- Email: [email protected]
Environmental effects on social interaction networks and male reproductive behaviour in guppies, Poecilia reticulata
Animal Behaviour 81 (2011) 551e558
Contents lists available at ScienceDirect
Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav
Environmental effects on social interaction networks and male reproductive behaviour in guppies, Poecilia reticulata M. Edenbrow a, b, *, S.K. Darden a, I.W. Ramnarine c,1, J.P. Evans d, 2, R. James e, 3, D.P. Croft a a
Centre for Research in Animal Behaviour, School of Psychology, Washington Singer Labs, University of Exeter School of Biological Sciences, College of Natural Sciences, Bangor University, U.K. c Department of Life Sciences, University of the West Indies d Centre for Evolutionary Biology, School of Animal Biology (M092), University of Western Australia e Department of Physics, University of Bath b
a r t i c l e i n f o Article history: Received 20 April 2010 Initial acceptance 6 July 2010 Final acceptance 26 November 2010 Available online 21 January 2011 MS. number: 10-00274R Keywords: guppy harassment network Poecilia reticulata predation risk reproductive strategy sexual network sexual selection social network strategy
In social species, the structure and patterning of social interactions have implications for the opportunities for sexual interactions. We used social network analysis to explore the effect of habitat structural complexity on the social and sexual behaviour of male Trinidadian guppies. We used replicated seminatural pools in which we quantified male social network structure and reproductive behaviour under simple and complex habitats. In addition, we compared two populations of guppies that differed in their evolutionary history of predation (one high, one low). The level of habitat complexity did not significantly affect social network structure. However, social networks differed significantly between populations, which we suggest is due to differences in predator experience. Males from the high-predation population had greater overall social network differentiation and fewer maleemale associations than their low-risk counterparts. Contrary to our prediction that males would associate more frequently with relatively large (more fecund) females, we observed a negative correlation between female size and the strength of maleefemale associations. We also found no effect of population or habitat complexity on either harassment or sexual network structures. There was, however, a significant interaction between habitat structure and population on the expression of reproductive strategies, with high-predation males expressing fewer sigmoid displays in the complex habitat and the opposite trend in low-predation males. We suggest this pattern is driven by population differences in maleemale competition. We discuss our results in the context of the evolution of social structure and male reproductive strategies. Ó 2010 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Differences between males and females in sexual strategies are largely determined by differences in reproductive investment (Bateman 1948; Trivers 1972; Clutton-Brock & Parker 1995). Males have evolved a number of precopulatory behaviours that are used by females during mate choice, such as courtship displays and territory defence (Houde 1997; Hoglund & Sheldon 1998). However, male mating strategies have also evolved to subvert female choice. For example, males may parasitize another male’s investment (Taborsky 1997) and/or coerce females to obtain matings (Gross 1996). The individual expression of mating tactics in * Correspondence: M. Edenbrow, Centre for Research in Animal Behaviour, School of Psychology, Washington Singer Labs, University of Exeter, Perry Road, Exeter EX4 4QG, U.K. E-mail address: [email protected] (M. Edenbrow). 1 I. W. Ramnarine is at the Department of Life Sciences, University of the West Indies, St Augustine, Trinidad & Tobago. 2 J. P. Evans is at the Centre for Evolutionary Biology, School of Animal Biology (M092), University of Western Australia, Crawley, WA 6009, Australia 3 R. James is at the Department of Physics, University of Bath, Bath BA2 7AY, U.K.
many species is plastic, with males alternating between strategies to maximize reproductive potential (Godin 1995). The adoption of one reproductive strategy over another is often dependent upon multiple social and environmental stimuli (Magurran & Seghers 1994b; Endler 1995; Godin 1995; Kokko & Rankin 2006). For example, in guppies individual males employ two tactics interchangeably to attain a successful mating (Godin 1995). First, during courtship males adopt an ‘S’ shape, known as a sigmoid display, to display their bright colour patterns to females. Second, males may attempt unsolicited matings by thrusting their intromittent organ, the gonopodium, towards the female’s genital pore without prior display. It has been shown that in high-predation localities males engage in gonopodial thrusts more than sigmoid displays (Luyten & Liley 1985; Godin 1995), which is likely to be caused by males exploiting the female’s preoccupation with predator avoidance under high-risk conditions (Evans et al. 2002). By contrast, in lowpredation localities males exhibit the opposite trend, engaging in significantly fewer gonopodial thrusts and more sigmoid displays that are longer in duration (Luyten & Liley 1985).
0003-3472/$38.00 Ó 2010 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.anbehav.2010.11.026
552
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
In social species, the structure and patterning of social interactions, that is, who meets whom, will have implications for the opportunity for sexual interactions and thus male and female reproductive strategies. For example, increased social mixing and maleemale interactions have the potential to increase competition for access to females and it has been shown that males adjust the frequency of their displays in response to competitors (Farr 1980; Houde 1988). Furthermore, the level of maleemale competition has the potential both to facilitate and to hinder precopulatory female choice (reviewed in Wong & Candolin 2005). Previous work suggests that habitat complexity may play an important role in influencing patterns of social interactions. For example, Orpwood et al. (2008) compared the shoaling behaviour of European minnows, Phoxinus phoxinus, between simple and complex habitats in the presence of a predator. Minnows formed larger shoals in simple habitats when exposed to predators, supporting the hypothesis that individuals shoal as a sheltering mechanism, that is, the selfish herd principle (Hamilton 1971) and/or the dilution effect (reviewed in Krause & Ruxton 2002). Habitat complexity may also influence the frequency and outcome of behavioural interactions between individuals. For example, Hibler & Houde (2006) demonstrated that the structural complexity of the environment plays an important role in sexual interactions of guppies. In the presence of visual barriers, male interference competition and courtship displays decreased while female responsiveness increased. This was attributed to structural complexity increasing privacy, benefiting both males, via reduced competition, and females, via increased ability to assess males in the absence of interference by other males. The structural complexity of the environment can therefore influence both social and sexual interactions, as well as reproductive strategies employed by males in a population. Currently, however, we know very little about how the ecological environment influences the fine-scale population social structure (i.e. distribution of social association/interactions) and what implications this structure has for male reproductive strategies. The ecological environment also typically covaries with the physical environment, and therefore any attempt to understand their relative influence and interactive effects demands careful experimentation. In this study, we used social network analysis to examine how habitat structural complexity, population differences and their interactive effects influence the population social structure (i.e. who meets whom, based on spatial proximity of individuals) and sexual interactions (courtship intensity and harassment; defined as sigmoid displays and combined scores of nips, chases and gonopodial thrust events, respectively) in guppies. We also examined the effect of habitat structural complexity on male reproductive strategies (the rate/min of sigmoid displays and gonopodial thrusts) allowing us to explore the relationship between the ecological environment, the social environment and male reproductive behaviour. Guppies have a promiscuous mating system (Houde 1997), categorized as female-based polygyny, in which males compete with one another for female mating access with no parental care contribution (Kodric-Brown 1985). In this mating system, females are sexually receptive for short periods (Liley & Wishlow 1974) and during this receptive phase females solicit copulations with multiple males via behavioural (Liley 1966), visual and chemical cues (Crow & Liley 1979; Guevara-Fiore et al. 2009), resulting in mixed-paternity broods (Neff et al. 2008). Previous work has shown that the structural complexity of the habitat can influence both the shoaling behaviour of fish (see Mikheev 2009 for a discussion) and reproductive strategies in male guppies (Hibler & Houde 2006). In our approach, we compared the structure of social and sexual interaction networks and male reproductive behaviour under two
environmental conditions: (1) a structurally simple habitat and (2) a structurally complex habitat. Furthermore, we examined these effects for two populations from the same river drainage that differ in their evolutionary history of predation risk. We hypothesized that the structural complexity of a habitat will have an effect on both social and sexual network structure, as well as male reproductive strategies. Previous studies have shown that visual obstructions can reduce predator detection abilities and increase risk sensitivity (Schooley et al. 1996; Whittingham et al. 2004; Devereux et al. 2006). By contrast, visual barriers may also present a ‘safe haven’ and reduce perceived risk (Candolin & Voigt 1998; Dzieweczynski & Rowland 2004). We therefore predicted that, in our structurally complex habitat, high-predation fish will move less freely within their environment, either because these structures present a refuge or because they limit the ability to detect risk. We therefore expected that social mixing, encounter frequencies and sexual interactions would be limited in this habitat type, with the opposite pattern expected in the simple habitat. Moreover, we predicted that low-predation individuals would have higher levels of social mixing and more evenly distributed sexual interactions regardless of habitat structural complexity. As our study was restricted to two populations, we lacked the replication needed to draw general conclusions regarding the effect of predation risk on social network structure or male reproductive behaviour. However, our approach did allow us to explore the potential for population differences in social network structure and male reproductive behaviour, and, importantly, to gain insights into how these traits respond to different levels of environmental structural complexity. To our knowledge, this is the first study to compare social network structure between different populations of conspecifics. Furthermore, as far as we are aware, this is also the first study to implement focal animal sampling for the construction of animal social networks. This approach is especially novel in the fact that, as well as association data, we could also extract data on behavioural interactions during these associations simultaneously, which would otherwise not be possible with the point sampling methods generally implemented in animal social network studies (Croft et al. 2008). METHODS Experiments were conducted in June 2007. We collected guppies from two sampling sites in the Aripo River, Northern Mountain Range of Trinidad: high predation (10 390 27N, 61130 34W) and low predation (10 400 49N, 61130 44W). Fish were collected using seine net (1 1 m, mesh size 3 mm) sweeps within single pools and transferred to covered 30-litre buckets containing 10 litres of river water (approximately 30e40 individuals per bucket). All of the pools in high- and low-predation localities were similar in their structural complexity. These pools were characteristically open bodies of water with few visual obstructions (such as vegetation and large rocks). They were selected to reduce the potential for confounding experiential effects owing to natural structural differences between populations. However, wild guppies were not restricted to the specific pools from which they were collected, and thus we were unable to control for an individual’s previous experience of habitat structural complexity. Adult guppies were transported to the University of the West Indies (within 45 min of capture) and, upon arrival, fish (125 individuals) were transferred immediately to an outdoor holding pool (diameter 244 cm, average water depth 20 cm) and allowed to acclimatize for 24 h. During the study 500 fish were collected (250 per population), of which 480 were used in trials. Although guppies are prolific and found in high densities in tributaries of the Northern Mountain Range, because of the large number of individuals used in
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
this study we collected fish from four pools (two per population/125 individuals from each pool) on 4 days separated by 1 week to lessen the impact on local population densities. The high-predation locality (lower Aripo) has multiple guppy predators including pike cichlids, Crenicichla sp., blue acara, Aequidens pulcher, and wolf fish, Hoplias malabaricus, while the low-predation site (upper Aripo) primarily features the killifish, Rivulus hartii, a minor guppy predator (Mattingly & Butler 1994; Botham et al. 2006; Croft et al. 2006b). Previous studies on these two populations have shown increased shoaling tendency in lower Aripo populations compared to their upper Aripo counterparts (Seghers 1974; Magurran & Seghers 1990). Populations also differ in productivity (Houde 1997; Magurran 2005) and these population-level differences in resource availability may influence sociality; for example, group formation has the potential to increase foraging success (Krause & Ruxton 2002). Two trial arenas, one defined as simple and one as complex, were constructed using artificial pools to simulate natural river structures (diameter 180 cm, water depth 12 cm; see Fig. 1). The simple structural habitat was assembled using six large flat stones (10 cm diameter) and eight small flat stones (5 cm diameter), collected from the Aripo River, and organized as shown in Fig. 1a. The complex structural habitat had the same assemblage of stones as the simple structural habitat, with the exception that visual barriers were also added to provide a series of repeated visual obstructions (Fig. 1b) similar to those described by Hibler & Houde (2006). Six replicates per habitat treatment (structurally complex and structurally simple) per population were performed giving a total of 24 trials. Each trial involved 10 male and 10 female guppies. Females were represented by three size classes, small, medium and large with a ratio of 3:4:3, and all males were size matched within trials (2 mm) and between populations. As adult body size is known to differ as a function of predation pressure, with larger females occurring under lower predation risk (Reznick & Endler 1982), size classes were chosen relative to the mean body size of sexually mature females in the population (high predation: large ¼ 29e32 mm, medium ¼ 26e28.9 mm, small ¼ 24e25.9 mm; low predation: large ¼ 32e36 mm, medium ¼ 29e31.9 mm, small ¼ 26e28.9 mm). Differences in body sizes between populations have been suggested to result from size-selective predation as a function of predator assemblage, with high-predation localities selecting for fast growth and smaller size at sexual maturity (Reznick & Endler 1982). All individuals within a trial were individually marked using visible implant elastomer (see Croft et al. 2003 for details). Following marking, fish were transferred to the test arena and allowed 36 h for acclimatization (2 nights and 1 day). Data Collection Procedure Prior to beginning the sampling period, the observer stood 0.5 m from the side of the pool for 10 min to allow fish to acclimatize to his presence (10 min was deemed to be sufficient for all fish to regain activity and freely move within the arena). For the rest of the sampling period, the observer remained in this position without moving to avoid startling the test fish. Previous work using this method has demonstrated that it does not have a significant effect on the behaviour of the test fish (e.g. Croft et al. 2006a; Darden et al. 2009). Using focal animal sampling (Altmann 1974), the observer sampled each male in a trial for 5 min, giving 50 min total observations per trial. During each focal follow, the observer recorded the number of social encounters, defined as fish that approached within four body lengths (Croft et al. 2004b) of the focal individual, and the identity of these individuals encountered. The frequency of sigmoid displays, gonopodial thrusts, chases and nips as well as the identity of the individual that the behaviour was directed towards were also recorded. Data were collected using a Dictaphone (Creative MuVo
553
(a)
(b) Figure 1. Schematic diagram of the trial arena structure. (a) Simple habitat structure and (b) complex habitat structure. In the complex habitat two sizes of barrier (large and small) were created and alternated around the arena. Each of the large barriers measured 55 cm long of which 20 cm was turned at 90 degrees creating an L shape. The small visual barriers measured 35 cm long with 15 cm being turned at 90 degrees. Both were 12 cm high.
