and Evolution, 9, 1255–1267. 

Reid, J.M., Arcese, P. & Keller, L.F. (2003) Inbreeding depresses immune response in song 
sparrows (Melospiza melodia): Direct and inter-generational effects. Proceedings of the Royal 
Society B: Biological Sciences, 270, 2151–2157. 

Reid, J.M., Arcese, P., Keller, L.F., Elliott, K.H., Sampson, L. & Hasselquist, D. (2007) 
Inbreeding effects on immune response in free-living song sparrows (Melospiza melodia). 
Proceedings of the Royal Society B: Biological Sciences, 274, 697–706. 

Renner, S.S. & Zohner, C.M. (2018) Climate change and phenological mismatch in trophic 
interactions among plants, insects, and vertebrates. Annual Review of Ecology, Evolution, and 
Systematics, 49, 165–182. 

Reznick, D.N. (1989) Life-history evolution in guppies: 2. Repeatability of field observations 
and the effects of season on life histories. Evolution, 43, 1285–1297. 

Reznick, D.N., Bassar, R.D., Travis, J. & Helen Rodd, F. (2012) Life-history evolution in 
guppies VIII: the demographics of density regulation in guppies (poecilia reticulata). Evolution, 
66, 2903–2915. 

Reznick, D.N., Butler IV, M.J. & Rodd, H. (2001) Life-history evolution in guppies. VII. The 
comparative ecology of high- and low-predation environments. American Naturalist, 157, 126–
140. 

Reznick, D.N., IV, M.J.B., Rodd, F.H. & Ross, P. (1996) Life-History Evolution in Guppies 
(Poecilia reticulata) 6. Differential Mortality as a Mechanism for Natural Selection. Evolution, 
50, 1651. 

 89 

Richardson, J. & Smiseth, P.T. (2017) Intraspecific competition and inbreeding depression: 
Increased competitive effort by inbred males is costly to outbred opponents. American 
Naturalist, 189, 539–548. 

Riitters, K., Wickham, J., Neill, R.O., Jones, B. & Smith, E. (2000) Global-Scale Patterns of 
Forest Fragmentation. Conservation Ecology, 4, 1–24. 

Rodd, F.H. & Reznick, D.N. (1997) Variation in the demography of guppy populations: The 
importance of predation and life histories. Ecology, 78, 405–418. 

Rodríguez-Cabal, M.A., Aizen, M.A. & Novaro, A.J. (2007) Habitat fragmentation disrupts a 
plant-disperser mutualism in the temperate forest of South America. Biological Conservation, 
139, 195–202. 

Ronce, O. (2007) How Does It Feel to Be Like a Rolling Stone? Ten Questions About Dispersal 
Evolution. Annual Review of Ecology, Evolution, and Systematics, 38, 231–253. 

Ross-Gillespie, A., O’Riain, M.J. & Keller, L.F. (2007) Viral epizootic reveals inbreeding 
depression in a habitually inbreeding mammal. Evolution, 61, 2268–2273. 

Rudgers, J.A., Afkhami, M.E., Bell-Dereske, L., Chung, Y.A., Crawford, K.M., Kivlin, S.N., et 
al. (2020) Climate Disruption of Plant-Microbe Interactions. Annual Review of Ecology, 
Evolution, and Systematics, 51, 561–586. 

Rutter, M.T. & Rausher, M.D. (2004) Natural selection on extrafloral nectar production in 
Chamaecrista fasciculata: The costs and benefits of a mutualism trait. Evolution, 58, 2657–2668. 

Sachs, J.L. & Simms, E.L. (2006) Pathways to mutualism breakdown. Trends in Ecology and 
Evolution, 21, 585–592. 

Samson, F. & Knopf, F. (1994) Prairie conservation in North America. BioScience, 44, 418–421. 

Sardans, J. & Peñuelas, J. (2012) The role of plants in the effects of global change on nutrient 
availability and stoichiometry in the plant-soil system. Plant Physiology, 160, 1741–1761. 

Schmitt, J. & Ehrhardt, D.W. (1990) Enhancement of inbreeding depression by dominance and 
suppression in Impatiens capensis. Evolution, 44, 269–278. 

Schtickzelle, N., Mennechez, G.G. & Baguette, M. (2006) Dispersal depression with habitat 
fragmentation in the bog fritillary butterfly. Ecology, 87, 1057–1065. 

Sexton, J.P., McIntyre, P.J., Angert, A.L. & Rice, K.J. (2009) Evolution and Ecology of Species 
Range Limits. Annual Review of Ecology, Evolution, and Systematics, 40, 415–436. 

Slatkin, M. (1987) Gene Flow and the Geographic Structure of Natural Populations. Science, 
236, 787–792. 

 90 

Smallbone, W., Oosterhout, C. van & Cable, J. (2016) The effects of inbreeding on disease 
susceptibility: Gyrodactylus turnbulli infection of guppies, Poecilia reticulata. Experimental 
Parasitology, 167, 32–37. 

Smith, K.F., Acevedo-Whitehouse, K. & Pedersen, A.B. (2009) The role of infectious diseases in 
biological conservation. Animal Conservation, 12, 1–12. 

Smith, P., House, J.I., Bustamante, M., Sobocká, J., Harper, R., Pan, G., et al. (2016) Global 
change pressures on soils from land use and management. Global Change Biology, 22, 1008–
1028. 

Soons, M.B., Messelink, J.H., Jongejans, E. & Heil, G.W. (2005) Habitat fragmentation reduces 
grassland connectivity for both short-distance and long-distance wind-dispersed forbs. Journal of 
Ecology, 93, 1214–1225. 

Sork, V.L. & Schemske, D.W. (1992) Fitness Consequences of Mixed-Donor Pollen Loads in 
the Annual Legume Chamaecrista fasciculata. American Journal of Botany, 79, 508–515. 

Spiegel, O., Leu, S.T., Bull, C.M. & Sih, A. (2017) What’s your move? Movement as a link 
between personality and spatial dynamics in animal populations. Ecology Letters, 20, 3–18. 

Spielman, D., Brook, B.W. & Frankham, R. (2004) Most species are not driven to extinction 
before genetic factors impact them. Proceedings of the National Academy of Sciences, 101, 
15261–15264. 

Spigler, R.B., Theodorou, K. & Chang, S.M. (2017) Inbreeding depression and drift load in 
small populations at demographic disequilibrium. Evolution, 71, 81–94. 

Springer, A.L., Messina, F.J. & Gompert, Z. (2020) Measuring the effect of environmental stress 
on inbreeding depression alone obscures the relative importance of inbreeding–stress interactions 
on overall fitness in Callosobruchus maculatus. Evolutionary Applications, 13, 2597–2609. 

Staudinger, C., Mehmeti-Tershani, V., Gil-Quintana, E., Gonzalez, E.M., Hofhansl, F., 
Bachmann, G., et al. (2016) Evidence for a rhizobia-induced drought stress response strategy in 
Medicago truncatula. Journal of Proteomics, 136, 202–213. 

Stevens, L., Yan, G. & Pray, L.A. (1997) Consequences of Inbreeding on Invertebrate Host 
Susceptibility to Parasitic Infection. Evolution, 51, 2032–2039. 

Taylor, P.D. (1988) An inclusive fitness model for dispersal of offspring. Journal of Theoretical 
Biology, 130, 363–378. 

Travis, J., Bassar, R.D., Coulson, T., Lopez-Sepulcre, A. & Reznick, D. (2023) Population 
Regulation and Density-Dependent Demography in the Trinidadian Guppy. American Naturalist, 
202, 413–432. 

 91 

Travis, J., Reznick, D., Bassar, R.D., López-Sepulcre, A., Ferriere, R. & Coulson, T. (2014) Do 
eco-evo feedbacks help us understand nature? Answers from studies of the Trinidadian guppy. 
Advances in Ecological Research. 

Trochet, A., Courtois, E.A., Stevens, V.M., Baguette, M., Chaine, A., Schmeller, D.S., et al. 
(2016) Evolution of sex-biased dispersal. The Quarterly Review of Biology, 91, 297–320. 

