Colony Size and Individual Fitness in the Social Spider Anelosimus

vol. 152, no. 3

the american naturalist

september 1998

Colony Size and Individual Fitness in the Social Spider Anelosimus eximius

Leticia Avile´s 1 and Paul Tufin˜o 2

1. Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona 85721; 2. Departamento de Biologı´a, Pontificia Universidad Cato´lica del Ecuador, Quito, Ecuador Submitted July 28, 1997; Accepted March 30, 1998

abstract: The effects of colony size on individual fitness and its components were investigated in artificially established and natural colonies of the social spider Anelosimus eximius (Araneae: Theridiidae). In the tropical rain forest understory at a site in eastern Ecuador, females in colonies containing between 23–107 females had a significantly higher lifetime reproductive success than females in smaller colonies. Among larger colonies, this trend apparently reversed. This overall fitness function was a result of the conflicting effects of colony size on different components of fitness. In particular, the probability of offspring survival to maturity increased with colony size while the probability of a female reproducing within the colonies decreased with colony size. Average clutch size increased with colony size when few or no wasp parasitoids were present in the egg sacs. With a high incidence of egg sac parasitoids, this effect disappeared because larger colonies were more likely to be infected. The product of the three fitness components measured—probability of female reproduction, average clutch size, and offspring survival—produced a function that is consistent with direct estimates of the average female lifetime reproductive success obtained by dividing the total number of offspring maturing in a colony by the number of females in the parental generation. Selection, therefore, should favor group living and intermediate colony sizes in this social spider. Keywords: sociality, group living, density dependence, cooperation, colony extinction.

One of the central questions in the study of social evolution is why organisms live in groups. In various social systems, group living has been shown to provide benefits in terms of foraging success (Caraco and Wolf 1975; Clark and Mangel 1986; Cash et al. 1993), predator protection (Alexander 1974; Caraco and Pulliam 1984; Tyler Am. Nat. 1998. Vol. 152, pp. 403–418.  1998 by The University of Chicago. 0003-0147/98/5203-0007$03.00. All rights reserved.

1995), competitive ability (Buss 1981), or some combination of the above (e.g., Pulliam and Caraco 1984). At the same time, group living has been shown to have costs resulting from unavoidable competition among group members (e.g., Michener 1964; Komdeur 1994; Booth 1995; and references above). A generally accepted hypothesis is that, if group living is to be maintained, the balance between the costs and benefits of sociality should result in higher individual fitness among social than among solitary individuals (Alexander 1974; Vehrencamp 1983; Emlen and Vehrencamp 1985). This hypothesis can be tested by investigating the shape of the function relating individual fitness to colony size. If there is no individual benefit to group living or if there is only a cost, individual fitness should be either a constant or a decreasing function of colony size. If group living has a positive effect on certain components of individual fitness and assuming that costs are unavoidable, individual fitness should be highest at intermediate colony sizes (see Wilson 1975; Rodman 1981; Sibly 1983; Giraldeau 1988). The goals of this study are twofold. First, we wish to determine how colony size affects the average lifetime reproductive success (LRS) of colony members in a model organism in which such direct measure of individual fitness is possible. And, second, we wish to investigate how different components of fitness (e.g., probability of reproduction, clutch size, offspring survival) interact to produce the overall fitness function. Estimating the effects of group size on the total number of mature offspring produced during the life of an individual—its LRS—has many practical difficulties (for two examples in birds, see Vehrencamp et al. 1988; Brown and Brown 1996; for a laboratory study on communally nursing house mice, see Konig 1994). These difficulties arise because individuals tend to have long life spans relative to the rate of change in size of their groups, may reproduce several times while spending only part of their life in their original groups, or may be at times subject to socially enforced reproductive suppression. Also, the range of existing group sizes and the number of groups available are often limited. These complications are largely

404 The American Naturalist absent in the nonterritorial permanent social or ‘‘cooperative’’ spiders, one of which, the Neotropical Anelosimus eximius Simon (Araneae: Theridiidae), is the subject of this study. The ‘‘cooperative,’’ or nonterritorial permanent-social spiders (D’Andrea 1987; Avile´s 1997), build communal nests in which colony members of succeeding generations cooperate in prey capture, feeding, and brood care. The 17 species known to exhibit this type of behavior occur in tropical and subtropical areas of the world and belong to nine genera in six spider families (Avile´s 1997). The colonies of these species may contain from one to thousands of individuals who are relatively short lived (6–12 mo) and tend to spend their entire lives within their natal nest. Because matings apparently take place among nest mates generation after generation, the colonies of these spiders are also self-sustaining populations that may grow, proliferate, and become extinct without mixing with one another (for exceptions, see Avile´s 1997). Parental and offspring generations tend to be relatively discrete so that the number of mature individuals produced per member of the parental generation can be estimated directly. Also, because in spiders females deposit their clutch within a sac, average clutch size and average number of clutches per individual can be estimated with relative ease. Finally, because colony members are typically all able to reproduce (e.g., Darchen and Delange-Darchen 1986; Lubin 1995), individual spiders could, at least in theory, leave their colony any time to breed on their own. Whether they do so or not at different colony sizes, therefore, may reflect the balance between the costs and benefits of group living. Here we present evidence that the average LRS in A. eximius increases with colony size from small to intermediate-sized colonies and that the overall fitness function is primarily the result of an increased probability of offspring survival and a decreased probability of female reproduction as the size of the colonies increase. Additionally, since the colonies of A. eximius are not only social groups but also relatively closed and self-sustaining populations, this study illustrates the effects of density dependence in the growth of populations. Material and Methods Study Organisms Anelosimus eximius occurs in lowland tropical rain forests from Panama to southern Brazil (Levi 1963). Our study area (0°2′S 76°20′W, 200–300 m elevation) was located in the Cuyabeno Nature Reserve in eastern Ecuador. Here the colonies of A. eximius occur either along the forest edge (along roads and rivers) or in the forest understory. Colonies containing a single female plus her

offspring occur in either habitat, but along the forest edge, the colonies may grow to contain tens of thousands of spiders, while, in the forest understory, they grow to contain at the most a few thousand individuals (Avile´s 1992; Venticinque et al. 1993). Our study was conducted in the forest understory, and, therefore, its quantitative predictions apply to this habitat only. The colonies of A. eximius may last for less than one to several generations, with parents overlapping with their offspring for part of their life cycle but usually dying before their offspring reach reproductive maturity (Avile´s 1986). Successive generations continue to occupy and enlarge the native nest until the colonies either proliferate or become extinct. Colony proliferation only takes place at relatively large colony sizes and typically involves gravid females that, either individually or in groups, disperse to establish new colonies in the vicinity of the parental colony (Vollrath 1982; Avile´s 1992; Venticinque et al. 1993). Colony extinction has been found to be relatively frequent (Vollrath 1982; Avile´s 1992; Venticinque et al. 1993), in one study estimated to occur at a rate of 21% per generation among forest understory colonies (Avile´s 1992). An even higher rate of colony extinction may be characteristic of forest edge colonies (Venticinque et al. 1993). Colony sex ratios are of the order of 10 females per male, a bias that has been confirmed to result from an overproduction of female embryos (Avile´s and Maddison 1991). Establishment of Artificial Colonies and Experimental Design We investigated the effect of colony size on the average female lifetime reproductive success, probability of female reproduction, offspring survival, and colony survival on a set of 34 colonies ranging in size from one to 165 individuals (table 1). Twenty-one of these colonies were artificially established for purposes of this study (see below); the remaining 13 colonies occurred naturally in nearby areas of the forest. The 13 natural colonies were included in the study to ensure representation of very small colonies that could not be established artificially. Natural and artificially established colonies were indistinguishable from each other once artificial foundation had been accomplished. Because we could not estimate clutch size without destroying egg sacs, we investigated this additional fitness component in the artificially established and in natural colonies in generations subsequent to the one used for the above mentioned estimates (F1 or F2, table 1). The 21 artificially established colonies used in the study were part of an initial set of 27 colonies in four size classes that were created in December 1993 using spiders

