1 Many studies of cooperatively breeding birds have found no effect of group size on reproductive success, contrary to predictions of most adaptive hypotheses. A model is proposed for variation in group-size effects: group size has a reduced effect on success when conditions for breeding are good, such as in good environmental conditions or in groups with older breeders. This hypothesis is tested with a case study of white-browed scrubwrens Sericornis frontalis and a review of the literature.
2 The scrubwren is a cooperatively breeding passerine with male helpers. Previous analyses revealed no effect of group size on reproductive success, but those analyses were restricted to groups with older females (Magrath & Yezerinac 1997). Here 7 years’ data are used to contrast the effect of group size on reproductive success for yearling and older females.
3 Yearling females breeding in groups had more than double the seasonal reproductive success than those breeding in pairs, even after controlling for territory quality. However, group size still had no effect on the reproductive success of older females. Yearling females tended to survive better in groups, but older females tended to survive better in pairs, emphasizing this pattern.
4 Yearlings breeding in pairs were more likely to be found on poor-quality territories than those breeding in groups, exaggerating the already-strong effect of group size on yearling success. Older females were not affected significantly by territory quality.
5 Group size, territory quality and female age affected different components of seasonal reproductive success. Group size increased the success of individual nesting attempts, while both territory quality and female age affected the length of the breeding season, and thus the number of breeding attempts.
6 A sample of the literature on cooperative breeders shows that group size has a larger effect on reproductive success in poorer conditions, caused either by younger, inexperienced breeders or poorer environmental conditions. Scrubwrens therefore illustrate a widespread pattern, which provides an explanation for much of the variation in group-size effects among and within species. Clearly single estimates of group-size effects for species can be inadequate to test ideas about the evolution of cooperative breeding.
In cooperatively breeding birds, more than a simple pair provide care to the young in a single brood (Brown 1987; Emlen 1997; Cockburn 1998). Breeding groups are often formed through natal philopatry of young previously reared by the group, and such groups usually consist of a dominant pair and related subordinates who provide alloparental care (Brown 1987; Stacey & Koenig 1990).
Although there are many potential benefits to helpers, the most widely cited is that they increase the reproductive success of relatives by increasing reproductive productivity of groups (Brown 1987; Emlen 1997; Mumme 1997). Mumme (1997), for example, notes that for both birds and mammals, there is both ‘extensive correlational evidence’ and experimental evidence that helpers increase the reproductive success of recipients. In some cases, larger groups can have more than double the success of pairs (e.g. Rabenold 1990; Emlen & Wrege 1991; Heinsohn 1992). This finding, in combination with evidence that helping is often preferentially directed to kin, suggests that indirect benefits are often crucial to explaining cooperative breeding (Emlen 1997; Mumme 1997).
Increasing a group’s reproductive success can also bring direct benefits to helpers, even if kinship is unimportant. Four of the seven potential direct benefits of helping listed by Emlen & Wrege (1989) require an increase in reproductive success with group size. Higher group success can be beneficial to helpers because: (1) increasing group size can enhance survival, (2) larger groups may allow territorial expansion and budding, (3) helpers may form coalitions with young produced in the group and (4) young produced in the group may later become helpers. Overall, it is important to quantify the relationship between group size and reproductive success in order to assess adaptive explanations for the evolution of cooperative breeding.
Although reproductive success is often higher in groups, this is not always true; about one-third of studies find no effect of group size (reviews by Cockburn 1998; Hatchwell 1999). The true proportion might be higher because correlations between group size and reproductive success are likely to be inflated, particularly in species in which groups form through natal philopatry (Brown et al. 1982; Brown 1987). Pairs on high quality territories may have higher reproductive success, leading to larger groups in later years, so that a correlation between group size and success may be confounded by territory quality, or the age or quality of breeders.
In addition to variation among species, there is also variation in the effect of group size on reproductive success within species. In an experimental study of Florida scrub jays Aphelocoma c. coerulescens Bosc, for example, Mumme (1992) found an effect of group size on reproductive success in only one of two years. Similarly, the effect of group size appears to be greater in poor years in some species (e.g. Acorn woodpeckers, Melanerpes formicivorus Swainson, Koenig & Stacey 1990; Discussion). Even within a population, there can be variation among groups in the same year. In Seychelles warblers, Acrocephalus sechellensis Oustalet, the effect of an additional helper depends on territory quality and the original group size; although a single helper raises reproductive success, further ‘helpers’ may depress reproductive success, especially on poor territories (Komdeur 1994).
One suggested cause of variation in the magnitude of group-size effects is that helpers could have a greater effect on reproductive success when nestling starvation is common (Emlen 1991; Magrath & Yezerinac 1997; Hatchwell 1999; Legge 2000). In this situation, the provision of food by helpers is likely to have a greater effect on reproductive success than when food is abundant. Hatchwell (1999) and Legge (2000) provide support for this idea, by showing that helpers increase the total rate of provisioning specifically in those species in which brood reduction is more common. Hatchwell also found some comparative evidence that helpers have a greater effect on fledging success in those species that suffer more brood reduction when breeding in pairs.
