Interactive effects of environmental stress and inbreeding on reproductive traits in a wild bird population


  • A. B. MARR,

    1. Centre for Applied Conservation Research, Forest Sciences, 3rd floor, 2424 Main Mall, University of British Columbia, Vancouver, BC, V6T 1Z4, Canada; Cornell Laboratory of Ornithology, 159 Sapsucker Woods Road, Ithaca, NY, 14850, USA;
    Search for more papers by this author
  • P. ARCESE,

    1. Centre for Applied Conservation Research, Forest Sciences, 3rd floor, 2424 Main Mall, University of British Columbia, Vancouver, BC, V6T 1Z4, Canada; Cornell Laboratory of Ornithology, 159 Sapsucker Woods Road, Ithaca, NY, 14850, USA;
    Search for more papers by this author

    1. Centre for Applied Conservation Research, Forest Sciences, 3rd floor, 2424 Main Mall, University of British Columbia, Vancouver, BC, V6T 1Z4, Canada; Cornell Laboratory of Ornithology, 159 Sapsucker Woods Road, Ithaca, NY, 14850, USA;
    Search for more papers by this author
  • J. M. REID,

    1. School of Biological Sciences, Zoology Building, University of Aberdeen, Tillydrone Avenue, Aberdeen, AB24 2TZ, Scotland, UK;
    Search for more papers by this author
  • L. F. KELLER

    1. Zoological Museum, University of Zürich, Winterthurerstrasse 190, CH-8057, Zürich, Switzerland
    Search for more papers by this author

Amy Marr, Centre for Applied Conservation Research, Forest Sciences, 3rd floor, 2424 Main Mall, University of British Columbia, Vancouver, BC, V6T 1Z4, Canada. Tel.: 207 221 2986. E-mail:


  • 1Conservation biologists are concerned about the interactive effects of environmental stress and inbreeding because such interactions could affect the dynamics and extinction risk of small and isolated populations, but few studies have tested for these interactions in nature.
  • 2We used data from the long-term population study of song sparrows Melospiza melodia on Mandarte Island to examine the joint effects of inbreeding and environmental stress on four fitness traits that are known to be affected by the inbreeding level of adult birds: hatching success, laying date, male mating success and fledgling survival.
  • 3We found that inbreeding depression interacted with environmental stress to reduce hatching success in the nests of inbred females during periods of rain.
  • 4For laying date, we found equivocal support for an interaction between parental inbreeding and environmental stress. In this case, however, inbred females experienced less inbreeding depression in more stressful, cooler years.
  • 5For two other traits, we found no evidence that the strength of inbreeding depression varied with environmental stress. First, mated males fathered fewer nests per season if inbred or if the ratio of males to females in the population was high, but inbreeding depression did not depend on sex ratio. Second, fledglings survived poorly during rainy periods and if their father was inbred, but the effects of paternal inbreeding and rain did not interact.
  • 6Thus, even for a single species, interactions between the inbreeding level and environmental stress may not occur in all traits affected by inbreeding depression, and interactions that do occur will not always act synergistically to further decrease fitness.


The offspring of closely related mates often show lower fitness and physiological efficiency than the offspring of distantly related mates (Falconer & Mackay 1996; Crnokrak & Roff 1999; Keller & Waller 2002). This phenomenon, called inbreeding depression, can affect the size and persistence of small, relatively isolated populations (Saccheri et al. 1998). Environmental stressors can also affect reproductive rates and the dynamics of populations (Lack 1954; Newton 1998). Environmental stress may exacerbate inbreeding depression if genes that affect stress tolerance are also affected by inbreeding (Pederson 1968; Kristensen et al. 2005) or if selection against deleterious mutations is more intense under stressful conditions (Kondrashov & Houle 1994).

The empirical evidence for interactions between inbreeding depression and environmental stress, however, is mixed. Armbruster & Reed (2005) reviewed mostly lab studies and those comparing lab and field conditions to find that inbreeding depression was greater in more stressful environments in 76% of cases, and significantly greater in 48% of cases. 24% of studies showed lower inbreeding depression in more stressful environments. Among studies of natural populations, there have been only a few testing the interactive effects of environmental stress and inbreeding, precluding reliable estimates of their magnitude or occurrence in nature. Inbred Soay sheep Ovis aries, for example, had higher parasite burdens and lower survival than more outbred sheep, particularly in years of high population density (Coltman et al. 1999). In cactus finches Geospiza scandens, inbreeding depression in survival of juveniles from ringing to age 1 was only present in dry years, and adult finches experienced greater inbreeding depression in survival when conditions were dry and intraspecific competition was high (Keller et al. 2002). In contrast, there were no interactions between average body condition of the population and inbreeding depression in collared flycatchers Ficedula albicollis (Kruuk, Sheldon & Merilä 2002).