Individual Fitness in a Social Spider 405 Table 1: Colony size range and number of Anelosimus eximius colonies used to estimate the effect of colony size on female lifetime reproductive success and the various fitness components measured

Generation and colony type

Number of colonies used in the analyses a

b

c

d

e

Colony size range

although, during foundation, they lost a fraction of their population due to dispersal and predation by wasps. Thus, we have used the number of individuals (adult/ subadult females) present in these colonies 2 wk after foundation (table 1) as the independent variable for estimating fitness components. Data Collection

P: NC AEC F1: NC AEC F2: NC AEC

13 21

11 17

⋅⋅⋅ ⋅⋅⋅

11 19

13 21

1–7 3–165

⋅⋅⋅ ⋅⋅⋅

⋅⋅⋅ ⋅⋅⋅

2 13

⋅⋅⋅ ⋅⋅⋅

⋅⋅⋅ ⋅⋅⋅

1–150 37–496

⋅⋅⋅ ⋅⋅⋅

⋅⋅⋅ ⋅⋅⋅

9 10

⋅⋅⋅ ⋅⋅⋅

⋅⋅⋅ ⋅⋅⋅

1–150 22–190

Note: a ⫽ female lifetime reproductive success; b ⫽ egg sacs laid per female (includes all colonies that survived the entire egg-laying period); c ⫽ average clutch size and incidence of egg sac parasitism (includes all colonies where at least one sac was laid); d ⫽ offspring survival; e ⫽ colony survival. NC ⫽ natural colonies; AEC ⫽ artificially established colonies. Colony size ranges correspond to the total number of subadult and adult females present in the colonies 2 wk after foundation (AEC) or at first sighting (NC). In adddition to females, the colonies contained approximately one male per 10 females, which is the usual sex ratio in this species (Avile´s and Maddison 1991).

from four large colonies collected along the banks of the Cuyabeno river. By allocating to each colony and treatment a similar proportion of spiders from each source colony, we created colonies containing 80 (10 colonies), 160 (eight colonies), 320 (six colonies), and 640 (three colonies) subadult and adult females plus males in a ratio of one male per 10 females. We created a larger number of colonies of the smaller size classes because we expected lower survival rates of smaller colonies and a less unbalanced design toward the end of the experiment. We combined in the artificial colonies spiders from all source colonies in order to control for source effects. Spiders from different colonies can be combined readily as no nestspecific recognition mechanisms are present in this species (Tapia and De Vries 1980; L. Avile´s, unpublished observations). The artificial colonies were established throughout an area of 130 m ⫻ 50 m on bushes of similar characteristics and located at a minimum distance of 5 m from each other. Treatments were allocated to the preselected sites in a completely random design. Six of the initial 27 colonies could not be used in the study because, during foundation, they moved to an inaccessible location (five colonies) or were attacked by ants (one colony). The remaining 21 colonies became established successfully,

Using a flashlight to illuminate hidden areas within the nests, every 15–20 d, we censused in situ the number of inhabitants of all colonies, including the number of egg sacs, adult (instar 6 for males, instar 7 for females) and subadult (instar 5 for males, instars 5 and 6 for females) individuals, and approximate number of juveniles (instars 1–4). These instars had been characterized in a previous study and are easily distinguishable by eye in the field (Avile´s 1986). We monitored the colonies from December 1993 until November 1994, a period during which two and a half spider generations elapsed within the nests. With the exception of occasionally removing dry leaves to improve visibility, we avoided any manipulation of the nests that could impair normal development of the first experimental generation (founder adult/subadult females to F1 adult/subadult females). Egg sacs were collected from the colonies only during their F1 or F2 generations. During our final visit to the field, we calibrated our error in scoring nest inhabitants by dissecting in the laboratory nests that had previously been censused in the field. We detected a tendency to increasingly underestimate nest contents as the size of the colonies increased (Tufin˜o 1997). This tendency, however, was extremely predictable (R 2 ⫽ 0.97). We, thus, have used the equation of the regression of field counts on actual nest contents (Y ⫽ 1.134X ⫹ 0.002X 2) to correct for this bias. Data Analysis We defined lifetime reproductive success (LRS) in A. eximius colonies as the number of subadult/adult females produced per subadult/adult female in the maternal generation. We considered three potential components of female LRS: the probability of reproducing in the colonies, mean clutch size, and the probability of offspring survival to maturity. These three components should determine LRS according to the equation LRS (x) ⫽ p(reproduction| x) ⫻ (clutchsize | x) ⫻ p(offspringsurv |x) ,

(1)

where x, colony size, is defined as the number of subadult/adult females in the parental generation. Our

406 The American Naturalist goal in this study was to determine the functional form of the LRS and of each of the factors on the right-hand side of the equation. We have done this as follows. We estimated the probability of female reproduction (factor 1, right-hand side of eq. [1]) as the average number of egg sacs laid per female in a colony (total egg sac production of the parental generation divided by the number of adult/subadult females in the parental generation). This measure of reproduction counts each egg-laying event as one reproductive event and, thus, estimates the probability of laying one sac. We estimated the total egg sac production of a generation from the periodic counts of egg sacs, assuming that each sac remained in a colony for 4 wk, the average period of development of the eggs (Avile´s 1986). This method assumes that no sacs were removed by predators or discarded by the spiders prior to completing their developmental period. If either of these events were common, our method would underestimate the number of sacs produced. In these estimates, we included only colonies that survived the entire egglaying period in order to avoid confounding survival with fertility effects (table 1). Mean clutch size in a colony (factor 2, right-hand side of eq. [1]) was estimated as the average number of eggs per sac in samples of one to 17 sacs collected from natural and artificially established colonies in their F1 and F2 generations (table 1). Because a fraction of the 177 egg sacs sampled contained parasitoids that had destroyed part of the clutch, we considered the effect of colony size on the average number of eggs per sac in unparasitized sacs only and in parasitized and unparasitized sacs combined. We excluded from these analyses three unparasitized egg sacs from the F1 generation that contained abnormally few eggs and a colony with close to 500 individuals that appeared as an outlier. We analyzed separately the data for the two generations because the slopes of the regressions of mean eggs per sac on colony size were not homogeneous across generations, whether all sacs (F ⫽ 9.6, df ⫽ 1, 28, P ⫽ .004) or only unparasitized sacs (F ⫽ 7.4, df ⫽ 1, 27, P ⫽ .01) were considered (P values are for the interaction between colony size and generation in a weighted least squares ANCOVA). We combined, on the other hand, the data for natural and experimental colonies because the slopes for these two data sets were not significantly different from each other (F ⫽ 0.24, df ⫽ 1, 28, P ⫽ .63 when all sacs are considered; F ⫽ 0.63, df ⫽ 1, 27, P ⫽ .44 when only unparasitized sacs are considered). Regressions were weighted by the number of sacs used to estimate mean clutch sizes because we considered that estimates based on a larger number of sacs were more precise (sample sizes reflected the number of sacs available in the colonies). Using contingency table analysis, we also investi-