TX) and extracted using a Palm pilot (Palm Tungsten E2). Behavioural data collection software, FIT version 3.1 coupled with FIT Manager version 3.1.1 (Held & Manser 2005), was used to extract association matrices, courtship and harassment matrices/frequencies as well as association duration data. Analysis Procedure Social networks We used a Newman weighted association index (Newman 2001) to quantify pairwise associations in which pairs of individuals
554
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
observed in a group of size g (within four body lengths of the focal individual) are given a weighting, 1/(g 1), reflecting the fact that a given pair in a small group is more likely to be interacting than a pair in a large group. We used this association index because although population size per pool was identical (i.e. 10 males and 10 females), groups of fish fragment into smaller shoals and thus shoal sizes differed both within pools and between trials. These association weights were then accumulated over the sampling period and the matrix of associations was used to construct social networks (Croft et al. 2008). For each network, we calculated mean weighted network degree (weighted by total number of interactions with conspecifics), mean network tie strength (calculated as the weighted degree divided by the number of association partners), and mean social network differentiation, calculated as the coefficient of variation, CV (mean associations per individual divided by the standard deviation of these associations). Each of these measures was averaged across all males per trial generating a network average. We further subdivided these network measures by both maleemale and maleefemale associations to explore intra- and intersexual associations. The effects of habitat and population on social network structure were explored using multivariate general linear model statistics (GLM) in which mean network measures were used as response variables. To compensate for the reduced test power resulting from multiple fixed factors in our GLM, we used backward elimination to remove nonsignificant independent variables using the criterion a < 0.1. In female guppies, fecundity is positively correlated with body size and previous laboratory work has shown that males prefer to mate with large females (Dosen & Montgomerie 2004). In the current investigation, we examined the relationship between female body length and maleefemale social and sexual interactions by correlating network measures with female body length using Spearman rank correlation. Sexual interaction networks We constructed two types of sexual interaction network based on either harassment behaviours or courtship behaviours. Harassment networks were constructed based on a combined score of intersexual interactions involving gonopodial thrusts, nips and chase events. The strength of the network tie was based on the number of such events between a given male and female. Courtship networks were constructed based on intersexual interactions during sigmoid displays. The strength of the network tie was based on the number of sigmoid displays a given male directed at a given female. To quantify the structure of the intersexual harassment and courtship networks, we calculated the mean weighted network degree, mean tie strength, mean differentiation and mean unweighted network exclusivity (calculated as the number of males in addition to the focal male interacting with a given female; this exclusivity measure has an inverse relationship, with higher values denoting lower exclusivity; Sih et al. 2009). The effects of habitat and population on social network structure and reproductive strategy were explored using GLM in which network measures and reproductive strategy rate/min were used as response variables. Backward elimination was again used where appropriate using the same criterion as in our social network analysis. Significance levels were held at 0.05, although marginally nonsignificant values within our simplified models (following backward elimination procedures) are discussed (see Stoehr 1999). To explore each dependent variable within these marginally nonsignificant models, we also calculated effect sizes and 95% confidence intervals (CIs) using Cohen’s d (see Nakagawa & Cuthill 2007 for a discussion). Reporting effect size and CIs is considered a preferable approach to retrospective power analysis (Steidl et al. 1997; Thomas 1997) because it allows effective interpretation of nonsignificant/ marginally nonsignificant results (Nakagawa & Cuthill 2007).
Ethical Note All practices adhered to the guidelines for ethical research of the University of the West Indies, Trinidad. One necessary aspect of social network studies is the ability to identify individuals accurately to explore fine-scale social interactions. Female guppies do not possess natural colour markings that could be used for identification. The free-ranging nature of the trial arenas coupled with observing interactions from above also made identification of males based upon colour variation impossible. It was therefore necessary to mark fish with unique identification marks (see above) following anaesthetization with MS-222 (Croft et al. 2003). Fish were observed regularly following the anaesthetization/marking procedure and normal swimming behaviour was rapidly resumed with no sign of adverse effects upon individual behaviour. Upon completion of experiments, all test fish were released into a large artificial pond at the University of the West Indies, Trinidad. Although nips and chases were recorded, no physical damage resulting from them was observed in any of the 240 females. These behaviours were not exhibited during aggressive interactions; instead, nips are often directed towards a female’s gonopore to determine her responsiveness via pheromonal cues (Herdman et al. 2004; Guevara-Fiore et al. 2010), whereas chasing may facilitate isolation of females from competitors (Magellan et al. 2005) and thus these behaviours represent assessment and/or coercive tactics and not direct aggression. In addition, fish were only housed in artificial pools for a maximum of 10 days, being released into a large artificial pond following each trial. Furthermore, guppies, as a species, are well documented as rarely showing aggressive behaviour in free-swimming situations; instead, individuals are often observed jockeying with one another without engaging in outright aggressive interactions (Houde 1997). Thus the potential for adverse effects and/or physical damage in our study was minimal. However, as a precaution, fish health during this study was monitored twice daily (morning and afternoon) as well as during trials. We also used sex ratios and stocking densities that are similar to those observed in natural pool formations in the wild and thus harassment levels are unlikely to have exceeded those occurring in natural conditions. During the study, no adverse effects of male harassment on test fish health were observed. RESULTS Social Network Structure We found a nonsignificant tendency for an effect of population on social network structure following backward elimination of nonsignificant fixed factors (GLM: F3,18 ¼ 2.828, P ¼ 0.068). This effect was driven by a significant difference in the social differentiation (CV) between populations (GLM: mean square, MS ¼ 0.037, F3,18 ¼ 5.067, P ¼ 0.036, d ¼ 0.96, 95%CI ¼ 0.90 to 1.12) with fish originating from the population experiencing high predation exhibiting greater levels of social differentiation (Fig. 2), a conclusion that is supported by the observed large effect size and narrow positive CI range. Weighted network degree (GLM: MS ¼ 1.382, F3,18 ¼ 0.781, P ¼ 0.387, d ¼ 0.37, 95%CI ¼ 0.43 to 1.12) and tie strength (GLM: MS ¼ 70.333, F3,18 ¼ 0.764, P ¼ 0.392, d ¼ 0.37, 95%CI ¼ 6.38 to 5.11) did not differ significantly between populations and exhibited weak effect sizes and CIs spanning zero (null hypothesis). To explore social network structure further, we subdivided the networks into inter- and intrasexual associations. No significant effect of population or habitat complexity was observed on the structure of maleefemale social networks (GLM: population: F3,15 ¼ 1.055, P ¼ 0.397; habitat: F3,15 ¼ 0.435, P ¼ 0.731; population*habitat: F3,15 ¼ 0.734, P ¼ 0.548). In contrast maleemale
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
Table 1 The results of Spearman rank correlations (rS) investigating the relationship between female body length and the number of male social partners with which she was observed to associate
Mean differentiation (CV)
1.2 1 0.8 0.6 0.4
Source: population/habitat structure
rS
N
High predation Low predation Complex Simple
0.157 0.365 0.242 0.411
100 120 110 110
P 0.118 <0.001 0.011 <0.001
The results are split as a function of population and habitat complexity.
0.2 0
555
High predation
Low predation
Male Reproductive Behaviour
Population
social network structure exhibited a nonsignificant tendency for an effect of population on social network structure following backward elimination of population*habitat interaction and habitat (GLM: F3,18 ¼ 2.578, P ¼ 0.086). This effect was driven by a significant difference in maleemale weighted degree between populations (GLM: MS ¼ 8.798, F3,18 ¼ 8.481, P ¼ 0.009, d ¼ 1.2, 95%CI ¼ 0.60 to 1.77) with males from the low-predation habitat having a higher average weighted degree, which is supported by the observed large effect size and a narrow positive CI range (Fig. 3). Tie strength (GLM: MS ¼ 16.392, F3,18 ¼ 0.146, P ¼ 0.706, d ¼ 0.16, 95%CI ¼ 7.71 to 4.97) and social differentiation (GLM: MS ¼ 0.001, F3,18 ¼ 0.168, P ¼ 0.696, d ¼ 0.17, 95%CI ¼ 0.12 to 0.22) did not differ significantly between populations and exhibited weak/near zero effect size. We observed significant negative correlations between maleefemale social network degree and female body size in both the simple and complex habitats (Table 1) demonstrating that males associated more frequently with smaller females. A similar pattern was observed when the data were split by predation risk, although this was only significant for low-predation fish (Table 1). Harassment and Courtship Networks In contrast to the structure of the social networks, we found no effect of population, habitat or the population*habitat interaction on the structure of harassment networks (population: F2,18 ¼ 2.109, P ¼ 0.139; habitat: F2,18 ¼ 1.117, P ¼ 0.371; population*habitat: F2,18 ¼ 0.835, P ¼ 0.494) or courtship networks (population: F2,18 ¼ 0.289, P ¼ 0.833; habitat: F2,18 ¼ 0.663, P ¼ 0.587; population*habitat: F2,18 ¼ 2.381, P ¼ 0.108).
Mean differentiation (CV)
1.2 1 0.8 0.6 0.4
We found that the duration of maleefemale associations differed significantly between populations (MS ¼ 236.36, F1,19 ¼ 5.353, P ¼ 0.032) and with the interaction between population and habitat structure (MS ¼ 187.20, F1,19 ¼ 4.204, P ¼ 0.053). Overall, the longest maleefemale association durations occurred in high-predation individuals, with the interaction between population and habitat resulting in high-predation males spending more time with females in complex habitats, and low-predation males spending more time with females in simple habitats (Fig. 4). However, when we examine the effect sizes for population effects (d ¼ 0.96, 95%CI ¼ 3.92 to 4.13) and habitat*population interactions (high predation/simpleecomplex habitat: d ¼ 0.83, 95%CI ¼ 7.07 to 4.68; low predation/simpleecomplex habitat: d ¼ 0.92, 95%CI ¼ 4.33 to 3.68; complex habitat/highelow predation: d ¼ 1.64, 95%CI ¼ 6.26 to 4.40; simple habitat/highelow predation: d ¼ 0.12, 95%CI ¼ 3.72 to 5.37), it is clear that although effects are generally large, CIs are wide and encompass zero (null hypothesis), which is probably the result of our small sample size. We found a nonsignificant tendency for an effect of the interaction between population and habitat on male reproductive strategies (population: F2,18 ¼ 1.775, P ¼ 0.198; habitat: F2,18 ¼ 0.244, P ¼ 0.786; population*habitat: F2,18 ¼ 3.008, P ¼ 0.075). This was due to an effect on the rate that males used sigmoid displays towards females (MS ¼ 0.246, F1,19 ¼ 4.705, P ¼ 0.043) which for males from the high-predation site was highest in the simple habitat and for males from the low-predation site was highest in the complex habitat (Fig. 5).When we examine the effect sizes and CIs for each of
30 High predation Mean male−female association duration (s)
Figure 2. Mean SE social network differentiation for whole networks separated by population of origin.