Here, we test for the influence of several potential environmental stressors on four reproductive traits known to be affected by inbreeding depression in song sparrows resident on Mandarte Island, Canada. Throughout, we follow the definition that ‘stressful’ environmental conditions are those that reduce mean trait values for fitness components relative to a more benign environment (Hoffmann & Parsons 1991; Armbruster & Reed 2005). We also test if stress interacts with inbreeding to affect fitness in this population. Because the breeding success and survival of all song sparrows on Mandarte Island has been monitored intensively since 1975, and given the species’ relatively short generation time, the social pedigree for this population is among the best known for a free-living vertebrate. The Mandarte song sparrow study is also unusual among studies of vertebrate populations because immigration rates are low and individuals show a wide range of inbreeding levels, including frequent cases of moderate to close inbreeding (Keller, Marr & Reid 2006).


field site and methods

Mandarte is a small, 6 ha island in the Haro Strait of British Columbia, Canada. The number of adult song sparrows that reside year-round on Mandarte is small and has fluctuated from a high of 159 adults in 1985 to a low of 11 adults in 1989. Each March–August from 1975 to 2004, except 1980, attempts were made to find and monitor all nests to record clutch size, hatching success and subsequent survival. Virtually all nests with young were found and all birds colour-ringed for later identification, most as 5–7-day-old nestlings. Since 1975, the population has received 0–4 immigrants per year (average = 1·2). Smith (2006) provides further details of the island, population, and study methods.

pedigree and microsatellite data

A social pedigree was constructed for the population by identifying mothers as the incubating female, and fathers as the male defending the territory in which a female nested (Keller 1998). We used this pedigree to calculate Wright's inbreeding coefficient, F, for each individual in the population using proc inbreed (SAS Institute 2003). F describes the probability that two alleles at a randomly chosen locus will be identical by descent from an ancestor shared by the mother and father (Falconer & Mackay 1996). Genetic data on offspring hatched from 1993 to 1996 suggest that intraspecific brood parasitism does not occur in this population, but that females do engage in extrapair matings, with 28% of 469 genotyped offspring sired outside the social pair (O’Connor et al. 2006). Because extrapair paternity (EPP) introduces error to inbreeding coefficients and estimates of inbreeding depression (Marr, Dallaire & Keller, in press), we revisit the implications of EPPs for our analyses in the Discussion.

To limit the potential for bias due to variation in pedigree length across years, we restricted our analyses to the period from 1990 to 2004 and we excluded individuals with unknown parents or grandparents unless they were the offspring of an immigrant. Immigrants to the population were assumed to be unrelated to each other and to Mandarte residents (Keller et al. 2006). In practice, however, excluding individuals with poorly known pedigrees had little influence on final statistical models and parameter estimates.

statistical analyses

To test for interactive effects of inbreeding and stress, we selected a priori four reproductive traits known from prior work (Keller 1998; Keller et al. 2006) to exhibit inbreeding depression in the Mandarte population, including: male mating success (inbred males tend to father fewer nests annually), fledgling survival (offspring of inbred males tend to survive poorly from age 12 to 24 days), laying date (inbred females tend to lay eggs later in spring) and hatching success (inbred females tend to hatch a smaller fraction of eggs laid in each clutch). For analyses here, we note that paternal F was the predictor of inbreeding depression in fledgling survival, rather than fledgling F. Likewise, maternal F was the predictor of inbreeding depression in hatching success, rather than embryo F.

We next reviewed prior studies on song sparrows, other songbirds, and our field observations to compile a list of environmental stressors that may have influenced the traits above and for which we had data (see below). Data were analysed for each trait using generalized linear mixed models with an appropriate error structure (log-linear models with Poisson error for analyses of male mating success and laying date, logistic models with binomial error for analyses of fledgling survival and hatching success). Following the advice of Draper & Smith (1998), we plotted residuals vs. fitted Yi values and examined normality test statistics generated by the univariate procedure in SAS to determine if response variables needed transformations. This approach showed that residuals in analyses of laying dates were not normally distributed because some birds each year bred much later than the average. A Poisson distribution provided the best fit to the data and was therefore used in the analyses. Statistical analyses were conducted in SAS using the nlmixed procedure (SAS Institute 2003) with initial parameter values derived from the output of the genmod procedure using only fixed effects.