gated whether there was an interaction between colony size and the proportion of parasitized egg sacs. We analyzed the F1 and F2 generations separately because of a significant difference in the level of egg sac parasitism in these two generations. The proportion of offspring that survived to maturity (factor 3, right-hand side of eq. [1]) was estimated indirectly by dividing the total number of adult/subadult females in the F1 generation by the total number of eggs estimated to have been produced by the parental generation. Since we did not have direct counts of eggs but only counts of sacs produced by the parental generation, we used colony size to predict expected number of eggs per sac for colonies of different sizes. For this purpose, we used two extreme cases of the previously established relationship between colony size and clutch size: a relatively flat function, as in the F1 generation when parasitized and unparasitized sacs are considered, and a more or less steeply increasing function, as in the F2 generation when only unparasitized sacs are considered. We tested for the effect of colony size on offspring survival using a nonparametric ANOVA of the data grouped in six size classes and linear and polynomial regressions on arcsine-transformed proportions. In these analyses, we included all colonies where at least one sac was laid (table 1). The net per capita rate of growth or average LRS of females in the nests (left-hand side, eq. [1]) was estimated directly by dividing the total number of offspring that became adult/subadult in a colony by the total number of adult/subadult females in the parental generation. Colonies that went extinct before their offspring became adult/subadult were counted as having a growth rate of 0. We used nonparametric ANOVA of the data grouped in six size classes to investigate the effect of colony size on LRS. All 34 natural and artificially established colonies were included in these analyses (table 1). The six size classes, which are the same as those used in the offspring survival analyses, were determined a priori using natural breaks in the distribution of colony sizes while trying to make the size classes equivalent in range and number of colonies they contained. We investigated the shape of the functions relating colony size to overall fitness and the different fitness components using both a model-free curve-fitting method (distance weighted least squares, DWLS; see SYSTAT 1992, p. 260, for further details) and least squares regressions with alternative linear, polynomial, and nonlinear models. The nonlinear models were derived from first principles with elements taken from standard population growth models to provide functions that would be biologically more realistic than linear or quadratic models that increase with no bound or become negative for some ranges of the independent variable. To discriminate

Individual Fitness in a Social Spider 407 between alternative models, we used the generalized likelihood ratio test (Borowiak 1989, p. 69) for nested nonlinear models or compared R 2 and residual mean square values for nonnested models (Bates and Watts 1988, p. 107). The exact form of the functions, however, is not critical, as long as they provide a reasonable fit to the data. Predicted values from the parameterized models were replaced in the right-hand side of equation (1) to obtain a predicted overall fitness function with which to compare the directly estimated LRS (see above). In all cases, ‘‘colony size’’ corresponds to the number of subadult and adult females present in the colonies at the start of the parental generation (2 wk after foundation in the case of the artificially established colonies) or at the time the egg sacs were collected. A secondary goal of this study was to investigate the effect of colony size on the probability of a colony surviving to the next generation and on the average life span of the colonies. A colony was considered to have survived until the next generation when at least one member of the F1 generation became adult, a process that required at least 24 wk. Colony life spans were estimated from the extinction date records (to 3-wk precision) of marked colonies that remained stationary in their prerecorded locality. We used logistic regression to assess the effect of colony size on the probability of a colony surviving to the next generation and nonparametric actuarial analyses (StatView; Abacus Concepts 1994) to estimate cumulative hazard and survival functions of the artificially established and natural colonies. We treated as right censored

the colonies that did not naturally go extinct during the study period. Trend rank tests were used to investigate whether different colony sizes had the same hazard (and, equivalently, survival) functions. For purposes of these tests, we grouped the colonies according to their size at foundation or first sighting. Results Effects of Colony Size on the Probability of Female Reproduction The average number of egg sacs laid per female consistently decreased with colony size (F ⫽ 66.6, df ⫽ 1, 25, P ⬍ .00001; fig. 1). While nine out of 10 solitary females laid an egg sac each, the average female in colonies of about 100 individuals produced only 0.26 sacs. This latter result implies that, at the most, one out of four females in these larger colonies reproduced—less if any female had laid more than one sac. Linear and nonlinear models were fit to these data, with the nonlinear model explaining a higher proportion of the variance (fig. 1). Effects of Colony Size on Clutch Size If we consider unparasitized sacs only, we found that the mean number of eggs per sac among colonies sampled in both generations significantly increased with colony size (F ⫽ 6.70, df ⫽ 1, 12, P ⫽ .02 for the F1; F ⫽ 11.33, df ⫽ 1, 17, P ⫽ .004 for the F2). As noted in the ‘‘Material and Methods’’ section, however, the slopes of the re-

Figure 1: Probability of female reproduction in Anelosimus eximius natural and artificially established colonies, as estimated from the ratio of egg sacs per female in colonies of different sizes. A, Raw data shown with linear (dashed line) (R 2 ⫽ 0.72; adjusted R 2 ⫽ 0.71, residual MS ⫽ 0.023) and model-free curve fitting (solid line) (DWLS, tension ⫽ 0.2). Open circle ⫽ nine data points; black circles ⫽ one data point. B, Mean ⫾ SEs (10,000 bootstraps) of the data grouped in six size classes, showing function obtained by fitting to the raw data the model y ⫽ a/(1 ⫹ bx), where a ⫽ 1.022 ⫾ 0.037 and b ⫽ 0.030 ⫾ 0.004 (raw R 2 ⫽ 0.98; corrected R 2 ⫽ 0.90, residual MS ⫽ 0.010).

408 The American Naturalist

Figure 2: Effect of colony size on the mean number of eggs per sac in colonies of Anelosimus eximius sampled in two separate generations. A, F1 generation. B, F2 generation. Mean ⫾ SEs (parametric) of unparasitized sacs only shown for each colony. Solid lines obtained by fitting the model y ⫽ a ⫹ bx to the mean number of eggs per sac in unparasitized sacs only. Dashed lines obtained by fitting the model y ⫽ a ⫹ bx to the mean number of eggs per sac in unparasitized and parasitized sacs combined.

gressions were not homogeneous across generations as the F2 had a steeper slope (fig. 2; note the different scales of the X-axis). If we include parasitized sacs in these analyses (fig. 2, dashed lines), colony size no longer had a significant effect in the F1 generation (F ⫽ 0.59, df ⫽ 1, 13, P ⫽ .46). This loss of significance apparently was a result of a high incidence of egg sac parasitoids in this generation (see below). With a low incidence of egg sac parasitoids, colony size continued to have a significant effect in the F2 generation even when both parasitized and unparasitized sacs were included in the analysis (F ⫽ 8.17, df ⫽ 1, 17, P ⫽ .01).

the sacs were parasitized in the F2 generation, 42.5% of the sacs were parasitized in the F1. After controlling for colony size, this difference was significant at the P ⫽ .002 level (Mantel-Haenszel χ 2 ⫽ 9.9 for the data grouped in seven colony size classes). In both generations, there was an abrupt switch from no parasitized egg sacs among small colonies (⬍140 individuals in the F1 and ⬍60 in F2) to some level of infestation among larger colonies, although not all large colonies were infested (fig. 3). The association between colony size and the proportion of parasitized sacs, however, was only significant in the F1—or more heavily parasitized—generation (likelihood ratio χ 2 ⫽ 15.6, df ⫽ 6, P ⫽ .02).