Low predation
25
20
15
10
5 0.2 0
High predation
Low predation Population
Figure 3. Mean SE social network degree during maleemale encounters separated by population of origin.
0
Complex
Simple Habitat
Figure 4. Mean SE duration that males spent associating with females separated by population of origin and habitat complexity (simple and complex).
556
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
these interactions (high predation/simpleecomplex habitat: d ¼ 1.0, 95%CI ¼ 0.76 to 1.11; low predation/simpleecomplex habitat: d ¼ 0.83, 95%CI ¼ 0.59 to 0.95; complex habitat/highelow predation: d ¼ 1.08, 95%CI ¼ 0.84 to 1.20; simple habitat/highelow predation: d ¼ 0.71, 95%CI ¼ 0.47 to 0.84) it is evident that these effects mirror the differences depicted in Fig. 5. Furthermore, the narrow and positive CI ranges further strengthen our confidence that interactive effects between habitat complexity and population of origin differentially affect male reproductive strategies. DISCUSSION Our approach has allowed us to study replicated social and sexual networks within a population and suggests that male guppies originating from sites characterized by high- and lowpredation risk exhibit differences in social network structure. Males originating from the high-predation site formed social networks that were more socially differentiated (i.e. the strength of interactions was less equally distributed) and had a tendency to associate with fewer other males. In addition, we observed that males from the high-predation area performed more sigmoid displays in the structurally simple habitat, whereas males from the low-predation area performed more sigmoid displays in the structurally complex habitat. Social and Sexual Interaction Networks Population differences in shoaling tendency are well documented in guppies originating from the same river used in our study (Seghers 1974; Magurran & Seghers 1990), with highpredation fish generally forming larger social groups (reviewed in Magurran 2005). These population differences in shoaling are suggested to provide individuals with multiple antipredator benefits (Neill & Cullen 1974; Fuiman & Magurran 1994; reviewed in Krause & Ruxton 2002; Tosh et al. 2006). In the current investigation, our results suggest that population differences, which are probably related to variation in predation risk between localities, were a significant factor influencing male social network structure. Overall, males from the high-predation site had more differentiated social interactions. This larger mean differentiation score in the
0.8
High predation Low predation
Mean courtship rate/min
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
Complex
Simple Habitat
Figure 5. Mean SE courtship interaction rates of males separated by population of origin and habitat complexity (simple and complex).
high-predation population indicates greater fracturing of the social network and less evenly distributed associations with conspecifics of both sexes. By contrast, low-predation fish (lower mean differentiation) had greater social mixing and more homogeneous social associations with conspecifics. Closer examination of the inter- and intrasexual social networks reveals some interesting patterns. In particular, we observed that maleemale weighted degree differed between populations, with males from the high-predation population having lower values. This result is surprising, especially when we consider the benefits of phenotypically matched shoal formations under increased predation risk (McRobert & Bradner 1998). There may, however, be benefits for males associated with avoiding other male conspecifics. For example, by reducing malee male competition, males may be better able to isolate females from competitors and reduce interference competition (Hibler & Houde 2006). Another factor that may contribute to the observed patterns is risk sensitivity. Males from high-predation locations may reduce movements in the habitat to avoid encounters with predators (reviewed in Lima & Dill 1990) thereby negatively influencing encounter rates with other males. In contrast, low-predation males exhibited greater social mixing, which is likely to increase maleemale competition and have implications for male reproductive success. We found that the degree of habitat complexity used in this study had no significant effect upon social network structure. This result is surprising when we consider that habitat complexity has multiple effects on fish behaviour. For example, structurally complex habitats are used as refuges from predators (Werner et al. 1983), with shoaling tendency and shoal sizes increasing in simple habitats (Orpwood et al. 2008). A possible explanation for the nonsignificant effect of habitat complexity on social network structure is that the level of complexity used in this study was not sufficient to result in a difference in interaction patterns (e.g. changes in shoal size). However, habitat complexity did have a significant effect on male reproductive behaviour, which we suggest is due to risk sensitivity (see below for a discussion). Future work examining the social network structure of fish populations under a wider range of habitat structural complexity would therefore be rewarding. While we observed a significant effect of population on the male social network structure, there was no effect of population or habitat complexity on the structure of the maleefemale social network or harassment or courtship networks. Again, this is surprising given the well-documented population differences in shoaling behaviour between these two populations (Seghers 1974; Magurran & Seghers 1990). Research has indicated that female mate choice is influenced by social factors (Dugatkin & Godin 1992; Doutrelant & McGregor 2000; reviewed in Galef & White 2000; Doutrelant et al. 2001; Godin et al. 2005). The structure of maleefemale social networks is therefore likely to have implications for the information available to females when choosing a mate (Sih et al. 2009). In the current investigation, however, maleefemale network structure did not differ between populations, suggesting that females have similar exposure to information on which to make such decisions. It would be interesting to explore the patterns of maleefemale associations across populations living in natural habitats. It is well documented that in populations that experience high-predation risk, male and female guppies use different habitats, with females more frequently being found in deeper water than males (Croft et al. 2004a, 2006b). No such segregation is observed in low-predation habitats (Croft et al. 2004a, 2006b). Thus it is likely that when females have the opportunity to segregate from males spatially, the structure of the maleefemale social networks may look very different.
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
In fish, female fecundity is generally positivity correlated with body size and previous work has shown that male guppies prefer to associate with larger females (Dosen & Montgomerie 2004). Despite this, the current results reveal that in free-swimming situations males associated more frequently with smaller females even when larger females were available. One possible explanation for this pattern is that the energetic costs for males associating with larger females over a prolonged duration may exceed those for smaller females. In fish, optimal foraging rates and swimming speeds are size dependent (Beamish 1978; Hjelm & Persson 2001) and individuals in a group of conspecifics with dissimilar body lengths may be forced to travel and forage at suboptimal speeds, potentially incurring an energetic cost, which may generate selection for group assortment by body size during shoaling (Ruckstuhl 2007). Reproductive Behaviour We observed that male guppies from the population characterized by high-predation risk associated with females for longer in the complex habitat, whereas those originating from the lowpredation locality exhibited opposite trends. This result, although significant, must be interpreted cautiously when generalizing trends shaping guppy associations. Even though effect sizes were large, indicating that population and habitat complexity had a substantial influence upon association duration, confidence intervals surrounding these effect sizes were wide and spanned zero, indicating that a null result is possible. The wide confidence intervals are likely to be the result of a small sample size and greater replication would be expected to reduce this range, as described by Nakagawa & Cuthill (2007). With this in mind we suggest a possible scenario that could explain the direction of the observed results and urge researchers to explore this further. We propose that population differences in male association durations with females may be underpinned by risk-sensitive behaviour (Seghers 1974; Breden et al. 1987; Magurran & Seghers 1994a). In some species, visual obstructions reduce predator detection abilities generating risk sensitivity (Schooley et al. 1996; Whittingham et al. 2004; Devereux et al. 2006) and in others, visual barriers present a ‘safe haven’ reducing perceived risk (Candolin & Voigt 1998; Dzieweczynski & Rowland 2004). Thus males that originate from areas of high predation may be less willing to leave a female because the probability of locating future mates in a complex habitat may be reduced (Candolin & Voigt 2001) while the potential for encountering a predator may be higher (Magnhagen 1991). Low-predation fish exhibited the reversed pattern, which we suggest is due to maleemale competition. In the social networks, low-predation fish were observed to have a higher weighted network degree, suggesting that levels of maleemale competition for access to females may be higher. In the complex habitat, groups of low-predation fish were observed to encounter females, followed by rapid interactions and subsequent fission (M. Edenbrow, personal observation), potentially limiting association duration for the majority of males. Further work investigating the direct mechanisms driving these population differences would certainly be rewarding and provide greater understanding of how malee female association duration differs as a function of population and habitat complexity. A similar populationehabitat interaction was observed with male reproductive strategies. Males originating from the highpredation population performed more sigmoid displays in the simple habitat whereas low-predation males performed more sigmoid displays in the complex habitat. The observed results for high-predation fish are consistent with previous laboratory findings on guppies by Hibler & Houde (2006) who observed that in the
557
presence of visual barriers, males remain with females until successfully isolated from conspecifics, which may increase female receptivity and reduce male courtship rates (Hibler & Houde 2006). However, in the current investigation, we observed the opposite pattern in males that originated from the low-predation population, which we suggest is driven by maleemale competition. As noted above, males from the low-predation site had a higher weighted social network degree; it is therefore likely that increased maleemale competition negatively influences the success of female isolation from male conspecifics. Previous work has demonstrated that when courtship is interrupted by male interference (Houde 1988), males restart the display sequence to regain female attention (Hibler & Houde 2006), potentially increasing average display rate as a consequence. We suggest that in lowpredation fish, groups of males were able to confine females in the complex habitat, leading to increased rates of maleemale competition and higher display rates. In conclusion, our results suggest that differences in selective factors between populations have the potential to influence the structure of male social networks and sexual behaviour directly. Social network structure is likely to have implications for multiple processes that shape populations, including disease and information transfer, social learning and the maintenance of cooperation (Newman 2003; Croft et al. 2005, 2006a, 2008; Krause et al. 2007; Sih et al. 2009). Our results suggest that predation risk may influence association and in turn interaction patterns in social networks and future work exploring the generality of this pattern and its ecological and evolutionary implications would be very rewarding. Our study was restricted to two populations and thus it lacks the replication needed to draw general conclusions regarding the effect of predation risk on behaviour. However, it does lay the foundations for future studies specifically aimed at testing the generality of these findings across a broader range of populations with differing predator regimes. Our approach demonstrates the potential of adopting the social network framework to study sexual interactions and understanding how these structures relate to individual fitness, and may provide new insights into sexual selection. Acknowledgments Funding was provided to D.P.C. by the NERC (NE/E001181/1). References Altmann, J. 1974. Observational study of animal behavior: sampling methods. Behaviour, 49, 227e267. Bateman, A. J. 1948. Intra-sexual selection in Drosophila. Heredity, 2, 349e368. Beamish, F. W. H. 1978. Swimming capacity. In: Fish Physiology (Ed. by W. S. Hoar & J. D. Randall), pp. 101e187. New York: Academic Press. Botham, M. S., Kerfoot, C. J., Louca, V. & Krause, J. 2006. The effects of different predator species on antipredator behavior in the Trinidadian guppy, Poecilia reticulata. Naturwissenschaften, 93, 431e439. Breden, F., Scott, M. A. & Michel, E. 1987. Genetic differentiation for antipredator behavior in the Trinidad guppy, Poecilia reticulata. Animal Behaviour, 35, 618e620. Candolin, U. & Voigt, H. R. 1998. Predator-induced nest site preference: safe nests allow courtship in sticklebacks. Animal Behaviour, 56, 1205e1211. Candolin, U. & Voigt, H. R. 2001. Correlation between male size and territory quality: consequence of male competition or predation susceptibility? Oikos, 95, 225e230. Clutton-Brock, T. H. & Parker, G. A. 1995. Sexual coercion in animal societies. Animal Behaviour, 49, 1345e1365. Croft, D. P., Arrowsmith, B. J., Bielby, J., Skinner, K., White, E., Couzin, I. D., Magurran, A. E., Ramnarine, I. & Krause, J. 2003. Mechanisms underlying shoal composition in the Trinidadian guppy (Poecilia reticulata). Oikos, 100, 429e438. Croft, D. P., Botham, M. S. & Krause, J. 2004a. Is sexual segregation in the guppy, Poecilia reticulata, consistent with the predation risk hypothesis? Environmental Biology of Fishes, 71, 127e133. Croft, D. P., Krause, J. & James, R. 2004b. Social networks in the guppy (Poecilia reticulata). Proceedings of the Royal Society B, 271, S516eS519.
