Behavioral Ecology and Sociobiology

, Volume 54, Issue 4, pp 416–422

Kinship and food availability influence cannibalism tendency in early-instar wolf spiders (Araneae: Lycosidae)


    • Department of Biological SciencesUniversity of Cincinnati
  • Phillip W. Taylor
    • Department of Biological SciencesUniversity of Cincinnati
    • Department of PsychologyMacquarie University
  • George W. Uetz
    • Department of Biological SciencesUniversity of Cincinnati
Original Article

DOI: 10.1007/s00265-003-0646-8

Cite this article as:
Roberts, J.A., Taylor, P.W. & Uetz, G.W. Behav Ecol Sociobiol (2003) 54: 416. doi:10.1007/s00265-003-0646-8


For potentially cannibalistic animals such as spiders, the ability to recognize and avoid kin and/or preferentially cannibalize non-relatives would permit exploiting conspecifics as prey while minimizing loss of inclusive fitness. We investigated the effects of relatedness and availability of alternative food on cannibalism tendency in pairs of juvenile Hogna helluo (Walckenaer), a North American wolf spider (Araneae: Lycosidae). For second-instar spiderlings (dispersing stage), cannibalism was more likely among pairs of non-sibs than pairs of sibs and, interestingly, was also more likely when other prey were available. We found no evidence of increased cannibalism in pairings involving broods of greatest average size disparity, indicating that size differences are unlikely to explain differences in cannibalism tendency. Additionally, the relative number of deaths from cannibalism or other causes did not increase with increasing risk of starvation. For third-instar spiderlings, which had lived independently of their mother and sibs following dispersal, cannibalism rates were very high in all treatments and there were no significant effects of relatedness or food availability. Our results suggest that spiders with predominantly solitary lifestyles may bias cannibalism toward non-kin during the juvenile associative period, and that this effect is lost in the subsequent instar. Results are discussed in the context of several potential mechanisms that might result in differential cannibalism.


Kin discriminationCannibalismHogna helluoSpiderlings


Animals from taxonomically diverse groups appear able to recognize kin and display favoritism to relatives (Komdeur and Hatchwell 1999; Hauber and Sherman 2000; Schausberger and Croft 2001). This nepotism is demonstrated in many different circumstances ranging from alloparental care to alarm calls (Fletcher 1987), but for many species is expressed as differential cannibalism of kin and non-kin (Pfennig 1997). Cannibalism is common across numerous animal taxa, and while it can occur in a number of contexts (e.g., filial, sexual, competitive) (Elgar and Crespi 1992), cannibalism is most often thought of as a foraging decision (Dong and Polis 1992; Pfennig 1997). Although consumption of conspecifics will tend to reduce competition and conspecifics may be a useful source of food that increase the rate of survival of cannibals in times of low prey availability, closely related conspecifics also provide inclusive fitness, reducing their "profitability" as prey (Jones 1982; Dong and Polis 1992; Sadler and Elgar 1994; Pfennig 1997). In cannibalistic species with an extended associative period, a high proportion of the conspecifics that an individual meets may be close relatives and, under these circumstances, we might expect animals to either avoid cannibalism of conspecifics—kin and non-kin alike—or to possess the ability to avoid eating kin (kin discrimination) as a means of preserving inclusive fitness.

There are a wide range of cannibalistic taxa that exhibit an extended associative period, either as a product of their environment (e.g., ephemeral ponds and streams) or lifestyle (e.g., dispersal limitations, communal brood care, eusociality). Many of these also recognize and discriminate kin (reviews in Hepper 1991; Elgar and Crespi 1992; Sadler and Elgar 1994; Pfennig 1997). Some notable, well-studied examples include rodents (Elwood 1992), larval amphibians (Walls and Roudebush 1991; Pfennig et al 1993, 1994), and social hymenopteran insects (Klahn and Gamboa 1983; Crespi 1992; Panek and Gamboa 2000). A large number of arthropod taxa are cannibalistic at some life stage; however, most studies have been limited to aspects of sexual cannibalism (Elgar 1992), and even fewer have explored the role of kin discrimination in cannibalism.