Urbas, P., Araújo, M.V.J., Leal, I.R. & Wirth, R. (2007) Cutting More from Cut Forests : Edge 
Effects on Foraging and Herbivory of Leaf-Cutting Ants in Brazil. Biotropica, 39, 489–495. 

Venable, D.L. & Brown, J.S. (1988) The Selective Interactions of Dispersal , Dormancy , and 
Seed Size as Adaptations for Reducing Risk in Variable Environments. The American Naturalist, 
131, 360–384. 

Vittecoq, M., Djiéto-Lordon, C., McKey, D. & Blatrix, R. (2012) Range expansion induces 
variation in a behavioural trait in an ant-plant mutualism. Acta Oecologica, 38, 84–88. 

Watson, J.R., Kendall, B.E., Siegel, D.A. & Mitarai, S. (2012) Changing seascapes, stochastic 
connectivity, and marine metapopulation dynamics. American Naturalist, 180, 99–112. 

Weese, D.J., Schwartz, A.K., Bentzen, P., Hendry, A.P. & Kinnison, M.T. (2011) Eco-
evolutionary effects on population recovery following catastrophic disturbance. Evolutionary 
Applications, 4, 354–366. 

Willi, Y., Buskirk, J. Van & Hoffmann, A.A. (2006) Limits to the adaptive potential of small 
populations. Annual Review of Ecology, Evolution, and Systematics, 37, 433–458. 

Wood, C.W., Pilkington, B., Vaidya, P., Biel, C. & Stinchcombe, J.R. (2018) Genetic conflict 
with a parasitic nematode disrupts the legume-rhizobia mutualism. Evolution Letters, 213876. 

Wright, S. (1966) Polyallelic random drift in relation to evolution. Proceedings of the National 
Academy of Sciences of the United States of America, 55, 1074–1081. 

Xu, P. & Wang, E. (2023) Diversity and regulation of symbiotic nitrogen fixation in plants. 
Current Biology. 

Zahavi, A. (1977) The cost of honesty (Further Remarks on the Handicap Principle). Journal of 
Theoretical Biology, 67, 603–605. 

Zajitschek, S.R.K. & Brooks, R.C. (2008) Distinguishing the effects of familiarity, relatedness, 
and color pattern rarity on attractiveness and measuring their effects on sexual selection in 
guppies (Poecilia reticulata). American Naturalist, 172, 843–854. 

 92 

 
 
 
APPENDIX A: 

Supporting information: Reproductive benefits associated with dispersal in headwater 
populations of Trinidadian guppies (Poecilia reticulata) 

Supplementary Appendix 2.I. 

We explored whether differences between the metrics of cumulative lifetime distance moved 

(“total distance” in the manuscript) and absolute distance moved (first vs. last capture locations) 

affected reproduction. To do so, we calculated absolute distance as the minimum distance 

between the pools where individuals were first and last captured in our study. We note that this 

metric cannot be interpreted as the distance between locations of birth and death, given that we 

capture fish at a minimum size of 14 mm and cannot determine timing or location of death.  

Absolute distance is very strongly correlated to cumulative distance (adjusted R-squared: 

0.5063; Figure S2.1) and range (adjusted R-squared: 0.7809; Figure S2.2). This is not surprising 

given that most individuals in our study move little or not at all, such that for many fish the 

movement from initial to final locations is the same as their total dispersal distance. 

Figure S2.1. Correlation between absolute and cumulative distance in dataset. 

 93 

 
 
 
 
 
 
We then reran all reproductive models using absolute distance as an independent variable 

in the place of cumulative distance (“total distance” in the manuscript) or range. Absolute 

distance did not significantly predict reproductive success (count or zero-inflation) for males. 

This result is in contrast to what we found for cumulative distance, where it significantly 

increased the chances of males having non-zero offspring. The dispersal status models currently 

in the manuscript also show a significant effect of disperser status on the chance of males having 

non-zero offspring and a marginally significant effect on their number of offspring. As such, it 

seems that dispersal status and cumulative distance are correlated to reproductive success for 

males in a way that absolute distance is not.  

For females, increases in absolute distance led to significant increases in total number of 

offspring, and no effect on the chances of having non-zero offspring. This is similar to the 

outcomes of using the "total distance" metric (as presented in the manuscript). That is, 

cumulative distance and dispersal status both have marginally significant effects on number of 

offspring, and no effect on zero-inflation. Thus, for females, the metric of absolute distance 

captures a similar trend as cumulative distance and status. 

When comparing absolute distance to range, and their effects on number of mates, we 

again find qualitatively similar patters between the metrics. Both range and absolute distance 

have significantly positive effects on number of mates for males. For females, the effect of 

absolute distance does not significantly alter number of mates, compared to a nearly significant 

effect of range. 

These results raise interesting questions about how the effects of absolute and cumulative 

distance differ from the perspective of a dispersing guppy, and what the potential underlying 

biological causes and consequences of those differences may be. For example, the correlation 

 94 

 
 
between cumulative and absolute distances is actually higher for males (R-squared = 0.5781) 

than females (R-squared = 0.4521), yet they have a more similar effect on reproduction in 

females. This could be due to differences in exploratory behavior and territoriality between the 

sexes, and their effects on reproduction. Though males travel farther cumulative distances than 

females, they do not significantly differ in their absolute distances. This suggests that males 

explore new pools more than females, moving more throughout the entirety of our study reach 

before reaching their location of last capture.  

Ultimately, the similarity in most of the results between these analyses that use absolute 

distance and the ones presented in the manuscript that use cumulative distance, in addition to the 

strong correlation between the different predictors, suggests that our current metrics of dispersal 

adequately capture the aspects of guppy dispersal that we are most interested in. Though this 

analysis does raise interesting questions about sex differences in guppy dispersal behavior, the 

main results linking reproductive benefits to dispersal still hold when considering either total or 

absolute distance. 

Supplementary Appendix 2.II.  

We are extremely grateful to the field assistants who contributed to the capture-mark-recapture 

data collection in Trinidad during 2009 − 2011. Their names are listed in alphabetical order with 

project managers highlighted in bold: Jeremy Austin, Brian Ayers, Ron Bassar, Nicole Bedford, 

Dale Broder, Ethan Brown, Elizabeth Brunner, Mark Burton, Nichole Bushey, Becky Chong, 

Jacob Dillon, Connor Fitzpatrick, Courtney Fitzpatrick, Dylan Fitzpatrick, Eva Fischer, 

Andrew Furness, Stephanie George, Cameron Ghalambor, Katherine Gleason, Ryan Godfrey, 

Michael Grundler, Guanapo, Corey Handelsman, Justa Heinen, Maria Hernandez, Emily 

 95 

 
Josephs, Sean Kelly, Erin Kenison, Danielle Koplinka-Loehr, Katie Lawry, Andrés Lopez-

Sepulcre, Kevin Mierbachtol, DJ Ozzo, Elsa Ordway, Rachel Paseka, Marbles, Emily Nash, 

Jim Nunnaly, Amber Masters, Will Roberts, Bret Robinson, Rob Rozeske, Emily Ruell, Jerry 

Schneider, Alisha Shah, Julian Torres-Dowdall, Mauricio Torres, Jack Torresdal, Alice 

Vislova, Kellen Watson, Bill White, Jean-Paul Zagarola. 

 96 

 
 
 A) 

B) 

Figure S2.3. Distributions of variables included in models of fitness and dispersal. A) Lifetime 
reproductive success and number of mater for males and females. B) Hybrid index and longevity 
for all fish, and male standard length. 

 97 

Males01020304050600400Females01020304050600500Lifetime reproductive success051015200400051015200400Number of individualsNumber of mates 
 
 
 
 
 
A) 

B) 

Figure S2.4. Effect of longevity on total dispersal distance (A) and lifetime reproductive success 
(B). Lines show negative binomial regression and shading as 95% confidence interval, colored 
by sex. 

 98 

01002003004005101520Longevity (months)Total distance (m)0501001505101520Longevity (months)Total number of offspringSexFM 
 
 
 
 
 
 
 
 
Figure S2.5. Effect of age (months since first capture) on monthly dispersal distance. Lines 
show loess fit with shading as 95% confidence intervals, colored by stream. 

