INBREEDING AND ITS FITNESS EFFECTS IN AN INSULAR POPULATION OF SONG SPARROWS (MELOSPlZA MELODIA)

Transcription

1 Evolution. 52(1) pp INBREEDING AND ITS FITNESS EFFECTS IN AN INSULAR POPULATION OF SONG SPARROWS (MELOSPlZA MELODIA) LUKAS F. KELLER 1 Department of Wildlife Ecology, 1630 Linden Drive, University of Wisconsin-Madison, Madison, Wisconsin Abstract.-Inbreeding depression is thought to be a major factor affecting the evolution of mating systems and dispersal. While there is ample evidence for inbreeding depression in captivity, it has rarely been documented in natural populations. In this study, I examine data from a long-term demographic study of an insular population of song sparrows (Melospiza melodia) and present evidence for inbreeding depression. Forty-four percent of all matings on Mandarte Island, British Columbia, were among known relatives. Offspring of a full-sib mating (f= 0.25) experienced a reduction in annual survival rate of 17.5% on average. Over their lifetime, females with f = 0.25 produced 48% fewer young that reached independence from parental care. In contrast, male lifetime reproductive success was not affected by inbreeding. Reduced female lifetime reproductive success was mostly due to reduced hatching rates of the eggs of inbred females. Relatedness among the parents did not affect their reproductive success. Using data on survival from egg stage to breeding age, I estimated the average song sparrow egg on Mandarte Island to carry a minimum of 5.38 lethal equivalents (the number of deleterious genes whose cumulative effect is equivalent to one lethal); 2.88 of these lethal equivalents were expressed from egg stage to independence of parental care. This estimate is higher than most estimates reported for laboratory populations and lower than those reported for zoo populations. Hence, the costs of inbreeding in this population were substantial and slightly above those expected from laboratory studies. Variability in estimates of lethal equivalents among years showed that costs of inbreeding were not constant across years. Key words.-inbreeding, inbreeding depression, lethal equivalents, lifetime reproductive success, pedigree, song sparrow, survival. Animal and plant breeders have long recognized the deleterious effects of inbreeding (inbreeding depression). In his book on cross-fertilization in plants, Darwin (1876) pointed out that many adaptations of plants could be seen in terms of the selective advantage of avoiding inbreeding depression. Ever since, the evolutionary consequences of inbreeding in natural populations have been of interest because of the presumed effects of inbreeding on the evolution of mating systems (e.g., Charlesworth and Charlesworth 1987), dispersal (e.g., Greenwood and Harvey 1982), and the amount of genetic variation maintained in populations (e.g., Wright 1969). More recently, inbreeding has become a major concern of conservation biologists because it is an unavoidable consequence of isolation in small populations (e.g., O'Brien et al. 1985; Soule 1986; Lande 1988). Data from plants, domestic and laboratory animals, and humans all show that inbreeding depression occurs commonly (e.g., Cavalli-Sforza and Bodmer 1971; Wright 1977; Charlesworth and Charlesworth 1987; Thornhi111993). However, inbreeding is rarely studied in wild populations of animals under natural conditions. This is because direct estimates of inbreeding require long-term studies of individually marked organisms. Although the occurrence of inbreeding has been reported for several wild animal populations (e.g., Ralls et al. 1986), only a few studies have quantified inbreeding depression in natural vertebrate populations. Table 1 summarizes those studies known to me and the traits for which inbreeding depression was reported. Ten of these studies reported significant inbreeding depression in some traits, while the four others did not find any effects of inbreeding 1 Present address: Department of Ecology and Evolutionary Biology, Princeton University, Princeton, New Jersey ; E mail: Ifkeller@princeton.edu The Society for the Study of Evolution. All rights reserved. Received February 10, Accepted September 29, on fitness. However, some of the studies were based on very small sample sizes, and others found inbreeding depression in some fitness traits but not in others (e.g., Greenwood et al. 1978; van Noordwijk and Scharloo 1981; Bensch et al. 1994; Kempenaers et al. 1996). Although hypotheses have been presented to explain the different outcomes (e.g., Gibbs and Grant 1989; Hedrick 1994; Keller et al. 1998), inconsistencies among these results have led to questions about whether inbreeding depression occurs at all in natural vertebrate populations (Shields 1993), or whether it is important in populations larger than 25 individuals (Caughley and Sinclair 1994). Clearly, more empirical data from natural populations are needed to establish how frequently inbreeding occurs when partner choice is natural and what the effects of inbreeding on fitness are. My aim in this paper is to examine the occurrence and fitness consequences of inbreeding in a well-studied population of song sparrows on Mandarte Island, British Columbia. The detailed pedigree going back to 1975 allowed me to address the following questions: (1) What are the levels of inbreeding in this population and how do they compare to estimates derived in studies of other bird species? (2) What are the effects of inbreeding on fitness? I evaluated the effects of inbreeding on two fitness traits: a bird's survival after it has reached independence of parental care, and a bird's lifetime reproductive success (LRS), estimated as the number of young that reach independence from parental care produced during a bird's lifetime. Independence from parental care was chosen as the cut-off point in both cases because the direct effects of the parents on the survival of their offspring end at that time. A detailed analysis of the effects of inbreeding on reproductive success needs to take into account both parental in-

2 FITNESS EFFECTS OF INBREEDING IN SONG SPARROWS 241 TABLE 1. Summary of previously published studies on inbreeding depression in natural vertebrate populations. I only included studies that addressed inbreeding depression in wild populations with natural mating patterns, that is, where inbreeding had not been achieved experimentally. The method used to measure levels of inbreeding is given together with the number of inbred matings or individuals used in the published analyses, and a list of the traits for which inbreeding depression was reported. "None" means that no inbreeding depression was found in any of the traits examined. No sample sizes for inbred matings are given for the studies that used DNA similarity to infer inbreeding because their approach does not allow a distinction between inbred and noninbred matings. Number of inbred Trait for which inbreeding Species Reference Method matings observed depression was found Great Tit Bulmer 1973 pedigree 7 recruitment Great Tit Greenwood et a! pedigree 13 nestling mortality (inc!. hatching) Great Tit van Noordwijk and Scharloo pedigree 245 hatch rate 1981 Blue Tit Kempenaers et a! DNA similarity hatch rate Medium Ground-Finch Gibbs and Grant 1989 pedigree 27 none Large Ground-Finch Grant and Grant 1995 pedigree 11 none Great Reed Warbler Bensch et al DNA similarity hatch rate and number of fledged young Song Sparrow Keller et al pedigree 137 survival through storm Olive Baboon Packer 1979 pedigree 8 (one male only) infant survival Chacma Baboon Bulger and Hamilton 1988 pedigree 20 (three males none only) Yellow Baboon Alberts and Altmann 1995 pedigree 3 survival from conception to 30 days Black-tailed Prairie Dog Hoogland 1992 pedigree 198 none Golden Lion Tamarin Dietz and Baker 1993 pedigree 14 survival to weaning Common Shrew Stockley et a! DNA similarity survival to maturity breeding coefficients and the kinship coefficient among the pair, because all