Egg Sac Parasites, Clutch Size, and Colony Size Forty-five out of the 177 sacs sampled were parasitized, containing from one to 12 (median ⫽ 6) larvae or pupae of a parasitoid wasp in the family Eulophidae. Infested egg sacs contained significantly fewer eggs as the wasps had apparently destroyed part of the clutch, while unparasitized sacs contained an average of 38.4 ⫾ 1.74 eggs (mean ⫾ 95% confidence interval; CI), parasitized sacs contained only 26.9 ⫾ 3.45 eggs (t ⫽ 6.23, df ⫽ 175, P ⬍ .0001). Further, among parasitized egg sacs, there was a significant trend toward fewer eggs per sac as the number of parasites in the sacs increased (F ⫽ 6.09, df ⫽ 1, 43, P ⫽ .02). A high incidence of egg sac parasitism, combined with a higher probability of infestation of larger colonies, probably explains why colony size no longer had a significant effect on clutch size in the F1 generation when parasitized and unparasitized sacs were included in the analysis (fig. 2A). While only 13.5% of

Effect of Colony Size on Offspring Survival The proportion of offspring that survived to maturity increased with colony size, whether we assumed that the average number of eggs per sac was independent of colony size, as in the F1 generation when both parasitized and unparasitized sacs were included (F ⫽ 14.2, df ⫽ 1, 28, P ⬍ .0001; fig. 4) or whether we assumed an increasing function of colony size, as in the F2 generation when only unparasitized sacs were included (F ⫽ 12.6, df ⫽ 1, 28, P ⫽ .001). While about 4% of the offspring survived in colonies containing less than 10 individuals, 18%– 24% of the offspring survived in colonies of 51–107 individuals (table 2). The effect was most pronounced among the smallest size classes, and it leveled off or even declined among the largest size classes (fig. 4), as suggested by the shape of the DWLS curve and a significant second order term in a quadratic regression (t ⫽ ⫺3.10, P ⫽

Individual Fitness in a Social Spider 409

Figure 3: Proportion of parasitized egg sacs in Anelosimus eximius colonies in their F1 (A) and F2 (B) generations. Open circles ⫽ two data points; black circles ⫽ one data point.

.005, under either assumption, N ⫽ 30 colonies). We have fit to these data a unimodal fitness function, which, unlike the quadratic, is positive for all values of N ⬎ 0 (fig. 4B). Effects of Colony Size on Colony Survival The probability that a colony as a whole survived until the next generation increased significantly as the number of subadult/adult females in the parental generation increased (logistic regression χ 2 ⫽ 22.5, P ⬍ .0001, N ⫽ 34

colonies). Only one out of 10 colonies founded by a single female survived the minimum of 24 wk required for at least some offspring to become adult, while all colonies containing close to 100 individuals survived for at least that long (fig. 5). Through time, this effect resulted in a lower cumulative risk of extinction and longer survival times of larger colonies among both natural and experimental colonies. Among natural colonies, the mean life span from first sighting of colonies with a single adult female was 3.7 ⫾ 1.1 wk (X ⫹ SE, N ⫽ 11), while that of colonies of two to four individuals was 21.3 ⫾ 4.7 wk

Figure 4: Effect of colony size on the proportion of offspring estimated to have survived to the subadult/adult stage in Anelosimus eximius natural and artificially established colonies. Estimates shown were obtained under the assumption that the number of eggs laid per sac was a relatively flat function of colony size, as in the F1 generation when parasitized and unparasitized sacs were included in the analyses. A, Linear (dashed line) regression (R 2 ⫽ 0.21, adjusted R 2 ⫽ 0.18) and model-free DWLS (solid line) curve (tension ⫽ 0.2). Open circle ⫽ eight data points; black circles ⫽ one data point. B, Means ⫾ SEs (10,000 bootstraps) of the data grouped in six size classes, showing the function obtained by fitting to the raw data the model y ⫽ (1 ⫺ d x )/(1 ⫹ bx), where d ⫽ 0.99 ⫾ 0.004, b ⫽ 0.018 ⫾ 0.010 (raw R 2 ⫽ 0.64, corrected R 2 ⫽ 0.32; the model with the 1/(1 ⫹ bx) factor is significantly better than the model without this factor using the generalized likelihood ratio test, χ 2 ⫽ 7.1, P ⫽ .01).

410 The American Naturalist Table 2: Mean ⫾ SEs of the proportion of offspirng estimated to have survived to the subadult/adult stage in colonies of Anelozimus eximius classified in six size

Range of colony sizes in size class 1–2 3–9 23–46 51–75 88–107 144–165

Number of colonies 9 2 5 7 5 2

Mean proportion of surviving offspring*

Mean proportion of surviving offspring†

X

⫾ SE

X

⫾ SE

.04 .04 .13 .24 .24 .17

.03 .03 .05 .07 .05 .05

.03 .04 .10 .18 .18 .12

.03 .03 .04 .05 .04 .03

Note: Mean proportions are averages among colonies within each size class; SEs were estimated by bootstrapping the data within the size classes 10,000 times. * Estimates are based on the assumption that the number of eggs per sac produced by the parental generation were a relatively flat function, as in the F1 generation when parasitized and unparasitized sacs were included. Kruskal-Wallis test statistic ⫽ 13.6, df ⫽ 5, P ⫽ .02. † Estimates are based on the assumption that the number of eggs per sac produced by the parental generation were an increasing function of colony size, as in the F2 generation when only unparasitized sacs were included. Kruskal-Wallis test statistic ⫽ 13.0, df ⫽ 5, P ⫽ .02.

(N ⫽ 5) and that of colonies of ⬎4 to over 100 individuals was 45.0 ⫾ 4.9 wk (N ⫽ 14). These differences were statistically significant at the P ⬍ .0001 level (Mantel-Cox trend rank test χ 2 ⫽ 20.4, df ⫽ 1). Among the artificially established colonies, colonies that initially contained 50– 165 individuals (N ⫽ 17) survived longer than colonies with less than 50 individuals (N ⫽ 10, χ 2 ⫽ 5.2, df ⫽ 1, P ⫽ .02).

Figure 5: Proportion of Anelosimus eximius colonies in six size classes that survived to the next generation as a function of colony size. The function shown was obtained by fitting to the data the model y ⫽ 1 ⫺ d x, where d ⫽ 0.958 ⫾ 0.04. Figures by each of the dots correspond to the number of colonies in each size class used in the estimates.