 99 

0501001500369AgeDistance movedStreamTaylorCaigualAll observations (N=7579) 
 
 
 
 
Figure S2.6. Effect of male standard length on total dispersal distance. Line shows negative 
binomial regression and shading as 95% confidence interval. 

 100 

 
 
 
 
Figure S2.7. Monthly seasonal dispersal over the course of the capture-mark-recapture study, 
wet season in blue and dry season in orange. Each point represents an individual capture, with 
positive values representing upstream movement and negative values showing downstream 
movement. Fish in Taylor (A) had decreased movement in the dry season, while Caigual (B) 
showed uniform dispersal throughout the year. 

 101 

 
 
Figure S2.8. Effect of total dispersal distance on total number of offspring. Panels represent the 
same analyses on subsets of the data that include only individuals of the same longevity. Despite 
decreasing sample sizes with increasing longevity, trends are qualitatively similar.  Line shows 
negative binomial regression and shading as 95% confidence interval, colored by sex. 

 102 

02040024Log total distanceTotal number of offspringSexFMAll fish (N=1357)010203040012345Log total distanceTotal number of offspringLongevity = 2 (N=495)05101520024Log total distanceTotal number of offspringLongevity = 3 (N=290)051015024Log total distanceTotal number of offspringLongevity = 4 (N=153)0204060024Log total distanceTotal number of offspringLongevity = 5 (N=137)05010015012345Log total distanceTotal number of offspringLongevity = 6 (N=85) 
 
 
 
 
Figure S2.9. Effect of dispersal status (moving ≥ 10 m) on total number of offspring. Dispersal 
status “FALSE” represents individuals that moved less than 10 m during the course of our study. 
Panels represent the same analyses on subsets of the data that include only individuals of the 
same longevity. Despite decreasing sample sizes with increasing longevity, trends are 
qualitatively similar.  

 103 

01020304050FALSETRUEDispersal statusNumber of offspringAll fish (N=1357)010203040FALSETRUEDispersal statusNumber of offspringLongevity = 2 (N=495)01020FALSETRUEDispersal statusNumber of offspringLongevity = 3 (N=290)010FALSETRUEDispersal statusNumber of offspringLongevity = 4 (N=153)01020304050FALSETRUEDispersal statusNumber of offspringLongevity = 5 (N=137)010203040FALSETRUEDispersal statusNumber of offspringLongevity = 6 (N=85) 
 
 
Table S2.1: Summary of dispersing and non-dispersing individuals by stream and sex.  

Taylor 

Caigual 

Both Streams 

Disperser 

Non-
Disperser 

All  Disperser 

Non-
Disperser 

All 

Disperser 

Non-
Disperser 

All 

F 

M 

100 
(27%) 

98 
(32%) 

276 (73%) 

376 

171 (45%) 

206 (55%) 

377 

210 (68%) 

308 

156 (53%) 

140 (47%) 

296 

271 
(36%) 

254 
(42%) 

482 (64%) 

753 

350 (58%) 

604 

All 

198 
(29%) 

486 
(71%) 

684 

327 (49%) 

346 
(51%) 

673 

525 
(39%) 

832 (61%) 

1357 

 104 

 
 
 
 
 
 
Table S2.2. Model output for monthly dispersal distance in response to age.  
All fish: 
Call:  glmmTMB.all.age.distance <- glmmTMB(min_dist ~ Age + I(Age^2) + Sex + Stream + 
(1|FishID) + (1|Capture_event), ziformula = ~ Age + I(Age^2) + Sex + Stream + (1|FishID) + 
(1|Capture_event), family=nbinom2) 

     AIC         BIC          logLik     deviance   df.resid  
 16055.7  16156.9     -8012.9     16025.7     6282  

Count model coefficients: 

(Intercept) 
Age 
I(Age^2) 
SexM 
StreamCaigual 

z value 

Std. Error 

Estimate 
2.490851                     0.145333            17.139                 < 2e-16 *** 
0.262037                     0.049465            5.297                   1.17e-07 *** 
-0.018425                    0.005784            -3.185                  0.00145 ** 
0.079378                     0.062582            1.268                    0.20466 
-0.590060                    0.068015            -8.675                  < 2e-16 *** 

Pr(>|z|) 

Zero-inflation model coefficients: 
(Intercept) 
Age 
I(Age^2) 
SexM 
StreamCaigual 

3.990996                     0.334676            11.925                 <2e-16 *** 
-1.110323                    0.065618            -16.921                <2e-16 *** 
0.126267                     0.009068            13.925                 <2e-16 *** 
-0.864999                    0.092355            -9.366                  <2e-16 *** 
-0.821609                    0.098619            -8.331                  <2e-16 *** 

 105 

 
 
 
 
 
 
 
 
Table S2.3. Model output for dispersal status and dispersal distance. Results presented 
separately for status and distance, and for all fish vs. males only.  
a) Dispersal status (moving ≥ 10 m): 
All fish: 
Call: glm(formula = disp.status ~ Sex.x + Stream * hindex + Longevity, family = "binomial") 

Deviance Residuals:  
    Min       1Q        Median       3Q      Max   
-1.9845  -0.9666  -0.5946   1.1534   2.2671  

Coefficients: 

(Intercept) 
Sex.xM 
StreamTaylor 
hindex 
Longevity 
StreamTaylor:hindex 

z value 

Std. Error 

Estimate 
-0.98609                       0.15184                   -6.494                8.35e-11 *** 
0.54703                        0.12362                   4.425                 9.64e-06 *** 
-0.09329                       0.16197                   -0.576                0.56462 
-1.27910                       0.44920                   -2.848                0.00441 ** 
0.18699                        0.02390                   7.825                 5.06e-15 *** 
-0.50577                       0.51282                   -0.986                0.32401 

Pr(>|z|) 

(Dispersion parameter for binomial family taken to be 1) 
Null deviance: 1811.1  on 1356  degrees of freedom 
Residual deviance: 1627.1  on 1351  degrees of freedom 
AIC = 1639.1 

Males only: 
Call:  glm(formula = disp.status ~ Stream * hindex + Longevity + Male_SL, family = "binomial") 

Deviance Residuals:  
    Min       1Q        Median       3Q      Max   
-2.4790  -0.9492  -0.5145   1.0221   2.3136 

Coefficients: 

(Intercept) 
StreamTaylor 
hindex 
Longevity 
Male_SL 
StreamTaylor:hindex 

z value 

Std. Error 

Estimate 
-7.48376                       1.57885               -4.740                  2.14e-06 *** 
-0.28026                       0.28297               -0.990                  0.322 
-0.11827                       0.58155               -0.203                  0.839 
0.31263                        0.05385                5.805                  6.42e-09 *** 
0.35313                        0.08525                4.142                  3.44e-05 *** 
-1.14974                       0.71108               -1.617                  0.106 

Pr(>|z|) 

(Dispersion parameter for binomial family taken to be 1) 
Null deviance: 733.89  on 535  degrees of freedom 
Residual deviance: 620.32  on 530  degrees of freedom 
(68 observations deleted due to missingness) 
AIC: 632.32 

 106 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table S2.3. (cont’d)  
b) Dispersal distance: 
All fish: 
Call:  glmmTMB(total_dist ~ Sex.x + Stream*hindex + log(Longevity), 
                                    ziformula = ~ Sex.x + Stream*hindex + log(Longevity), 
                                    family=nbinom2,data=d) 

     AIC        BIC      logLik   deviance   df.resid  
  7896.1   7963.8  -3935.0   7870.1        1344  

Count model coefficients: 

(Intercept) 
Sex.xM            
StreamTaylor      
hindex                
log(Longevity)       
StreamTaylor:hindex  

z value 

Std. Error 

Estimate 
Pr(>|z|) 
2.03783                       0.12447              16.372                  < 2e-16 *** 
0.18125                       0.07394               2.451                   0.0142 *   
1.13730                       0.09713              11.709                  < 2e-16 *** 
0.03830                       0.27245               0.141                   0.8882     
0.68927                       0.07130               9.667                   < 2e-16 *** 
-1.27645                      0.30052               -4.247                  2.16e-05 *** 