three have been found to affect fitness in laboratory populations (e.g., Sittmann et al. 1966). Since a large fraction of song sparrow pairs on Mandarte Island (43%) are not paired for life, an analysis of the effects of all three parameters on LRS is not possible. Hence, I also analyzed inbreeding effects on several measures of seasonal reproductive success. METHODS Study Population and Field Methods In 1974, J. N. M. Smith started an ongoing detailed study of the population of song sparrows (Melospiza melodia) resident on Mandarte Island (e.g., Smith 1988, Arcese et al. 1992). Mandarte is a small (six hectare), rocky island in Haro Strait approximately 25 km NE ofvictoria, British Columbia. It is covered by grass and shrubs and is the largest seabird colony in southern British Columbia. The song sparrows primarily inhabit the dense shrubbery dominated by Rosa nutkana and Symphoricarpos albus. Because these shrub patches are narrow and divided by trails, observations of individual birds can be made relatively easily. For more details on Mandarte Island, see Tompa (1964), Drent et al. (1964), and Smith (1988). Starting in 1974, all song sparrows on Mandarte Island were individually color-banded with one metal and up to three colored plastic bands. Smith (1981) and Arcese et al. (1992) give details of the study methods. In brief, the general field methods on Mandarte Island include finding almost all nests during incubation and banding nestlings, usually six days after hatching. Shortly before the young gain independence from their parents at days of age, we determined which young had survived by watching the parents feed the nearly independent young. In the following "independence" refers to the time when young are no longer under parental care and "independent young" refers to birds that have survived to this point. Survival to breeding age was estimated by censusing the breeding population regularly. Censuses of territorial adults in the spring were exact because territorial song sparrows are extremely conspicuous on Mandarte Island at this time of the year. However, some juveniles that survived the winterremainedin the population as nonterritorialfloaters (Smith and Arcese 1989). These individuals are secretive and occasionally obtain breeding status by replacing territorial birds. I did not include floaters in the analyses here, unless they bred at some time. Pedigree Construction Detailed behavioral observations were used to determine parentage for each brood. Song and territorial aggression in male song sparrows occur throughout the year and make it easy to identify the owner of a particular territory (Arcese 1987, 1989a,b). Each territory was observed every two to seven days. During each visit we recorded the identity of the territorial male and female. We also noted proximity and interactions between any male and the female. After the young hatched we noted who was feeding the nestlings and who was uttering alarm calls when we approached the nest to band the young. A combination of all these observations allowed an identification of the putative parents in 99.5% of all 2109 breeding attempts. Uncertainties in assigning paternity arose when floaters replaced a territorial male during the presumed egg-laying period. In these cases (0.9% of all breeding attempts), we assigned paternity based on feeding behavior. The new male was assumed to be the father of a brood if he fed the young.

3 242 LUKAS F. KELLER In several cases the new male was never seen feeding during several observationperiods and the former male was assumed to have sired the brood. In a few cases the former male was observed sneaking into his old territory to feed the young (Keller, Hatch, and Arcese, unpubl. obs.). If there was no observation indicating either one of the males to be the likely father (four cases in the whole dataset), I entered the male as unknown and assumed for the pedigree that the father was completely unrelated to any bird in the population. This will likely have led to an underestimation of the true inbreeding coefficient of individuals in that pedigree. While a consistent underestimation of the inbreeding coefficients could lead to an overestimate of the severity of inbreeding depression, it is not a problem in this case because of the rarity of this event (four of 2109 breeding attempts). The information on parentage allowed the construction of a pedigree spanning 16 generations between 1974 and Inbreeding Estimates I used Wright's (1969) coefficient of inbreeding,j, to quantify the degree of inbreeding based on the pedigree. Inbreeding and kinship coefficients were calculated using PEDSYS (Southwest Foundation for Biomedical Research, San Antonio, TX) and the Stevens-Boyce algorithm option (Boyce 1983). The kinship coefficient between a pair is equal to the inbreeding coefficient of the offspring resulting from that mating. To facilitate comparisons with previously published studies, mean inbreeding in the population was calculated as the mean kinship coefficient of all pairs. Inbreeding and kinship coefficients are given relative to the founder population in 1975, assuming that all adults in 1975 and all subsequent immigrants were unrelated. An individual was inbred if its parents were known to be related. If the parents had no detected common ancestor, I assumed that an individual had an inbreeding coefficient of zero. Hence, the estimates of inbreeding are minima, since the likelihood of detecting a common ancestor also depends on the depth of the pedigree. The depth of the pedigree varied considerably through time: the smallest incidence of inbreeding we could have detected assuming a common ancestor in 1975 ([1I2]2g-1, with g being the number of generations in the pedigree) dropped from f = in 1978 to f = 1.8 X in Hence, early in the study, pairings may have been assumed to be noninbred simply because their ancestry was unknown. To avoid such incorrect assignments, and to accommodate for the varying depth of the pedigree, I only used data from individuals whose grandparents were known for the following analyses. Because immigrants were assumed to be unrelated to the breeding population, their offspring's inbreeding coefficients were set to zero. Figure 1 shows the effect of excluding pairs in which not all eight grandparents were known on estimates of the mean level of inbreeding in the Mandarte population. Not correcting for the varying depth of the pedigree would have resulted in an underestimate of the mean level of inbreeding, particularly during the early parts of the study. For there were no breeding pairs for which all grandparents were known. Incomplete pedigree information for 1980 led to a gap in the pedigree and resulted in very low sample sizes of birds with E C.. c; 0.08 ;:-.. g 0.08 at C.. ~ 0.04.a c 0.02 c : I \ I E year FIG. 1. Mean coefficients of inbreeding on Mandarte Island during the course of the study. Mean inbreeding was calculated as the mean of the kinship coefficients of all breeding pairs in a particular year. The solid line represents the estimates based on the entire dataset. The estimates represented by the broken line are based on the subset of individuals for which at least all eight grandparents were known. known grandparents in 1981 and 1982 and in a mean f of zero in Because the pedigree was based on observations, extrapair fertilizations will lead to inaccurate assignments of parentage. Preliminary results from DNA analyses of91 broods and 193 young suggest that approximately 15% of all young on Mandarte Island are sired by extrapair males (Keller 1996). However, there was no relation between the likelihood of a female engaging in extrapair fertilizations and the degree of relatedness between her and her mate, or their respective inbreeding coefficients. Therefore, while the extrapair fertilizations introduced error variance in our estimates of inbreeding, they were unlikely to bias them. Unbiased error in the inbreeding coefficients will tend to make the estimates of inbreeding depression more conservative. In fact, the one study that suffered from extremely high levels of extrapair fertilizations reported no inbreeding depression in that population (Rowley et al. 1986). It is possible, therefore, that inbreeding depression was more pronounced than what I calculated based on the pedigree. However, this effectcan only be quantifiedonce the extrapair males have been identified. Fitness Traits Information on which individuals survived to independence was only available from 1981 onward. Hence, I restricted the analyses of the fitness traits to cohorts born in 1981 or later. The number of independent young produced is correlated well with the number of young produced that reach breeding age (r = 0.87; Hochachka et al. 1989), but the distribution of number of independent young includes fewer zeros and makes statistical procedures more straightforward. The Mandarte song sparrow population shows marked density dependence in reproductive success and survival (Arcese et al. 1992). Changes in population size also lead to year-to-