Effect of Colony Size on Female Lifetime Reproductive Success There was a significant difference in the mean reproductive success of females in the six colony size classes (Kruskal-Wallis test statistic ⫽ 14.5, df ⫽ 5, P ⫽ .01). While solitary females (N ⫽ 10) and female pairs (N ⫽ 1) produced on average 0.82 ⫾ 0.77 adult/subadult offspring per capita (X ⫾ bootstrap SE), females in colonies of 51–75 individuals produced 2.55 ⫾ 0.83 offspring per capita, the largest average growth among the colony size classes (table 3). Females in the largest size class (144– 165 individuals), however, had a lower LRS than females in intermediate-sized colonies (23–107 individuals), suggesting a reversal of the effect of colony size on individual fitness beyond a certain colony size. Such a reversal is also suggested by a significant second-order term in a quadratic regression (t ⫽ ⫺2.40, P ⫽ .02 using square root–transformed values) and a unimodal DWLS curve (fig. 6). With only two colonies in the largest size class, however, this result should be considered tentative. Additional studies that include colonies larger than considered here would be necessary to confirm this reversal and to determine its location, shape, and magnitude. For heuristic purposes, we have used the raw LRS data to fit a function that contains negative (e⫺cx, x ⫽ colony size, as previously defined) and positive (x γ, 0 ⱕ γ ⱕ 1) densitydependent factors representing, respectively, the effects of competition and cooperation on individual fitness (Avile´s, in press). A unimodal fitness function results

Individual Fitness in a Social Spider 411 Table 3: Nonparametric ANOVA for the effect of colony size on the mean per capita production of grown offspring (LRS) in 13 natural and 21 artificially established colonies of Anelosimus eximius at the Cuyabeno Nature Reserve Range of colony sizes in size class 1–2 3–9 23–46 51–75 88–107 144–165

Number of colonies

Mean per capita growth rate

SE

Rank sum

11 4 5 7 5 2

.82 .61 1.81 2.55 2.19 1.32

.77 .53 .87 .83 .85 .41

109.0 50.5 102.5 172.0 118.0 43.0

Note: Mean net per capita growth rates are averages among colonies within each size class. Standard errors were estimated by bootstrapping the data 10,000 times within each size class. Kruskal-Wallis test statistic ⫽ 14.54, P ⫽ .01, assuming a χ 2 distribution with df ⫽ 5.

when the positive density-dependent factor is larger than 1 (or γ ⬎ 0) (fig. 6). Discussion Consequences of Colony Size on Individual Fitness and Its Components The results of this study indicate that group living increases the average LRS in the social spider Anelosimus eximius. At a tropical rain forest site in eastern Ecuador, solitary females and female pairs in forest understory nests raised an average of 0.8 offspring per capita, while females in larger colonies raised an increasing number of offspring, up to a maximum of two and a half offspring

per capita in colonies of about 60 individuals (table 3 and fig. 6). Beyond this maximum, this trend apparently reversed, although, due to the paucity of data in the large colony-size range, this result should be considered tentative. Assuming that this result is confirmed, we predict by extrapolation (using the function shown in fig. 6) that the average female LRS would reach levels equivalent to that of solitary individuals in colonies containing between 300 and 400 spiders This apparently unimodal fitness function was a consequence of the conflicting effects of colony size on different components of fitness (fig. 7). The primary benefit of group living appeared to be an increase in the probability of offspring survival. While, on average, less than

Figure 6: Effect of colony size on the mean per capita production of grown offspring (LRS) in Anelosimus eximius colonies containing from one to 165 individuals. A, Model-free curve fitting using DWLS, tension ⫽ 0.2. Open circle ⫽ 10 data points; black circles ⫽ one data point. B, Data grouped in six colony size classes, with X ⫾ SEs (10,000 bootstraps) shown for each size class. The function in B was obtained by fitting the model y ⫽ Re⫺cx x γ to the ungrouped data shown in A, where R ⫽ 0.46 ⫾ 0.67, c ⫽ 0.01 ⫾ 0.01, and g ⫽ 0.55 ⫾ 0.53 (raw R 2 ⫽ 0.39; corrected R 2 ⫽ 0.10).

412 The American Naturalist

Figure 7: Summary of the functions estimated to relate colony size to three components of female fitness in Anelosimus eximius forest understory colonies. The bottom panels show the point-by-point product of the predicted values of each of these components, as in the right-hand side of equation (1). Note that the resulting function is of a shape and magnitude comparable to the average LRS function estimated directly, as shown in figure 6B. The offspring survival functions were estimated using two alternative assumptions of the relationship between colony size and average number of eggs per sac: a relatively flat function, as in the F1 generation when parasitized and unparasitized sacs were included in the analyses ( panels in the left column) and an increasing function, as in the F2 generation when only unparasitized sacs were included.

5% of the offspring survived in colonies of less than 10 individuals, around 20% of the offspring survived in colonies containing 50–100 individuals. The probability of female reproduction, however, decreased with colony size. Nine out of 10 females that established nests on their own laid an egg sac each, while at the most one out of four females in nests of about 100 individuals reproduced. Clutch size was an increasing function of colony size in a generation in which the incidence of egg sac parasitism was low, suggesting that females, although having a smaller chance to reproduce in large colonies, can expect to produce a larger brood. This effect, however, was nullified by a large incidence of egg sac parasitoids among large colonies in the other generation sampled. According to equation (1), the three fitness components estimated in this study—probability of female reproduction, mean clutch size, and offspring survival— should be responsible for the shape and magnitude of the overall LRS function. Figure 7 shows that a point-bypoint product of the predicted values of these three fitness components for colonies ranging in size from one to 200 individuals produces a function (fig. 7, lower panels) of a shape and magnitude closely resembling the separately derived function for the LRS (fig. 6).

In addition to its effect on individual fitness, colony size was found to have an effect on the survival probability of whole colonies. Only one out of 10 colonies founded by a single female survived until their offspring reached reproductive maturity, while all colonies with more than 100 individuals survived at least one generation. This result, however, should not be extrapolated to colonies larger than those included in this study as the probability of survival to the next generation may again decrease among extremely large colonies (L. Avile´s, personal observations). An important point to emphasize is that the particular fitness measures presented here are a result of the interaction between individuals in groups of different sizes and the particular environmental conditions prevailing when and where this study was conducted. Therefore, the exact shape and magnitude of the various functions may not be the same had this study been conducted at a different time or place. This is illustrated by the different slopes of mean clutch size on colony size in the F1 and F2 generations (fig. 2), the different levels of egg sac parasitism in these two generations, the failure of a first set of artificially established colonies (see next section), and the wider size range of forest edge colonies that suggests

Individual Fitness in a Social Spider 413 a wider and perhaps taller fitness function in this habitat. The functions presented here, therefore, are a sample in time and space of the possible fitness functions of this social spider. Estimates of the effect of colony size on only certain aspects of individual fitness are available for other social spiders. The most complete study is that of Uetz and Hieber (1997, and papers cited therein) in the colonial Metepeira incrassata. In this species, which forms nests with common supporting lines but individual webs, Uetz and Hieber found that the production of live hatchlings was maximum at intermediate colony sizes. In contrast, hatchling production was independent of colony size in a different colonial spider studied by Smith (1982). Among species of a similar level of sociality as A. eximius, several authors have also reported on a lower probability of female reproduction (Seibt and Wickler 1988b; Henschel 1991–1992) and increased probability of colony survival (Riechert et al. 1986; Lubin 1991; J. Henschel, personal communication) as the size of the colonies increases. In contrast to the findings reported here, fewer eggs per sac were laid in larger colonies in an African social spider (Seibt and Wickler 1988b). No data are available on the effect of colony size on the probability of offspring survival in these other species. Two authors (Vollrath 1986; Rypstra 1993) have interpreted the lack of universal reproduction in A. eximius as evidence that this species has crossed the threshold toward eusociality, the level used to characterize the social insects with reproductive castes. We disagree with this interpretation. The implication behind the term eusociality is that sterility or subfertility has been selected as a socially adaptive trait (Crespi and Choe 1997). The fact that in A. eximius the probability of female reproduction is a decreasing function of colony size suggests that differences in reproductive success are merely a side product of crowding and competition within the growing colonies. With no other evidence, therefore, no special selective forces need to be invoked to explain a situation typical of animal populations where individuals are subject to negative density-dependent effects. Labeling the social spiders as ‘‘eusocial’’ only distracts attention from the fundamental differences between these organisms and the eusocial insects. While the social spider colonies constitute relatively closed intrabreeding populations, the insects have formed colonies that are extended family groups whose members disperse to find mates outside their nests. We believe that explaining why spiders and insects have given rise to such fundamentally different systems is one of the current challenges in the study of animal sociality. Overall, this study serves to illustrate the need to consider all factors likely to affect LRS. A failure to measure