Here a general model is proposed for variation in group-size effects among and within species (Fig. 1a). In good breeding conditions, helpers may have little or no effect on the groups’ reproductive success, because there is no major limiting factor that they can ameliorate. ‘Good breeding conditions’ includes both environmental conditions, such as food supply and predator abundance, and the quality, age or experience of breeders. For example, if food is abundant, a pair may be able to provide easily for optimal growth of offspring; or when predators are absent, there may be no benefit to extra vigilance. In poorer conditions, however, provisioning by helpers could reduce the risk of starvation of young or vigilance might reduce the risk of predation, and so increase the reproductive success of the group. In extremely poor conditions, however, helpers may be of little benefit, since their activities may rarely be sufficient to allow successful breeding. This model suggests that the failure to find group-size effects in many species might be because studies have been conducted at benign locations, or because averaging effects over individuals of different age or quality may obscure important variation.
An alternative model for failure to find an effect of group size on reproductive success is that there truly is no effect, regardless of conditions (Fig. 1b). Another possibility is that instead of having a reduced effect in benign conditions, groups might increase reproductive success by a constant increment, regardless of environmental conditions (Fig. 1c). In this case, groups will none the less have a proportionately smaller effect in better conditions, which might make it more difficult to detect an effect of group size.
One uncertainty in making predictions about group-size effects is that helpers might affect the survival and thus future reproductive success of breeders rather than, or in addition to, current reproductive success. Again, the magnitude of future benefits to breeders may depend on the conditions for breeding. Thus it is relevant to assess breeder survival as well as current reproductive success.
Testing these models requires examining the magnitude of group-size effects under different conditions for breeding. Because the demographic models that form the basis of cooperative breeding theory require detailed and long-term studies of marked individuals, it is rarely possible to conduct simultaneous studies at several sites differing in environmental conditions (Reyer 1990 provides an important exception; see below). However, it may be possible to examine group size effects in good and poor years (e.g. Woolfenden & Fitzpatrick 1984; Koenig & Stacey 1990), or on territories of different quality (e.g. Komdeur 1994, 1996a).
I suggest that focusing on breeder age may be a particularly powerful way to test these models, because individuals become more proficient at breeding as they age and gain experience (Sæther 1990; Forslund & Pärt 1995). Therefore, one can test the prediction that group size will have a greater effect on reproductive success in poorer conditions by comparing younger, inexperienced breeders with older, experienced individuals. Furthermore, an effect of group size on the reproductive success of younger individuals is important because effects early in life will have a disproportionate effect on lifetime reproductive success.
The white-browed scrubwren, Sericornis frontalis Vigors & Horsfield, is one of the species in which there appears to be no effect of group size on reproductive success Magrath & Yezerinac (1997). However, analyses were restricted to groups in which females were at least 2 years old, since the sample of yearlings was too small to include. In this paper, the effect of group size on reproductive success is examined for yearling females compared with older females, using data gathered over 7 years. Given that group size may covary with territory quality and breeder quality (above), these potentially confounding variables are also examined. It is concluded that yearlings do have greater reproductive success when breeding in groups rather than pairs, and the lack of benefit from breeding in groups for older females is confirmed. Finally, a survey of the literature shows that the pattern found in scrubwrens is of widespread importance in cooperatively breeding birds.
Materials and methods
The white-browed scrubwren is a small (11–15 g) passerine, endemic to Australia, placed either in the family Acanthizidae (Schodde & Mason 1999), or included in the subfamily Acanthizinae in the Pardalotidae (Christidis & Schodde 1991; Christidis & Boles 1994). Scrubwrens are largely sedentary as adults and breed in diverse habitats with thick vegetation, from coastal rain forest to alpine heath (Blakers, Davies & Reilly 1984). They feed primarily on arthropods found on or near the ground, often under leaf litter, but also search under bark and in thick foliage above the ground (Keast 1978; Ambrose 1985; Cale 1994; personal observation). Adults can be sexed by plumage.
Female scrubwrens build domed nests, usually on or near the ground and hidden under low vegetation or leaf litter, in which they usually lay three eggs at 2-day intervals (Magrath et al. 2000). The birds are multibrooded, commonly laying up to four clutches and raising up to two broods in a breeding season that extends from July (mid-winter) to January (mid-summer; Magrath et al. 2000).
The female alone incubates the eggs, for a period of 17–21 days, after which both sexes feed young. The nestlings fledge after about 15 days, and are usually fed by adults for a further 6–7 weeks (Magrath et al. 2000).
The study was conducted on a colour-banded population resident in and adjacent to the Australian National Botanic Gardens, in Canberra (35°16′S 149°6′E), over the seven breeding seasons 1992–98. All birds were colour-banded for these years and many were banded in a pilot year in 1991, and so were of known age in 1992. Scrubwrens bred both in the Gardens and in the adjacent Canberra Nature Park (c. 9 km2), and there was dispersal both into and out of the Gardens. The Gardens occupy an area of 40 ha, of which 27 ha are planted with native Australian plants. Most of the remainder is natural woodland, which is contiguous with the woodland of Canberra Nature Park.
The birds are resident throughout the year, and territories were visited at least three times a week during the breeding season to document group composition and reproductive attempts. A complete record was available of fledging success throughout the breeding season for between 37 and 47 resident females per year.
Success of a nesting attempt was measured as the number of young fledged or the probability of fledging any young. Number fledging was estimated as the number banded (when 9 or 10 days old), less the number found dead in the nest. If the nest was damaged, or found empty before the expected fledging date, only those seen alive during intensive searches of the territory were counted as having fledged. Seasonal reproductive success of females was measured as the total number of young fledging over the whole breeding season or the probability of fledging any young.