Zero-inflation model coefficients: 
(Intercept) 
Sex.xM 
StreamTaylor 
hindex 
log(Longevity) 
StreamTaylor:hindex 

0.7509                          0.1937                  3.877                  0.000106 *** 
-0.5877                         0.1239                  -4.744                 2.09e-06 *** 
0.3969                          0.1670                   2.376                 0.017499 *   
1.8705                          0.4571                   4.092                 4.27e-05 *** 
-0.9323                         0.1195                   -7.803                6.06e-15 *** 
-1.0837                         0.5058                   -2.143                0.032147 *   

Males only: 
Call: glmmTMB(total_dist ~ Male_SL + Stream*hindex + log(Longevity),  
                                      ziformula = ~ Male_SL + Stream*hindex + log(Longevity), 
                                      family=nbinom2) 

     AIC        BIC      logLik   deviance   df.resid  
  3394.0   3449.7   -1684.0   3368.0       523  

Count model coefficients: 

(Intercept) 
Male_SL 
StreamTaylor 
hindex 
log(Longevity) 
StreamTaylor:hindex 

z value 

Std. Error 

Pr(>|z|) 
Estimate 
-2.7612                     0.9054                    -3.050                      0.00229 **  
0.2633                      0.0485                    5.429                       5.67e-08 *** 
1.0921                      0.1533                    7.122                       1.06e-12 *** 
0.4356                      0.2988                    1.458                       0.14491     
0.6945                      0.1049                    6.618                       3.65e-11 *** 
-1.4285                     0.3581                    -3.989                      6.63e-05 *** 

Zero-inflation model coefficients: 
(Intercept) 
Male_SL 
StreamTaylor 
hindex 
log(Longevity) 
StreamTaylor:hindex 

2.89838                     1.56897                   1.847                      0.0647 .   
-0.12482                    0.08528                   -1.464                     0.1433     
0.63306                     0.28889                    2.191                     0.0284 *   
1.08094                     0.60279                    1.793                     0.0729 .   
-1.33409                    0.23015                   -5.797                     6.77e-09 *** 
-0.76240                    0.70513                   -1.081                     0.2796  

 107 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table S2.4. Model output for lifetime reproductive success (LRS). Results presented separately 
for status and distance, and for males and females. 
a) Dispersal status (moving ≥ 10 m): 
Male LRS by dispersal status: 
Call: glmmTMB(LRS ~ disp.status + Stream*hindex + Longevity + Male_SL,  
        ziformula = ~ disp.status + Stream*hindex + Longevity + Male_SL, family=nbinom2) 

     AIC      BIC         logLik        deviance      df.resid  
  1409.5   1470.9     -689.8          1379.5         428 

Count model coefficients: 

(Intercept) 
disp.statusTRUE 
StreamTaylor 
hindex 
Longevity 
Male_SL 
StreamTaylor:hindex 

z value 

Std. Error 

Estimate 
Pr(>|z|) 
-1.01081                        1.24334            -0.813                       0.41623     
0.29813                         0.17608             1.693                       0.09042 .   
0.76258                         0.30297             2.517                       0.01183 *   
3.13134                         0.46996             6.663                       2.68e-11 *** 
0.10351                         0.03638             2.846                       0.00443 **  
0.05261                         0.06767             0.777                       0.43690     
-2.53136                        0.56414             -4.487                      7.22e-06 *** 

Zero-inflation model coefficients: 
(Intercept) 
disp.statusTRUE 
StreamTaylor 
hindex 
Longevity 
Male_SL 
StreamTaylor:hindex 

6.93705                         2.10281              3.299                       0.00097 *** 
-0.81742                        0.28415              -2.877                      0.00402 **  
-0.40665                        0.42960              -0.947                      0.34385     
-1.56677                        0.77881              -2.012                      0.04425 *   
-0.16920                        0.07276              -2.325                      0.02005 *   
-0.28269                        0.11392              -2.481                      0.01308 *   
0.72034                         0.91030               0.791                      0.42875    

Female LRS by dispersal status: 
Call: glmmTMB(LRS ~ disp.status + Stream*hindex + Longevity,  
          ziformula = ~ disp.status + Stream*hindex + Longevity, family=nbinom2) 

     AIC      BIC         logLik        deviance      df.resid  
  1675.1   1732.7     -824.5          1649.1         610 

Count model coefficients: 

(Intercept) 
disp.statusTRUE 
StreamTaylor 
hindex 
Longevity 
StreamTaylor:hindex 

z value 

Std. Error 

Estimate 
-0.20817                       0.19590              -1.063                      0.287941     
0.25069                        0.13447               1.864                      0.062285 .   
0.93491                        0.20469               4.567                      4.94e-06 *** 
1.85030                        0.55458               3.336                      0.000849 *** 
0.12263                        0.01868               6.565                      5.21e-11 *** 
-1.63847                       0.59084               -2.773                     0.005552 **  

Pr(>|z|) 

Zero-inflation model coefficients: 
(Intercept) 
disp.statusTRUE 
StreamTaylor 
hindex 
Longevity 
StreamTaylor:hindex 

 2.11033                       0.33952               6.216                       5.11e-10 *** 
-0.36899                       0.24166               -1.527                      0.1268     
-0.30166                       0.31628               -0.954                      0.3402     
-1.58791                       0.93383               -1.700                      0.0891 .  
-0.24491                       0.04682               -5.230                      1.69e-07 *** 
0.73478                        0.98627                0.745                       0.4563 

 108 

 
 
 
 
 
 
 
 
 
 
 
 
 
Table S2.4. (cont’d)  
b) Dispersal distance: 
Male LRS by dispersal distance: 
Call: glmmTMB(LRS ~ total_dist + Stream*hindex + Longevity + Male_SL,  
           ziformula = ~ total_dist + Stream*hindex + Longevity + Male_SL, family=nbinom2) 

     AIC      BIC         logLik        deviance      df.resid  
  1409.5   1470.9     -689.8         1379.5          428 

Count model coefficients: 

(Intercept) 
total_dist 
StreamTaylor 
hindex 
Longevity 
Male_SL 
StreamTaylor:hindex 

Estimate 
-1.088168    
0.002548    
0.674386    
3.149473    
0.106751   
0.061621   
-2.480057    

Zero-inflation model coefficients: 
(Intercept) 
total_dist 
StreamTaylor 
hindex 
Longevity 
Male_SL 
StreamTaylor:hindex 

 6.056602    
-0.017379    
-0.193640    
-1.590957    
-0.170592    
-0.243500    
0.725659    

Std. Error 
1.319158  
0.002542 
0.305067   
0.477491    
0.037071    
0.070457    
0.568054   

2.145451    
0.007748   
0.447262   
0.777393   
0.073341   
0.115808   
0.914829   

z value 
-0.825  
1.002 
2.211   
6.596 
2.880   
0.875   
-4.366  

2.823   
-2.243   
-0.433   
-2.047   
-2.326   
-2.103   
0.793   

Female LRS by dispersal distance: 
Call: glmmTMB(LRS ~ total_dist + Stream*hindex + Longevity,  
             ziformula = ~ total_dist + Stream*hindex + Longevity, family=nbinom2) 

     AIC      BIC         logLik        deviance      df.resid  
  1676.6   1734.3      -825.3        1650.6          610 

Pr(>|z|) 
0.40943  
0.31619 
0.02706 * 
4.23e-11 *** 
0.00398 ** 
0.38179 
1.27e-05 *** 

0.00476 ** 
0.02490 * 
0.66505 
0.04070 * 
0.02002 * 
0.03550 * 
0.42765 

Count model coefficients: 

(Intercept) 
total_dist 
StreamTaylor 
hindex 
Longevity 
StreamTaylor:hindex 

Estimate 
-0.107259    
0.003079    
0.872491    
1.785193    
0.120411    
-1.586753    

Zero-inflation model coefficients: 
(Intercept) 
total_dist 
StreamTaylor 
hindex 
Longevity 
StreamTaylor:hindex 