offspring survival, for instance, would have led us to the mistaken conclusion that group living decreases individual fitness in this social spider. In a frequently cited paper, Michener (1964) noted that the per capita female reproductivity in several hymenopteran species was a decreasing function of colony size (but, see Itoˆ 1993; Jeanne and Nordheim 1996). In that respect, Michener’s data closely resembles the data presented here in figure 1. Michener’s studies, however, did not follow the eventual fate of the eggs or larvae on which the estimates were based. It is possible that if offspring survival had been factored in, the LRS of colony members in those hymenopteran species would have been found to be highest at some intermediate colony size. There is a further component of individual fitness that was not considered in this or Michener’s study that may further tilt the balance in favor of group living. That is the probability of successful dispersal. The LRS estimates presented here apply only to individuals that are already established in a nest; no attempt was made to locate dispersing individuals. If the risks of dispersal are taken into account, it is likely that the fitness of individuals that leave their colony to breed on their own may be considerably lower than the fitness of solitary individuals reported here. It is also likely that the probability of successful dispersal may be higher for individuals that disperse as a group, further discouraging a solitary lifestyle. Environmental Factors Responsible for the Observed Fitness Patterns While this study provides information regarding the fitness consequences of group living in this social spider, it does not, with the exception of the egg sac parasitoids, identify the particular environmental factors responsible for the observed patterns. Suggested benefits of group living in various social spider species include increased efficiency in the capture of large prey (Nentwig 1985; Rypstra 1990; Pasquet and Krafft 1992), increased protection from predators (J. Henschel, personal communication), and reduced per capita investment in silk (Riechert 1985; Tietjen 1986; Jakob 1990). Costs have been suggested to arise because prey availability does not keep pace with the increase in spider numbers as nest size increases (Ward 1986) and, in A. eximius, because competition over large prey may set up a dominant-subordinant dichotomy that may cause the reproductive failure of some females (Rypstra 1993). Increased predator protection and improved ability of groups to acquire resources may be responsible for the increased offspring survival probability in larger colonies. Behaviors such as egg sac guarding, prey sharing with the

414 The American Naturalist spiderlings, and regurgitation feeding would directly influence the proportion of surviving offspring, in particular because individuals do not discriminate in favor of their own offspring when performing these behaviors (Kullmann 1972; Krafft 1979; Christenson 1984; Seibt and Wickler 1988a; Avile´s 1993). As the size of the colonies continues to increase, however, crowding and competition may eventually undermine the ability of individuals to care for their offspring and may lead to lower offspring survival probabilities in colonies larger than those considered in this study. An abiotic factor likely to affect the well being of the spiders in tropical rainforest habitats is the amount of precipitation that larger colonies are apparently better able to withstand (Riechert et al. 1986). Heavy rains may have been responsible for the failure of a first set of colonies that we artificially established in August 1993, at the start of a period of heavy precipitation in the area (1,155 mm of rain during the following 3 mo). These colonies went extinct at a rate per 3-wk period that was correlated with the amount of precipitation in the preceding 3 wk (Spearman ρ ⫽ 0.93, P ⫽ .02; see Tufin˜o 1997, for further details). All but one of these colonies were extinct 4 mo after foundation. In contrast, the second set of artificially established colonies, which were established at the start of the drier season (664 mm of rain in 3 mo), persisted into their F2 generation. The amount of precipitation apparently interacted with colony size to cause the extinction of this first set of colonies, as larger colonies survived longer (Tarone-Ware trend rank test χ 2 ⫽ 4.6, P ⫽ .03). Because the rain can cause extensive damage to the webs (L. Avile´s, personal observation), it would force the spiders to spend extra time and energy in web repair activities. Under these circumstances, larger colonies may be able to repair equivalent amounts of damage with less per capita expenditure in silk. Group Size Evolution This study opens the question of whether the size distribution of naturally occurring colonies of A. eximius reflects the consequences of different group sizes on individual fitness. Figure 8 shows that colonies both smaller and larger than the size that maximizes individual fitness are the rule rather than the exception among understory colonies of this species. A broad range of colony sizes appears characteristic of other social spiders as well (Avile´s 1997). Why is this so? Why are the colonies not maintained at a size that maximizes individual fitness? Why are propagules that contain a single female still being produced? The existence of colonies larger than the optimum size can be addressed by considering models of group size

evolution (Rodman 1981; Sibly 1983; Giraldeau 1988; Higashi and Yamamura 1993; Rannala and Brown 1994). According to these models, the size that maximizes individual fitness (the ‘‘optimum’’ group size) is not stable because outsiders should continue to join groups in which insider fitness has not yet dropped to the level of solitary individuals. The paradoxical situation that, when group size finally stabilizes (at the ‘‘stable’’ group size), group living is no longer an evolutionarily stable strategy (Rannala and Brown 1994) can be resolved by considering the indirect fitness benefits of allowing relatives to join the groups and the existence of mechanisms to prevent outsiders from joining (Giraldeau and Caraco 1993; Higashi and Yamamura 1993). Given these considerations, the groups are expected to attain a compromise size that is intermediate between the optimum and the stable group size. The assumption that individuals are free to leave or to join a group at any time, however, makes these models of only limited applicability to social spiders. In social spiders, and other animal societies that have any demographic stability, colony growth tends to occur through the recruitment of individuals produced within the groups. This demographic nature of the groups may explain the existence of a wide range of colony sizes in these organisms (fig. 8). If the groups are founded at a relatively small size, for instance, it may take several generations before they attain the optimum or the stable size. Similarly, instantaneous dispersal may not be possible because individuals may need to reach a particular age or stage before they can disperse. In A. eximius, for instance, active dispersal apparently only involves inseminated adult females (Vollrath 1982; L. Avile´s, personal observation). As individuals mature to the dispersal age, the groups may on occasion become overcrowded, fail to acquire sufficient resources, and crash rather than disperse. Additionally, these models do not take into account the costs of dispersal that should further discourage individuals from leaving their groups. If dispersal costs are taken into account, then the group size that triggers dispersal may be even larger than the stable group size. It is clear that models are needed that take into account the costs of dispersal and the demographic nature of many animal societies. The question of why single females are still establishing groups on their own when they only have a slim chance of success is more difficult to address. It is likely, however, that there is a large stochastic component to the success of solitary dispersers and that a small fraction of them may sometimes do really well. In fact, the only colony established by a single female that succeeded in reaching the next generation had the largest per capita offspring production of any colony in the study (fig. 6).