The date the first egg was laid was used as the measure of the beginning of the breeding season for a female. The ‘end of the breeding season’ was estimated as the date of failure or fledging of the final attempt, but females were excluded that died before they had the opportunity to lay again that season. It was assumed that females had had the opportunity to lay again if they survived more than 19 days after a failed attempt or 41 days after a successful attempt, as only 5% of females that survived beyond these periods initiated another clutch if they had not already done so. (Failed and successful attempts were treated separately following analyses in Magrath et al. 2000.)
Social groups and helping behaviour
Social groups are territorial throughout the year, and usually consist of a single breeding pair, or a trio of a female and two males (about 10% of groups had more males, usually three; Magrath & Whittingham 1997). Males in a group form a dominance hierarchy, with older males being more dominant. Groups larger than pairs generally form through natal philopatry of males, although occasionally males immigrate into groups as subordinates (Magrath & Whittingham 1997). Females always disperse from their natal group before or at the onset of the following breeding season, and attempt to fill vacancies with territorial males. Thus yearling females can breed in pairs, if they join a single male, or groups, if they join two or more males.
In 94% of 317 cases, females bred either in pairs or groups throughout the season. In those cases where group size changed within a season, the mean size over individual breeding attempts was used. The census date for each nest was the date of hatching or failure, whichever came first. Females breeding in ‘pairs’ had a mean of less than 1·5 males per attempt; those in ‘groups’ had a mean of 1·5 or more.
Females were classified as ‘yearlings’ or ‘older’. Females were classed as ‘yearlings’ throughout the breeding season following that in which they were hatched. In 70% of cases (47/67) a yearling’s age was known from banding records or because she was caught with traces of juvenile plumage. However, it can be difficult to age birds from plumage when more than about 3 months old, so 30% of ‘yearlings’ were classified as such simply because they were immigrants to the study population. It was assumed that immigrants were yearlings because only 9% of older females moved territories between years once they had bred, and in 15/16 of these cases they moved to an adjacent territory (Magrath, unpublished data). Furthermore, during the breeding season, breeding vacancies were never filled by unbanded immigrants until the date at which juvenile females began to disperse, so there was no evidence of a ‘floater’ population of older birds. Females were classified as ‘older’ birds if they were known to be older than ‘yearlings’. Some females were of unknown age in a given year, but were classified as ‘older’ females in later years.
Scrubwrens are most common in wet areas with dense cover (Blakers et al. 1984; Ambrose & Davies 1989; Christidis & Schodde 1991; personal observation), so it was assumed that these habitats were of higher ‘quality’ and territories were ranked according to three criteria that reflected these features. The classification was carried out before any analyses of the effect of habitat on reproductive success. First, each territory was classified as being in a gully (score = 1), or not in a gully (0); gullies are wetter and have thicker vegetation. Secondly, each territory was classified as being primarily in rain forest (1), cultivated garden beds (0·5) or uncultivated areas (0); the rain forest is densely vegetated and heavily irrigated, cultivated beds are regularly irrigated and uncultivated areas irregularly or not irrigated. Thirdly, territories were classified as being in the south-east (1), north-east (0·66), south-west (0·33) or north-west (0·0); the ground slopes down primarily from west to east, and secondarily from north to south, and lower areas receive water running from upper areas and are more heavily vegetated. Finally, a sum of scores from the three criteria was used as an estimate of ‘territory quality’. In practice, about one-third of territories had a sum of exactly 1·5, one-third had higher scores and one-third had lower scores. Given this distribution, ‘territory quality’ was classified as ‘low’ (sum < 1·5), ‘medium’ (1·5) or ‘high’ (> 1·5) before analyses were carried out.
Survival was recorded until the next breeding season (1 August of the following year) for all females from the 1992–97 breeding seasons. It was assumed that a female that ‘disappeared’ had died, given that most females were site-faithful once they had found a breeding vacancy (above). Breeding groups were not followed in 1999, but a census of 2-year-old females was carried out in August 1999, so that survival of 1998 yearlings could be estimated. However, given the different methods used in 1999, and the lack of data on older females, data are also presented on survival that excludes the 1998 yearlings.
Analyses of seasonal reproductive success and number of breeding attempts per season could only entail a single season for an individual as a yearling, but could potentially entail multiple years for an ‘older’ female. To avoid repeated measures from older females, and to make data directly comparable to those from yearlings, the random number generator in SPSS 9·0 (SPSS Inc. 1999a) was used to select one season if there were data for more than one season. Given that there was only one season for an individual as a yearling and one as an older bird, the frequency distributions for the two classes of female were directly comparable. Some females were represented in both the ‘yearling’ and ‘older’ class, which enabled pairwise comparisons (Results), but many appeared only once. The subsample used for analysis included one complete breeding season from each of 62 yearlings (38 in pairs, 24 in groups) and 73 older females (33 in pairs, 40 in groups).
In the analysis of seasonal reproductive performance, normally distributed variables (e.g. number of breeding attempts per season) were analysed using the general linear modelling procedure of SPSS 9·0 (SPSS Inc. 1999b), while dichotomous dependent variables were analysed using the logistic regression or log-linear modelling procedures of SPSS (SPSS Inc. 1999c). Non-parametric tests were used when a continuous dependent variable had a non-normal distribution.
Analyses of the success of individual nesting attempts within a season were carried out on those females and seasons used in the random selection described above. In this case, some females did poorly throughout a season while others did well, so it was not appropriate to use ‘nests’ as the unit of analysis. In this case the proportion of nests that were successful was modelled, using number of successful nests out of the number of attempts for each female, assuming a binomial distribution (a binomial distribution and logit link function in the generalized linear modelling procedure of Genstat 5, Release 4·1 for Windows [Genstat 5 Committee 1993, p. 352; Baird & Hunt 1998]).