 1.998890    
-0.003880    
-0.214694    
-1.388891    
-0.247760    
 0.553283    

z value 
-0.561  
1.845  
4.208  
3.269  
6.384  
-2.715  

6.157  
-1.204     
-0.667     
-1.528     
-5.196  
0.568     

Pr(>|z|) 
0.574865    
0.064965 . 
2.57e-05 ***   
0.001081 ** 
1.73e-10 *** 
0.006636 ** 

7.40e-10 *** 
0.229 
0.505 
0.127 
2.03e-07 *** 
0.570 

Std. Error 
0.191226   
0.001669    
0.207320    
0.546174    
0.018863    
0.584534   

0.324641    
0.003224   
0.321891   
0.909127   
0.047679   
0.973838    

 109 

  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table S2.5. Model output for number of mates as a response to net range. Results presented 
separately for males and females. 
Mates by range for males: 
Call: glmmTMB(NumMates ~ net_range + Stream*hindex + Longevity, 
             ziformula = ~ net_range + Stream*hindex + Longevity, family=nbinom2) 

     AIC      BIC         logLik        deviance      df.resid  
  1009.3   1063.8      -491.7         983.3           474 

Count model coefficients: 

(Intercept) 
net_range 
StreamTaylor 
hindex 
Longevity 
StreamTaylor:hindex 

Estimate 
-1.860020    
0.007493    
0.397865    
4.004606    
0.142510    
-2.715788    

Zero-inflation model coefficients: 
(Intercept) 
net_range 
StreamTaylor 
hindex 
Longevity 
StreamTaylor:hindex 

0.93879     
-0.06256   
-0.17207     
0.31275     
-0.34331     
-0.40510     

Std. Error 
0.423770   
0.002721    
0.398920    
0.556801    
0.038921    
0.659690   

1.00815    
0.06451   
1.09044   
1.68424    
0.20066   
1.90798   

z value 
-4.389  
2.754  
0.997  
7.192  
3.661  
-4.117  

0.931    
-0.970    
-0.158    
0.186    
-1.711    
-0.212    

Mates by range for females: 
Call: glmmTMB(NumMates ~ net_range + Stream*hindex + Longevity, 
               ziformula = ~ net_range + Stream*hindex + Longevity, family=nbinom2) 

     AIC      BIC         logLik        deviance      df.resid  
    1164.8 1222.1     -569.4         1138.8           597 

Count model coefficients: 

(Intercept) 
net_range 
StreamTaylor 
hindex 
Longevity 
StreamTaylor:hindex 

Estimate 
-1.343364    
0.005223    
0.899758    
3.127215    
0.094350    
-3.043065    

Zero-inflation model coefficients: 
(Intercept) 
net_range 
StreamTaylor 
hindex 
Longevity 
StreamTaylor:hindex 

3.420125    
0.011949    
0.642681    
3.083049    
-1.364230    
-4.491290    

Std. Error 
0.246658   
0.002887    
0.240047    
0.610870    
0.022458    
0.662271   

0.983719    
0.009765    
0.753834    
2.130647    
0.455567   
2.325537   

z value 
-5.446  
1.809  
3.748  
5.119  
4.201  
-4.595  

3.477  
1.224  
0.853  
1.447  
-2.995  
-1.931  

Theta = 2.3921  
Number of iterations in BFGS optimization: 71; Log-likelihood: -569.4 on 13 Df 

Pr(>|z|) 
1.14e-05 *** 
0.005881 ** 
0.318591   
6.38e-13 *** 
0.000251 *** 
3.84e-05 *** 

0.3518 
0.3321 
0.8746 
0.8527 
0.0871 . 
0.8319 

Pr(>|z|) 
5.14e-08 *** 
0.070442 .   
0.000178 *** 
3.07e-07 *** 
2.65e-05 *** 
4.33e-06 *** 

0.000508 *** 
0.221056   
0.393909   
0.147897 
0.002748 ** 
0.053447 . 

 110 

 
 
 
 
 
 
    
 
 
 
 
 
Table S2.6. Output from models of total distance and upstream and downstream distances in 
response to season. 
a) Generalized linear mixed models of distance moved in each capture occasion, with individual 
and month as random effects. Analyses done separately for each stream. 
Taylor 

Family: nbinom2  ( log ) 

Formula:  min_dist ~ Season + (1 | FishID_nodash) + (1 | month) 

Conditional model: 

Groups 

Name 

Variance 

Std.Dev. 

FishID_nodash 

(Intercept) 

0.41736 

month 

(Intercept) 

0.03679 

0.6460 

0.1918 

Number of obs: 469, groups:  FishID_nodash, 334; month, 11 
Overdispersion parameter for nbinom2 family (): 1.45  

Conditional model: 

(Intercept) 

SeasonW 

Estimate 

Std. Error 

2.3071 

1.0994 

0.1358 

0.1692 

z value 

16.993 

6.499 

Pr(>|z|) 

< 2e-16 *** 

8.07e-11 *** 

Log-likelihood: -2006.9; AIC = 4023.9 

Caigual 

Family: nbinom2  ( log ) 

Formula:  min_dist ~ Season + (1 | FishID_nodash) + (1 | month) 

Conditional model: 

Groups 

Name 

Variance 

Std.Dev. 

FishID_nodash 

(Intercept) 

0.145766 

Month 

(Intercept) 

0.006882 

0.38179 

0.08296 

Number of obs: 651, groups: FishID_nodash, 376; month, 12 
Overdispersion parameter for nbinom2 family (): 3.24 

Conditional model: 

(Intercept) 

SeasonW 

Estimate 

2.53176 

-0.03655 

Std. Error 

z value 

Pr(>|z|) 

0.07835 

0.09012 

32.31 

-0.41 

<2e-16 *** 

0.685 

Log-likelihood: -2265.6; AIC = 4541.1 

 111 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table S2.6. (cont’d)  
b) Negative binomial models of upstream and downstream movement in response to season. 
Analyzes done separately for each direction in each stream. 
Taylor upstream 

Call: glm.nb(formula = upstream ~ Season, data = t.disp, init.theta = 0.8605303544, link = log) 

Null deviance: 282.05  on 227  degrees of freedom 

Residual deviance: 259.48  on 226  degrees of freedom 

Coefficients: 

(Intercept) 

SeasonW 

Estimate 

Std. Error 

2.7874 

0.8066 

0.1351 

0.1602 

z value 

20.626 

5.036 

Pr(>|z|) 

< 2e-16 *** 

4.75e-07 *** 

2 x log-likelihood: -1992.1410 

AIC = 1998.1 

Taylor downstream 
Call: glm.nb(formula = downstream * -1 ~ Season, data = t.disp, init.theta = 1.126641728, link = 
log) 

Null deviance: 373.10  on 241  degrees of freedom 

Residual deviance: 266.47  on 240  degrees of freedom 

Coefficients: 

Estimate 

Std. Error 

z value 

Pr(>|z|) 

(Intercept) 

SeasonW 

2.1959 

1.5235 

0.1110 

0.1341 

19.77 

11.36 

<2e-16 *** 

<2e-16 *** 

2 x log-likelihood: -2048.1020 

AIC = 2054.1 

 112 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table S2.6. (cont’d)  
b) Negative binomial models of upstream and downstream movement in response to season. 
Analyzes done separately for each direction in each stream. 
Caigual upstream 

Call: glm.nb(formula = upstream ~ Season, data = c.disp, init.theta = 2.184139022, link = log) 

Null deviance: 278.02  on 261  degrees of freedom 

Residual deviance: 277.53  on 260  degrees of freedom 

Coefficients: 

Estimate 

Std. Error 

(Intercept) 

2.64309 

SeasonW 

-0.07814 

0.09990 

0.11198 

z value 

26.456 

-0.698 

Pr(>|z|) 

<2e-16 *** 

0.485 

2 x log-likelihood:  -1837.824 

AIC = 1843.8 

Caigual downstream 
Call: glm.nb(formula = downstream * -1 ~ Season, data = c.disp, init.theta = 2.163640678, link = 
log) 

Null deviance: 419.89  on 388  degrees of freedom 

Residual deviance: 419.82  on 387  degrees of freedom 

Coefficients: 

(Intercept) 

SeasonW 

Estimate 

2.59027 

-0.02581   

Std. Error 

0.08299 

0.09285 

z value 

31.213 

-0.278 

Pr(>|z|) 

<2e-16 *** 

0.781 

2 x log-likelihood:  -2725.339 

AIC = 2731.3 

 113 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
APPENDIX B: 

Supporting information: Rollin’ on the river: Local census size, sex ratio, and relatedness drive 
dispersal behavior in the Trinidadian guppy (Poecilia reticulata) 

Supplemental Tables: 

Table S3.1. Model output for best model of dispersal probability in response to census size in 
Caigual. 