558
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
Croft, D. P., James, R., Ward, A. J. W., Botham, M. S., Mawdsley, D. & Krause, J. 2005. Assortative interactions and social networks in fish. Oecologia, 143, 211e219. Croft, D. P., James, R., Thomas, P. O. R., Hathaway, C., Mawdsley, D., Laland, K. N. & Krause, J. 2006a. Social structure and co-operative interactions in a wild population of guppies (Poecilia reticulata). Behavioral Ecology and Sociobiology, 59, 644e650. Croft, D. P., Morrell, L. J., Wade, A. S., Piyapong, C., Ioannou, C. C., Dyer, J. R. G., Chapman, B. B., Yan, W. & Krause, J. 2006b. Predation risk as a driving force for sexual segregation: a cross-population comparison. American Naturalist, 167, 867e878. Croft, D. P., James, R. & Krause, J. 2008. Exploring Animal Social Networks. Princeton, New Jersey: Princeton University Press. Crow, R. T. & Liley, N. R. 1979. Sexual pheromone in the guppy, Poecilia reticulata (Peters). Canadian Journal of Zoology, 57, 184e188. Darden, S. K., James, R., Ramnarine, I. W. & Croft, D. P. 2009. Social implications of the battle of the sexes: sexual harassment disrupts female sociality and social recognition. Proceedings of the Royal Society B, 276, 2651e2656. Devereux, C. L., Whittingham, M. J., Fernandez-Juricic, E., Vickery, J. A. & Krebs, J. R. 2006. Predator detection and avoidance by starlings under differing scenarios of predation risk. Behavioral Ecology, 17, 303e309. Dosen, L. D. & Montgomerie, R. 2004. Female size influences mate preferences of male guppies. Ethology, 110, 245e255. Doutrelant, C. & McGregor, P. K. 2000. Eavesdropping and mate choice in female fighting fish. Behaviour, 137, 1655e1669. Doutrelant, C., McGregor, P. K. & Oliveira, R. F. 2001. The effect of an audience on intrasexual communication in male Siamese fighting fish, Betta splendens. Behavioral Ecology, 12, 283e286. Dugatkin, L. A. & Godin, J. G. J. 1992. Reversal of female mate choice by copying in the guppy (Poecilia reticulata). Proceedings of the Royal Society B, 249, 179e184. Dzieweczynski, T. L. & Rowland, W. J. 2004. Behind closed doors: use of visual cover by courting male three-spined stickleback, Gasterosteus aculeatus. Animal Behaviour, 68, 465e471. Endler, J. A. 1995. Multiple-trait coevolution and environmental gradients in guppies. Trends in Ecology & Evolution, 10, 22e29. Evans, J. P., Kelley, J. L., Ramnarine, I. W. & Pilastro, A. 2002. Female behaviour mediates male courtship under predation risk in the guppy (Poecilia reticulata). Behavioral Ecology and Sociobiology, 52, 496e502. Farr, J. A. 1980. Social behavior patterns as determinants of reproductive success in the guppy, Poecilia reticulata Peters (Pisces: Poeciliidae) an experimental study of the effects of intermale competition, female choice, and sexual selection. Behaviour, 74, 38e90. Fuiman, L. A. & Magurran, A. E. 1994. Development of predator defences in fishes. Reviews in Fish Biology and Fisheries, 4, 145e183. Galef, B. G. & White, D. J. 2000. Evidence of social effects on mate choice in vertebrates. Behavioural Processes, 51, 167e175. Godin, J. G. J. 1995. Predation risk and alternative mating tactics in male Trinidadian guppies (Poecilia reticulata). Oecologia, 103, 224e229. Godin, J. G. J., Herdman, E. J. E. & Dugatkin, L. A. 2005. Social influences on female mate choice in the guppy, Poecilia reticulata: generalized and repeatable traitcopying behaviour. Animal Behaviour, 69, 999e1005. Gross, M. R. 1996. Alternative reproductive strategies and tactics: diversity within sexes. Trends in Ecology & Evolution, 11, 92e98. Guevara-Fiore, P., Skinner, A. & Watt, P. J. 2009. Do male guppies distinguish virgin females from recently mated ones? Animal Behaviour, 77, 425e431. Guevara-Fiore, P., Stapley, J. & Watt, P. 2010. Mating effort and female receptivity: how do male guppies decide when to invest in sex? Behavioral Ecology and Sociobiology, 64, 1665e1672. Hamilton, W. D. 1971. Geometry for the selfish herd. Journal of Theoretical Biology, 31, 295e311. Held, J. & Manser, T. 2005. A PDA-based system for online recording and analysis of concurrent events in complex behavioral processes. Behavior Research Methods, 37, 155e164. Herdman, E. J. E., Kelly, C. D. & Godin, J. G. J. 2004. Male mate choice in the guppy (Poecilia reticulata): do males prefer larger females as mates? Ethology, 110, 97e111. Hibler, T. L. & Houde, A. E. 2006. The effect of visual obstructions on the sexual behaviour of guppies: the importance of privacy. Animal Behaviour, 72, 959e964. Hjelm, J. & Persson, L. 2001. Size-dependent attack rate and handling capacity: inter-cohort competition in a zooplanktivorous fish. Oikos, 95, 520e532. Hoglund, J. & Sheldon, B. C. 1998. The cost of reproduction and sexual selection. Oikos, 83, 478e483. Houde, A. E. 1988. The effects of female choice and maleemale competition on the mating success of male guppies. Animal Behaviour, 36, 888e896. Houde, A. E. 1997. Sex, Color, and Mate Choice in Guppies. Princeton, New Jersey: Princeton University Press. Kodric-Brown, A. 1985. Female preference and sexual selection for male coloration in the guppy (Poecilia reticulata). Behavioral Ecology and Sociobiology, 17, 199e205. Kokko, H. & Rankin, D. J. 2006. Lonely hearts or sex in the city? Density-dependent effects in mating systems. Philosophical Transactions of the Royal Society B, 361, 319e334.
Krause, J. & Ruxton, G. D. 2002. Living in Groups. Oxford: Oxford University Press. Krause, J., Croft, D. P. & James, R. 2007. Social network theory in the behavioural sciences: potential applications. Behavioral Ecology and Sociobiology, 62, 15e27. Liley, N. R. 1966. Ethological isolating mechanisms in four sympatric species of Poeciliid fishes. Behaviour, Supplement, 13, 1e197. Liley, N. R. & Wishlow, W. 1974. Interaction of endocrine and experiential factors in regulation of sexual-behaviour in female guppy Poecilia reticulata. Behaviour, 48, 185e214. Lima, S. L. & Dill, L. M. 1990. Behavioural decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology, 68, 619e640. Luyten, P. H. & Liley, N. R. 1985. Geographic variation in the sexual behaviour of the guppy, Poecilia reticulata (Peters). Behaviour, 95, 164e179. McRobert, S. P. & Bradner, J. 1998. The influence of body coloration on shoaling preferences in fish. Animal Behaviour, 56, 611e615. Magellan, K., Pettersson, L. B. & Magurran, A. E. 2005. Quantifying male attractiveness and mating behaviour through phenotypic size manipulation in the Trinidadian guppy, Poecilia reticulata. Behavioral Ecology and Sociobiology, 58, 366e374. Magnhagen, C. 1991. Predation risk as a cost of reproduction. Trends in Ecology & Evolution, 6, 183e185. Magurran, A. E. 2005. Evolutionary Ecology: The Trinidadian Guppy. Oxford: Oxford University Press. Magurran, A. E. & Seghers, B. H. 1990. Population differences in the schooling behaviour of newborn guppies, Poecilia reticulata. Ethology, 84, 334e342. Magurran, A. E. & Seghers, B. H. 1994a. Predator inspection behaviour covaries with schooling tendency amongst wild guppy, Poecilia reticulata, populations in Trinidad. Behaviour, 128, 121e134. Magurran, A. E. & Seghers, B. H. 1994b. Sexual conflict as a consequence of ecologyeevidence from guppy, Poecilia reticulata, populations in Trinidad. Proceedings of the Royal Society B, 255, 31e36. Mattingly, H. T. & Butler, M. J. I. V. 1994. Laboratory predation on the Trinidadian guppy: implications for the size-selective predation hypothesis and guppy life history evolution. Oikos, 69, 54e64. Mikheev, V. 2009. Combined effects of predators and parasites on shoaling behavior of fishes. Journal of Ichthyology, 49, 1032e1041. Nakagawa, S. & Cuthill, I. C. 2007. Effect size, confidence interval and statistical significance: a practical guide for biologists. Biological Reviews, 82, 591e605. Neff, B. D., Pitcher, T. E. & Ramnarine, I. W. 2008. Inter-population variation in multiple paternity and reproductive skew in the guppy. Molecular Ecology, 17, 2975e2984. Neill, S. R. J. & Cullen, J. M. 1974. Experiments on whether schooling by their prey affects the hunting behavoiur of cephalopods and fish predators. Journal of Zoology, 172, 549e569. Newman, M. E. J. 2001. Scientific collaboration networks II. Shortest paths, weighted networks and centrality. Physical Review E, 64 016132, DOI: 10.1103/ PhysRevE.64.016132. Newman, M. E. J. 2003. Mixing patterns in networks. Physical Review E, 67 026126. Orpwood, J. E., Magurran, A. E., Armstrong, J. D. & Griffiths, S. W. 2008. Minnows and the selfish herd: effects of predation risk on shoaling behaviour are dependent on habitat complexity. Animal Behaviour, 76, 143e152. Reznick, D. & Endler, J. A. 1982. The impact of predation on life-history evolution in Trinidadian guppies (Poecilia reticulata). Evolution, 36, 160e177. Ruckstuhl, K. E. 2007. Sexual segregation in vertebrates: proximate and ultimate causes. Integrative and Comparative Biology, 47, 245e257. Schooley, R. L., Sharpe, P. B. & VanHorne, B. 1996. Can shrub cover increase predation risk for a desert rodent? Canadian Journal of Zoology, 74, 157e163. Seghers, B. H. 1974. Schooling behaviour in the guppy (Poecilia reticulata): an evolutionary response to predation. Evolution, 28, 486e489. Sih, A., Hanser, S. F. & McHugh, K. A. 2009. Social network theory: new insights and issues for behavioral ecologists. Behavioral Ecology and Sociobiology, 63, 975e988. Steidl, R. J., Hayes, J. P. & Schauber, E. 1997. Statistical power analysis in wildlife research. Journal of Wildlife Management, 61, 270e279. Stoehr, A. M. 1999. Are significance thresholds appropriate for the study of animal behaviour? Animal Behaviour, 57, F22eF25. Taborsky, M. 1997. Bourgeois and parasitic tactics: do we need collective, functional terms for alternative reproductive behaviours? Behavioral Ecology and Sociobiology, 41, 361e362. Thomas, L. 1997. Retrospective power analysis. Conservation Biology, 11, 276e280. Tosh, C. R., Jackson, A. L. & Ruxton, G. D. 2006. The confusion effect in predatory neural networks. American Naturalist, 167, E52eE65. Trivers, R. L. 1972. Parental investment and sexual selection. In: Sexual Selection and the Descent of Man (Ed. by B. Campbell), pp. 136e179. Chicago: Aldine-Atherton. Werner, E. E., Gilliam, J. F., Hall, D. J. & Mittelbach, G. G. 1983. An experimental test of the effects of predation risk on habitat use in fish. Ecology, 64, 1540e1548. Whittingham, M. J., Butler, S. J., Quinn, J. L. & Cresswell, W. 2004. The effect of limited visibility on vigilance behaviour and speed of predator detection: implications for the conservation of granivorous passerines. Oikos, 106, 377e385. Wong, B. B. M. & Candolin, U. 2005. How is female mate choice affected by male competition? Biological Reviews, 80, 559e571.