Modelling in Genstat or SPSS was begun with a full model including all factors, their interactions and covariates, and then non-significant terms were progressively eliminated. In logistic regression or log-linear modelling, the significance of terms was determined by the change in deviance, which follows a chi-square distribution, when the term was dropped from the model. Thus the change in ‘likelihood ratio chi-square’ values, not Wald statistics, was used to assess effects in logistic regression models. The values reported here refer to the change in deviance at the stage the term was dropped from the model for non-significant terms, or the effect of dropping the term from the final model for significant terms. F-ratios and significance values are reported for similar analyses of normally distributed dependent variables.
Group size, reproductive success and survival
Yearlings in pairs fledged fewer young per season than older females or yearlings in groups (Fig. 2; Kruskal–Wallis; χ = 18·3, d.f. = 3, P < 0·001; medians and IQR in Fig. 2 legend). Overall means ± SE (n) were: yearling in pair 0·8 ± 0·3 (38), yearling in group 2·5 ± 0·4 (24), older female in pair 2·5 ± 0·4 (33), older female in group 2·1 ± 0·3 (40). These differences may reflect differences in the probability of total failure, differences in the number of fledglings if successful, or both. Restricting analysis to females that did not fail totally revealed no differences among females. Mean numbers fledging ± SE (n) if the female did not fail totally were: yearling in pair 3·3 ± 0·6 (9), yearling in group 3·5 ± 0·3 (17), older female in pair 3·7 ± 0·4 (22), older female in group 3·1 ± 0·3 (27); grand mean 3·4 ± 0·2 (75); Fig. 2; Kruskal–Wallis, χ2= 2·4, d.f. = 3, P = 0·5; median 3 for all group types. Because the differences among females relate to the probability of total failure, and because of the bimodal frequency distributions, subsequent analyses use a dichotomous dependent variable – the probability of success at producing any fledglings over the season.
Yearlings in pairs had a much lower probability of producing any fledglings over the season (24%), compared with yearlings in groups (71%; log-linear model χ2= 13·8, d.f. = 1, P < 0·001; Fig. 3). By contrast, older females in pairs had the same seasonal success regardless of whether they bred in pairs or groups (both 67%). Furthermore, the statistical effect of group size was strongly dependent on the female’s age (interaction, χ2= 7·1, d.f. = 1, P = 0·008).
Yearlings in pairs did not have a different probability of surviving until the next breeding season than yearlings in groups (pair 70%, n = 33; group 85%, n = 20; log-linear model χ2= 1·7, d.f. = 1, P = 0·2). Similarly, there was no significant difference for older females (pair 83%, n = 36; group 67%, n = 36; χ2= 2·7, d.f. = 1, P = 0·1). However, the differences in survival observed were in opposite directions for yearlings and older females, and the interaction was just significant (χ2= 4·0, d.f. = 1, P = 0·05). If there are real differences, yearlings appear to survive better in groups and older females in pairs, reinforcing the relative benefit to yearlings of breeding in groups.
Does territory quality confound the relationship?
Yearlings in pairs were typically found on lower-quality territories than those in groups (log-linear model χ2= 6·4, d.f. = 2, P = 0·04; data shown as sample sizes in Fig. 4a). By contrast, there was no difference for older females (log-linear model χ2= 0·02, d.f. = 2, P = 1·0; shown as sample sizes in Fig. 4b). Given this distribution, territory quality might confound the relationship between group size and reproductive success for yearling females, so it is necessary to consider the joint effects of group size and territory quality.
Yearlings in groups had a higher seasonal reproductive success than those in pairs, even while controlling for territory quality (log-linear model including group size and territory quality; χ2= 13·7, d.f. = 1, P < 0·001; Fig. 4a). Territory quality was also important; yearlings breeding on higher quality territories had higher success in both pairs and groups (χ2= 10·8, d.f. = 2, P = 0·005), and the effect of group size remained constant over territories of different quality (three-way interaction, χ2= 0·7, d.f. = 2, P = 0·7).
In contrast to yearlings, there was no discernible effect of group size or territory quality on seasonal reproductive success for older females (log-linear model: group size, χ2= 0·0; territory quality, χ2= 1·2, d.f. = 2, P = 0·5; interaction, χ2= 0·9, d.f. = 2, P = 0·6; Fig. 4b).
It is possible that there are differences in the quality of breeding locations unrelated to the ‘territory quality’ index. To assess this possibility, pairwise comparisons were used of seasonal reproductive success of the same female, breeding with the same male in the same location for all breeding attempts as a yearling and then in the following year as an older bird. If the low success of yearling females in pairs is related to age and group size, rather than location, then seasonal reproductive success should be higher in the second year (a directional hypothesis). By contrast, if the poor performance of yearling females in pairs is related to location, and the overall increase in success observed is due to most females moving to a better location, then there should be no change in success when breeding conditions are the same. Yearlings that bred in groups should show no change between years.