 114 

 moved in next stepPredictorsOdds RatiosCIp(Intercept)0.780.14 – 4.240.773census scaled0.150.07 – 0.32<0.001Season [W]3.510.58 – 21.240.172Sex [M]0.540.40 – 0.72<0.001Weight0.080.03 – 0.19<0.001census scaled * Season[W]2.831.28 – 6.260.010Random Effectsσ23.29τ00 FishID0.63τ00 Month0.67ICC0.28N Month7N FishID823Observations2367Marginal R2 / Conditional R20.170 / 0.404 
 
 
Table S3.2. Model output for best model of dispersal probability in response to sex ratio in 
Caigual. 

 115 

 moved in next stepPredictorsOdds RatiosCIp(Intercept)0.710.13 – 3.840.688sex ratio99.7529.86 – 333.28<0.001Sex [M]0.640.34 – 1.180.152Season [W]1.560.27 – 9.150.623Weight0.030.01 – 0.07<0.001sex ratio * Sex [M]0.170.03 – 0.880.035Random Effectsσ23.29τ00 FishID1.08τ00 Month0.67ICC0.35N Month7N FishID817Observations2336Marginal R2 / Conditional R20.110 / 0.419 
 
Table S3.3. Model output for best model of dispersal probability in response to relatedness in 
Caigual. 

 116 

 moved in next stepPredictorsOdds RatiosCIp(Intercept)1.280.16 – 10.110.814relatedness jaccard8404.6063.56 – 1111356.20<0.001relatedness jaccard^20.000.00 – 0.00<0.001Season [W]1.960.24 – 16.190.531Sex [M]0.490.35 – 0.70<0.001Weight0.020.01 – 0.04<0.001Random Effectsσ23.29τ00 FishID1.52τ00 Month0.96ICC0.43N Month7N FishID807Observations2313Marginal R2 / Conditional R20.075 / 0.474 
 
Table S3.4. Model output for best model of dispersal probability in response to census size in 
Taylor. 

 117 

 moved in next stepPredictorsOdds RatiosCIp(Intercept)0.220.07 – 0.690.009Season [W]2.080.69 – 6.320.196census scaled0.660.49 – 0.890.006Sex [M]1.350.76 – 2.380.308Weight0.530.04 – 7.040.630Random Effectsσ23.29τ00 FishID_nodash0.00τ00 Month0.65N FishID_nodash331N Month11Observations460Marginal R2 / Conditional R20.064 / NA 
 
Table S3.5. Model output for best model of dispersal probability in response to sex ratio in 
Taylor. 

 118 

 moved in next stepPredictorsOdds RatiosCIp(Intercept)0.100.02 – 0.590.011Season [W]1.830.45 – 7.390.394sex ratio1.430.23 – 8.910.704Sex [M]1.290.28 – 5.930.740Weight4.930.19 – 129.810.339sex ratio * Sex [M]1.490.13 – 16.420.746Random Effectsσ23.29τ00 FishID_nodash0.00τ00 Month0.96ICC0.23N Month11N FishID_nodash267Observations335Marginal R2 / Conditional R20.027 / 0.247 
 
 
Table S3.6. Model output for best model of dispersal probability in response to relatedness in 
Taylor. 

 119 

 moved in next stepPredictorsOdds RatiosCIp(Intercept)0.140.02 – 0.820.030Season [W]3.190.53 – 19.300.207relatedness jaccard0.690.00 – 1201.260.922relatedness jaccard^20.000.00 – 328.020.220Sex [M]1.180.52 – 2.710.691Weight3.230.10 – 103.220.507Random Effectsσ23.29τ00 FishID_nodash0.00τ00 Month1.66ICC0.33N Month11N FishID_nodash250Observations314Marginal R2 / Conditional R20.112 / 0.410 
 
 
APPENDIX C: 

Supporting information: Mutualism benefits are lost at high inbreeding 

Supplemental Methods 

Crosses: 

To obtain seeds for this experiment, we manipulated two generations of plant crosses for C. 

fasciculata individuals collected from two natural populations in the US Midwest (Westland, MI, 

and Sandusky, OH). The Ohio population is a large population located at the Erie Sand Barrens 

Preserve, while the Michigan population was a small marginal road-side population. Maternal 

seeds used to initiate crosses for this experiment were the result of several generations of 

greenhouse crossings of field-collected seeds. Four maternal families from each population were 

used to generate all crosses described below. Our first generation of crosses was organized into 

four different cross types: between the two populations, between maternal lines within the same 

population, within maternal lines, and selfed (Figure 4.2). Flowers of C. fasciculata are buzz 

pollinated and open for approximately one day, with plants flowering for approximately 4 

months in the greenhouse environment. Germination rate for the parental generation was low, 

with only one third (101 out of 303) of individual seeds surviving to the seedling stage. During 

their flowering period, plants were checked for flowers daily, and every open flower was hand-

pollinated using a vibrating device and a toothpick to obtain pollen and place it on anthers, 

respectively. We attempted to maximize cross variety by randomly choosing pollen donor and 

recipients from the available open flowers each day, and conducting reciprocal crosses for each 

plant pair that produced more than one flower on any given day. Given the lower fecundity of 

selfed crosses, we also attempted to self-pollinate at least one flower per plant per flowering day. 

Because C. fasciculata has explosive seed dispersal, all maturing seed pods were bagged with 

 120 

 
mesh bags to prevent seed loss, and seeds were collected once pods had opened within bags. We 

obtained over 300 F1 seeds from the initial parental plants. 

F1 seeds were then germinated in the greenhouse in the following summer (2020). 249 

F1 individuals germinated, with 146 (58%) surviving to flowering. We applied the same protocol 

of crosses above for this generation: once plants started flowering we checked them for flowers 

daily and pollinated all open flowers with randomly chosen individuals that also had open 

flowers on that date, and selfed all plants each day when possible. We obtained a total of 3423 

seeds from the F1 crosses, which varied from twice-selfed to outbred. These F2 seeds were then 

used in the common garden experiment described below. 

Germination: 

The germination protocol was the same for plants in the parental, F1, and F2 generations. All C. 

fasciculata seeds were geminated by first being submerged in a solution of 10% bleach for one 

minute, then scarified with a sterile razor blade and put in petri dish with deionized water. Petri 

dishes contained up to 12 scarified seeds, always from the same seed pod (assumed full siblings), 

and were placed on a bench in a greenhouse on a 90-75F, 16 light-8 dark cycle. Every day for 15 

days, seeds were checked for cotyledon and radicle development. Individuals were planted once 

cotyledons were green and radicles were longer than 0.5 cm, into 2.5 x 10 inch “cone-tainers” 

(Stuewe & Sons, Inc.) filled with Premier Pro-Mix FPX 2.8cf soil from BFG Supply Co. Seeds 

that did not sprout cotyledons or radicles within 15 days were discarded and counted as failed 

germinants. Parental and F1 generations were kept in the greenhouse for the duration of their 

lifecycle. The experimental F2 plants remained in the greenhouse only until rhizobia inoculations 

were completed, then were moved to an outdoor common garden. 