Contents lists available at ScienceDirect
Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav
Environmental effects on social interaction networks and male reproductive behaviour in guppies, Poecilia reticulata M. Edenbrow a, b, *, S.K. Darden a, I.W. Ramnarine c,1, J.P. Evans d, 2, R. James e, 3, D.P. Croft a a
Centre for Research in Animal Behaviour, School of Psychology, Washington Singer Labs, University of Exeter School of Biological Sciences, College of Natural Sciences, Bangor University, U.K. c Department of Life Sciences, University of the West Indies d Centre for Evolutionary Biology, School of Animal Biology (M092), University of Western Australia e Department of Physics, University of Bath b
a r t i c l e i n f o Article history: Received 20 April 2010 Initial acceptance 6 July 2010 Final acceptance 26 November 2010 Available online 21 January 2011 MS. number: 10-00274R Keywords: guppy harassment network Poecilia reticulata predation risk reproductive strategy sexual network sexual selection social network strategy
In social species, the structure and patterning of social interactions have implications for the opportunities for sexual interactions. We used social network analysis to explore the effect of habitat structural complexity on the social and sexual behaviour of male Trinidadian guppies. We used replicated seminatural pools in which we quantified male social network structure and reproductive behaviour under simple and complex habitats. In addition, we compared two populations of guppies that differed in their evolutionary history of predation (one high, one low). The level of habitat complexity did not significantly affect social network structure. However, social networks differed significantly between populations, which we suggest is due to differences in predator experience. Males from the high-predation population had greater overall social network differentiation and fewer maleemale associations than their low-risk counterparts. Contrary to our prediction that males would associate more frequently with relatively large (more fecund) females, we observed a negative correlation between female size and the strength of maleefemale associations. We also found no effect of population or habitat complexity on either harassment or sexual network structures. There was, however, a significant interaction between habitat structure and population on the expression of reproductive strategies, with high-predation males expressing fewer sigmoid displays in the complex habitat and the opposite trend in low-predation males. We suggest this pattern is driven by population differences in maleemale competition. We discuss our results in the context of the evolution of social structure and male reproductive strategies. Ó 2010 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Differences between males and females in sexual strategies are largely determined by differences in reproductive investment (Bateman 1948; Trivers 1972; Clutton-Brock & Parker 1995). Males have evolved a number of precopulatory behaviours that are used by females during mate choice, such as courtship displays and territory defence (Houde 1997; Hoglund & Sheldon 1998). However, male mating strategies have also evolved to subvert female choice. For example, males may parasitize another male’s investment (Taborsky 1997) and/or coerce females to obtain matings (Gross 1996). The individual expression of mating tactics in * Correspondence: M. Edenbrow, Centre for Research in Animal Behaviour, School of Psychology, Washington Singer Labs, University of Exeter, Perry Road, Exeter EX4 4QG, U.K. E-mail address: [email protected] (M. Edenbrow). 1 I. W. Ramnarine is at the Department of Life Sciences, University of the West Indies, St Augustine, Trinidad & Tobago. 2 J. P. Evans is at the Centre for Evolutionary Biology, School of Animal Biology (M092), University of Western Australia, Crawley, WA 6009, Australia 3 R. James is at the Department of Physics, University of Bath, Bath BA2 7AY, U.K.
many species is plastic, with males alternating between strategies to maximize reproductive potential (Godin 1995). The adoption of one reproductive strategy over another is often dependent upon multiple social and environmental stimuli (Magurran & Seghers 1994b; Endler 1995; Godin 1995; Kokko & Rankin 2006). For example, in guppies individual males employ two tactics interchangeably to attain a successful mating (Godin 1995). First, during courtship males adopt an ‘S’ shape, known as a sigmoid display, to display their bright colour patterns to females. Second, males may attempt unsolicited matings by thrusting their intromittent organ, the gonopodium, towards the female’s genital pore without prior display. It has been shown that in high-predation localities males engage in gonopodial thrusts more than sigmoid displays (Luyten & Liley 1985; Godin 1995), which is likely to be caused by males exploiting the female’s preoccupation with predator avoidance under high-risk conditions (Evans et al. 2002). By contrast, in lowpredation localities males exhibit the opposite trend, engaging in significantly fewer gonopodial thrusts and more sigmoid displays that are longer in duration (Luyten & Liley 1985).
0003-3472/$38.00 Ó 2010 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.anbehav.2010.11.026
552
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
In social species, the structure and patterning of social interactions, that is, who meets whom, will have implications for the opportunity for sexual interactions and thus male and female reproductive strategies. For example, increased social mixing and maleemale interactions have the potential to increase competition for access to females and it has been shown that males adjust the frequency of their displays in response to competitors (Farr 1980; Houde 1988). Furthermore, the level of maleemale competition has the potential both to facilitate and to hinder precopulatory female choice (reviewed in Wong & Candolin 2005). Previous work suggests that habitat complexity may play an important role in influencing patterns of social interactions. For example, Orpwood et al. (2008) compared the shoaling behaviour of European minnows, Phoxinus phoxinus, between simple and complex habitats in the presence of a predator. Minnows formed larger shoals in simple habitats when exposed to predators, supporting the hypothesis that individuals shoal as a sheltering mechanism, that is, the selfish herd principle (Hamilton 1971) and/or the dilution effect (reviewed in Krause & Ruxton 2002). Habitat complexity may also influence the frequency and outcome of behavioural interactions between individuals. For example, Hibler & Houde (2006) demonstrated that the structural complexity of the environment plays an important role in sexual interactions of guppies. In the presence of visual barriers, male interference competition and courtship displays decreased while female responsiveness increased. This was attributed to structural complexity increasing privacy, benefiting both males, via reduced competition, and females, via increased ability to assess males in the absence of interference by other males. The structural complexity of the environment can therefore influence both social and sexual interactions, as well as reproductive strategies employed by males in a population. Currently, however, we know very little about how the ecological environment influences the fine-scale population social structure (i.e. distribution of social association/interactions) and what implications this structure has for male reproductive strategies. The ecological environment also typically covaries with the physical environment, and therefore any attempt to understand their relative influence and interactive effects demands careful experimentation. In this study, we used social network analysis to examine how habitat structural complexity, population differences and their interactive effects influence the population social structure (i.e. who meets whom, based on spatial proximity of individuals) and sexual interactions (courtship intensity and harassment; defined as sigmoid displays and combined scores of nips, chases and gonopodial thrust events, respectively) in guppies. We also examined the effect of habitat structural complexity on male reproductive strategies (the rate/min of sigmoid displays and gonopodial thrusts) allowing us to explore the relationship between the ecological environment, the social environment and male reproductive behaviour. Guppies have a promiscuous mating system (Houde 1997), categorized as female-based polygyny, in which males compete with one another for female mating access with no parental care contribution (Kodric-Brown 1985). In this mating system, females are sexually receptive for short periods (Liley & Wishlow 1974) and during this receptive phase females solicit copulations with multiple males via behavioural (Liley 1966), visual and chemical cues (Crow & Liley 1979; Guevara-Fiore et al. 2009), resulting in mixed-paternity broods (Neff et al. 2008). Previous work has shown that the structural complexity of the habitat can influence both the shoaling behaviour of fish (see Mikheev 2009 for a discussion) and reproductive strategies in male guppies (Hibler & Houde 2006). In our approach, we compared the structure of social and sexual interaction networks and male reproductive behaviour under two
environmental conditions: (1) a structurally simple habitat and (2) a structurally complex habitat. Furthermore, we examined these effects for two populations from the same river drainage that differ in their evolutionary history of predation risk. We hypothesized that the structural complexity of a habitat will have an effect on both social and sexual network structure, as well as male reproductive strategies. Previous studies have shown that visual obstructions can reduce predator detection abilities and increase risk sensitivity (Schooley et al. 1996; Whittingham et al. 2004; Devereux et al. 2006). By contrast, visual barriers may also present a ‘safe haven’ and reduce perceived risk (Candolin & Voigt 1998; Dzieweczynski & Rowland 2004). We therefore predicted that, in our structurally complex habitat, high-predation fish will move less freely within their environment, either because these structures present a refuge or because they limit the ability to detect risk. We therefore expected that social mixing, encounter frequencies and sexual interactions would be limited in this habitat type, with the opposite pattern expected in the simple habitat. Moreover, we predicted that low-predation individuals would have higher levels of social mixing and more evenly distributed sexual interactions regardless of habitat structural complexity. As our study was restricted to two populations, we lacked the replication needed to draw general conclusions regarding the effect of predation risk on social network structure or male reproductive behaviour. However, our approach did allow us to explore the potential for population differences in social network structure and male reproductive behaviour, and, importantly, to gain insights into how these traits respond to different levels of environmental structural complexity. To our knowledge, this is the first study to compare social network structure between different populations of conspecifics. Furthermore, as far as we are aware, this is also the first study to implement focal animal sampling for the construction of animal social networks. This approach is especially novel in the fact that, as well as association data, we could also extract data on behavioural interactions during these associations simultaneously, which would otherwise not be possible with the point sampling methods generally implemented in animal social network studies (Croft et al. 2008). METHODS Experiments were conducted in June 2007. We collected guppies from two sampling sites in the Aripo River, Northern Mountain Range of Trinidad: high predation (10 390 27N, 61130 34W) and low predation (10 400 49N, 61130 44W). Fish were collected using seine net (1 1 m, mesh size 3 mm) sweeps within single pools and transferred to covered 30-litre buckets containing 10 litres of river water (approximately 30e40 individuals per bucket). All of the pools in high- and low-predation localities were similar in their structural complexity. These pools were characteristically open bodies of water with few visual obstructions (such as vegetation and large rocks). They were selected to reduce the potential for confounding experiential effects owing to natural structural differences between populations. However, wild guppies were not restricted to the specific pools from which they were collected, and thus we were unable to control for an individual’s previous experience of habitat structural complexity. Adult guppies were transported to the University of the West Indies (within 45 min of capture) and, upon arrival, fish (125 individuals) were transferred immediately to an outdoor holding pool (diameter 244 cm, average water depth 20 cm) and allowed to acclimatize for 24 h. During the study 500 fish were collected (250 per population), of which 480 were used in trials. Although guppies are prolific and found in high densities in tributaries of the Northern Mountain Range, because of the large number of individuals used in
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
this study we collected fish from four pools (two per population/125 individuals from each pool) on 4 days separated by 1 week to lessen the impact on local population densities. The high-predation locality (lower Aripo) has multiple guppy predators including pike cichlids, Crenicichla sp., blue acara, Aequidens pulcher, and wolf fish, Hoplias malabaricus, while the low-predation site (upper Aripo) primarily features the killifish, Rivulus hartii, a minor guppy predator (Mattingly & Butler 1994; Botham et al. 2006; Croft et al. 2006b). Previous studies on these two populations have shown increased shoaling tendency in lower Aripo populations compared to their upper Aripo counterparts (Seghers 1974; Magurran & Seghers 1990). Populations also differ in productivity (Houde 1997; Magurran 2005) and these population-level differences in resource availability may influence sociality; for example, group formation has the potential to increase foraging success (Krause & Ruxton 2002). Two trial arenas, one defined as simple and one as complex, were constructed using artificial pools to simulate natural river structures (diameter 180 cm, water depth 12 cm; see Fig. 1). The simple structural habitat was assembled using six large flat stones (10 cm diameter) and eight small flat stones (5 cm diameter), collected from the Aripo River, and organized as shown in Fig. 1a. The complex structural habitat had the same assemblage of stones as the simple structural habitat, with the exception that visual barriers were also added to provide a series of repeated visual obstructions (Fig. 1b) similar to those described by Hibler & Houde (2006). Six replicates per habitat treatment (structurally complex and structurally simple) per population were performed giving a total of 24 trials. Each trial involved 10 male and 10 female guppies. Females were represented by three size classes, small, medium and large with a ratio of 3:4:3, and all males were size matched within trials (2 mm) and between populations. As adult body size is known to differ as a function of predation pressure, with larger females occurring under lower predation risk (Reznick & Endler 1982), size classes were chosen relative to the mean body size of sexually mature females in the population (high predation: large ¼ 29e32 mm, medium ¼ 26e28.9 mm, small ¼ 24e25.9 mm; low predation: large ¼ 32e36 mm, medium ¼ 29e31.9 mm, small ¼ 26e28.9 mm). Differences in body sizes between populations have been suggested to result from size-selective predation as a function of predator assemblage, with high-predation localities selecting for fast growth and smaller size at sexual maturity (Reznick & Endler 1982). All individuals within a trial were individually marked using visible implant elastomer (see Croft et al. 2003 for details). Following marking, fish were transferred to the test arena and allowed 36 h for acclimatization (2 nights and 1 day). Data Collection Procedure Prior to beginning the sampling period, the observer stood 0.5 m from the side of the pool for 10 min to allow fish to acclimatize to his presence (10 min was deemed to be sufficient for all fish to regain activity and freely move within the arena). For the rest of the sampling period, the observer remained in this position without moving to avoid startling the test fish. Previous work using this method has demonstrated that it does not have a significant effect on the behaviour of the test fish (e.g. Croft et al. 2006a; Darden et al. 2009). Using focal animal sampling (Altmann 1974), the observer sampled each male in a trial for 5 min, giving 50 min total observations per trial. During each focal follow, the observer recorded the number of social encounters, defined as fish that approached within four body lengths (Croft et al. 2004b) of the focal individual, and the identity of these individuals encountered. The frequency of sigmoid displays, gonopodial thrusts, chases and nips as well as the identity of the individual that the behaviour was directed towards were also recorded. Data were collected using a Dictaphone (Creative MuVo
553
(a)
(b) Figure 1. Schematic diagram of the trial arena structure. (a) Simple habitat structure and (b) complex habitat structure. In the complex habitat two sizes of barrier (large and small) were created and alternated around the arena. Each of the large barriers measured 55 cm long of which 20 cm was turned at 90 degrees creating an L shape. The small visual barriers measured 35 cm long with 15 cm being turned at 90 degrees. Both were 12 cm high.