The mean probability of success over a season for a female in her second breeding season was more than double that of breeding as a yearling in a pair, as expected if group size affected the success of yearlings (compare with Fig. 4a). The probability of success was 0·25 as a yearling in a pair to 0·58 as an older bird, despite all else being held constant (McNemar matched-pairs test, n = 12 females, one-tailed P = 0·06; Siegel & Castellan 1988; implemented in SPSS 9·0). There were similar results for the number of fledglings produced over the season (means 1·0 and 1·92 fledglings per season; Wilcoxon matched-pairs test, n = 14, one-tailed P = 0·06). In contrast to yearlings breeding in pairs, those that bred in groups did not have higher probability of seasonal success in the subsequent year, and the mean success was actually lower (probability of success 0·82 as a yearling, 0·55 as an older bird; McNemar matched-pairs test, n = 11, P = 0·4). Similarly, the number of fledglings produced tended to be lower the following year (means 2·8 fledglings as a yearling and 1·6 as an older bird; Wilcoxon matched-pairs test, n = 11, P = 0·08).
Does breeder quality confound the relationship?
A relationship between group size and reproductive success could be confounded by breeder quality (Brown 1987). This issue is addressed by using two indirect measures that are likely to be correlated with female quality, and using age as a possible correlate of ‘quality’ for dominant (or pair) males.
First, if yearling females that breed in pairs are lower quality birds than yearlings that breed in groups, this should be revealed in lower reproductive success in subsequent breeding seasons. Therefore the seasonal reproductive success of older females that had bred in pairs as yearlings was compared with older females that had bred in groups as yearlings. No difference was found in the proportion of females producing any fledglings over the season (previously bred in: pair, 0·60, n = 20; group, 0·63, n = 16; log-linear model χ2= 0·02, d.f. = 1, P = 0·9). Similarly, there was no difference in the number of fledglings produced over the season (previously in: pair, mean 1·9, median 2, IQR 0–3, n = 20; group, mean 1·6, median 2, IQR 0–3, n = 16; Kolmogorov–Smirnov Z = 0·26, P = 1).
Secondly, the reproductive performance of yearling females was examined according to whether they died before the following breeding season. The previous analysis could be biased if poorer-quality yearlings were more likely to die, and so were excluded from the comparison. Yearlings in groups had higher success than those in pairs regardless of whether they subsequently survived or died (log-linear model: survived, χ2= 8·3, d.f. = 1, P = 0·004, n = 43; died, χ2= 5·5, d.f. = 1, P = 0·02, n = 19; Table 1a). Furthermore, the strength of the group size did not differ between the two classes of females (interaction, χ2= 0·3, d.f. = 1, P = 0·6). If the data from 1998 are excluded (see Methods), there are still strong effects for each class of female (log-linear model: survived, χ2= 8·0, d.f. = 1, P = 0·005, n = 37; died, χ2= 13·5, d.f. = 1, P < 0·001, n = 12; Table 1b). However, in this sample the effect of group size was stronger if the female had subsequently died than if she had survived (interaction, χ2= 5·6, d.f. = 1, P = 0·02; Table 1b).
Table 1. Seasonal reproductive success of yearlings in pairs and groups according to whether they survived until the next breeding season (1 August). (a) Includes yearlings breeding from 1992 to 1998; (b) excludes 1998 yearlings, because the population was not followed closely in 1999 (see Methods)
Fate of female
Yearling females in pairs bred typically with younger males than other females and so male age might confound the relationship between group size and reproductive success of yearlings. The median age of the male was 3 years (range 1–10) for yearling-female pairs, 5 years (2–11) for yearling-female groups, 4 years (2–11) for older-female pairs and 5 years (1–11) for older-female groups; Kruskal–Wallis, χ2= 13·6, d.f. = 3, P = 0·004; male age is the minimum age if not known exactly.
To assess whether male age did confound the relationship between group size and reproductive success for yearling females, a logistic regression of seasonal reproductive success (failure or success) was used on group size and territory quality (categorical variables), and male age and age squared (continuous variables). Age and age-squared were used in case the effect of male age was not a monotonic increase. There was no effect of male age (χ2= 0·8, P = 1, P = 0·4) or age-squared (χ2= 0·0) on the reproductive success of yearling females. (As in previous analyses, both territory quality and group size had significant effects.)
Contrasting effects of group size, territory quality and female age
Group size, female age and territory quality affected seasonal reproductive success in different ways, further isolating the ‘group size’ effect.
Group size affected the success of individual nesting attempts, not the number of attempts per season. The proportion of nests producing any fledglings was much lower for yearlings in pairs than for other females (Fig. 5; binomial model of number of successful attempts out of total attempts, interaction of female age and group size: χ2= 8·6, d.f. = 1, P = 0·003, n = 126). By contrast, group size did not affect the number of breeding attempts per season either alone (F = 2·5, d.f. = 1, 130, P = 0·12), in interaction with female age (F = 0·15, d.f. = 1, 129, P = 0·7), territory quality or both (all P > 0·2).
Female age affected the number of breeding attempts, and only affected the success of individual attempts through the interaction with group size (above). Older females had more attempts (F = 22·5, d.f. = 1, 131, P < 0·001; Fig. 6).
Yearling females had fewer breeding attempts because they started breeding over 4 weeks later than older females. The median date that the first eggs of the season were laid was 27 September (IQR 9 September–7 October) for yearlings and 26 August (IQR 20 August–4 September) for older females (Mann–Whitney U, Z = 6·4, P < 0·001; the date within years was first adjusted to the grand median over years). When date of the first egg was included in a model of number of breeding attempts, female age was no longer significant (F = 0·3, d.f. = 1, 120, P = 0·6), while date was highly significant (F = 43·7, d.f. = 1, 122, P < 0·001). There was still no effect of group size alone or in interaction with any other variable; the only other significant factor was territory quality (F = 5·6, d.f. = 2, 122, P = 0·005).