 121 

 
 
Inoculations: 

For the common garden experiment, we attempted to geminate 3209 seeds, resulting in 2571 live 

seedlings following the germination protocol described above. Out of those experimental plants, 

1715 C. fasciculata F2 individuals of all crossing backgrounds were inoculated with 

Bradyrhizobium sp., while 856 plants did not receive any rhizobia. We used two rhizobia strains 

which were obtained from previous collections of naturally occurring C. fasciculata at the 

Kellogg Biological Station (KBS, Hickory Corners, MI), which did not differ in their ability to 

nodulate with experimental plants or the number of nodules they produced (Figure S4.1). The 

strains were collected in 2017 from nodules in C. fasciculata grown at the Kellogg Biological 

Station, such that both experimental plant populations should be naïve to the strains used in the 

common garden. The strains had been maintained at -80C until 2021, and were revived first on 

solid tryptone yeast media. They were incubated on that solid media for 2 days at 30ºC degrees, 

then re-plated on fresh solid media and incubated again for 2 days at 30ºC degrees. The isolates 

were then transferred to a liquid modified arabinose gluconate media where they were incubated 

at 30ºC degrees on a MaxQ 4000 Orbital Shaker (Thermo Scientific). On the 3rd day of growth 

on liquid media, we measured optical density of each culture with a spectrophotometer set to 

OD600 (Eppendorf BioPhotometer plus) and diluted all cultures to the concentration of the least 

concentrated sample. We then added 2ml of liquid culture to each plant, with control plants 

receiving the same volume of sterile liquid arabinose gluconate media that had been incubated 

for the same amount of time.  

We conducted a pilot experiment using these inoculation protocols, rhizobia strains, F2 

seeds, and soil mixture in the greenhouse in January of 2021 to confirm that the strains would 

produce nodules on these plants. In this pilot, we also tested for the effects of autoclaving soil on 

 122 

rhizobia contamination to inform the design of our main experiment. This pilot confirmed that 

our rhizobia strains did generally nodulate in our F2 plants. Out of 137 pilot plants in non-

autoclaved soil, half of which were inoculated with rhizobia, all inoculated plants formed 

nodules and only one uninoculated plant was contaminated (produced one nodule). Based on that 

pilot, we decided to not autoclave soil for our common garden experiment, as that would alter 

nutrient content and availability for plants (Alphei & Scheu 1993; Endlweber & Scheu 2006).  

Given the large number of experimental plants, inoculations were conducted with the 

methods described above on two separate dates, June 14th for individuals that were planted 

between May 1st and May 28th, and June 21st for individuals planted between May 29th and June 

13th. We organized experimental plants into racks so that individuals with the same inoculation 

treatment were placed together, to minimize contamination between pots of different treatments 

while plants were watered in the greenhouse. Plants were later reshuffled when moved to the 

field. We also measured the height of each plant at the time of inoculation. We included planting 

date on all statistical models of mutualism and fitness traits described below. 

Common garden: 

All experimental (F2) plants were germinated, planted, and inoculated in the greenhouse. 

On July 1st, all individuals were moved to a fenced outdoor common garden at the Kellogg 

Biological Station in Hickory Corners, MI, USA. Outdoors, plants were organized into 18 

blocks, each containing 9 racks. Each rack contained 12 plants of the same rhizobia treatment, 

and each block contained 3 racks of each treatment. Every 3 weeks, all racks were rotated to 

minimize microenvironmental effects of block placement. The common garden allowed for 

natural levels of insect herbivory and pollination, while preventing mammalian herbivory. Given 

a history of vole infestations in our field site, we also used deterrents (Solar Powered Mole 

 123 

Groundhog Repellent Stakes Sonic Gopher Chaser Vole Deterrent and Shake Away 5006258 

Fox Urine Granules) in addition to the fencing to avoid destructive rodent herbivory. Six sonic 

repellent stakes were placed in regular intervals along a corridor between blocks of experimental 

plants and left running for the duration of the experiment, while the urine granules were applied 

monthly around the perimeter of the experiment, approximately 3m away from the edge of the 

blocks. We did not observe any herbivory from rodents, deer, or other mammals in our 

experiment. 

We watered plants daily to avoid desiccation in the common garden, except on days with 

natural precipitation. We scored the day of first flower for each individual, and height at 

flowering on that date. To study the plant-ant mutualism, 58 individuals of C. fasciculata that 

spanned our inbreeding gradient were selected to undergo weekly ant surveys. For the month 

after plants were placed outside, we conducted two surveys of the number of extra-floral 

nectaries (EFN) on each of those 58 individuals and quantified insect herbivory as the percentage 

of plant leaves with any herbivore damage. We also conducted seven ant surveys on those plants, 

were we counted the number of ants on each plant, then calculated an average number of ants per 

survey per plant. Once plants began to set seed (late August), we manually collected seed pods 

daily before seed pods dehisced and seeds were dispersed into the environment. Plants started 

being harvested in October, once they senesced and no longer had any green leaves or flowers, 

with harvest being completed in November. At time of harvest, we separated above and 

belowground biomass for weighting and counting rhizobia root nodules. Plant fitness was 

quantified as the total number of viable seeds produced throughout the growing season, as well 

as above and belowground biomass at senescence. Seed viability was visually estimated, any 

seeds were very thin, small, or hollow (indications that they did not have developed endosperm) 

 124 

 
were not counted or weighted. We also estimated the strength of plants’ interactions with 

rhizobia by counting the number of nitrogen-fixing nodules on the roots of each plant at the end 

of the experiment.  

 125 

 
Supplemental Figures: 

Figure S4.1. Variation in nodule number based on inoculation treatment, with red lines on 
histograms representing average nodule number for that treatment. Plants that did not receive 
rhizobia (“blank” treatment, or “B”) had a majority of zero values, but some contamination made 
the average number of nodules non-zero. Plants inoculated with “M” and “O” strains produced a 
similar distribution of nodules. Therefore, all analyses in the main text do not consider strain 
variation, instead grouping “M” and “O” treatments into “inoculated with rhizobia,” while the 
blank treatment plants are considered “not inoculated with rhizobia.” 

 126 

 
 
 
Figure S4.2. Predicted relationships based on best models. A) There was no effect of rhizobia 
inoculation on plant height at flowering, but height did decrease with increased plant inbreeding 
B) Plants that flowered at intermediate ages (in days) were taller. B) Plants with higher 
proportion of Ohio ancestry (higher values) were shorter. 

Figure S4.3. Visualization of actual data, corresponding to model results in Figure 4.5A in the 
main text. Different lines show different rhizobia inoculation treatments. A) Quadratic lines of 
best fit for complete dataset. B) Quadratic lines of best fit for only the subset plants that 
produced non-zero seeds. C) Boxplot of inbreeding values for plants that produced no seeds 
(“0”) vs. plants that produced at least 1 seed (“1”).  

 127 

30354045−10123Pedigree inbreedingHeight at flowering (cm)A10203040406080Age at floweringHeight at flowering (cm)rhizobiaNYB253035404550−10123Pedigree inbreedingHeight at flowering (cm)prop.E00.250.50.751C01002003000.000.250.500.75InbreedingTotal seed set (all)A01002003000.000.250.500.75InbreedingTotal seed set (counts only)rhizobiaNYB●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●0.000.250.500.7501NASeed productionPedigree inbreedingC 
 
 
 
 
Figure S4.4. Visualization of actual data, corresponding to model results in Figure 4.5C and 
4.5D in the main text. Different lines show different rhizobia inoculation treatments. A) 
Quadratic lines of best fit for inbreeding effects on aboveground biomass. B) Quadratic lines of 
best fit for ancestry effects on aboveground biomass. 

Figure S4.5. Visualization of actual data, corresponding to model results in Figure 4.6A. 
Quadratic lines of best fit for inbreeding effects on belowground biomass, different lines show 
different rhizobia inoculation treatments. 

 128 

012340.000.250.500.75InbreedingAboveground biomassA012340.00.40.8Ancestry proportionAboveground biomassrhizobiaNYB0.02.55.07.510.00.000.250.500.75InbreedingBelowground biomassrhizobiaNY 
 
 
 
 
 
Figure S4.6. Visualization of actual data, corresponding to model results in Figure 4.6B. 
Different lines show different rhizobia inoculation treatments.  A) Quadratic lines of best fit for 
inbreeding effects on height at flowering. B) Quadratic lines of best fit for ancestry effects on 
height. C) Quadratic lines of best fit for age effects on height. 

Figure S4.7. Visualization of actual data, corresponding to model results in Figure 4.6B. 
Different lines show different rhizobia inoculation treatments. A) Quadratic lines of best fit for 
inbreeding effects on individual seed mass. B) Quadratic lines of best fit for ancestry effects on 
individual seed mass. 