4 FITNESS EFFECTS OF INBREEDING IN SONG SPARROWS 243 year variation in the average level of inbreeding (Keller et al. 1994, 1998). Thus it was necessary to include cohort or year effects in all analyses presented here. All statistical analyses were performed with SAS, version 6.07 (SAS Institute 1990), and all tests reported are two tailed. Survival Analysis My estimates of survival are estimates of local survival since an individual that disappeared from the island could have died or dispersed. No adult bird is known to have dispersed to or from Mandarte Island, despite intensive censuses on eight islands surrounding Mandarte during several years (~rcese 1989a; Smith et al. 1996). Similarly, only seven yearhngs settled on the surrounding islands between 1988 and 1995 (Smith, Keller, and Arcese, unpubl. data) of 815 young tha~ reached independence on Mandarte Island during this penod. Hence, while a few juveniles are known to have dispersed, dispersal is unlikely to have occurred often enough to bias my results. Survival following independence was analyzed using a discrete-time proportional-hazards model developed by Heisey (1992). Proportional-hazards models analyze the influence of covariates on time until death, assuming that some function of the covariates acts in a multiplicative fashion on a baseline hazard function. The approach proposed by Heisey (1992) assumes that the probability of an individual k surviving year i is given by: where (Xi is the baseline hazard for year i, Zk is a vector of covariates, and 13 is a vector of regression coefficients. This approach is essentially nonparametric, that is, it makes no assumption about the dependence of the hazard on time (Agresti 1990, p. 194), and accommodates left-truncateddata. The latter was important for my analyses because the birds surviving to independence (the starting point of my survival analysis) represent a left-truncated sample, that is, some inbreeding effects may have occurred before the starting point. I. stratified the data by cohort and age. Hence, the proportional-hazards model quantified the decrease in annual survival due to inbreeding among same-aged birds alive in the same year. Strata in which either all birds died or survived were excluded since they did not contribute any information. Birds still alive in 1996 were right-censored, that is, the likelihood function was adjusted for the fact that these birds contribute to our knowledge of the survivor function, but nothing to our knowledge of the age at death. The model was implemented using Proc Probit (SAS Institute 1990). Reproductive Success The number of independent young produced over a lifetime (LRS) was calculated for each breeding bird. I used an analysis of covariance to evaluate the effects of inbreeding, with the inbreeding coefficient of the breeding bird, life span (in years), and life span squared as the covariates and with cohort as a categorical variable. Including life span as covariates in the models means that the measure of LRS was corrected for longevity, which otherwise is a component of LRS (Smith 1988). This was done because survival was analyzed sepa- (1) rately. LRS data were transformed using the square root to achieve a normal error structure and homoscedasticity. Adults from cohorts born between 1981 and 1992 were used for these analyses. Ten females and 17 males that hatched in or before 1992 were still alive at the time of analysis. However, all were at least four years old, with a low expected future reproductive success (No1and Smith 1987). Thus their future reproductive success is unlikely to alter the results presented here. A feeding experiment in 1985 greatly changed patterns of reproductive success on the island (Arcese and Smith 1988). For all the analyses ofreproductive success, I therefore excluded the year I analyzed inbreeding effects on the following measures of seasonal reproductive success (SRS): number of nesting attempts, number of eggs laid, proportion of eggs hatched, number of independent young produced, and proportion of hatched young that reached independence. All these variables w~re yearly averages for pairs. An ANCOVA was employed With maternal and paternal inbreeding coefficients and the pair's kinship coefficient as covariates and with year as the categorical variable. Number of eggs laid and number of independent young produced were In-transformed. Percentage data were transformed using the standard transformation for percentage data with unequal sample sizes: 2CVN)arcsin CVP) (Draper and Smith 1981, p. 239), where P is the percentage and N is the sample size upon which P is based, for example, the number of eggs laid for the hatch rate. Since 43% of all pairs are represented more than once in this dataset of SRS the least-squares standard errors of the estimates are deflated and cannot be used for hypothesis testing. Instead, I used a bootstrap approach with the pair as the bootstrap unit. The distribution of the coefficients under the null hypothesis was obtained by bootstrapping the residuals produced by the original ANCOVA and then applying the same ANCOVA model on these bootstrap samples (e.g., Tibshirani 1.992). The bootstrap routine chose pairs (not single observations) at random. If a particular pair was chosen, the observations from all years that pair was breeding were included in the bootstrap sample. Significance on a two-tailed basis was assessed by comparing the actual coefficients with the frequency distribution of the coefficients obtained from the 1000 bootstrap runs; 95% confidence intervals for the parameter estimates were calculated using bias-corrected percentile bootstrapping (Dixon 1993). Estimates of Lethal Equivalents Morton et al. (1956) developed a method for estimating the costs of inbreeding from mortality data. Their method allows for an easy comparison of the costs of inbreeding across studies and is based on the concept of lethal equivalents, which is the number of deleterious genes whose cumulative effect is the equivalent of one lethal. Estimates of the number of lethal equivalents were obtained using the regression equation: -InS = A + Bf, where S is the probability of survival, f is the inbreeding coefficient, A is the load in a random-mating population both due to genetic and environmental causes, and B is a measure (2)