TX) and extracted using a Palm pilot (Palm Tungsten E2). Behavioural data collection software, FIT version 3.1 coupled with FIT Manager version 3.1.1 (Held & Manser 2005), was used to extract association matrices, courtship and harassment matrices/frequencies as well as association duration data. Analysis Procedure Social networks We used a Newman weighted association index (Newman 2001) to quantify pairwise associations in which pairs of individuals
554
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
observed in a group of size g (within four body lengths of the focal individual) are given a weighting, 1/(g 1), reflecting the fact that a given pair in a small group is more likely to be interacting than a pair in a large group. We used this association index because although population size per pool was identical (i.e. 10 males and 10 females), groups of fish fragment into smaller shoals and thus shoal sizes differed both within pools and between trials. These association weights were then accumulated over the sampling period and the matrix of associations was used to construct social networks (Croft et al. 2008). For each network, we calculated mean weighted network degree (weighted by total number of interactions with conspecifics), mean network tie strength (calculated as the weighted degree divided by the number of association partners), and mean social network differentiation, calculated as the coefficient of variation, CV (mean associations per individual divided by the standard deviation of these associations). Each of these measures was averaged across all males per trial generating a network average. We further subdivided these network measures by both maleemale and maleefemale associations to explore intra- and intersexual associations. The effects of habitat and population on social network structure were explored using multivariate general linear model statistics (GLM) in which mean network measures were used as response variables. To compensate for the reduced test power resulting from multiple fixed factors in our GLM, we used backward elimination to remove nonsignificant independent variables using the criterion a < 0.1. In female guppies, fecundity is positively correlated with body size and previous laboratory work has shown that males prefer to mate with large females (Dosen & Montgomerie 2004). In the current investigation, we examined the relationship between female body length and maleefemale social and sexual interactions by correlating network measures with female body length using Spearman rank correlation. Sexual interaction networks We constructed two types of sexual interaction network based on either harassment behaviours or courtship behaviours. Harassment networks were constructed based on a combined score of intersexual interactions involving gonopodial thrusts, nips and chase events. The strength of the network tie was based on the number of such events between a given male and female. Courtship networks were constructed based on intersexual interactions during sigmoid displays. The strength of the network tie was based on the number of sigmoid displays a given male directed at a given female. To quantify the structure of the intersexual harassment and courtship networks, we calculated the mean weighted network degree, mean tie strength, mean differentiation and mean unweighted network exclusivity (calculated as the number of males in addition to the focal male interacting with a given female; this exclusivity measure has an inverse relationship, with higher values denoting lower exclusivity; Sih et al. 2009). The effects of habitat and population on social network structure and reproductive strategy were explored using GLM in which network measures and reproductive strategy rate/min were used as response variables. Backward elimination was again used where appropriate using the same criterion as in our social network analysis. Significance levels were held at 0.05, although marginally nonsignificant values within our simplified models (following backward elimination procedures) are discussed (see Stoehr 1999). To explore each dependent variable within these marginally nonsignificant models, we also calculated effect sizes and 95% confidence intervals (CIs) using Cohen’s d (see Nakagawa & Cuthill 2007 for a discussion). Reporting effect size and CIs is considered a preferable approach to retrospective power analysis (Steidl et al. 1997; Thomas 1997) because it allows effective interpretation of nonsignificant/ marginally nonsignificant results (Nakagawa & Cuthill 2007).
Ethical Note All practices adhered to the guidelines for ethical research of the University of the West Indies, Trinidad. One necessary aspect of social network studies is the ability to identify individuals accurately to explore fine-scale social interactions. Female guppies do not possess natural colour markings that could be used for identification. The free-ranging nature of the trial arenas coupled with observing interactions from above also made identification of males based upon colour variation impossible. It was therefore necessary to mark fish with unique identification marks (see above) following anaesthetization with MS-222 (Croft et al. 2003). Fish were observed regularly following the anaesthetization/marking procedure and normal swimming behaviour was rapidly resumed with no sign of adverse effects upon individual behaviour. Upon completion of experiments, all test fish were released into a large artificial pond at the University of the West Indies, Trinidad. Although nips and chases were recorded, no physical damage resulting from them was observed in any of the 240 females. These behaviours were not exhibited during aggressive interactions; instead, nips are often directed towards a female’s gonopore to determine her responsiveness via pheromonal cues (Herdman et al. 2004; Guevara-Fiore et al. 2010), whereas chasing may facilitate isolation of females from competitors (Magellan et al. 2005) and thus these behaviours represent assessment and/or coercive tactics and not direct aggression. In addition, fish were only housed in artificial pools for a maximum of 10 days, being released into a large artificial pond following each trial. Furthermore, guppies, as a species, are well documented as rarely showing aggressive behaviour in free-swimming situations; instead, individuals are often observed jockeying with one another without engaging in outright aggressive interactions (Houde 1997). Thus the potential for adverse effects and/or physical damage in our study was minimal. However, as a precaution, fish health during this study was monitored twice daily (morning and afternoon) as well as during trials. We also used sex ratios and stocking densities that are similar to those observed in natural pool formations in the wild and thus harassment levels are unlikely to have exceeded those occurring in natural conditions. During the study, no adverse effects of male harassment on test fish health were observed. RESULTS Social Network Structure We found a nonsignificant tendency for an effect of population on social network structure following backward elimination of nonsignificant fixed factors (GLM: F3,18 ¼ 2.828, P ¼ 0.068). This effect was driven by a significant difference in the social differentiation (CV) between populations (GLM: mean square, MS ¼ 0.037, F3,18 ¼ 5.067, P ¼ 0.036, d ¼ 0.96, 95%CI ¼ 0.90 to 1.12) with fish originating from the population experiencing high predation exhibiting greater levels of social differentiation (Fig. 2), a conclusion that is supported by the observed large effect size and narrow positive CI range. Weighted network degree (GLM: MS ¼ 1.382, F3,18 ¼ 0.781, P ¼ 0.387, d ¼ 0.37, 95%CI ¼ 0.43 to 1.12) and tie strength (GLM: MS ¼ 70.333, F3,18 ¼ 0.764, P ¼ 0.392, d ¼ 0.37, 95%CI ¼ 6.38 to 5.11) did not differ significantly between populations and exhibited weak effect sizes and CIs spanning zero (null hypothesis). To explore social network structure further, we subdivided the networks into inter- and intrasexual associations. No significant effect of population or habitat complexity was observed on the structure of maleefemale social networks (GLM: population: F3,15 ¼ 1.055, P ¼ 0.397; habitat: F3,15 ¼ 0.435, P ¼ 0.731; population*habitat: F3,15 ¼ 0.734, P ¼ 0.548). In contrast maleemale
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
Table 1 The results of Spearman rank correlations (rS) investigating the relationship between female body length and the number of male social partners with which she was observed to associate
Mean differentiation (CV)
1.2 1 0.8 0.6 0.4
Source: population/habitat structure
rS
N
High predation Low predation Complex Simple
0.157 0.365 0.242 0.411
100 120 110 110
P 0.118 <0.001 0.011 <0.001
The results are split as a function of population and habitat complexity.
0.2 0
555
High predation
Low predation
Male Reproductive Behaviour
Population
social network structure exhibited a nonsignificant tendency for an effect of population on social network structure following backward elimination of population*habitat interaction and habitat (GLM: F3,18 ¼ 2.578, P ¼ 0.086). This effect was driven by a significant difference in maleemale weighted degree between populations (GLM: MS ¼ 8.798, F3,18 ¼ 8.481, P ¼ 0.009, d ¼ 1.2, 95%CI ¼ 0.60 to 1.77) with males from the low-predation habitat having a higher average weighted degree, which is supported by the observed large effect size and a narrow positive CI range (Fig. 3). Tie strength (GLM: MS ¼ 16.392, F3,18 ¼ 0.146, P ¼ 0.706, d ¼ 0.16, 95%CI ¼ 7.71 to 4.97) and social differentiation (GLM: MS ¼ 0.001, F3,18 ¼ 0.168, P ¼ 0.696, d ¼ 0.17, 95%CI ¼ 0.12 to 0.22) did not differ significantly between populations and exhibited weak/near zero effect size. We observed significant negative correlations between maleefemale social network degree and female body size in both the simple and complex habitats (Table 1) demonstrating that males associated more frequently with smaller females. A similar pattern was observed when the data were split by predation risk, although this was only significant for low-predation fish (Table 1). Harassment and Courtship Networks In contrast to the structure of the social networks, we found no effect of population, habitat or the population*habitat interaction on the structure of harassment networks (population: F2,18 ¼ 2.109, P ¼ 0.139; habitat: F2,18 ¼ 1.117, P ¼ 0.371; population*habitat: F2,18 ¼ 0.835, P ¼ 0.494) or courtship networks (population: F2,18 ¼ 0.289, P ¼ 0.833; habitat: F2,18 ¼ 0.663, P ¼ 0.587; population*habitat: F2,18 ¼ 2.381, P ¼ 0.108).
Mean differentiation (CV)
1.2 1 0.8 0.6 0.4
We found that the duration of maleefemale associations differed significantly between populations (MS ¼ 236.36, F1,19 ¼ 5.353, P ¼ 0.032) and with the interaction between population and habitat structure (MS ¼ 187.20, F1,19 ¼ 4.204, P ¼ 0.053). Overall, the longest maleefemale association durations occurred in high-predation individuals, with the interaction between population and habitat resulting in high-predation males spending more time with females in complex habitats, and low-predation males spending more time with females in simple habitats (Fig. 4). However, when we examine the effect sizes for population effects (d ¼ 0.96, 95%CI ¼ 3.92 to 4.13) and habitat*population interactions (high predation/simpleecomplex habitat: d ¼ 0.83, 95%CI ¼ 7.07 to 4.68; low predation/simpleecomplex habitat: d ¼ 0.92, 95%CI ¼ 4.33 to 3.68; complex habitat/highelow predation: d ¼ 1.64, 95%CI ¼ 6.26 to 4.40; simple habitat/highelow predation: d ¼ 0.12, 95%CI ¼ 3.72 to 5.37), it is clear that although effects are generally large, CIs are wide and encompass zero (null hypothesis), which is probably the result of our small sample size. We found a nonsignificant tendency for an effect of the interaction between population and habitat on male reproductive strategies (population: F2,18 ¼ 1.775, P ¼ 0.198; habitat: F2,18 ¼ 0.244, P ¼ 0.786; population*habitat: F2,18 ¼ 3.008, P ¼ 0.075). This was due to an effect on the rate that males used sigmoid displays towards females (MS ¼ 0.246, F1,19 ¼ 4.705, P ¼ 0.043) which for males from the high-predation site was highest in the simple habitat and for males from the low-predation site was highest in the complex habitat (Fig. 5).When we examine the effect sizes and CIs for each of
30 High predation Mean male−female association duration (s)
Figure 2. Mean SE social network differentiation for whole networks separated by population of origin.