 129 

2040600.000.250.500.75InbreedingHeight at floweringA2040600.00.40.8Ancestry proportionHeight at floweringB204060406080Age at floweringHeight at floweringrhizobiaNYC0.0050.0100.0150.0200.000.250.500.75InbreedingSeed massA0.0050.0100.0150.0200.00.40.8Ancestry proportionSeed massrhizobiaNYB 
 
 
 
Supplemental Tables: 

Table S4.1. Model output for best model of total seed set. 

 130 

 total seed numberPredictorsEstimatesCIpCount Model(Intercept)52.9349.62 – 56.46<0.001ped inbreeding c0.840.80 – 0.88<0.001rhizobia [Y]2.152.03 – 2.27<0.001prop E1.271.16 – 1.39<0.001ped inbreeding c^20.960.95 – 0.98<0.001ped inbreeding c *rhizobia [Y]1.091.03 – 1.160.002(Intercept)720.03Zero-Inflated Model(Intercept)0.210.17 – 0.25<0.001prop E2.361.62 – 3.45<0.001ped inbreeding c0.930.84 – 1.030.161Observations2566 
 
 
Table S4.2. Model output for best model of total seed set, including only individuals for which 
we conducted ant surveys. 

 131 

 total seed numberPredictorsEstimatesCIpCount Model(Intercept)69.5952.59 – 92.08<0.001T ant1.211.04 – 1.400.014ped inbreeding c0.850.73 – 0.990.038rhizobia [Y]1.471.18 – 1.840.001prop E1.491.06 – 2.120.023ped inbreeding c^20.960.90 – 1.020.204ped inbreeding c *rhizobia [Y]1.241.03 – 1.490.023(Intercept)68.96Zero-Inflated Model(Intercept)0.050.01 – 0.300.001prop E1.050.07 – 14.710.973ped inbreeding c2.200.99 – 4.900.053Observations57 
 
 
 
Table S4.3. Model output for best model of aboveground biomass. 

 132 

 aboveground biomassPredictorsEstimatesCIp(Intercept)1.231.10 – 1.35<0.001prop E-0.29-0.40 – -0.17<0.001ped inbreeding c-0.09-0.14 – -0.040.001rhizobia [Y]0.520.46 – 0.58<0.001ped inbreeding c^2-0.03-0.05 – -0.010.001ped inbreeding c *rhizobia [Y]-0.06-0.11 – 0.000.054Random Effectsσ20.33τ00 planting_date_julian0.05ICC0.14N planting_date_julian21Observations1619Marginal R2 / Conditional R20.182 / 0.301 
 
 
Table S4.4. Model output for best model of belowground biomass. 

Table S4.5. Model output for best model of height at flowering. 

 133 

 belowground biomassPredictorsEstimatesCIp(Intercept)1.731.67 – 1.79<0.001ped inbreeding c1.031.01 – 1.050.002rhizobia [Y]0.900.87 – 0.94<0.001ped inbreeding c^21.021.01 – 1.030.003Observations1436R2 conditional / R2 marginalNA / 0.006 height at floweringPredictorsEstimatesCIp(Intercept)-42.34-52.64 – -32.03<0.001age at flowering in days2.882.56 – 3.19<0.001age at flowering indays^2-0.02-0.03 – -0.02<0.001rhizobia [Y]0.40-0.16 – 0.970.163ped inbreeding c-0.90-1.22 – -0.59<0.001prop E-8.93-10.01 – -7.85<0.001ped inbreeding c^2-0.16-0.33 – 0.010.066Random Effectsσ235.14τ00 planting_date_julian15.66ICC0.31N planting_date_julian21Observations1973Marginal R2 / Conditional R20.266 / 0.492 
 
 
 
Table S4.6. Model output for best model of individual seed mass. 

 134 

 seed ratio by envelope cPredictorsEstimatesCIp(Intercept)0.300.14 – 0.46<0.001prop E^20.940.19 – 1.700.014prop E-1.39-2.10 – -0.69<0.001ped inbreeding c-0.19-0.25 – -0.14<0.001ped inbreeding c^2-0.05-0.09 – -0.000.044rhizobia [Y]0.08-0.02 – 0.190.127ped inbreeding c^2 *rhizobia [Y]0.060.01 – 0.110.019Observations1951R2 / R2 adjusted0.040 / 0.037 
 
 
Table S4.7. Model output for best model of nodule number. 

 135 

 nodule numberPredictorsIncidence Rate RatiosCIpCount Model(Intercept)25.7520.86 – 31.79<0.001prop E0.930.76 – 1.130.462ped inbreeding c0.950.90 – 1.000.069rhizobia [Y]1.641.43 – 1.89<0.001ped inbreeding c^20.980.95 – 1.020.315(Intercept)4.920.01 – 0.03Zero-Inflated Model(Intercept)1.551.09 – 2.200.014prop E1.110.57 – 2.150.760ped inbreeding c1.321.08 – 1.600.005ped inbreeding c^21.080.98 – 1.200.139rhizobia [Y]0.020.01 – 0.03<0.001Random Effectsσ20.62τ00 harvesting_date0.17ICC0.21N harvesting_date36Observations1432Marginal R2 / Conditional R20.071 / 0.269 
 
 
Table S4.8. Model output for best model of contamination. 

Table S4.9. Model output for best model of nodulation failure. 

 136 

 contaminationPredictorsOdds RatiosCIp(Intercept)0.120.08 – 0.18<0.001prop E1.000.52 – 1.930.994ped inbreeding c0.850.71 – 1.020.084ped inbreeding c^20.980.88 – 1.080.643Random Effectsσ23.29τ00 harvesting_date0.19ICC0.05N harvesting_date39Observations1546Marginal R2 / Conditional R20.010 / 0.063 no nodPredictorsOdds RatiosCIp(Intercept)0.010.01 – 0.03<0.001prop E0.820.26 – 2.550.728ped inbreeding c1.691.05 – 2.730.031ped inbreeding c^21.070.87 – 1.310.534Random Effectsσ23.29τ00 harvesting_date1.28τ00 planting_date_julian0.21ICC0.31N harvesting_date38N planting_date_julian21Observations1512Marginal R2 / Conditional R20.075 / 0.363 
 
 
Table S4.10. Model output for best model of average ant number. 

Table S4.11. Model output for best model of EFN count. 

 137 

 T antPredictorsIncidence Rate RatiosCIp(Intercept)0.070.02 – 0.23<0.001EFN Count 7 91.071.03 – 1.12<0.001Observations58R2 conditional / R2 marginalNA / 0.242 EFN Count 7.9PredictorsIncidence Rate RatiosCIp(Intercept)3.261.81 – 5.87<0.001aboveground biomass9.274.24 – 20.28<0.001aboveground biomass^20.510.39 – 0.66<0.001ped inbreeding c0.890.82 – 0.960.004ped inbreeding c^20.930.85 – 1.020.144rhizobia [Y]0.890.69 – 1.140.345ped inbreeding c^2 *rhizobia [Y]1.111.00 – 1.240.056Random Effectsσ20.07τ00 planting_date_julian0.12ICC0.63N planting_date_julian16Observations50Marginal R2 / Conditional R20.577 / 0.845 
  
 
Table S4.12. Model output for best model of herbivory. 

 138 

 herbivory percPredictorsEstimatesCIpCount Model(Intercept)0.070.04 – 0.12<0.001T ant1.020.64 – 1.630.940ped inbreeding c1.211.01 – 1.460.043ped inbreeding c^21.231.08 – 1.410.002rhizobia [Y]1.080.59 – 2.000.800ped inbreeding c^2 *rhizobia [Y]0.860.75 – 0.990.038(Intercept)13276649987.500.41 – 1.13Zero-Inflated Model(Intercept)0.320.15 – 0.650.002T ant1.320.51 – 3.400.567ped inbreeding c0.680.41 – 1.130.135Random Effectsσ20.28τ00 planting_date_julian0.02ICC0.07N planting_date_julian17Observations58Marginal R2 / Conditional R20.472 / 0.509