5 244 LUKAS F. KELLER TABLE 2. Levels of inbreeding on Mandarte Island and in three other natural bird populations. For each study the number of pairs, the percentage that were inbred and the mean inbreeding coefficient are given for the entire dataset and for the subset of pairs whose grandparents were all known. In addition, the number of pairs with f 2: in each study are given. A pair was considered inbred if > 0) if a common ancestor was known from the pedigrees. Mean inbreeding wascalculated as the meankinshipcoefficient of all breeding pairs. The three other studies are: vn&s = van Noordwijk and Scharloo (1981), G&G = Gibbs and Grant (1989), K. et al. = Kempenaers et al. (1996). See Table 1 for more details on those studies. This study vn&s G&G K. et al. Total number of pairs % inbred meanf Ql Number of pairs with known grandparents % inbred meanf Total number of pairs with f2: of the increased deaths caused by increased inbreeding. Hence, the lower limit for the number of lethal equivalents per gamete is given by B, whereas the upper limit is (A + B). Cavalli-Sforza and Bodmer (1971) give a detailed description of the concept of lethal equivalents. I estimated values for A and B using the above regression equation for two sets of survival data: from egg to independence and from independence to breeding age. For the former analysis, each nest contributed one data point (because the fate of nestlings in the same nest is highly correlated), whereas each independent young contributed a data point to the latter analysis. I classified the inbreeding coefficients into seven categories, calculated percent survival in each category, and ran a weighted regression for each cohort independently, with the sample size in each inbreeding category as the weighting factor. To be able to assess the total effects of inbreeding on survival, I also estimated lethal equivalents for overall survival from egg to breeding age. I used the same approach as above, with each nest contributing one data point. RESULTS AND DISCUSSION Observed Levels of Inbreeding The average inbreeding coefficient (j) in the years was f = for all pairs and f = for the pairs with known grandparents (Table 2). However, meanfvaried greatly over time. After the initial increase in mean inbreeding due to the increased pedigree depth, mean f ranged between 0.02 and 0.04 from 1983 to 1990 (Fig. 1). In February 1989, a winter storm killed all but 12 of the song sparrows alive on Mandarte Island (Rogers et al. 1991; Keller et al. 1994). This population bottleneck led to a strong increase in mean f to levels ranging from 0.06 to A total of 671 unique pairings were observed during this study, but all eight grandparents were known for only 285 of them (Table 2). Of those matings, 205 (72%) were between detectably related birds. Fifty-one parings (7.6% of all matings TABLE 3. Results of a discrete proportional hazards model estimating the effects of inbreeding on annual survival. A positive coefficient indicates a decrease in annual survival probability with increasing inbreeding coefficientf. The analysis was stratified by year and age with data from 1981 to Hence, a separate intercept was estimated for each age-year class, but is not given here for clarity. The log-likelihood is given for the complimentary loglog link function. See Methods for details of the model employed. df Estimate SE X 2 P Year and age f Log likelihood = N = 1324 individuals or 17.2% of all inbred matings) were between half-sibs and full-sibs or between parents and offspring or grandparents and grandchildren if? 0.125). Hence, matings between more distant kin (first-cousins or more distant,f< 0.125) made up the vast majority of all inbred pairings (82.8%). These matings alone lead to a mean inbreeding coefficient at the population level of f = compared to if all matings are included. Hence, 61% of the overall inbreeding in the population is caused by matings among distant relatives. The detected levels of close inbreeding are, therefore, not necessarily a good approximation of the population level of inbreeding as suggested, for example, by Greenwood et al. (1978). Comparing the levels of inbreeding observed in different populations requires that one accounts for differences in pedigree depth (see Fig. 1). One way to achieve this is to compare subsamples of the datasets that include only pairs with some chosen minimum number of known ancestry, that is, only pairs for which at least all grandparents were known. Based on this subset, the mean inbreeding coefficient and the proportion of matings among known relatives is higher on Mandarte Island than reported for any other bird population (Table 2). This difference is likely attributable to the different effective population sizes and different levels of immigration. On Mandarte Island, less than 3% of all breeding birds were immigrants (almost all females) while on Vlieland Island 20% of the females and 8% of the males were immigrants (van Tienderen and van Noordwijk 1988). An even higher fraction of the breeding birds immigrated into the blue tit study area (Kempenaers et al. 1996). While the great tits, the blue tits, and the song sparrows had similar average population sizes (40 to 60 pairs), the medium ground-finches had a much higher average population size (approximately 200 individuals; Grant and Grant 1992). Hence, we expect to find fewer matings among relatives in the finches. In addition, the generation time of the medium ground-finches is approximately three times longer than that of song sparrows (Grant and Grant 1992; Keller, unpubl. data). Hence, pedigrees of medium ground-finches were shallower than those for song sparrows assembled over the same number of years and are thus less likely to detect distant common ancestors. Survival Inbreeding significantly decreased annual survival after independence after accounting for the effects of age and year (Table 3). An offspring of a full-sib mating if = 0.25) was on average 17.5% (95% CI: 3-33%) less likely to survive a

6 FITNESS EFFECTS OF INBREEDING IN SONG SPARROWS 245 TABLE 4. ANCOVA of the effects of cohort, life span, life span-, and inbreeding on lifetime reproductive success in females and males. LRS data were square-root transformed. N refers to the number of males and females, respectively, used in this analysis. Females (N = 161) Males (N = 157) df Type 11\ SS Estimate SE F P df Type 11\ SS Estimate SE F P Cohort I 7.41 < Lifespan < < < Lifespan f year than a noninbred bird. This confirms earlier findings from this population (Keller et al. 1994, 1998) that inbreeding reduced the likelihood of survival through a sever winter storm in First winter (juvenile) survival is typically much lower than survival following the first breeding season (adult survival), leading to a higher opportunity for selection (Crow 1989). Therefore, one might expect more pronounced inbreeding depression in juvenile survival. I reran the proportional-hazards model with separate slopes for the effect of inbreeding on survival of juveniles and adults. The estimates for the inbreeding effects with this model were 1.05 (SE = 0.62) for juveniles and 1.68 (SE = 0.93) for adults. These slopes are not statistically different from each other, suggesting that inbreeding depression in survival was similar across life stages. If anything, inbreeding depression was marginally more pronounced in adults than juveniles. Few studies of captive populations report inbreeding depression in both juvenile and adult survival. However, in Japanese quail (Sittmann et al. 1966) inbreeding depression in birds aged zero to five weeks was approximately twice that of birds aged five to 16 weeks. Of several other studies on inbreeding depression and juvenile survival, most found no effect (Table 1). In birds, only Bulmer (1973) reported reduced survival postfledging, but this result was not upheld in a subsequent and more extensive analysis of the same population (Greenwood et al. 1978). On Viieland Island, inbred juvenile great tits even showed a slight tendency for enhanced recruitment into the breeding population (van Noordwijk and Scharloo 1981). Hence, despite the ubiquity of inbreeding depression in juvenile survival in domestic birds (e.g., Sittmann et al. 1966) and in a variety of species in laboratories (e.g., Wright 1977) and zoos (e.g., Ralls et al. 1979; 1988, Lacy et al. 1993), only Stockley et al. (1993) and this study have demonstrated the same pattern in natural populations. With the exception of studies by van Noordwijk and Schar 100 (1981) and Hoogland (1992), small sample size