Low predation
25
20
15
10
5 0.2 0
High predation
Low predation Population
Figure 3. Mean SE social network degree during maleemale encounters separated by population of origin.
0
Complex
Simple Habitat
Figure 4. Mean SE duration that males spent associating with females separated by population of origin and habitat complexity (simple and complex).
556
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
these interactions (high predation/simpleecomplex habitat: d ¼ 1.0, 95%CI ¼ 0.76 to 1.11; low predation/simpleecomplex habitat: d ¼ 0.83, 95%CI ¼ 0.59 to 0.95; complex habitat/highelow predation: d ¼ 1.08, 95%CI ¼ 0.84 to 1.20; simple habitat/highelow predation: d ¼ 0.71, 95%CI ¼ 0.47 to 0.84) it is evident that these effects mirror the differences depicted in Fig. 5. Furthermore, the narrow and positive CI ranges further strengthen our confidence that interactive effects between habitat complexity and population of origin differentially affect male reproductive strategies. DISCUSSION Our approach has allowed us to study replicated social and sexual networks within a population and suggests that male guppies originating from sites characterized by high- and lowpredation risk exhibit differences in social network structure. Males originating from the high-predation site formed social networks that were more socially differentiated (i.e. the strength of interactions was less equally distributed) and had a tendency to associate with fewer other males. In addition, we observed that males from the high-predation area performed more sigmoid displays in the structurally simple habitat, whereas males from the low-predation area performed more sigmoid displays in the structurally complex habitat. Social and Sexual Interaction Networks Population differences in shoaling tendency are well documented in guppies originating from the same river used in our study (Seghers 1974; Magurran & Seghers 1990), with highpredation fish generally forming larger social groups (reviewed in Magurran 2005). These population differences in shoaling are suggested to provide individuals with multiple antipredator benefits (Neill & Cullen 1974; Fuiman & Magurran 1994; reviewed in Krause & Ruxton 2002; Tosh et al. 2006). In the current investigation, our results suggest that population differences, which are probably related to variation in predation risk between localities, were a significant factor influencing male social network structure. Overall, males from the high-predation site had more differentiated social interactions. This larger mean differentiation score in the
0.8
High predation Low predation
Mean courtship rate/min
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
Complex
Simple Habitat
Figure 5. Mean SE courtship interaction rates of males separated by population of origin and habitat complexity (simple and complex).
high-predation population indicates greater fracturing of the social network and less evenly distributed associations with conspecifics of both sexes. By contrast, low-predation fish (lower mean differentiation) had greater social mixing and more homogeneous social associations with conspecifics. Closer examination of the inter- and intrasexual social networks reveals some interesting patterns. In particular, we observed that maleemale weighted degree differed between populations, with males from the high-predation population having lower values. This result is surprising, especially when we consider the benefits of phenotypically matched shoal formations under increased predation risk (McRobert & Bradner 1998). There may, however, be benefits for males associated with avoiding other male conspecifics. For example, by reducing malee male competition, males may be better able to isolate females from competitors and reduce interference competition (Hibler & Houde 2006). Another factor that may contribute to the observed patterns is risk sensitivity. Males from high-predation locations may reduce movements in the habitat to avoid encounters with predators (reviewed in Lima & Dill 1990) thereby negatively influencing encounter rates with other males. In contrast, low-predation males exhibited greater social mixing, which is likely to increase maleemale competition and have implications for male reproductive success. We found that the degree of habitat complexity used in this study had no significant effect upon social network structure. This result is surprising when we consider that habitat complexity has multiple effects on fish behaviour. For example, structurally complex habitats are used as refuges from predators (Werner et al. 1983), with shoaling tendency and shoal sizes increasing in simple habitats (Orpwood et al. 2008). A possible explanation for the nonsignificant effect of habitat complexity on social network structure is that the level of complexity used in this study was not sufficient to result in a difference in interaction patterns (e.g. changes in shoal size). However, habitat complexity did have a significant effect on male reproductive behaviour, which we suggest is due to risk sensitivity (see below for a discussion). Future work examining the social network structure of fish populations under a wider range of habitat structural complexity would therefore be rewarding. While we observed a significant effect of population on the male social network structure, there was no effect of population or habitat complexity on the structure of the maleefemale social network or harassment or courtship networks. Again, this is surprising given the well-documented population differences in shoaling behaviour between these two populations (Seghers 1974; Magurran & Seghers 1990). Research has indicated that female mate choice is influenced by social factors (Dugatkin & Godin 1992; Doutrelant & McGregor 2000; reviewed in Galef & White 2000; Doutrelant et al. 2001; Godin et al. 2005). The structure of maleefemale social networks is therefore likely to have implications for the information available to females when choosing a mate (Sih et al. 2009). In the current investigation, however, maleefemale network structure did not differ between populations, suggesting that females have similar exposure to information on which to make such decisions. It would be interesting to explore the patterns of maleefemale associations across populations living in natural habitats. It is well documented that in populations that experience high-predation risk, male and female guppies use different habitats, with females more frequently being found in deeper water than males (Croft et al. 2004a, 2006b). No such segregation is observed in low-predation habitats (Croft et al. 2004a, 2006b). Thus it is likely that when females have the opportunity to segregate from males spatially, the structure of the maleefemale social networks may look very different.
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
In fish, female fecundity is generally positivity correlated with body size and previous work has shown that male guppies prefer to associate with larger females (Dosen & Montgomerie 2004). Despite this, the current results reveal that in free-swimming situations males associated more frequently with smaller females even when larger females were available. One possible explanation for this pattern is that the energetic costs for males associating with larger females over a prolonged duration may exceed those for smaller females. In fish, optimal foraging rates and swimming speeds are size dependent (Beamish 1978; Hjelm & Persson 2001) and individuals in a group of conspecifics with dissimilar body lengths may be forced to travel and forage at suboptimal speeds, potentially incurring an energetic cost, which may generate selection for group assortment by body size during shoaling (Ruckstuhl 2007). Reproductive Behaviour We observed that male guppies from the population characterized by high-predation risk associated with females for longer in the complex habitat, whereas those originating from the lowpredation locality exhibited opposite trends. This result, although significant, must be interpreted cautiously when generalizing trends shaping guppy associations. Even though effect sizes were large, indicating that population and habitat complexity had a substantial influence upon association duration, confidence intervals surrounding these effect sizes were wide and spanned zero, indicating that a null result is possible. The wide confidence intervals are likely to be the result of a small sample size and greater replication would be expected to reduce this range, as described by Nakagawa & Cuthill (2007). With this in mind we suggest a possible scenario that could explain the direction of the observed results and urge researchers to explore this further. We propose that population differences in male association durations with females may be underpinned by risk-sensitive behaviour (Seghers 1974; Breden et al. 1987; Magurran & Seghers 1994a). In some species, visual obstructions reduce predator detection abilities generating risk sensitivity (Schooley et al. 1996; Whittingham et al. 2004; Devereux et al. 2006) and in others, visual barriers present a ‘safe haven’ reducing perceived risk (Candolin & Voigt 1998; Dzieweczynski & Rowland 2004). Thus males that originate from areas of high predation may be less willing to leave a female because the probability of locating future mates in a complex habitat may be reduced (Candolin & Voigt 2001) while the potential for encountering a predator may be higher (Magnhagen 1991). Low-predation fish exhibited the reversed pattern, which we suggest is due to maleemale competition. In the social networks, low-predation fish were observed to have a higher weighted network degree, suggesting that levels of maleemale competition for access to females may be higher. In the complex habitat, groups of low-predation fish were observed to encounter females, followed by rapid interactions and subsequent fission (M. Edenbrow, personal observation), potentially limiting association duration for the majority of males. Further work investigating the direct mechanisms driving these population differences would certainly be rewarding and provide greater understanding of how malee female association duration differs as a function of population and habitat complexity. A similar populationehabitat interaction was observed with male reproductive strategies. Males originating from the highpredation population performed more sigmoid displays in the simple habitat whereas low-predation males performed more sigmoid displays in the complex habitat. The observed results for high-predation fish are consistent with previous laboratory findings on guppies by Hibler & Houde (2006) who observed that in the
557
presence of visual barriers, males remain with females until successfully isolated from conspecifics, which may increase female receptivity and reduce male courtship rates (Hibler & Houde 2006). However, in the current investigation, we observed the opposite pattern in males that originated from the low-predation population, which we suggest is driven by maleemale competition. As noted above, males from the low-predation site had a higher weighted social network degree; it is therefore likely that increased maleemale competition negatively influences the success of female isolation from male conspecifics. Previous work has demonstrated that when courtship is interrupted by male interference (Houde 1988), males restart the display sequence to regain female attention (Hibler & Houde 2006), potentially increasing average display rate as a consequence. We suggest that in lowpredation fish, groups of males were able to confine females in the complex habitat, leading to increased rates of maleemale competition and higher display rates. In conclusion, our results suggest that differences in selective factors between populations have the potential to influence the structure of male social networks and sexual behaviour directly. Social network structure is likely to have implications for multiple processes that shape populations, including disease and information transfer, social learning and the maintenance of cooperation (Newman 2003; Croft et al. 2005, 2006a, 2008; Krause et al. 2007; Sih et al. 2009). Our results suggest that predation risk may influence association and in turn interaction patterns in social networks and future work exploring the generality of this pattern and its ecological and evolutionary implications would be very rewarding. Our study was restricted to two populations and thus it lacks the replication needed to draw general conclusions regarding the effect of predation risk on behaviour. However, it does lay the foundations for future studies specifically aimed at testing the generality of these findings across a broader range of populations with differing predator regimes. Our approach demonstrates the potential of adopting the social network framework to study sexual interactions and understanding how these structures relate to individual fitness, and may provide new insights into sexual selection. Acknowledgments Funding was provided to D.P.C. by the NERC (NE/E001181/1). References Altmann, J. 1974. Observational study of animal behavior: sampling methods. Behaviour, 49, 227e267. Bateman, A. J. 1948. Intra-sexual selection in Drosophila. Heredity, 2, 349e368. Beamish, F. W. H. 1978. Swimming capacity. In: Fish Physiology (Ed. by W. S. Hoar & J. D. Randall), pp. 101e187. New York: Academic Press. Botham, M. S., Kerfoot, C. J., Louca, V. & Krause, J. 2006. The effects of different predator species on antipredator behavior in the Trinidadian guppy, Poecilia reticulata. Naturwissenschaften, 93, 431e439. Breden, F., Scott, M. A. & Michel, E. 1987. Genetic differentiation for antipredator behavior in the Trinidad guppy, Poecilia reticulata. Animal Behaviour, 35, 618e620. Candolin, U. & Voigt, H. R. 1998. Predator-induced nest site preference: safe nests allow courtship in sticklebacks. Animal Behaviour, 56, 1205e1211. Candolin, U. & Voigt, H. R. 2001. Correlation between male size and territory quality: consequence of male competition or predation susceptibility? Oikos, 95, 225e230. Clutton-Brock, T. H. & Parker, G. A. 1995. Sexual coercion in animal societies. Animal Behaviour, 49, 1345e1365. Croft, D. P., Arrowsmith, B. J., Bielby, J., Skinner, K., White, E., Couzin, I. D., Magurran, A. E., Ramnarine, I. & Krause, J. 2003. Mechanisms underlying shoal composition in the Trinidadian guppy (Poecilia reticulata). Oikos, 100, 429e438. Croft, D. P., Botham, M. S. & Krause, J. 2004a. Is sexual segregation in the guppy, Poecilia reticulata, consistent with the predation risk hypothesis? Environmental Biology of Fishes, 71, 127e133. Croft, D. P., Krause, J. & James, R. 2004b. Social networks in the guppy (Poecilia reticulata). Proceedings of the Royal Society B, 271, S516eS519.