Some individuals contributed multiple observations to the dataset in analyses for each trait. For example, for analyses of male nesting attempts, each adult male contributed an observation for each year it was alive. For analyses of female laying date, each female contributed an observation every year that she laid at least one egg. For analyses of fledgling survival and hatching success, all breeding attempts with at least one fledgling or one egg, respectively, were included. To prevent any lack of independence among observations from causing standard errors and test statistics to be under- or overestimated, respectively, we included bird identity as a random variable in all analyses. Sample sizes ranged from 371 to 680 observations for 155–262 birds. Sample sizes varied among analyses because nests that fail at an earlier life-history stage do not contribute data for later stages.

For each trait, we used a multimodel inference approach to explore a suite of statistical models estimating the effects of inbreeding, potential stressors, and the interaction of these factors. A complete rationale for this approach is provided with examples by Burnham & Anderson (2002). We selected this approach because we identified several potential stressors for each trait and we wanted to illustrate the degree of model selection uncertainty in each case (Burnham & Anderson 2002; p. 271). In brief, multimodel inference involves: (1) creating a limited set of predetermined statistical models based on prior knowledge of the system; (2) estimating fits of these models and calculating their relative support based on Akaike's Information Criterion (AIC); and (3) interpreting results based on the rule of thumb that all models within an AIC value of 4 from the best-supported model are biologically plausible descriptions of the true relationships in nature. Support for particular models was judged using the bias-corrected version of Akaike's Information Criterion (AICC, Burnham & Anderson 2002), calculated from the maximum-likelihood derived log-likelihoods generated by the nlmixed procedure. Lower AICC scores indicate better model support. All models included age as a categorical variable, because we knew that reproductive success varies nonlinearly with age in this population (Smith, Marr & Hochachka 2006a).

To estimate effect sizes for each fitness trait, we used our model output to calculate the adjusted average fitness of age 2 outbred birds (i.e. birds with F = 0), age 2 moderately inbred (F = 0·06), and age 2 highly inbred birds (F = 0·25) from the lowest to highest levels of environmental stress observed for each trait. Age 2 birds were chosen for graphical comparisons because this is the median age of breeding birds in the population. Inbreeding levels of F = 0 and F = 0·25 were used for descriptive purposes, because these are the levels most often reported in experimental studies (e.g. Crnokrak & Roff 1999). An individual with F = 0·25 could be the offspring of a mating between full sibs or between a parent and offspring with no previous relationship deeper in the pedigree (Falconer & Mackay 1996). We also plotted an inbreeding level of F = 0·06 because inbreeding coefficients of adult song sparrows on Mandarte have ranged from F = 0–0·31 and averaged F ≈ 0·06 in recent years. An inbreeding coefficient of F = 0·0625 corresponds to a mating between first cousins.

Male mating success

Male song sparrows on Mandarte devote considerable time to territory defence and mate attraction and also help females feed young (Arcese 1989). Females typically produce two or three nests annually (range 0–6; Smith et al. 2006a) and, since 1975, about 40% of females had more than one social mate within a season. Mate switches mainly occurred after males were evicted from their territories or females dispersed between nesting attempts. Most successful males attracted one female to their territory at a time, but about 4% had two or more females nesting in their territories simultaneously. Thus, we measured male mating success as the number of nests initiated in a male's territory. Males without territories or females were included in analyses and assigned a count of zero nests. Two measures of competitive stress were explored: the total number of adult males alive and the ratio of adult males to females (adult sex ratio) at the end of April each year.

Fledgling survival

Song sparrows fledge at 9–11 days after hatching but depend on their parents for food to about 24–30 days of age (Smith 2006). To assess the survival of fledglings, offspring were thus counted at days 12 and 24 by listening for begging calls and observing parents delivering food. Our field observations suggest that fledglings beg more often and intensely during rainy periods. We thus explored several measures of rainfall including: the daily average rainfall when fledglings were 12–24 days, as well as the average daily rainfall during the rainiest 2-, 3- or 4-day interval when fledglings were 12–24 days. Weather data came from the nearest Environment Canada weather station, located at Victoria International Airport, 11 km west of Mandarte. We assessed models with rain intervals of 2–4 days because exploratory analyses showed that 98% of rain intervals (consecutive days with more than 2 mm day−1) lasted 4 days or less. Analyses were conducted using the number of day 24 offspring per day 12 offspring as the response variable, with the unit of analysis being the brood.

Laying date

Many studies show that temperature can influence the onset of breeding in birds (e.g. Newton 1998). Female song sparrows on Mandarte that began breeding earlier had more nests, fledglings, and recruits than females initiating breeding later the same year (Smith et al. 2006a). Most females laid a first egg between late March and early May (median: 15 April) but did so later in years when temperatures were cool from February to April (Smith et al. 2006a). We therefore tested for the influence of daily average air temperature during six time intervals: February, March, April, February–March, March–April, and February–April.