cannot be ruled out as the reason for the absence of inbreeding depression in juvenile survival. Given the typically high environmental variability in natural mortality patterns, fairly large sample sizes are required to obtained sufficient statistical power. Even in the song sparrow dataset, significant inbreeding depression in survival was only detected after age effects and year-to-year variation in survival were accounted for. Interestingly, few studies have analyzed inbreeding depression in adult survival in a natural environment. Jimenez et al. (1994) reported effects of inbreeding on adult survival in white-footed mice that had been released into natural habitat. They reported a 44% reduction in survival of inbred mice if = 0.25). In comparison, survival of song sparrows with f = 0.25 was reduced by 24% (using the estimate for adult birds only, b = 1.68). Hence, inbreeding can substantially reduce adult survival. Even the data available for humans focus on the effects of inbreeding in children (e.g., Cavalli-Sforza and Bodmer 1971; Bittles and Neel 1994), although Bashi (1977) showed that inbred children had lower cognitive abilities than children born to unrelated parents. Effects on cognitive performance would be expected to prevail in adults. Yet I am not aware of any data on the effects in adults. The paucity of data on inbreeding depression in adult survival could be taken as evidence that inbreeding depression in later life-history stages is weak. However, Husband and Schemske (1996) found substantial inbreeding depression late in the life cycle in a literature survey of inbreeding depression in plants. Other than the higher opportunity for selection in juveniles, there is also no a priori genetic or ecological reason to expect inbreeding depression to be generally more pronounced in early life-history stages. The evolutionary theory of senescence, for example, predicts that deleterious mutations would be expressed more in late life-history stages because selection against late acting mutations would be weakest (e.g., Stearns 1992). Hence, at present there seems to be no theoretical or empirical reason to expect inbreeding depression to be reduced late in the life cycle. Lifetime Reproductive Success Reproductive Success Inbreeding in females significantly reduced the number of independent young produced over a lifetime (Table 4), independent of the effect through the reduced life span expected from the results above. Females born to full-sib if = 0.25) or first-cousin pairs if = ) had on average only 52% (95% CI: 25-89%) and 87% (95% CI: 77-97%), respectively, of the LRS of a noninbred female. Thus, inbreeding can drastically reduce overall fitness in closely inbred individuals: a female with f = 0.25 produced on average only 2.7 independent young over her lifetime instead of the 5.3 of a noninbred female. In contrast, inbreeding depression in LRS was less in males and not statistically significant. The ANCOVA explained the same amount of the variance in LRS in males and in females (57% in both cases) suggesting that the lack of inbreeding depression in male LRS is not due to higher variance in LRS in males than in females. In fact, Hochachka

7 246 LUKAS F. KELLER et al. (1989) found no difference in the distribution of LRS between males and females in any given cohort. Three factors may contribute to the different findings for males and females. First, DNA analyses have not revealed any instances of egg dumping on Mandarte Island but approximately 15% of all young were sired by extrapair males. Thus, measures of reproductive success in males have a considerably larger biological error than in females, potentially masking significant inbreeding depression. Second, males do not necessarily breed in all years that they are alive, whereas only one single female is known not to have bred in a year. Male nonbreeders either hold a territory but do not acquire a mate, or are nonterritorial floaters (Arcese 1989b; Smith and Arcese 1989). Inbreeding does not seem to influence whether or not a male acquires a mate (Keller and Arcese, unpubl. data) and this might help to explain why LRS in males is not affected by inbreeding. Third, as shown below, male inbreeding only affects the egg stage and environmental factors such as predation that influence reproductive success afterward thus tend to reduce the impact of male inbreeding on later stages. No other study to date has documented inbreeding depression in LRS in a natural population. In contrast, van Noordwijk and Scharloo (1981) found that offspring from pairs where either the mother or the father were inbred experienced higher survival rates to breeding age. Seasonal Reproductive Success The results for seasonal reproductive success mirrored the results for LRS in direction and magnitude (Table 5). The number of independent young produced per year was only reduced in inbred females. Inbreeding depression in females in this measure of SRS was similar to that found in LRS: females withf= 0.25 produced 45% (95% CI: 1-61%) fewer independent young per season. In addition, Table 5 shows that relatedness among the parents did not affect the number of independent young produced. The detailed analysis of SRS also allowed me to quantify at what stage in the nesting cycle most of the inbreeding depression occured. In both sexes inbreeding depression occurred in the earlier stages of reproduction. For inbred males, most of the reduction in SRS occurred at the egg stage: the total number of eggs laid in a season was significantly reduced in pairs with an inbred male, but not in pairs that were inbred or where the female was inbred (Table 5). This effect was caused by the females of inbred males laying fewer clutches rather than smaller clutches. This can be seen from the significant negative effect of male inbreeding on the number of nesting attempts (Table 5) and the fact that male inbreeding was unrelated to average clutch size (b = -0.36, P = 0.67). The fewer eggs laid do not, however, translate into fewer independent young. Pairs with an inbred male had fewer nesting attempts not because they had longer internest intervals, but because they started the first nesting attempt later and stopped breeding earlier (effect of male inbreeding coefficient on duration of breeding season: b = , P = 0.016). One possible explanation for this might be that inbred males are less adept at defending their females from intruders, but at present I do not have the data to investigate this. TABLE 5. ANCOVA results of the effects of inbreeding on various measures of seasonal reproductive success: number of nesting attempts, number of eggs laid, proportion of eggs hatched, number of independent young produced, and proportion of hatched young that reached independence. Each data point (N) is the mean of all attempts in a year of a particular pair. N varies slightly for the different measures of seasonal reproductive success because not all nests were found at the same stage. Year was included as a categorical variable in all analyses but only the regression coefficients for the three covariates of interest are reported here: kincoef (coefficient of kinship between the mates), maleic (inbreeding coefficientof the male), and femic (inbreedingcoefficientof the female). "Bootstrap P" refers to the significance level obtained from a bootstrap with the pair as the bootstrap unit. Percent hatching was analyzed separately for the whole dataset and the subset of attempts that were not depredated, to separate the effects of inbreeding on hatchability from the effects on predation rates. Bootstrap Trait Covariate Coefficient P N Number of attempts kincoef maleic Ql8 femic Number of eggs kincoef maleic fernie Percent hatched kincoef maleic fernie Number of independent kincoef young maleic femic Percent independent kincoef that hatched maleic femic Attempts that were not depredated only Percent hatched kincoef maleic femic Inbreeding seems not to affect later stages of male