Territory quality affected the number of breeding attempts, and not the success of individual attempts. Birds breeding on higher quality territories had more breeding attempts (F = 5·4, d.f. = 2, 131, P = 0·007; Fig. 6). By contrast, territory quality did not affect the proportion successful, either as a main effect (log-linear model: χ2= 0·1, d.f. = 2, P = 0·9) or in interactions with female age (χ2= 0·3, d.f. = 2, P = 0·8), group size (χ2= 0·6, d.f. = 2, P = 0·5) or both (χ2= 0·5, d.f. = 2, P = 0·6).
Higher quality territories allowed more breeding attempts because breeding continued longer. Median dates of failure or fledgling of the last nest of the season were: low quality 20 November (30 October–6 December, n = 35); medium quality 5 December (20 November–17 December, n = 31); high quality 18 December (28 November–6 January, n = 27); Kruskal–Wallis: χ2= 13·3, d.f. = 2, P = 0·001. Furthermore, territory quality became non-significant in any model of number of attempts when the date of failure or fledging of the final attempt for the season was included in the model.
Age, group size, territory quality and reproductive success
Group size had a dramatic effect on the reproductive success of yearling females, but no detectable effect on the success of older females. Yearling females were more than twice as likely to fledge young in the breeding season if they bred in groups compared with pairs, and there was a trend for females breeding in groups to have higher annual survival. By contrast, older females were equally likely to fledge young in pairs and groups, supporting a previous study by Magrath & Yezerinac (1997). Furthermore, there was a trend for survival to be lower in groups compared with pairs.
The magnitude of variation in the effect of group size on reproductive success in this population approaches that among cooperatively breeding birds as a whole. For example, in summarizing effect of helpers on reproductive success in 19 species, Smith (1990) characterizes the strength of the effect from ‘none’ to ‘extreme’, the latter including bicolored wrens, Campylorhynchus griseus Swainson, and pied kingfishers, Ceryle rudis L. In those two species, a single helper is correlated with an increase of 3·3 times and 2·1 times, respectively, the reproductive success of pairs, similar to the magnitude of the association in yearling female scrubwrens.
Territory quality did partially confound the relationship between group size and reproductive success for yearling females, but only accounted for a small proportion of the difference. Yearling females breeding in pairs were found disproportionately on poor-quality territories, on which seasonal reproductive success was lower, so exaggerating the apparent effect of group size. Overall, yearling females were 3·0 times more likely to fledge young in a season if they bred in groups, while if yearlings in both pairs and groups had been uniformly distributed across the three categories of territory quality, the difference would have been 2·4 times greater. Thus the confounding effect of territory quality contributed about 20% of the observed increase in success of groups. The ranking of ‘territory quality’ used was indirect, so it is possible that a more direct estimate of quality would reveal it to be quantitatively more important. For example, the effect of territory quality would be underestimated if yearlings in pairs were found on poorer-quality territories within categories of territory quality. However, there should still be a large effect of group size, given that the probability of success is higher even for groups on low quality territories compared with pairs on high quality territories.
Territory quality and group size affected reproduction in different ways, again suggesting that finer-scale measurement of quality will not ‘explain’ the group-size effect. While group size affected the success of individual breeding attempts, territory quality affected the number of breeding attempts by allowing birds to continue breeding longer. Prolonged breeding may be possible because better territories were wetter and more densely vegetated, so drying out more slowly with the onset of hot weather in December (mean maximum 26 °C).
Pairwise comparisons of the same female breeding in the same location with the same male still showed that yearlings breeding in pairs tended to have higher success in the following year, while those breeding in groups did not. The ratio of increase in the probability of success was 2·3 (0·58/0·25), which is exactly the magnitude of increase from yearling pairs to older birds if ‘territory quality’ is held constant statistically (0·67/0·29). Thus, although only 12 females were available for this pairwise test and the difference was not quite significant at P = 0·06, the result is that expected with a real effect of group size. This result argues against the possibility that there is unmeasured variation in territory quality that confounds the relationship with group size and that is uncorrelated with the ‘territory quality’ index (e.g. perhaps variation in the risk of predation).
Female quality did not confound the relationship between group size and reproductive success of yearlings. The relationship between group size and success remained for both yearlings that survived until the next breeding season and those that did not survive. Thus differences in female quality, if reflected by survival, did not explain the greater success in groups. Subsequently, among those yearlings that did survive to breed as an older female, there was no difference in the reproductive success according to the group size in which the female bred as a yearling. Thus there was no evidence that yearling females that bred in pairs were of lower quality than those that bred in groups.
Male age, as a potential measure of breeding competence, also did not affect reproductive success. Male age may be unimportant to reproductive success in scrubwrens because males almost never bred as yearlings, and when they did become breeders were likely to have had experience as helpers. Specifically, although males in pairs with yearling females were younger than those in groups, none the less 95% were at least 2 years old, and 66% at least 3 years old.
It is difficult to assess the importance of the apparent effects of group size on survival with current sample sizes, as is often the case in studies of cooperative breeders. There was no significant effect of group size for either yearlings alone or older females alone, but there was a weak statistical interaction suggesting that yearlings survive relatively better in groups. Regardless of the true magnitude of effects, these data reinforce the results of reproductive success showing that yearling females benefit from breeding in groups, while older females do not. Thus the true benefit of breeding in groups for yearlings may be underestimated from seasonal reproductive success alone.
Why do only yearlings benefit from grouping?