Spiders are obligate predators and many species include conspecifics in their diet, but most have received little attention outside the bounds of sexual cannibalism. Females of some spider species may kill and eat males during courtship (Elgar 1992; Arnqvist and Henriksson 1997; Taylor and Jackson 1999; Johnson 2001), while copulating (Andrade 1996, 1998) or after decoupling (Sasaki and Iwahashi 1995). Males of Stegodyphus lineatus may kill a female's brood so that she will reproduce again using their sperm (Schneider and Lubin 1996, 1997). Sexually immature females of Portia labiata and P. shultzi aggressively mimic adult females to attract males as prey (Jackson and Pollard 1997). Cannibalism in spiders can also occur in contexts unrelated to sexual interaction. Wagner (1995) demonstrated that unmated and post-reproductive female Schizocosa ocreata readily accept conspecific spiderlings as prey, while females carrying eggsacs or young do not. In a few species, juveniles eat their mother before dispersing (Seibt and Wickler 1987; Kim and Horel 1998; Kim et al. 2000). Edgar (1969) reported that approximately 25% of the diet of the wolf spider Lycosa lugubris consisted of other spiders, and that 17% of prey were smaller conspecifics. Similar results have been found in field observations of other wolf spiders (Hallander 1970; Yeargan 1975) and, in some spiders, cannibalism among juveniles may be a major limit on the species' abundance (Wagner and Wise 1996, 1997).

Despite providing potential avenues for testing altruism and the evolution of sociality, studies combining kin discrimination and cannibalism in spiders are few. Evans (1999) found that juveniles of Diaea ergandros, a social crab spider, preferentially eat non-relatives when other food is scarce. Similarly, Bilde and Lubin (2001) detected a bias toward cannibalizing non-kin in juveniles of Stegodyphus lineatus, a subsocial spider. However, there is very little evidence whether such nepotism is also evident in solitary wandering spiders. In the wolf spider, Pardosa milvina, females whose young have recently dispersed exhibit low levels of cannibalism on juveniles, and while there was no control for the possibility of cannibalism by siblings, females placed with groups of their own or unrelated offspring may be more likely to cannibalize unrelated juveniles (Anthony 2003). In this study, we test whether juveniles of Hogna helluo (Araneae, Lycosidae), a large North American wolf spider, differentially cannibalize kin and non-kin. We also investigate the effects of feeding state and instar (age) on the incidence of cannibalism.


Adult female H. helluo (Walckenaer) were collected at the Cincinnati Nature Center, Rowe Woods (Clermont County, Ohio) in September and October 2000. Spiders were maintained individually in the laboratory in opaque plastic containers (160×160×50 mm) with clear lids. Each spider was provided ad libitum access to water via a cotton wick that protruded from a water reservoir, and was fed two to three sub-adult crickets (Acheta domestica) twice weekly. Laboratory conditions consisted of a 14 h:10 h light:dark cycle, temperature range of 24–26°C and approximately 65% RH. Females that had mated in nature produced eggsacs in the laboratory in November/December, which were assessed daily for spiderling emergence. The offspring that emerged from these eggsacs were used in the experiments.

Female wolf spiders carry their eggsacs attached to their spinnerets. Offspring hatch within this eggsac and remain inside throughout their first instar. When the spiderlings enter their second instar, they exit the eggsac and climb onto their mother's abdomen where they are carried for up to 10 days (unpublished data). During the last few days of this associative period, the juveniles make brief forays from their mother's abdomen, indicating the beginning of dispersal (Foelix 1996; personal observation). When the spiderlings of two females' broods (all of the spiderlings of a given eggsac) emerged on the same day, the two broods were designated as a same-aged "brood pair". In total, 15 "brood pairs" were established for this experiment (i.e., 30 broods as 15 same-aged brood pairs). Within each of the brood pairs, one brood was randomly chosen to serve as the "focal brood", while the "non-focal brood" served to provide spiderlings for non-sib pairings (see below).

In this study, we tested whether second-instar spiderlings (dispersing stage), or spiderlings separated at the dispersing stage and maintained individually until the third instar, were more likely to cannibalize non-sibs than sibs. We also tested whether the tendency for cannibalism was influenced by the presence of alternative prey. Seven of the 15 available brood pairs were randomly chosen to provide second-instar spiderlings for the first experiment. When spiderlings of both broods within each brood pair entered the dispersal stage, 40 individuals of the focal brood were assigned to 4 different treatments (i.e., 10 spiderlings/treatment): (1) paired with a sib with alternative food available ("sib/fed"), (2) paired with a sib without food available ("sib/unfed"), (3) paired with a non-sib from the "non-focal brood" of that brood pair with alternative food available ("non-sib/fed"), and (4) paired with a non-sib without alternative food available ("non-sib/unfed"). Hence, within each brood pair, this experiment used a total of 60 spiderlings from the focal brood (used for sib and non-sib pairs) and 20 spiderlings from the non-focal brood (used for non-sib pairs only).