reproductive success. In particular, the hatching rate was not influenced by the male's degree of inbreeding. This finding corresponds with all other studies in wild bird populations that have addressed this issue (van Noordwijk and Scharloo 1981; Kempenaers et al. 1996), but contrasts with data from Japanese quail in the laboratory (Sittmann et al. 1966) in which complete infertility in males was observed. The occurrence of extrapair fertilizations in natural populations may partly explain the general absence of reduced hatch rates among inbred males in the wild and the difference between the findings for laboratory and natural populations. It is possible that wild females that experience low hatching rates because of male infertility compensate by seeking extrapair fertilizations, thus masking the effect of male inbreeding on hatching rates. In females, inbreeding depression was most pronounced in hatchability (Table 5). The eggs of females with f = 0.25 had a 36% (95% CI: 13-60%) lower probability of hatching than eggs of outbred females. Up to 36% of nest failure prior to hatching is due to predation (Hatch and Arcese, unpubl. data). Hence, reduced hatching rates could result either from increased depredation or a reduced probability ofeggs hatch-

8 FITNESS EFFECTS OF INBREEDING IN SONG SPARROWS 247 TABLE 6. Estimates of lethal equivalents for two successive life-history stages, survival from egg to independence and survival from independence to breeding age, and for overall survival from egg to breeding age. A is the load in a randomly mating population and B is the estimate of the number of lethal equivalents per gamete. N refers to the number of nests (egg to independence and overall survival) and individuals (independence to breeding age), respectively, used in the regressions. Estimates of A and B were derived for each cohort separately. The median and the weighted mean (weighted by sample size N) of all cohorts are given at the bottom. See text for details. Survival to independence Survival after independence Overall survival Cohort A B N A B N A B N Median Weighted average ing even when not depredated. To test which mechanism may account for the reduced hatching rate in inbred females, I examined the hatching success of only those nests not depredated. The results (Table 5) were similar to those obtained with the entire dataset, suggesting that eggs laid by inbred females tend to hatch less successfully. Even though female inbreeding did not influence reproduction in the later stages, this strong effect on hatchability led to a reduction in the number of independent young reared. van Noordwijk and Scharloo (1981) also reported an effect of maternal inbreeding on hatchability in great tits, but the same seems not to apply to blue tits (Kempenaers et al. 1996). Sittmann et al. (1966) concluded that effects of maternal inbreeding on egg quality were responsible for the reduced hatchability of eggs laid by inbred Japanese quail hens. Overall, while inbreeding in the male and the female expressed itself early in the nesting cycle, inbreeding in the offspring (the parent's kinship coefficient) affected survival after independence from parental care (Tables 3, 5). Relatedness between the parents, the zygote's inbreeding coefficient, did not reduce the hatch rate. This is surprising because reduced hatchability of eggs with inbred embryos is the most common and often most pronounced effect of inbreeding in birds (Table 1). In quail, apart from male infertility, reduced hatchability contributed the most to the total genetic load (Sittmann et al. 1966). And while song sparrows, great tits, and Japanese quail show maternal inbreeding depression in hatchability, the effect of kinship was stronger than the maternal effects in Japanese quail and great tits. The opposite holds for song sparrows. In fact, related parents had a slight tendency to lay eggs that hatched at a higher rate (Table 5). There seems to be no general rule for the relative impact of parental versus zygotic inbreeding. Hollingsworth and Maynard Smith (1955) found in Drosophila subobscura that the majority of failures to hatch were caused by infertility of the inbred males, with inbreeding ofthe female and the zygote contributing to a lesser degree to hatching failure. In mammals, experiments by Wright (1922) on guinea pigs showed that the number of young born alive (the physiological equivalent of hatch rate in birds) was more strongly affected by maternal inbreeding then by the inbreeding of the zygote. Brewer et al. (1990) found the opposite for Peromyscus species. These differences in the relative importance of maternal or paternal versus zygotic inbreeding within taxa are puzzling and warrant more detailed studies of the underlying mechanisms. Lethal Equivalents Estimates of A and B for survival up to independence, from there to first breeding, and for overall survival are given in Table 6, together with the median and the weighted mean (weighed by sample size) of all cohorts. The weighted mean and the median are fairly similar. Because the median is lower than the weighted averge, I will use the median in the following. Since the estimates of lethal equivalents are per gamete, the average song sparrow egg on Mandarte Island carried an estimated minimum of 2B = 5.38 lethal equivalents; 2.88 of these lethal equivalents were expressed from egg to independence and 2.64 from independence to breeding age. The upper limit of these estimates of lethal equivalents is given by 2(A + B) = 9.48, 4.92, and 4.06, respectively. However, environmental causes of mortality are certainly very important on Mandarte and they are included in A. For this reason I assume that the true number of lethal equivalents is much closer to 5.38, 2.88, and 2.64, than to 9.48, 4.92, and It can be seen that the estimate of A reflects mortality due to environmental factors. In the winter of a storm killed 89% of the birds on Mandarte Island (Rogers et al. 1991; Keller et al. 1994). This unusually high mortality was reflected in the very high estimate of A for survival after independence in the 1988 cohort. In the following years, estimates of A were low because population

9 248 LUKAS F. KELLER density was low and density-dependent effects therefore were at a minimum. Estimates of the number of lethal equivalents varied greatly among cohorts for both measures of survival. This variability suggests that the costs of inbreeding in song sparrows were not constant across years. Although some of this variance may be caused by low sample size in some cohorts, there is evidence that variation in environmental severity might play a role (Keller et al. 1998). Growing evidence suggests that the degree of inbreeding depression is related to environmental conditions (e.g., Miller 1994; Pray et al. 1994; Hauser and Loeschcke 1996; but see Mitchell-Olds and Waller 1985). There was no correlation between the estimated number of lethal equivalents for the two successive stages of survival (Spearman rank correlation, r s = 0.37, P = 0.21), indicating that the loci affecting survival at the two stages acted independently. This independence is also reflected in the fact that the number of lethal equivalents for overall survival (B = 2.69) was very close to the sum of the two separate lethal equivalent estimates ( = 2.76). Evidence for the independence of genes causing inbreeding depression in various life-history stages was also reported for plants (Husband and Schemske 1996). It follows from equation (2) that the reduction in survival due to inbreeding of a certain degree can be estimated as 1 - e - Bf (Ralls et al. 1988). Hence, eggs produced by matings between first-degree relatives if = 0.25) experienced on average a 1 - e-2.69xo.25 = 49% reduction in their survival to breeding age. Thus, the costs of inbreeding on Mandarte Island are substantial. A comparison of the number of lethal equivalents is a useful way to contrast the costs of inbreeding among populations or species. In