Hatching success

Like most open-cup nesting songbirds in North America, female song sparrows work alone to incubate eggs (Smith et al. 2006a). Field observations on Mandarte Island suggested that poor hatching success sometimes coincided with rain storms. Thus, we explored several measures of rainfall including: the daily average rainfall during the 13 days that typically span incubation and the daily average rainfall during the rainiest 2-, 3-, or 4-day interval within the incubation period. Analyses were conducted using the number of hatchlings per eggs in a clutch as the response variable, thus making the clutch the unit of analysis.

Brown-headed cowbirds Molothrus ater are brood parasites that also can influence hatching success (Arcese, Smith & Hatch 1996). From 1990 to 2004, cowbirds were absent in 9 years, rare in 3 years (1–3% of song sparrow nests parasitized), and common in 3 years (13–32% of song sparrow nests parasitized). When cowbirds were common, associations between climate and hatching success can be obscured because rainfall declines over the breeding period but nest failure related to cowbird activity increases. Cowbirds typically begin egg-laying 2–3 weeks after song sparrows, and begin puncturing song sparrow eggs up to 9 days before beginning to parasitize nests (Arcese, Smith & Hatch 1996; Smith et al. 2006b). Thus, for 3 years when cowbirds were common (1993, 1994, 2000), we omitted from analyses of hatching success all nests with eggs during the period 9 days prior to cowbird laying to the end of breeding.


male mating success

Males fathered fewer nests annually when there were more males in the population overall and when there were more males per female in the population (Fig. 1A), but of these two measures of competition, we found much greater support for models using the number of males per female (adult sex ratio) as a predictor (Table 1). Males also fathered fewer nests seasonally if they were inbred (Tables 1 and 2). Based on AICC scores for all models examined, the best supported model predicted the number of nests fathered from a male's age, his inbreeding coefficient, and the adult sex ratio (Tables 1 and 2). Adding a term for the interaction between inbreeding and adult sex ratio yielded a model with only 36% of the support of the best model, based on Akaike weights, and the estimated interaction effect was small (Table 1, Model 2). Highly inbred males fathered 32·0% fewer nests than outbred males when the sex ratio was at its lowest level (i.e. low stress, M/F = 0·96; F=0 = 3·41, F=0·25 = 2·32 nests), and a nearly identical 31·1% difference existed between inbred and outbred males when males greatly outnumbered females (i.e. high stress, M/F = 3·07; F=0 = 0·61, F=0·25= 0·42 nests; Table 3; Fig. 1A).

Figure 1.

Model estimated effects of inbreeding and stress on four reproductive traits. Graphs show the predicted values for age 2 outbred (F = 0) birds, moderately inbred (F = 0·06), and highly inbred (F = 0·25) birds based on the best-supported models that included terms describing the interaction between inbreeding and stress, even if the interaction term was not statistically significant. Including even a non-significant interaction term serves to illustrate that for all traits examined except hatching success (D), this interaction had a marginal biological impact. Moving from left to right, the x-axis of each graph depicts the range of stress levels observed from lowest to highest. The y-axes are arranged such that lower performance (fewer nests, lower survival, later breeding, or lower hatching success) is lower on the page. Therefore, both axes for the graph of laying date (C) are reversed so that later breeding is lower on the page and cooler temperatures are further to the right.