For the second experiment, spiderlings from the eight remaining brood pairs were separated at the dispersal stage and maintained individually under conditions of ad libitum prey until they had molted into the third instar. Spiderling pairs were then established (within 72 h post-molt) by assignment to the same treatments as for second-instar spiderlings (20 spiderlings of each focal brood assigned to the 4 treatments—sib/fed, sib/unfed, non-sib/fed, non-sib/unfed). In this experiment, we used a total of 30 spiderlings from the focal brood and 10 from the non-focal brood within each brood pair.

To assess spiderling size, we preserved non-experimental representatives of each brood (15 second instar, 10 third instar) in 70% ethanol and measured them later. Cephalothorax length proved to be the easiest body part to measure. Spiderlings were digitally photographed (Pixera Pro camera, 1.2 megapixels) through a dissecting microscope (Wild M5) while pinned to a flat soft surface under 70% ethanol. Measurements were made using Image Tools software (ver. 2.25), a shareware program available from the University of Texas.

"Spiderling pairs" were housed together in plastic specimen containers similar to those used in a previous study of this species (Fisherbrand 4 oz/118 ml, see Walker et al. 1999). These containers are sufficiently small that frequent interactions were likely (Schneider 1996; Toft and Wise 1999a). Fed spiderlings had ad libitum access to Collembola as prey, from cultures used for over 5 years to successfully raise wolf spiders. Spiderlings in all treatments received water from a moist cotton dental wick on the container floor. We checked each pair for cannibalism or death from other causes (e.g., starvation, desiccation, drowning) each day until one spiderling died. Each dead spiderling was inspected using a dissecting microscope (Wild M5) and cannibalism was evident from feeding damage on the body. For each brood pair, we assessed the proportion of spiderling pairs within each treatment that died by cannibalism, thereby reducing the ten spiderling pairs (five for third-instar spiderlings) for each treatment to a single data point for each brood pair. Hence, each brood pair yielded a single data point for each of the four treatments. Effects of treatment (relatedness of spiderling pair and feeding regime) on the proportion of spiderling pairs in which cannibalism occurred (arcsine transformed) were investigated by ANOVA, with brood pair included as a blocking variable to account for potential brood effects. All analyses were carried out using JMP ver. 4.02 (SAS Institute).

When results indicated a greater tendency for cannibalism among non-sib spiderling pairs, we considered the possibility that this resulted from greater size differences between- rather than within-broods. We first tested for brood differences in spiderling size using one-way ANOVA. If there was significant size variation among broods then, on average, non-sib pairs would have greater size disparity than sib pairs. If greater average size disparity explained a higher tendency for cannibalism in non-sib spiderling pairs, then this effect might also be evident as an increased tendency for cannibalism among non-sib spiderling pairs with greater average size disparity compared with non-sib spiderling pairs with smaller average size disparity. For non-sib pairings, we then tested the relationship between proportion cannibalized and size difference between broods (|average size of spiderlings in focal brood-average size of spiderlings in non-focal brood|). This relationship was investigated by ANCOVA, including "treatment" (non-sib/fed, non-sib/unfed) as a categorical predictor and "difference between average size of spiderlings of the two broods within each brood pair" as a covariate.

An increase in cannibalism with increased risk of starvation was tested by adding a time component to the analyses of the unfed treatment of second-instar spiderlings (starvation risk should not increase in food-supplemented treatments). As the effect of brood pair as a blocking variable was not significant in our initial analysis (see Results), we pooled data across the seven brood pairs for the unfed treatments and subjected the number of deaths from cannibalism or other causes each week of the experiment to a 3-way contingency analysis, with relatedness (kin/nonkin), cause of death (cannibalism/other), and week (1–4) as factors (Zar 1999).