Individual Fitness in a Social Spider 415

Figure 8: Size distribution of Anelosimus eximius colonies (number of individuals estimated from the area of the cross section of the nest) in the forest understory at the Cuyabeno Nature Reserve. A colony containing about 5,000 individuals not shown in the figure.

At the colony level, producing many small propagules may be a risk-prone strategy that maximizes dispersal distance and the number of sites colonized, while producing a few large propagules may be a risk-averse strategy to ensure propagule survival. As with individual-level life-history strategies (e.g., Stearns 1992), which strategy characterizes particular species may depend on the prevailing conditions of its typical environment. Again, further study is needed in this area. Why Are Social Spiders and Other Social Organisms Social? Finally, looking at the LRS success of solitary individuals of A. eximius may suggest why this species is social. With an average production of 0.8 offspring per capita (probably less if dispersal risks are considered), solitary individuals and populations consisting entirely of solitary individuals would be unable to replace themselves in the environment they inhabit. The colonization of the tropical rain forest by this species or its ancestor, therefore, would have required the evolution of group living and cooperation. In other social organisms, the situation could be even more extreme. In the Australian cooperative breeding choughs, for instance, groups of at least four individuals are necessary for the successful fledging

of a single chick a year (Heinsohn 1992). In the naked mole rats, groups of less than a dozen individuals apparently do not exist (Brett 1991). Models for the evolution of group size and cooperation, therefore, should consider that groups smaller than a certain size may not even be a dynamically feasible alternative (Avile´s, in press). We conclude by arguing that group living and cooperation may have arisen in many instances because individuals of certain species would be unable to inhabit particular environments without the help of a group. In the case of the social spiders, this is a somewhat different explanation from the most commonly stated argument that their social behavior arose to take advantage of the ecological opportunity that large prey items represent (Nentwig 1985; Rypstra 1990; Uetz 1992). The latter argument has also been used to explain sociality in vertebrate social carnivores (Macdonald 1983). In the case of the spiders, at least, the results of this study suggest that access to larger prey may only be a secondarily derived benefit of group living as a basic survival strategy. Acknowledgments This study formed part of the Licenciatura thesis of P.T. at the Pontificia Universidad Cato´lica del Ecuador (PUCE) in Quito and was funded in part by the Research

416 The American Naturalist Training Group for the Analysis of Diversification at the University of Arizona, a small grant from the University of Arizona Foundation, and National Science Foundation grant DEB-9707474 to L.A. We wish to thank L. Arcos Tera´n, T. De Vries, and A. Padilla from the Department of Biology at the PUCE for making this collaboration possible; the Instituto Ecuatoriano Forestal y de Areas Naturales y Vida Silvestre for permission to conduct research at the Cuyabeno Nature Reserve; S. Cornejo of Nuevo Mundo Travel; F. Herrero and Don˜a Zulema for their hospitality; and G. Este´vez for field assistance in the early stages of this project. Special thanks to E. Dyresson and L. Roseman for invaluable statistical advice and to M. Hunter for identifying the parasitoid wasps. For comments on the manuscript we wish to thank W. Maddison, D. Miller, L. Roseman, Y. Ziv, and, in particular, J. Kingsolver, L. Rayor, and an anonymous reviewer. Literature Cited Abacus Concepts. 1994. Survival tools for StatView, version 4.1. Abacus Concepts, Berkeley, Calif. Alexander, R. D. 1974. The evolution of social behavior. Annual Review of Ecology and Systematics 5:325–383. Avile´s, L. 1986. Sex-ratio bias and possible group selection in the social spider Anelosimus eximius. American Naturalist 128:1–12. ———. 1992. Metapopulation biology, levels of selection and sex ratio evolution in social spiders. Ph.D. diss. Harvard University, Cambridge, Mass. ———. 1993. Newly-discovered sociality in the neotropical spider Aebutina binotata Simon (Dictynidae?). Journal of Arachnology 21:184–193. ———. 1997. Causes and consequences of cooperation and permanent sociality in spiders. Pages 476–498 in J. Choe and B. Crespi, eds. The evolution of social behavior in insects and arachnids. Cambridge University Press, Cambridge. ———. In press. Cooperation and nonlinear dynamics: an ecological perspective on the evolution of sociality. Evolutionary Ecology. Avile´s, L., and W. P. Maddison. 1991. When is the sex ratio biased in social spiders? chromosome studies of embryos and male meiosis in Anelosimus species (Araneae, Theridiidae). Journal of Arachnology 19:126– 135. Bates, D. M., and D. G. Watts. 1988. Nonlinear regression analysis and its applications. Wiley, New York. Booth, D. J. 1995. Juvenile groups in a coral-reef damselfish: density-dependent effects on individual fitness and population demography. Ecology 76:91–106. Borowiak, D. S. 1989. Model discrimination for nonlinear regression models. Marcel Dekker, New York.

Brett, R. A. 1991. The population structure of naked mole-rat colonies. Pages 97–136 in P. W. Sherman, J. U. M. Jarvis, and R. D. Alexander, eds. The biology of the naked mole-rat. Princeton University Press, Princeton, N.J. Brown, C. R., and M. B. Brown. 1996. Coloniality in the cliff swallow. University of Chicago Press, Chicago. Buss, L. 1981. Group living, competition, and the evolution of cooperation in a sessile invertebrate. Science (Washington, D.C.) 213:1012–1014. Caraco, T., and H. R. Pulliam. 1984. Sociality and survivorship in animals exposed to predation. Pages 279– 309 in P. W. Price, C. M. Slobodchikoff, and W. S. Gaud, eds. A new ecology: novel approaches to interactive systems. Wiley, New York. Caraco, T., and L. L. Wolf. 1975. Ecological determinants of groups sizes of foraging lions. American Naturalist 109:343–352. Cash, K., M. McKee, and F. Wrona. 1993. Short- and long-term consequences of grouping and group foraging in the free-living flatworm Dugesia tigrina. Journal of Animal Ecology 62:529–535. Christenson, T. E. 1984. Behaviour of colonial and solitary spiders of the Theridiid species Anelosimus eximius. Animal Behaviour 32:725–734. Clark, C. W., and M. Mangel. 1986. The evolutionary advantages of group foraging. Theoretical Population Biology 30:45–75. Crespi, B. J., and J. C. Choe. 1997. Explanation and evolution of social systems. Pages 449–524 in J. Choe and B. Crespi, eds. The evolution of social behavior in insects and arachnids. Cambridge University Press, Cambridge. D’Andrea, M. 1987. Social behaviour in spiders. Italian Journal of Zoology, Monograph 3:1–156. Darchen, R., and B. Delange-Darchen. 1986. Societies of spiders compared to the societies of insects. Journal of Arachnology 14:227–238. Emlen, S. T., and S. L. Vehrencamp. 1985. Cooperative breeding strategies among birds. Pages 359–374 in B. Ho¨lldobler and M. Lindauer, eds. Experimental behavioral ecology. Verlag, Stuttgart. Giraldeau, L.-A. 1988. The stable group and the determinants of foraging group size. Pages 33–53 in C. N. Slobodchikoff, ed. The ecology of social behavior. Academic Press, San Diego, Calif. Giraldeau, L.-A., and T. Caraco. 1993. Genetic relatedness and group size in an aggregation economy. Evolutionary Ecology 7:429–438. Heinsohn, R. G. 1992. Cooperative enhancement of reproductive success in white-winged choughs. Evolutionary Ecology 6:97–114. Henschel, J. R. 1991–1992. Is solitary life an alternative