Table 1.  Comparison of eight models examining the effects of a male's age, his inbreeding level (F), and competition on the number of nests that a male fathered in a year (his mating success). The column titled F × Competition refers to the interaction between the male's inbreeding level and either the total number of males in the population or the number of males per female, depending on which measure of competition was included in the model. Models are listed in order from best to worst based on the bias-corrected version of Akaike's Information Criterion (AICC). AICC scores come from generalized linear mixed models. The relative support of the data for each model, as a proportion between 0 and 1 (and summing to 1 across all models), is indicated by the Akaike weight. Separate coefficients were estimated for each of five age classes. Analyses were based on a dataset comprised of 680 observations of male seasonal reproductive success
Model no.AgeFMalesMales per femaleF× CompetitionParametersAICCAkaike weight
1XX X 71958·30·687
2XX XX81960·40·246
3X  X 61963·00·067
4XXX  72064·60·000
5XXX X82066·50·000
6X X  62068·30·000
7XX   62077·60·000
8X    52082·30·000
Table 2.  Statistical models of the effect of a breeding bird's age, inbreeding and stress on (A) mating success, (B) fledgling survival, (C) laying date and (D) hatching success. Models presented are those with the best support based on the AICC scores for each reproductive trait (i.e. Model 1 in each of Tables 1, 4–6). Separate coefficients were estimated for each of five age classes because reproductive success varies nonlinearly with a breeding bird's age. Age-effects are given as deviations from the mean value of age class 5. The labels ‘paternal’ and ‘maternal’ are used for the fledgling survival and hatching success traits to emphasize that the age and inbreeding level of the parent are the predictors, not the age or inbreeding of the fledgling or egg. Coefficients are negative for terms that reduce male mating success in model A, fledgling survival in model B, and hatching success in model D, but positive for terms that delay laying (i.e. increase laying date) in model C. For (A) and (C), coefficients and the random effect variances with standard errors are presented on the log scale; for (B) and (D) these statistics are presented on the logit scale
(A) Male mating success
Intercept 2·00720·164212·22< 0·001
Age 1−0·67630·1146−5·90< 0·001
Age 2 0·00440·11080·040·969
Age 3 0·08950·11840·760·450
Age 4−0·01220·1296−0·090·925
Age 5+ 0·0000
Males per female−0·81750·0761−10·74< 0·001
Random effect variance 0·01880·02190·860·393
   Nobservations = 680, Nmales = 262
(B) Fledgling survival
Intercept 2·01250·23648·51< 0·001
Paternal age 1−0·65500·2441−2·680·008
Paternal age 2−0·49940·2437−2·050·042
Paternal age 3−0·37690·2591−1·450·148
Paternal age 4−0·50870·2718−1·870·063
Paternal age 5+ 0·0000
Paternal F−2·49341·1437−2·180·031
Rain (3 day)−0·07360·0234−3·140·002
Random effect variance 0·24300·10132·400·018
   Nobservations = 650, Nmales = 179
(C) Female laying date
Intercept 5·00590·0437114·65< 0·001
Age 1 0·02820·02181·290·198
Age 2−0·01600·0228−0·700·484
Age 3−0·01830·0241−0·760·448
Age 4 0·02740·02621·040·298
Age 5+ 0·0000
F 0·45000·10454·31< 0·001
Temp. (February–March)−0·06210·0063−9·82< 0·001
Random effect variance 0·00130·00062·310·022
   Nobservations = 371, Nfemales = 166
(D) Hatching success
Intercept 2·28020·30747·42< 0·001
Maternal age 1−1·00870·2729−3·70< 0·001
Maternal age 2−0·23340·2828−0·830·410
Maternal age 3−0·64580·2849−2·270·025
Maternal age 4−0·87970·2922−3·010·003
Maternal age 5+ 0·0000
Maternal F−0·70072·0201−0·350·729
Rain (4 day)−0·01960·0349−0·560·576
Maternal F× Rain (4 day)−1·29280·4471−2·890·004
Random effect variance 0·87760·19114·59< 0·001
   Nobservations = 640, Nfemales = 155
Table 3.  Effects of inbreeding, various environmental stresses, and their interaction on four reproductive traits: male mating success (measured as number of nests fathered in a season), survival of fledglings of inbred males (day 24 offspring per day 12 offspring), laying date (date of first egg in spring, January 1 = 1), and hatching success of inbred females (hatchlings per egg). Reproductive trait estimates are the predicted values for outbred birds (F = 0) and highly inbred birds (F = 0·25) at the lowest and highest stress levels observed. Estimates are derived from the best-supported models in Tables 1, 4–6 that included a term for the interaction between inbreeding and environmental stress. The last four columns of the table show the effect on fitness of: (1) stress for birds with F = 0; (2) stress for birds with F = 0·25; (3) inbreeding at low stress levels; and (4) inbreeding at high stress levels. For mating success, fledgling survival, and hatching success, the effect of stress was reported as a percentage by calculating the difference in trait value estimates between the highest and lowest stress levels and then dividing by the trait value estimate for the lowest stress level (i.e. (Xhigh stress − Xlow stress)/Xlow stress). Similarly, the effect of inbreeding was reported as a percentage by calculating the difference in trait value estimates between birds with F = 0·25 and birds with F = 0 and then dividing by the trait value estimate for birds with F = 0 (i.e. (XF=0·25 − XF=0)/XF=0). For laying date, earlier laying (i.e. a smaller Julian date) indicates better performance. Therefore, for laying date, the effect of stress was calculated as the difference in Julian date between the highest and lowest stress levels (i.e. Xhigh stress − Xlow stress); the effect of inbreeding was calculated as the difference in Julian date between birds with F = 0·25 and birds with F = 0 (i.e. XF=0·25 − XF=0)
Table no., model no.Reproductive traitStressorReproductive trait estimateEffect of stress on F = 0Effect of stress on F = 0·25Effect of inbreeding at low stressEffect of inbreeding at high stress
F = 0 low stressF = 0 high stressF = 0·25 low stressF = 0·25 high stress
1, 5Male mating successMales 2.62  1·50  1·46  1·03−42·7 %−29·5 %−44·3 %−31·3 %
1, 2 Males per female 3.41  0·61  2·32  0·42−82·1 %−81·9 %−32·0 %−31·1 %
4, 4Fledgling survivalRainiest 2-day interval 0.822  0·525  0·704  0·415−36·1 %−41·1 %−14·4 %−21·0 %
4, 3 Rainiest 3-day interval 0.819  0·564  0·710  0·404−31·1 %−43·1 %−13·3 %−28·4 %
4, 9 Rainiest 4-day interval 0.811  0·654  0·705  0·454−19·4 %−35·6 %−13·1 %−30·6 %
4, 11 Rain for entire period 0.809  0·680  0·704  0·496−15·9 %−29·5 %−13·0 %−27·1 %
5, 10Female laying dateTemp. February92107107115 15 days  8 days 15 days  8 days
5, 8 Temp. March91110107119 19 days 12 days 16 days  9 days
5, 17 Temp. April92103111112 11 days  1 days 19 days  9 days
5, 2 Temp. February–March88111105118 23 days 13 days 17 days  7 days
5, 12 Temp. March–April92108109116 16 days  7 days 17 days  8 days
5, 4 Temp. February–April88111106118 23 days 12 days 18 days  7 days
6, 3Hatching successRainiest 2-day interval 0.886  0·852  0·864  0·123 −3·8 %−85·8 % −2·5 %−85·6 %
6, 2 Rainiest 3-day interval 0.886  0·851  0·865  0·112 −4·0 %−87·1 % −2·4 %−86·8 %
6, 1 Rainiest 4-day interval 0.886  0·856  0·867  0·062 −3·4 %−92·8 % −2·1 %−92·8 %
6, 7 Rain for entire period 0.878  0·889  0·858  0·115  1·3 %−86·6 % −2·3 %−87·1 %