558
M. Edenbrow et al. / Animal Behaviour 81 (2011) 551e558
Croft, D. P., James, R., Ward, A. J. W., Botham, M. S., Mawdsley, D. & Krause, J. 2005. Assortative interactions and social networks in fish. Oecologia, 143, 211e219. Croft, D. P., James, R., Thomas, P. O. R., Hathaway, C., Mawdsley, D., Laland, K. N. & Krause, J. 2006a. Social structure and co-operative interactions in a wild population of guppies (Poecilia reticulata). Behavioral Ecology and Sociobiology, 59, 644e650. Croft, D. P., Morrell, L. J., Wade, A. S., Piyapong, C., Ioannou, C. C., Dyer, J. R. G., Chapman, B. B., Yan, W. & Krause, J. 2006b. Predation risk as a driving force for sexual segregation: a cross-population comparison. American Naturalist, 167, 867e878. Croft, D. P., James, R. & Krause, J. 2008. Exploring Animal Social Networks. Princeton, New Jersey: Princeton University Press. Crow, R. T. & Liley, N. R. 1979. Sexual pheromone in the guppy, Poecilia reticulata (Peters). Canadian Journal of Zoology, 57, 184e188. Darden, S. K., James, R., Ramnarine, I. W. & Croft, D. P. 2009. Social implications of the battle of the sexes: sexual harassment disrupts female sociality and social recognition. Proceedings of the Royal Society B, 276, 2651e2656. Devereux, C. L., Whittingham, M. J., Fernandez-Juricic, E., Vickery, J. A. & Krebs, J. R. 2006. Predator detection and avoidance by starlings under differing scenarios of predation risk. Behavioral Ecology, 17, 303e309. Dosen, L. D. & Montgomerie, R. 2004. Female size influences mate preferences of male guppies. Ethology, 110, 245e255. Doutrelant, C. & McGregor, P. K. 2000. Eavesdropping and mate choice in female fighting fish. Behaviour, 137, 1655e1669. Doutrelant, C., McGregor, P. K. & Oliveira, R. F. 2001. The effect of an audience on intrasexual communication in male Siamese fighting fish, Betta splendens. Behavioral Ecology, 12, 283e286. Dugatkin, L. A. & Godin, J. G. J. 1992. Reversal of female mate choice by copying in the guppy (Poecilia reticulata). Proceedings of the Royal Society B, 249, 179e184. Dzieweczynski, T. L. & Rowland, W. J. 2004. Behind closed doors: use of visual cover by courting male three-spined stickleback, Gasterosteus aculeatus. Animal Behaviour, 68, 465e471. Endler, J. A. 1995. Multiple-trait coevolution and environmental gradients in guppies. Trends in Ecology & Evolution, 10, 22e29. Evans, J. P., Kelley, J. L., Ramnarine, I. W. & Pilastro, A. 2002. Female behaviour mediates male courtship under predation risk in the guppy (Poecilia reticulata). Behavioral Ecology and Sociobiology, 52, 496e502. Farr, J. A. 1980. Social behavior patterns as determinants of reproductive success in the guppy, Poecilia reticulata Peters (Pisces: Poeciliidae) an experimental study of the effects of intermale competition, female choice, and sexual selection. Behaviour, 74, 38e90. Fuiman, L. A. & Magurran, A. E. 1994. Development of predator defences in fishes. Reviews in Fish Biology and Fisheries, 4, 145e183. Galef, B. G. & White, D. J. 2000. Evidence of social effects on mate choice in vertebrates. Behavioural Processes, 51, 167e175. Godin, J. G. J. 1995. Predation risk and alternative mating tactics in male Trinidadian guppies (Poecilia reticulata). Oecologia, 103, 224e229. Godin, J. G. J., Herdman, E. J. E. & Dugatkin, L. A. 2005. Social influences on female mate choice in the guppy, Poecilia reticulata: generalized and repeatable traitcopying behaviour. Animal Behaviour, 69, 999e1005. Gross, M. R. 1996. Alternative reproductive strategies and tactics: diversity within sexes. Trends in Ecology & Evolution, 11, 92e98. Guevara-Fiore, P., Skinner, A. & Watt, P. J. 2009. Do male guppies distinguish virgin females from recently mated ones? Animal Behaviour, 77, 425e431. Guevara-Fiore, P., Stapley, J. & Watt, P. 2010. Mating effort and female receptivity: how do male guppies decide when to invest in sex? Behavioral Ecology and Sociobiology, 64, 1665e1672. Hamilton, W. D. 1971. Geometry for the selfish herd. Journal of Theoretical Biology, 31, 295e311. Held, J. & Manser, T. 2005. A PDA-based system for online recording and analysis of concurrent events in complex behavioral processes. Behavior Research Methods, 37, 155e164. Herdman, E. J. E., Kelly, C. D. & Godin, J. G. J. 2004. Male mate choice in the guppy (Poecilia reticulata): do males prefer larger females as mates? Ethology, 110, 97e111. Hibler, T. L. & Houde, A. E. 2006. The effect of visual obstructions on the sexual behaviour of guppies: the importance of privacy. Animal Behaviour, 72, 959e964. Hjelm, J. & Persson, L. 2001. Size-dependent attack rate and handling capacity: inter-cohort competition in a zooplanktivorous fish. Oikos, 95, 520e532. Hoglund, J. & Sheldon, B. C. 1998. The cost of reproduction and sexual selection. Oikos, 83, 478e483. Houde, A. E. 1988. The effects of female choice and maleemale competition on the mating success of male guppies. Animal Behaviour, 36, 888e896. Houde, A. E. 1997. Sex, Color, and Mate Choice in Guppies. Princeton, New Jersey: Princeton University Press. Kodric-Brown, A. 1985. Female preference and sexual selection for male coloration in the guppy (Poecilia reticulata). Behavioral Ecology and Sociobiology, 17, 199e205. Kokko, H. & Rankin, D. J. 2006. Lonely hearts or sex in the city? Density-dependent effects in mating systems. Philosophical Transactions of the Royal Society B, 361, 319e334.
Krause, J. & Ruxton, G. D. 2002. Living in Groups. Oxford: Oxford University Press. Krause, J., Croft, D. P. & James, R. 2007. Social network theory in the behavioural sciences: potential applications. Behavioral Ecology and Sociobiology, 62, 15e27. Liley, N. R. 1966. Ethological isolating mechanisms in four sympatric species of Poeciliid fishes. Behaviour, Supplement, 13, 1e197. Liley, N. R. & Wishlow, W. 1974. Interaction of endocrine and experiential factors in regulation of sexual-behaviour in female guppy Poecilia reticulata. Behaviour, 48, 185e214. Lima, S. L. & Dill, L. M. 1990. Behavioural decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology, 68, 619e640. Luyten, P. H. & Liley, N. R. 1985. Geographic variation in the sexual behaviour of the guppy, Poecilia reticulata (Peters). Behaviour, 95, 164e179. McRobert, S. P. & Bradner, J. 1998. The influence of body coloration on shoaling preferences in fish. Animal Behaviour, 56, 611e615. Magellan, K., Pettersson, L. B. & Magurran, A. E. 2005. Quantifying male attractiveness and mating behaviour through phenotypic size manipulation in the Trinidadian guppy, Poecilia reticulata. Behavioral Ecology and Sociobiology, 58, 366e374. Magnhagen, C. 1991. Predation risk as a cost of reproduction. Trends in Ecology & Evolution, 6, 183e185. Magurran, A. E. 2005. Evolutionary Ecology: The Trinidadian Guppy. Oxford: Oxford University Press. Magurran, A. E. & Seghers, B. H. 1990. Population differences in the schooling behaviour of newborn guppies, Poecilia reticulata. Ethology, 84, 334e342. Magurran, A. E. & Seghers, B. H. 1994a. Predator inspection behaviour covaries with schooling tendency amongst wild guppy, Poecilia reticulata, populations in Trinidad. Behaviour, 128, 121e134. Magurran, A. E. & Seghers, B. H. 1994b. Sexual conflict as a consequence of ecologyeevidence from guppy, Poecilia reticulata, populations in Trinidad. Proceedings of the Royal Society B, 255, 31e36. Mattingly, H. T. & Butler, M. J. I. V. 1994. Laboratory predation on the Trinidadian guppy: implications for the size-selective predation hypothesis and guppy life history evolution. Oikos, 69, 54e64. Mikheev, V. 2009. Combined effects of predators and parasites on shoaling behavior of fishes. Journal of Ichthyology, 49, 1032e1041. Nakagawa, S. & Cuthill, I. C. 2007. Effect size, confidence interval and statistical significance: a practical guide for biologists. Biological Reviews, 82, 591e605. Neff, B. D., Pitcher, T. E. & Ramnarine, I. W. 2008. Inter-population variation in multiple paternity and reproductive skew in the guppy. Molecular Ecology, 17, 2975e2984. Neill, S. R. J. & Cullen, J. M. 1974. Experiments on whether schooling by their prey affects the hunting behavoiur of cephalopods and fish predators. Journal of Zoology, 172, 549e569. Newman, M. E. J. 2001. Scientific collaboration networks II. Shortest paths, weighted networks and centrality. Physical Review E, 64 016132, DOI: 10.1103/ PhysRevE.64.016132. Newman, M. E. J. 2003. Mixing patterns in networks. Physical Review E, 67 026126. Orpwood, J. E., Magurran, A. E., Armstrong, J. D. & Griffiths, S. W. 2008. Minnows and the selfish herd: effects of predation risk on shoaling behaviour are dependent on habitat complexity. Animal Behaviour, 76, 143e152. Reznick, D. & Endler, J. A. 1982. The impact of predation on life-history evolution in Trinidadian guppies (Poecilia reticulata). Evolution, 36, 160e177. Ruckstuhl, K. E. 2007. Sexual segregation in vertebrates: proximate and ultimate causes. Integrative and Comparative Biology, 47, 245e257. Schooley, R. L., Sharpe, P. B. & VanHorne, B. 1996. Can shrub cover increase predation risk for a desert rodent? Canadian Journal of Zoology, 74, 157e163. Seghers, B. H. 1974. Schooling behaviour in the guppy (Poecilia reticulata): an evolutionary response to predation. Evolution, 28, 486e489. Sih, A., Hanser, S. F. & McHugh, K. A. 2009. Social network theory: new insights and issues for behavioral ecologists. Behavioral Ecology and Sociobiology, 63, 975e988. Steidl, R. J., Hayes, J. P. & Schauber, E. 1997. Statistical power analysis in wildlife research. Journal of Wildlife Management, 61, 270e279. Stoehr, A. M. 1999. Are significance thresholds appropriate for the study of animal behaviour? Animal Behaviour, 57, F22eF25. Taborsky, M. 1997. Bourgeois and parasitic tactics: do we need collective, functional terms for alternative reproductive behaviours? Behavioral Ecology and Sociobiology, 41, 361e362. Thomas, L. 1997. Retrospective power analysis. Conservation Biology, 11, 276e280. Tosh, C. R., Jackson, A. L. & Ruxton, G. D. 2006. The confusion effect in predatory neural networks. American Naturalist, 167, E52eE65. Trivers, R. L. 1972. Parental investment and sexual selection. In: Sexual Selection and the Descent of Man (Ed. by B. Campbell), pp. 136e179. Chicago: Aldine-Atherton. Werner, E. E., Gilliam, J. F., Hall, D. J. & Mittelbach, G. G. 1983. An experimental test of the effects of predation risk on habitat use in fish. Ecology, 64, 1540e1548. Whittingham, M. J., Butler, S. J., Quinn, J. L. & Cresswell, W. 2004. The effect of limited visibility on vigilance behaviour and speed of predator detection: implications for the conservation of granivorous passerines. Oikos, 106, 377e385. Wong, B. B. M. & Candolin, U. 2005. How is female mate choice affected by male competition? Biological Reviews, 80, 559e571.