Second-instar spiderlings

Occurrence of cannibalism by second-instar spiderlings was more frequent in pairs of non-sibs than in pairs of sibs (ANOVA, F1,18=6.742, p<0.02; Fig. 1a). Cannibalism was also influenced by the feeding treatment, being more common in the pairings that had alternative prey available (ANOVA, F1,18=61.405, p<0.001; Fig. 1a). Alternatively stated, mortality from other sources was greater for pairs of sibs and for unfed spiders. There was no significant interaction between relatedness and feeding regime (ANOVA, F1,18=0.735, p>0.4) and there was no significant effect of brood pair (ANOVA, F6,18=0.259, p>0.9).
Fig. 1.

a Second-instar spiderlings—mean proportion (+SE) of spiderling pairs in which cannibalism was observed for the seven brood pairs in each of the four treatment classes. b Third-instar spiderlings—mean proportion (+SE) of spiderling pairs in which cannibalism was observed for the eight brood pairs in each of the four treatment classes

There were significant size differences among broods (ANOVA: F13,194=7.103, p<0.001); however, we found no evidence of higher frequency of cannibalism in non-sib pairings involving broods of the greatest average size disparity (ANCOVA: treatment F1,3=41.967, p<0.001, size difference F1,3=0.003, p=0.960, treatment*size difference F1,3=1.368, p=0.269). This suggests that differences in cannibalism rates are unlikely to be explained by greater size disparity between non-sib pairs than sib pairs.

Subsequent analysis of the unfed treatment with a 3-way contingency analysis revealed that relatedness, cause of death, and week were not mutually independent (G=19.457, df=10, p<0.035). While cause of death was not independent of relatedness and week (G=18.312, df=7, p<0.016), and relatedness was not independent of cause of death and week (G=15.519, df=7, p<0.030), week was independent of relatedness and cause of death (G=10.091, df=9, p>0.340). This suggests that the relative number of deaths from cannibalism or other causes did not change over the course of the experiment (Fig. 2). Dropping the time component (week) from the analysis, the 2-way contingency with relatedness and cause of death as factors demonstrates that cause of death, as in the ANOVA results above, was not independent of relatedness (G=14.373, df=1, p<0.001).
Fig. 2.

Number of deaths from cannibalism or other causes for sib and non-sib pairings in the unfed treatments across the first 4 weeks of the experiment (in which 95% of deaths occurred)

Third-instar spiderlings

Cannibalism rates were very high across all treatments in this experiment compared with second-instar spiderlings (Fig. 1b). Alternatively stated, mortality from other sources was comparatively low. As for the second-instar spiderlings, there was no significant effect of brood pair (ANOVA, F7,21=0.9616, p>0.48). Unlike the previous experiment, the proportion of pairings in which cannibalism was observed did not vary with either relatedness of spiderlings (ANOVA, F1,21=0.055, p>0.8) or feeding regime (ANOVA, F1,21=1.245, p>0.25, Fig. 1b).


Kinship effects

Closely related conspecifics serve as a source of inclusive fitness but can also serve as prey for potentially cannibalistic animals such as spiders. Given that the cost/benefit balance of cannibalism may be quite different for spiders confronted by sibs rather than non-relatives, the ability to bias cannibalism toward non-relatives would seem to be favored. Evidence of kin discrimination has been found in several spider species that have extended communal lifestyles (Diaea ergandros, Evans 1999; Stegodyphus lineatus, Bilde and Lubin 2001). The present study appears to be the first evidence that solitary hunting spiders may also be less prone to cannibalizing their relatives, at least as juveniles during the period immediately following dispersal.

There are a variety of mechanisms that might underlie non-kin-biased cannibalism. Here we briefly consider the evidence as it relates to two likely mechanistic pathways: kin recognition and variation (e.g., size) between broods.

Kin recognition

When the outcome of intraspecific interactions varies with relatedness of the parties involved, abilities of kin recognition immediately arise as a potential explanation. The possibility of kin recognition in spiders has rarely been investigated, but is a viable explanation for the results of this study. Four basic mechanisms have been proposed for kin recognition: spatial distribution, familiarity, phenotypic matching, and recognition alleles (reviewed in Hepper 1991; Olsen 1992; Tang-Martinez 2001). While the current study is not designed to address the heritable components of the last two mechanisms, we can potentially address spatial distribution and association, particularly as they relate to chemical cues. So far, there appear to be only two studies providing evidence of chemically mediated kin recognition in spiders (Miller 1989; Evans 1999). However, in many spider species, information about species and sex may be evident in olfactory pheromones (Tietjen 1979a; Searcy et al. 1999), or less volatile chemical compounds associated with the spider's body, web, or draglines (Tietjen 1979b; Jackson 1987; Miller 1989; Evans and York-Main 1993; Taylor 1998), making chemical communication a likely pathway for kin recognition.