fledgling survival

Fledgling survival was lower during rainy periods (Fig. 1B). Of all measures of rainfall examined, the average daily rainfall during the rainiest 3-day interval explained the most variation in fledgling survival, but models including rainfall during the rainiest 2-day interval were nearly as well-supported (Table 4). Longer intervals of rain were also associated with lower fledgling survival (Tables 3 and 4). Fledglings experienced lower survival rates if their father was inbred (Tables 2 and 4), but there was only weak support for models that included an interaction between a father's F and rain (22% summed across all models; Table 4). Nevertheless, the predicted effects were consistent with higher inbreeding depression with greater stress. For example, in the absence of rain we estimate that fledglings with inbred fathers survived 13·3% worse than fledglings with outbred fathers (day 24 offspring per day 12 offspring: F=0 = 0·819, F=0·25 = 0·710); whereas during the rainiest 3-day interval, fledglings with highly inbred fathers survived 28·4% worse than fledglings with outbred fathers (F=0 = 0·564, F=0·25 = 0·404; Table 3, Fig. 1B).

Table 4.  Comparison of 14 generalized logistic regressions examining the effects of a breeding male's age (Paternal age), his inbreeding level (Paternal F), and various measures of rain on the survival of his fledgling offspring from day 12 to day 24. The column titled Paternal F × Rain refers to the interaction between the father's inbreeding coefficient and the daily average rain during the rainiest 2-, 3- or 4-day interval within the period that fledglings were day 12–24, or the entire fledgling period. Analyses are based on data from 650 nests with at least one fledgling. Models are listed in order from best to worst based on the AICC scores
Model no.Paternal agePaternal FRainiest 2-day intervalRainiest 3-day intervalRainiest 4-day intervalRain for entire periodPaternal F × RainParametersAICCAkaike weight
1XX X   71319·10·266
2XXX    71319·20·253
3XX X  X81321·10·095
4XXX   X81321·20·091
5X  X   61321·60·076
6X X    61321·60·075
7XX  X  71322·40·050
8XX   X 71323·20·034
9XX  X X81324·40·018
10X   X  61325·10·013
11XX   XX81325·20·012
12X    X 61325·80·009
13XX     61326·60·006
14X      51329·70·001