Individual Fitness in a Social Spider 417 for the social spider Stegodyphus dumicola? Namibia Wissenschaftliche Gesellschaft 43:71–79. Higashi, M., and N. Yamamura. 1993. What determines animal group size? insider-outsider conflict and its resolution. American Naturalist 142:553–563. Itoˆ, Y. 1993. Behavior and social evolution of wasps. Oxford University Press, Oxford. Jakob, E. M. 1990. Costs and benefits of group living for pholcid spiderlings: losing food, saving silk. Animal Behaviour 41:711–722. Jeanne, R. L., and E. V. Nordheim. 1996. Productivity in a social wasp: per capita output increases with swarms size. Behavioral Ecology 7:43–48. Komdeur, J. 1994. Experimental evidence for helping and hindering by previous offspring in the cooperativebreeding Seychelles warbler Acrocephalus sechellensis. Behavioral Ecology and Sociobiology 34:175–186. Konig, B. 1994. Components of lifetime reproductive success in communally and solitarily nursing house mice—a laboratory study. Behavioral Ecology and Sociobiology 34:275–283. Krafft, B. 1979. Organisation et e´volution des socie´te´s d’araigne´es. Psychology 1:23–51. Kullmann, E. J. 1972. Evolution of social behavior in spiders (Araneae, Eresidae and Theridiidae). American Zoologist 12:419–426. Levi, H. W. 1963. The American spiders of the genus Anelosimus (Araneae, Theridiidae). Transactions of the American Microscopical Society 82:30–48. Lubin, Y. D. 1991. Patterns of variation in female-biased colony sex ratios in a social spider. Biological Journal of the Linnean Society 43:297–311. ———. 1995. Is there division of labour in the social spider Achaearanea wau (Theridiidae)? Animal Behaviour 49:1315–1323. Macdonald, D. W. 1983. The ecology of carnivore social behavior. Nature (London) 301:379–384. Michener, C. D. 1964. Reproductive efficiency in relation to colony size in hymenopterous societies. Insectes Sociaux 11:317–342. Nentwig, W. 1985. Social spiders catch larger prey: a study of Anelosimus eximius (Araneae: Theridiidae). Behavioral Ecology and Sociobiology 17:79–85. Pasquet, A., and B. Krafft. 1992. Cooperation and prey capture efficiency in a social spider, Anelosimus eximius (Araneae, Theridiidae). Ethology 90:121–133. Pulliam, H. R., and T. Caraco. 1984. Living in groups: is there an optimal group size? Pages 122–147 in J. R. Krebs and M. B. Davies, eds. Behavioral ecology: an evolutionary approach. Blackwell Scientific, Oxford. Rannala, B. H., and C. R. Brown. 1994. Relatedness and conflict over optimal group size. Trends in Ecology & Evolution 9:117–119.

Riechert, S. E. 1985. Why do some spiders cooperate? Agelena consociata, a case study. Florida Entomologist 68:105–116. Riechert, S. E., R. M. Roeloffs, and A. C. Echternacht. 1986. The ecology of the cooperative spider Agelena consociata in equatorial Africa. Journal of Arachnology 14:175–191. Rodman, P. S. 1981. Inclusive fitness and group size with a reconsideration of group sizes in lions and wolves. American Naturalist 118:275–283. Rypstra, A. L. 1990. Prey capture and feeding efficiency of social and solitary spiders: a comparison. Acta Zoologica Fennica 190:339–343. ———. 1993. Prey size, social competition, and the development of reproductive division of labor in social spider groups. American Naturalist 142:868–880. Seibt, U., and W. Wickler. 1988a. Bionomics and social structure of ‘‘family spiders’’ of the genus Stegodyphus, with special references to the African species S. dumicola and S. mimosarum (Araneida, Eresidae). Verhandlungen des naturwissenschaftlichen Vereins in Hamburg 30:255–303. ———. 1988b. Why do ‘‘family spiders,’’ Stegodyphus (Eresidae), live in colonies? Journal of Arachnology 16: 193–198. Sibly, R. M. 1983. Optimal group size is unstable. Animal Behaviour 31:947–948. Smith, D. R. 1982. Reproductive success of solitary and communal Philoponella oweni (Araneae: Uloboridae). Behavioral Ecology and Sociobiology 11:149–154. Stearns, S. C. 1992. The evolution of life histories. Oxford University Press, Oxford. SYSTAT. 1992. SYSTAT: graphics, version 5.2 edition. SYSTAT, Evanston, Ill. Tapia, Y., and T. De Vries. 1980. Tolerancia y cooperacio´n en la aran˜a social Anelosimus jucundus del bosque tropical Rı´o Palenque, Ecuador. Revista de la Universidad Cato´lica 27:51–74. Tietjen, W. J. 1986. Effects of colony size on web structure and behavior of the social spider Mallos gregalis (Araneae, Dictynidae). Journal of Arachnology 14: 145–157. Tufin˜o, P. A. 1997. Efecto de la cooperacio´n en la supervivencia y tasa de crecimiento per ca´pita en colonias de la aran˜a social Anelosimus eximius, Cuyabeno, Amazonı´a Ecuatoriana. Licenciatura thesis. Pontificia Universidad Cato´lica del Ecuador, Quito. Tyler, W. A. 1995. The adaptive significance of colonial nesting in a coral-reef fish. Animal Behaviour 49:949– 966. Uetz, G. 1992. Foraging strategies of spiders. Trends in Ecology & Evolution 7:155–159. Uetz, G. W., and C. S. Hieber. 1997. Colonial web-build-

418 The American Naturalist ing spiders: balancing the costs and benefits of groupliving. Pages 458–475 in J. Choe and B. Crespi, eds. The evolution of social behavior in insects and arachnids. Cambridge University Press, Cambridge. Vehrencamp, S. L. 1983. A model for the evolution of despotic versus egalitarian societies. Animal Behaviour 31:667–682. Vehrencamp, S. L., R. R. Koford, and B. S. Bowen. 1988. The effect of breeding-unit size on fitness components in groove-billed Anis. Pages 291–304 in T. H. CluttonBrock, ed. Reproductive success. University of Chicago Press, Chicago. Venticinque, E. M., H. G. Fowler, and C. A. Silva. 1993. Modes and frequencies of colonization and its relation to extinctions, habitat and seasonality in the social spi-

der Anelosimus eximius in the Amazon (Araneidae: Theridiidae). Psyche (Cambridge) 100:35–41. Vollrath, F. 1982. Colony foundation in a social spider. Zeitschrift fu¨r Tierpsychologie 60:313–324. ———. 1986. Eusociality and extraordinary sex ratios in the spider Anelosimus eximius (Araneae: Theridiidae). Behavioral Ecology and Sociobiology 18:283–287. Ward, P. I. 1986. Prey availability increases less quickly than nest size in the social spider Stegodyphus mimosarum. Behaviour 97:213–225. Wilson, E. O. 1975. Sociobiology: the new synthesis. Belknap, Cambridge, Mass. Associate Editor: Donald H. Feener, Jr. Editor: Joel G. Kingsolver