The spatial distribution mechanism does not imply direct recognition per se; relatives are distributed predictably due to limited dispersal and most individuals in a particular place would be related. To avoid cannibalism of kin, second-instar spiderlings that had not lived independently prior to the start of the experiment (experienced limited dispersal) would be expected to suppress cannibalism of all conspecifics and/or to increase cannibalism only as risk of starvation increases. Neither of these predictions are supported by the results of our study. The significant effect of kinship on the overall proportion of spiders cannibalized suggests direct assessment of relatedness and differential treatment of kin instead of all conspecifics. In the non-fed treatment of second-instar spiderlings where starvation risk increased over time, week was independent of kinship and cause of death and therefore the relative number of deaths from cannibalism to deaths from other causes did not increase with increasing risk of starvation (Fig. 2).

The second, and more likely, mechanism to address is familiarity (previous association). The fact that kinship only influenced cannibalism rates in the second-instar spiderlings, those recently dispersed from their mother, may indicate that cannibalism of kin is inhibited by a chemical "signature" that is gleaned from the mother or sibs (familiarity or association), or is produced endogenously and gradually wears off after dispersal, or is lost during the molt to the next instar. In Tegenaria atrica, increased aggression among dispersing spiderlings and by mothers toward young is associated with marked changes in cuticular chemistry of spiderlings (Trabalon et al. 1996; Pourie and Trabalon 1999). However, it is presently unknown whether these are causal or incidental associations.

Variation between broods

In addition to having different parents, spiderlings from different broods likely vary somewhat in a variety of traits such as size, developmental rate, energetic reserves, and aggressiveness (e.g., due to heredity or environment). It is reasonable that there would be greater differences in these traits between the two spiderlings in non-sib pairs than in sib pairs. If spiderlings are more prone to cannibalize rivals that they have a distinct advantage over, then this could explain effects of kinship in our study.

Size is the one factor that is most often cited as indicative of an individual's susceptibility to cannibalism (Polis 1981; Dong and Polis 1992), and is the most readily quantified trait that might vary among broods. While members of individual pairings were not measured, measurements of representative samples demonstrate that the broods in the first experiment differed in size. Hence size disparity would have been greater (on average) in non-sib pairings than in sib pairings. Evans (1999) found that size was an important effect in spiders, as cannibalism was usually by larger individuals. However, we found no evidence of increased cannibalism when average size differences among broods were large. Although it appears that these size differences among broods did not exert much influence on cannibalism tendency other, less readily measured, differences might still have been more pronounced in non-sib pairings. These possible differences remain a potential explanation of increased cannibalism in non-sib pairings.

Feeding effects

Most studies have reported an increase in cannibalism frequency when alternative food is scarce (Elgar and Crespi 1992; Lindstrom and Sargent 1997; Dou et al. 2000), and the same is true for spiders (Andrade 1998; Evans 1999; Samu et al. 1999; Bilde and Lubin 2001). Studies of subsocial spiders suggest that abundance of prey tends to induce prolonged tolerance and delayed dispersal of juveniles (Krafft et al. 1986; Ruttan 1990; Gundermann et al. 1993; Schneider 1995; Kim 2000; Jones and Parker 2002). In the solitary spider Achaearanea tepidariorum, high prey abundance results in increased tolerance (Rypstra 1986). Interestingly, we found the opposite effect for a solitary hunting spider in the present study: cannibalism was more common when alternative food was available. There are a number of potential factors, not all mutually exclusive, that may provide an explanation for this seemingly contradictory finding.

Inhibition of ecdysis

Wolf spiders maintained on diets of extremely poor quality (including starvation) are unlikely to molt prior to death (Toft and Wise 1999a). In the event that cannibalism is inhibited by a chemical "signature" gleaned from the mother or sibs during the associative period prior to dispersal, then this compound may persist and remain active until lost in a molt. Molting was not observed in any of the unfed treatments involving second-instar spiderlings, and the low proportion of cannibalism observed (Fig. 1a) could be due to the persistence of a chemical compound that would otherwise be lost during normal growth.