laying date

Females laid their first eggs later in spring if they were inbred and when temperatures were cool (Fig. 1C). The best model predicted laying date from a female's age, her inbreeding coefficient, and the average daily temperature in February and March (Tables 2 and 5). The second best model, which included the terms of the best model and, additionally, an interaction between inbreeding and temperature, received only slightly less support (82% of the best model based on Akaike weights; Table 5). Thus, there was evidence of a possible interaction between the effects of inbreeding and temperature on laying date. Based on the model including this interaction (Fig. 1C), we estimate that highly inbred females laid their first eggs 17 days later on average than outbred females when temperatures were warm in February and March (F=0 = 88, F=0·25 = 105); whereas, when temperatures were cool, the estimated difference in laying date between highly inbred and outbred females was only 7 days (F=0 = 111, F=0·25 = 118; Table 3). Therefore, if anything, the difference in laying date between inbred and outbred birds was greater under the less stressful conditions of warmer springs.

Table 5.  Comparison of 20 generalized linear mixed models examining the effects of a female's age, her inbreeding level (F), and air temperature on the date that she lays her first egg in spring. Temperature was measured for each year as the daily average temperature (°C) during February, March, April, February–March, March–April, or February–April. Each analysis includes data from 371 nests. Models are listed in order from best to worst based on the AICC scores
Model no.AgeFTemp. FebruaryTemp. MarchTemp. AprilTemp. February–MarchTemp. March–AprilTemp. February–AprilF ×  Temp.ParametersAICCAkaike weight
1XX   X   72770·40·378
2XX   X  X82770·80·311
3XX     X 72772·10·165
4XX     XX82772·30·146
5X    X   62785·70·000
6X      X 62789·10·000
7XX X     72804·30·000
8XX X    X82805·70·000
9XXX      72814·20·000
10XXX     X82815·10·000
11XX    X  72816·10·000
12XX    X X82817·10·000
13X  X     62823·10·000
14X X      62826·80·000
15X     X  62835·10·000
16XX  X    72845·10·000
17XX  X   X82845·70·000
18X   X    62863·10·000
19XX       62863·30·000
20X        52878·90·000

hatching success

Unlike the three traits above, hatching success showed a strong interaction between inbreeding and environmental stress. Hatching success was lower in the eggs laid by inbred females, particularly when incubation coincided with periods of rain (a summed 95% support over all models was for those containing inbreeding by rain interactions; Tables 2, 3 and 6). Of all measures of rainfall considered, the rainiest 4-day interval during the incubation period was the best predictor of hatching success (Table 6), but hatching success was also reduced for all other intervals considered. According to our best model (Table 2), outbred females reared about 0·886 hatchlings per egg in the absence of rain, whereas highly inbred females (F = 0·25) had hatching success of 0·856 in the absence of rain; during periods of heavy rain, outbred females had only slightly diminished hatching success (hatchlings per egg: F=0 = 0·867), whereas highly inbred females were affected severely, performing less than 1/10th as well as outbred birds on average (F=0·25= 0·062; Table 3, Fig. 1D).

Table 6.  Comparison of 14 generalized logistic regressions examining the effects of a female's age (maternal age), her inbreeding level (maternal F), and various measures of rain on the hatching success of the eggs that she produces. The column titled Maternal F × Rain refers to the interaction between the mother's inbreeding coefficient and the daily average rain during the rainiest 2-, 3- or 4-day interval in the incubation period, or the entire incubation period. Analyses used data from 640 nests. Models are listed in order from best to worst based on the AICC scores
Model no.Maternal ageMaternal FRainiest 2-day intervalRainiest 3-day intervalRainiest 4-day intervalRain for entire periodMaternal F × RainParametersAICCAkaike weight
1XX  X X81836·50·769
2XX X  X81840·00·129
3XXX   X81842·00·047
4XX  X  71843·70·021
5XX X   71845·20·010
6X   X  61845·60·008
7XX   XX81846·30·006
8XXX    71846·80·004
9X  X   61847·00·004
10X X    61848·70·002
11XX   X 71853·80·000
12X    X 61855·60·000
13XX     61862·50·000
14X      51864·40·000


For the small and relatively isolated population of song sparrows on Mandarte Island, we found clear evidence that environmental stress affected four fitness traits that were also affected by inbreeding depression (Tables 1 and 4–6). We also found that inbreeding interacted with environmental stress for one of the four fitness traits; hatching success was lower in the nests of inbred females, and this deficit increased greatly during periods of rain (Fig. 1D). For laying date, it is less clear whether the effects of inbreeding and environmental stress interacted (Table 5). The average laying date for all females was delayed in cooler years, as expected if low temperatures increased thermal stress, but, if anything, inbreeding depression in laying date was reduced in cooler years (Fig. 1C). For two other traits, mating success and fledgling survival, the effects of inbreeding and environmental stress acted independently (Fig. 1A,B). Overall therefore our findings are generally consistent with the handful of other studies of natural populations that have found that inbreeding depression is sometimes exacerbated by environmental stress but the relationship between inbreeding and stress varies across traits. Our findings also fit with the general pattern identified in an in-depth review by Armbruster & Reed (2005). They showed that on average across a range of taxa, inbreeding depression tends to increase with stress, but 24% of the traits that they investigated and one of four traits (laying date) in our study exhibited less inbreeding depression under more stressful environmental conditions.