A striking feature of the results on scrubwrens is that, although yearlings are much more successful when breeding in groups, older females do not benefit at all. How could breeding in a group provide a major benefit to yearlings but no benefit to older females? There are three types of explanation relating to: (1) female competence at breeding, (2) behaviour of males and (3) female reproductive effort. Analyses of breeder quality (Results) show that the difference is not due to differences in female quality, another general explanation of age effects in birds (Forslund & Pärt 1995). The first explanation is that yearlings may be less competent at breeding, and so benefit from being in groups in which older females would not benefit. In this case, the number and behaviour of males may be identical, but the benefit nonetheless differs. For example, if yearlings are poor at detecting or identifying potential predators, then they would benefit from being warned by others that a predator is near. By contrast, an older female would not benefit by being informed of a predator that she had already identified. Secondly, males may behave differently when breeding with yearling females than with older females. In this case it is not necessarily true that yearlings are less competent at breeding. For example, males might be extra-vigilant when breeding with yearlings, or work harder at bringing food to the nest. Thirdly, it is possible that the reduced performance of yearlings in pairs is due to a reduced optimal reproductive effort for these birds. These hypotheses are not mutually exclusive, but each is considered in turn.
Yearlings do appear to be less competent than older birds, and are compensated by breeding in groups. Yearlings start the breeding season more than 4 weeks after older females, regardless of group size. This is consistent with yearlings being less competent at skills relevant to breeding than older birds, as has been found in many other species (Forslund & Pärt 1995). For example, blackbird Turdus merula L. yearlings are less good at foraging than older females, start breeding later and therefore have lower seasonal reproductive success (Desrochers 1992a). However, if they are provided with a food supplement, they start breeding at the same time as older females and have similar reproductive success (Desrochers 1992b). Furthermore, the effect of group size in scrubwrens arises primarily because yearling females do exceptionally badly, not because yearlings in groups do exceptionally well. This suggests that being in a group compensates a yearling for her lack of skill.
Subordinate males do behave differently towards yearling females, on average, than towards older females, but this cannot explain why only yearlings benefit. Subordinate males are more likely to provision nestlings if they are unrelated to the breeding female (Magrath & Whittingham 1997), and yearling females are all immigrants into breeding groups and so are unrelated to the subordinates. By contrast, about 40% of the older females in the sample in this paper are in groups with sons (3/40 of unknown relatedness), who only provision nestlings in about 50% of cases. However, whether a subordinate ‘helped’ or not had no effect on the seasonal reproductive success of older females (Magrath & Yezerinac 1997), and so this does not explain why only yearling females benefit from breeding in groups.
Finally, it is possible that yearling females in pairs have a lower optimal reproductive effort than those in groups, but this is at best a partial explanation. Yearling females in pairs might be selected to put little effort into reproduction, perhaps starting breeding at a later date, if high effort jeopardizes reproductive success in future (Forslund & Pärt 1995). The trend for lower survival of yearling females in pairs compared to groups is inconsistent with this hypothesis, but it is possible that lower effort emphasizes a pre-existing effect of competence and group size. For example, general incompetence at breeding may mean that a yearling female in a pair will have relatively poor success compared to later years when she is more competent, so it may pay to cut losses as a yearling to improve the probability of breeding in future (Forslund & Pärt 1995). Higher effort might lead to an even lower probability of survival.
Importance of age effects for cooperative breeders
The finding that yearling females gain a large benefit from breeding in groups, while older females do not benefit at all, has two major implications for the evolution of cooperative breeding in scrubwrens and other species.
First, group size can have a substantial effect on the lifetime reproductive success of females even if there is no effect on reproductive success for most females in most years. On average in the scrubwren population, 75% of females were older birds who do not benefit from being in groups. To assess the importance of reproductive success in a female’s first year to lifetime success, the following values were used: (1) a constant female annual survival of 75% for females once they become breeders; (2) a probability of producing any fledglings of 0·29 for yearlings in pairs and 0·69 for yearlings in groups and older females; and (3) a seasonal production of 3·4 fledglings if they are successful. All values are means from the scrubwren population, with probability of producing fledglings controlled for territory quality. Given these assumptions, the expected lifetime reproductive success is 15% less for females that bred as yearlings in a pair (8·0 fledglings) compared with a group (9·4 fledglings).
The effect of group size on lifetime success of females is sensitive to estimates of annual survival, but the effect of group size on yearlings is always potentially important. To assess the importance of estimates of survival, observed values were also used from the scrubwren population of 70% for yearlings in pairs, 85% for yearlings in groups, 83% for older females in pairs, and 67% for older females in groups. In this case, females have an expected lifetime success of 10·6 fledglings breeding only in pairs but only 8·4 breeding in groups, reflecting the lower survival of older females in groups. None the less, in each case, females always do better if breeding in groups rather than pairs as yearlings. Breeding in a pair as a yearling, but otherwise in groups, would result in 6·0 fledglings in a lifetime (29% lower than the 8·4 for groups alone); breeding in group as a yearling, but otherwise in pairs, would result in 14·1 fledglings (25% higher than the 10·6 for pairs alone).
Overall, scrubwren females benefit by breeding in groups as yearlings assuming this has no effect on group size in later years, but may do better if breeding exclusively in pairs compared to groups and possibly best of all breeding first in a group, and subsequently in pairs. Precise estimates of lifetime group-size effects require complete measures of social life-history combined with more precise estimates of annual survival.