Maintenance of a mixed diet

Most spiders are polyphagous (Riechert and Harp 1987), and several studies have demonstrated that a mixed diet can be advantageous (Uetz et al. 1992; Toft and Wise 1999a; Oelbermann and Scheu 2002), which could also explain a higher rate of cannibalism among spiderlings with a monotypic diet of Collembola as alternative prey. Wolf spiders (including Hogna) fed on monotypic diets exhibit reduced growth and survival compared to those reared on a mixed diet (Uetz et al. 1992; Toft and Wise 1999a; Oelbermann and Scheu 2002), and spiders fed low-quality prey insects switch to alternative prey over time, or avoid the same prey items with increasing frequency (Toft and Wise 1999b; but see Oelbermann and Scheu 2002). The wolf spider Pardosa ramulosa maintaines a mixed-prey diet, without switching to more abundant species as prey density changes (Greenstone 1978). Prey choice in Pardosa ramulosa was, however, consistent with maintaining a diet including an optimal blend of amino acids (Greenstone 1979). While mechanisms that result in spiders maintaining a mixed diet are not well studied, it seems possible that spiderlings that have fed upon several Collembola in sequence might be more likely to hunt conspecifics to increase prey diversity.

Hunger-induced lethargy/reduction in fighting ability

Spiders that fail to feed for an extended period may reduce their activity levels as a means of saving energy. If H. helluo spiderlings respond to hunger with reduced activity, then the tendency for lower cannibalism when heterospecific prey is unavailable may stem in part from reduced encounter frequency between inactive spiders. However, Anderson (1974) demonstrated that adult H. (Lycosa) lenta experience little reduction in foraging activity during extended periods of starvation. Moreover, recent studies of larger juvenile H. helluo suggest that these spiders actually become more active when prey are scarce (Walker et al. 1999), rendering this explanation unlikely. As a variation on the lethargy hypothesis, spiders that have not fed may be too weak to risk attacking a conspecific and instead invest all in the hope of finding alternative, safer prey. In studies of some solitary spiders, intraspecific predation among adults does not increase in response to periods of food deprivation (Jackson 1980a; Riechert 1981; Nossek and Rovner 1984). We know of no studies that directly address how feeding state influences spider fighting ability but it seems likely that prolonged hunger would cause a decline.

Superfluous killing

Conover (1966) introduced the term "superfluous feeding", referring to predators that, in the presence of abundant food, will continue to capture prey even after basic food requirements have been met, abandoning or only partially consuming prey items. This has been demonstrated in a number of predatory species including wolf spiders, and has been suggested as a general indicator of aggression (review in Maupin and Riechert 2001). Confronted with ad libitum prey, juvenile H. helluo may exhibit higher levels of aggression, increasing capture and partial consumption of all available prey items, including conspecifics. Superfluous killing of conspecifics resulting from aggression may be considered a consequence of overall (gross) density of individuals in the housing containers (Dong and Polis 1992). Fed treatments are, by necessity, higher in overall density of individuals and cannibalism would be expected to be higher if juvenile H. helluo exhibit higher levels of aggression at these higher density levels.

Starvation before cannibalism

There were two possible outcomes for termination of each trial: cannibalism and death from other causes. There were dramatic differences between second- and third-instar spiderlings that might be explained largely by the relatively greater sensitivity of second-instar spiderlings to starvation. That is, the much smaller, never-fed second-instar spiderlings may have been much more prone to dying of starvation before cannibalism might have occurred. In the absence of heterospecific prey, highly social Mallos gregalis will starve rather than attack each other (Jackson 1980b). Similarly, H. helluo juveniles in the first few weeks of life may be so reluctant to attack conspecifics that they eventually starve. The presence of alternative prey might have increased the chances of surviving through a period of cannibalism inhibition.

Previous studies have presented evidence that spiders with social lifestyles may bias cannibalism toward non-kin. This is the first study to suggest that spiders having solitary lifestyles may also possess this ability, at least during the phase immediately following dispersal. It is during this period that spiderlings are most likely to encounter close relatives. The ability to avoid eating relatives would thus increase a spiderling's inclusive fitness, while maintaining the direct fitness benefits that might accrue from eating non-relatives.


For financial support during this research, we thank the National Science Foundation (grant IBN 9906446 to G.W.U.); P.W.T. was awarded a University of Cincinnati post-doctoral fellowship. We are grateful to the Cincinnati Nature Center, Rowe Woods, for permission to collect spiders on their property. Kelly Roberts provided invaluable assistance with rearing and maintenance of spiders and staging of experiments. We are especially grateful to Trine Bilde and Yael Lubin for access to their previously unpublished data. We appreciate the editorial comments of Mark Elgar, Casey Harris, Jerald Hinn, Bruce Jayne, Scott Sakaluk, Sean Walker, and three anonymous reviewers. The work presented here adheres to all relevant laws and guidelines pursuant to the use of invertebrates in research in the United States of America.

Copyright information

© Springer-Verlag 2003