Hatching success, the one trait for which we found clear evidence of an interaction between inbreeding and environmental conditions, has received much recent interest. Several studies of birds demonstrate inbreeding depression or the effects of reduced genetic diversity on hatching success (van Noordwijk & Scharloo 1981; Bensch, Hasselquist & Schantz 1994; Kempenaers et al. 1996; Westemeier et al. 1998; Daniels & Walters 2000; Kruuk et al. 2002; Briskie & Mackintosh 2004; Spottiswoode & Møller 2004). It is unclear, however, if such effects are noted so consistently because inbreeding depression in hatching success is usually high or because it is easier to detect inbreeding depression in hatching success than in traits measured with less precision. For example, eggs and nestlings will often be counted with more precision than other fitness measures, and the largest sample sizes of nests are often available at this early stage of reproduction.

A recent study of Mandarte song sparrows provides another example of increased inbreeding depression during periods of poor environmental conditions: male annual reproductive success depends on spring temperatures (Keller et al. 2006). After warm springs, inbred males produced as many independent young as their outbred contemporaries. However, when the months January through April were cold, inbred males (F = 0·25) produced 46% fewer independent young on average (Keller et al. 2006).

Other past studies on Mandarte song sparrows mirror our conclusions for male mating success and fledgling survival (Tables 1 and 4; Fig. 1A,B) in suggesting that inbreeding depression may act independently from certain environmental stressors on some traits. While there was clear evidence for inbreeding depression in Mandarte song sparrows during a severe winter storm in February 1989 (Keller et al. 1994), subsequent analyses revealed that the level of inbreeding depression in juvenile survival was not unusually elevated during this period of extreme stress (Table 6 in Keller 1998).

In some cases, interactions between inbreeding depression and stress may be obscured by problems with the available data. For example, in our study we may have been unable to account for key explanatory variables, such as variation in predation rates or climatic factors other than rainfall or temperature. Second, paternity errors due to extra-pair fertilization will have introduced error into our estimates of inbreeding depression, typically causing inbreeding depression to be underestimated (Marr et al., in press). Finally, our statistical power to detect inbreeding by stress interactions may have been reduced by the distribution of the data. Although inbreeding coefficients of adult song sparrows on Mandarte Island ranged from F = 0 to 0·31, most birds had relatively low inbreeding coefficients. For example, less than 10% of our observations of mating success concerned males with F  0·125. Nevertheless, the omission of important environmental variables, paternity errors, and limited statistical power are unlikely to explain all cases where inbreeding depression and environmental stress have independent effects on life-history traits. In fact, the absence of inbreeding by environment interactions would be expected when genes mediating tolerance to environmental stress are relatively little influenced by inbreeding or when inbreeding and environmental factors affect different components of a trait or affect a trait at different times.

Studies of the relationship between inbreeding depression and stress have now shown variable responses among populations of a single species, among inbred lineages within a population (see Armbruster & Reed 2005), and among traits within a single, well studied system (this study). Given this large amount of unexplained variance, it is clear that we cannot yet make robust predictions about the modification of inbreeding depression during episodes of environmental stress. However, taken together, the available evidence does seem to suggest a general pattern: on average across a wide range of taxa, inbreeding depression tends to increase with environmental stress.


This paper is dedicated to the memory of Jamie Smith who initiated the Mandarte study. Jamie loved Mandarte and its song sparrows, and shared his enthusiasm with us all. We thank S. N. Aitken, M. C. Whitlock, D. Schluter, K. Ritland, and two anonymous reviewers for helpful comments on earlier drafts. The Tsawout and Tseycum First Nations bands kindly allowed us to work on Mandarte Island. Many people helped collect the field data presented here, including most recently S. D. Wilson, J. N. M. Smith, A. Johnston, D. Dagenais, R. M. Landucci, K. D. O’Connor, S. E. Runyan, and C. A. Saunders. Support for this research was provided by grants to A. B. Marr from NSF and to P. Arcese from NSERC of Canada.