general, my estimates of B agree with figures reported for other species (Table 7). One difficulty in comparing the results, however, is that the various estimates of lethal equivalents were taken over different parts of the life cycle. The second column in Table 7 therefore specifies over which part of the life cycle survival was measured. The genetic load in song sparrows measured from the egg stage to independence was higher than reported for the two great tit populations. Part of the difference stems from the fact that the estimate for the song sparrows includes more life stages than the two other estimates, in particular that by van Noordwijk and Scharloo (1981). The estimated number of lethal equivalents in the Mandarte song sparrows is similar to those reported for Japanese quail but above what was reported for other laboratory populations and below many of those reported for zoo populations. Hence, although some studies have reported more pronounced inbreeding depression in a natural environment than in the laboratory (e.g., Chen 1993; Jimenez et al. 1994), the comparison of the estimates of genetic load of zoo and wild populations indicates that other factors also influence the degree of inbreeding depression. In particular, lower selection pressures might lead to the maintenance of higher frequencies of deleterious alleles in zoos. ID... cf <"l o Q) <''IN I>Il N'<tl:: oo~ Conclusions My results show that inbreeding depression occurred in a wild population of song sparrows. Estimates of the number

10 of lethal equivalents indicate that inbreeding depression was substantial in this population: inbreeding atf = 0.25 reduced survival from egg to breeding age by 49%. Adult survival was reduced by 24% and lifetime reproductive success in females by 48%. A comparison of the number of lethal equivalents indicates that the levels of inbreeding depression observed in this island population were roughly similar to those reported for other natural populations. The estimates of inbreeding depression for different lifehistory stages can be combined into an index of overall loss of fitness due to inbreeding. Starting with an egg, one can calculate the expected number of eggs produced by this egg two generations later, as a function of the degree of inbreeding. Evaluating total fitness over two instead of only one entire life cycle was important to include the maternal effects of inbreeding on hatching success (Table 5). Using this approach, the overall fitness effects of inbreeding can be estimated as: (effects on survival from egg to breeding age) X (effects on adult survival) X (effects on reproductive success) X (same effects in second life cycle). On Mandarte Island, an egg with an inbreeding coefficient off = 0.25 thus experienced on average a total loss of fitness of 79%. In comparison, measures of lifetime reproductive success and survival varied by a factor four and three, respectively, among the cohorts born on Mandarte Island (Hochachka et al. 1989). Hence, although inbreeding depression was substantial, environmental factors contribute most to variation in fitness. The considerable effect of environmental factors does not mean that inbreeding depression is unimportant for the persistence of small, natural populations. Keller et al. (1994) have shown that inbreeding strongly reduced survival of song sparrows over a severe winter storm that killed 89% of the population. In fact, if inbreeding depression is most pronounced when environmental conditions are poorest, inbreeding depression may have a major impact on persistence times of populations. ACKNOWLEDGMENTS This work would have been impossible without the generous help of many people. P. Arcese, J. N. M. Smith, and S. C. Stearns gave helpful guidance and support at all stages of this project. The Tseycum and Tsawout Indian bands generously let us use their island for our research. I thank the numerous people who collected the data presented here, most recently P. Arcese, A. Cassidy, M. Hatch, I. Heaven, W. Hochachka, G. Jongejan, J. Leary, C. Rogers, J. N. M. Smith, and T. Sullivan. L. Freeman-Shade from Southwest Foundation for Biochemical Research, San Antonio, Texas, provided Pedsys and along with it crucial help in its use. (Pedsys can be obtained from D. Heisey spent hours explaining survival analyses to me and helped me run them. Discussions with and comments by P. Arcese, J. Crow, C. Denniston, D. Heisey, W. Hochachka, J. N. M. Smith, S. Stearns, D. Waller, E. Zouros, and two anonymous reviewers greatly improved this work. Funding provided by the Priority Programme Environment/Biodiversity of the Swiss Nationalfonds ( to S. Stearns), the Roche Research Foundation, and the US National Science Foundation (IBN to P. Arcese) is gratefully acknowledged. FITNESS EFFECTS OF INBREEDING IN SONG SPARROWS 249 LITERATURE CITED AGRESTI, A Categorical data analysis. John Wiley and Sons, New York. ARCESE, P Age, intrusion pressure and defense against floaters by territorial male song sparrows. Anim. Behav. 35: a. Intrasexual competition and the mating system in primarily monogamous birds: the case of the song sparrow. Anim. Behav. 38: b. Territory acquisition and loss in male song sparrows. Anim. Behav. 37: ARCESE, P., AND J. N. M. SMITH Effects of population density and supplemental food on reproduction in song sparrows. J. Anim. Ecol. 57: ARCESE, P., J. N. M. SMITH, W. M. HOCHACHKA, C. M. ROGERS, AND D. LUDWIG Stability, regulation, and the determination of abundance in an insular song sparrow population. Ecology 73: BASHI, J Effects of inbreeding on cognitive performance. Nature 266: BENSCH, S., D. HASSELQUlST, AND T. VON SCHANTZ Genetic similarity between parents predicts hatching failure: nonincestuous inbreeding in the great reed warbler? Evolution 48: BITTLES, A H., AND J. V. NEEL The costs of human inbreeding and their implications for variations at the DNA level. Nature Genet. 8: BOYCE, A J Computation of inbreeding and kinship coefficients on extended pedigrees. J. Hered. 74: BREWER, B. A, R. C. LACY, M. L. FOSTER, AND G. ALAKS Inbreeding depression in insular and central populations of Peromyscus mice. J. Hered. 81: BULGER, J., AND W. J. HAMILTON Inbreeding and reproductive success in a natural chacma baboon, Papio cynocephalus ursinus, population. Anim. Behav. 36: BULMER, M. G Inbreeding in the great tit. Heredity 30: CAUGHLEY, G., AND A R. E. SINCLAIR Wildlife ecology and management. Blackwell Scientific, Oxford. CAVALLI-SFORZA, L. L., AND W. R BODMER The genetics of human populations. Freeman, San Francisco; CA. CHARLESWORTH, D., AND B. CHARLESWORTH Inbreeding depression and its evolutionary consequences. Annu. Rev. Ecol. Syst. 18: CHEN, X Comparison of inbreeding and outbreeding in hermaphroditic Arianta arbustorum (L.) (land snail). Heredity 71: CROW, J. R Fitness variation in natural populations. Pp ill W. G. Hill and T. R C. Mackay, eds. Evolution and animal breeding: reviews on molecular and quantitative approaches in honour of Alan Robertson, CAB International, Wallingford, UK. DARWIN, C The effects of crossing and self-fertilization in the vegetable kingdom. John Murray, London. DIXON, P. M The bootstrap and the jackknife: describing the precision of ecological indices. Pp in S. M. Scheiner and J. Gurevitch, eds. Design and analysis of ecological experiments. Chapman and Hall, New York. \ DRAPER, N., AND H. SMITH Applied regression analysis. 2d ed. John Wiley and Sons, New York. DRENT, R., G. R VAN TETS,R TOMPA, AND K. VERMEER The breeding birds of Mandarte Island, British Columbia. Can. Field Nat. 78: GIBBS, H. L., AND P. R. GRANT Inbreeding in Darwin's medium ground finches (Geospiza fortis). Evolution 43: GRANT, P. R., AND B. R. GRANT Demography and the genetically effective population sizes of two populations of Darwin's finches. Ecology 73: The founding of a new population of Darwin's finches. Evolution 49: GREENWOOD, P.J., AND P.H. HARVEY The natal and breeding dispersal of birds. Annu. Rev. Ecol. Syst. 13:1-21.

11 250 LUKAS F. KELLER GREENWOOD, P. J., P. H. HARVEY, AND C. M. PERRINS Inbreeding and dispersal in the great tit. Nature 271: HAUSER, T. P., AND V. LOESCHCKE Drought stress and inbreeding depression in Lychnis jlos-cuculi (Caryophyllaceae). Evolution 50: HEDRICK, P. W Purging inbreeding depression and the probability of extinction: full-sib mating. Heredity 73: HEISEY, D Proportional hazards analysis of left-truncated survival data. Pp in SAS Institute, Inc. Proceedings of the 17th annual SAS user's group international conference. SAS Institute, Inc., Cary, NC. HOCHACHKA, W. M., J. M. N. SMITH, AND P. ARCESE Song sparrow. Pp in 1. Newton, ed. Lifetime reproduction in birds. Academic Press, London. HOLLINGSWORTH, M. J., AND J. MAYNARD SMITH The effects of inbreeding on the rate of development and on fertility in Drosophila subobscura. J. Genet. 53: HOOGLAND, J. L Levels of inbreeding among prairie dogs. Am. Nat. 139: HUSBAND, B. C., AND D. W. SCHEMSKE Evolution of the magnitude and timing of inbreeding depression in plants. Evolution 50: JIMENEZ, J. A., K. A. HUGHES, G. ALAKS, L. GRAHAM, AND R. C. LACY An experimental study of inbreeding depression in a natural habitat. Science 266: KELLER, L. E Inbreeding in an insular population of song sparrows. Ph.D. diss. Univ. of Wisconsin-Madison. KELLER, L. E, P. ARCESE, J. N. M. SMITH, W. M. HOCHACHKA, AND S. C. STEARNS Selection against inbred song sparrows during a natural population bottleneck. Nature 372: KELLER, L. E, P. ARCESE, W. M. HOCHACHKA, AND J. N. M. SMITH Natural selection and inbreeding in an insular population of song sparrows. Koninklijke Nederlandse Akademie van Wetenschappen, Verhandelingen, Afdeling Natuurkunde, II:94:in press. KEMPENAERS, B., E ADRIAENSEN, A. J. VAN NOORDWIJK, AND A. A. DHONDT Genetic similarity, inbreeding and hatching failure in blue tits: are unhatched eggs infertile? Proc. R. Soc. Lond. B BioI. Sci. 263: LACY, R. C., A. PETRIC, AND M. WARNEKE Inbreeding and outbreeding in captive populations of wild animal species. Pp in N. W. Thornhill, ed. The natural history ofinbreeding and outbreeding. Univ. of Chicago Press, Chicago. LANDE, R Genetics and demography in biological conservation. Science 241: MILLER, P. S Is inbreeding depression more severe in a stressful environment? Zoo BioI. 13: MITCHELL-OLDS, T., AND D. M. WALLER Relative performance of selfed and outcrossed progeny in Impatiens capensis. Evolution 39: MORTON, N. E., J. E CROW, AND H. J. MULLER An estimate of the mutational damage in man from data on consanguineous marriages. Proc. Nat. Acad. Sci. USA 42: NOL, E., AND J. N. M. SMITH Effects of age and breeding experience on seasonal reproductive success in the song sparrow. J. Anim. Ecol. 56: O'BRIEN, S. J., M. E. ROELKE, L. MARKER, A. NEWMAN, C. A. WINKLER, D. MELTZER, L. COLLY, J. E EVERMANN, M. BUSH, AND D. E. WILDT Genetic basis for species vulnerability in the cheetah. Science 227: PACKER, C Inter-troop transfer and inbreeding avoidance in Papio anubis. Anim. Behav. 27:1-36. PRAY, L. A., J. M. SCHWARTZ, C. J. GOODNIGHT, AND L. STEVENS Environmental dependency of inbreeding depression: implications for conservation biology. Conserv. BioI. 8: RALLS, K., K. BRUGGER, AND J. BALLOU Inbreeding and juvenile mortality in small populations of ungulates. Science 206: RALLS, K., P. H. HARVEY, AND A. M. LYLES Inbreeding in natural populations of birds and mammals. Pp in M. E. Soule, ed. Conservation biology, the science of scarcity and diversity. Sinauer, Sunderland, MA. RALLS, K., J. D. BALLOU, AND A. TEMPLETON Estimates of lethal equivalents and the cost of inbreeding in mammals. Conservo BioI. 2: ROGERS, C. M., J. N. M. SMITH, W. M. HOCHACHKA, A. L. E. V. CASSIDY, M. J. TAITT, P. ARCESE, AND D. SCHLUTER Spatial variation in winter survival of song sparrows Melospiza melodia. Ornis Scand. 22: ROWLEY, 1., E. RUSSELL, AND M. BROOKER Inbreeding: benefits may outweigh costs. Anim. Behav. 34: SAS INSTITUTE SAS/STAT user's guide. Verso 6, 4th ed. SAS Institute Inc., Cary, NC. SHIELDS, W. M The natural and unnatural history of inbreeding and outbreeding. Pp in N. W. Thornhill ed. The natural history of inbreeding and outbreeding. Univ. of Chicago Press, Chicago. SITTMANN, K., H. ABPLANALP, AND R. A. FRASER Inbreeding depression in Japanese quail. Genetics 54: SMITH, J. N. M Cowbird parasitism, host fitness, and age of the host female in an island song sparrow population. Condor 83: Determinants of lifetime reproductive success in the song sparrow. Pp in T. H. Clutton-Brock, ed. Reproductive success: studies of individual variation in contrasting breeding systems. Univ. of Chicago Press, Chicago. SMITH, J. N. M., AND P. ARCESE How fit are floaters? Consequences of alternative territorial behaviors in a nonmigratory sparrow. Am. Nat. 133: SMITH, J. N. M., M. J. TAITT, C. M. ROGERS, P.ARCESE, L. E KELLER, A. L. E. V. CASSIDY, AND W. M. HOCHACHKA A metapopulation approach to the population biology of song sparrows, Melospiza melodia. Ibis 138: SOULE, M. E., ED Conservation biology: the science of scarcity and diversity. Sinauer, Sunderland, MA. STEARNS, S. C The evolution of life histories. Oxford Univ. Press, Oxford. STOCKLEY, P., J. B. SEARLE, D. W. MACDONALD, AND C. S. JONES Female multiple mating behaviour in the common shrew as a strategy to reduce inbreeding. Proc. R. Soc. Lond. B BioI. Sci. 254: TEMPLETON, A. R., AND B. READ Factors eliminating inbreeding depression in a captive herd of Speke's gazelle (Gazella spekei). Zoo BioI. 3: THORNHILL, N. W., ED The natural history of inbreeding and outbreeding. Univ. of Chicago Press, Chicago. TIBSHIRANI, R Bootstrap hypothesis testing. Biometrics 48: TOMPA, E S Factors determining numbers of song sparrows Melospiza melodia (Wilson) on Mandarte Island, B.C., Canada. Acta Zool. Fenn. 109:1-73. VAN NOORDWIJK, A. J., AND W. SCHARLOO Inbreeding in an island population of the great tit. Evolution 35: VAN TIENDEREN, P.H., AND A. J. VAN NOORDWIJK Dispersal, kinship and inbreeding in an island population of the great tit. J. Evol. BioI. 1: WRIGHT, S The effects of inbreeding and crossbreeding on guinea pigs. III. Crosses between highly inbred families. USDA Bull. 1121: Evolution and the genetics ofpopulations: the theory of gene frequencies. Vol. 2. Univ. of Chicago Press, Chicago Evolution and the genetics of populations: experimental results and evolutionary deductions. Vol. 3. Univ. of Chicago Press, Chicago. Corresponding Editor: E. Zouros