The second major implication for cooperative breeding is that the sexes could differ in the benefits of producing philopatric sons. While scrubwren yearling females benefit by joining groups, through increased seasonal reproductive success and possibly higher survival, they do not benefit from philopatric sons in later years, because by then females will be older and will not benefit from breeding in larger groups. There might even be a cost of retaining sons, if their presence reduces female survival. A reduction in female survival in the presence of sons, if confirmed with more data, could potentially be offset by an increase in indirect fitness from later successful reproduction of those sons. In other words, any reduced survival might be due to the cost of ‘parental facilitation’ of offspring success (Brown & Brown 1984).
In contrast to females, breeding males could benefit from philopatric sons. The seasonal reproductive success of a group is not affected by the female’s age, whereas a pair’s success is much lower if the female is a yearling. Thus, from the perspective of a dominant male, philopatric sons act as ‘insurance’ against breeding in future with yearling females. The benefit to males of being in groups will then depend in part on the mean number of yearlings a male breeds with in his lifetime. Measuring the benefit precisely, however, would also require knowing how much paternity is gained by sons and extra-group males in social groups with yearlings and older females. Based on a small sample which does not allow analysis by female age, beta males sired about 20% of young when in groups with their fathers and unrelated females, but paternity by extra-group males was less frequent in groups than pairs (Whittingham et al. 1997).
Overall, the relationship between age, group size and reproductive success is potentially of widespread importance to cooperatively breeding birds, but has rarely been the focus of study. An effect of group size on reproductive success or survival early in life may be common and will have disproportionate importance to lifetime reproductive success. Scrubwrens have a similar high adult survival to many cooperative breeders, yet the difference in reproductive success between yearling pairs and groups could account for 15–29% of expected lifetime reproductive success of recruits.
Variation in group-size effects in birds
The stronger effect of group size on yearling compared with older female scrubwrens supports the general model that group size will have greater effects in poorer conditions for breeding (Fig. 1a). In addition to this comparison, two of three comparisons involving territory quality were supportive. The ratio of increase was greater for yearling females on poor quality territories (4·5) compared with medium (1·9) or high (2·3) quality territories; the one inconsistency was that high quality territories had a slightly higher ratio than medium quality territories.
To test whether it is generally true that group size has a greater effect in poor conditions, similar values were calculated for other cooperatively breeding birds in which comparisons could be made of ‘good’ and ‘poor’ conditions. To gain a representative sample, I attempted to include all species represented in Stacey & Koenig (1990). Data presented in that book, if available, were used, otherwise publications were searched on those species appearing before or after the book. Two other species were also included that were studied more recently. White-winged choughs, Corcorax melanorhamphos Vieillot, were included because they were the subjects of a food-supplementation experiment (Boland, Heinsohn & Cockburn 1997); Seychelles warblers were included because of unusually detailed measurement of territory quality (Komdeur 1994). Reproductive success of pairs and groups was calculated in poor conditions and good conditions. Depending on available data, ‘poor conditions’ could mean young, inexperienced breeders, subordinate breeders in plural-breeding species, sites or territories known to be poor, years with low prey density, or a series of years during which reproductive performance was below average. ‘Good conditions’ were the opposite. In some cases, the authors provided the required data, but in most cases values were calculated from tables, figures or other data in the text. In some species, it was possible to make comparisons using more than one classification into ‘poor’ and ‘good’ conditions; for example, using both female breeding experience and, separately, poor and good territories. In such cases the complete data are presented, but species’ means are also calculated.
Most species showed a greater effect of group size in poor conditions, as predicted (Table 2). On average, groups produced 2·3 times as many young as pairs in poor conditions, compared with 1·3 times as many in good conditions. Furthermore, all 11 species showed a greater ratio in poor compared with good conditions, as shown by the ‘ratio of change (worse/better)’ – values all being greater than 1·0 in Table 2 (one-tailed binomial probability < 0·001).
Table 2. Effect of group size on reproductive success under ‘worse’ and ‘better’ conditions for breeding
The small mean effect of group size in good conditions means that studies under those conditions would require large samples to detect a real difference. There may even be no benefit at all: 5/11 species showed no increase in success of groups over pairs for at least one comparison in ‘good’ circumstances and another (the splendid fairy-wren, Malurus splendens, Quoy & Gaimard), revealed an effect so small (1·1 increase) it was not statistically significant even after a long-term study (Russell & Rowley 1993). Furthermore, given that confounding variables are likely to overestimate group-size effects (Introduction), the real benefit of breeding in groups in good conditions may be even lower. It is therefore not surprising that many studies have not detected group-size effects.
A greater ratio of effect in poor conditions does not necessarily mean that there is a greater absolute (incremental) effect of groups in poor conditions compared with good conditions. For example, if pairs can produce one fledgling in poor conditions but two in good conditions, and groups result in one extra fledgling in all conditions, the ratio will be 2·0 in poor conditions and 1·5 in good conditions (also see Fig. 1c). To address this issue, the incremental increase was also examined in groups compared to pairs (Table 2, ‘Increment’ columns). In this case 9/11 species showed a greater increment in poor conditions (positive values in the ‘Incremental change’ column of Table 2), one showed no change, and one showed a decreased increment in poor conditions (one-tailed binomial probability = 0·01).
In support of the usefulness of focusing on younger breeders to explore the effects of breeding in ‘poorer’ conditions, Table 2 suggests that there are similar effects regardless of whether ‘poor’ conditions are defined by environmental conditions or breeder attributes.
It is concluded that groups usually do both relatively and incrementally better in poorer conditions for breeding within a species, supporting the generality of the data from scrubwrens. Given that evolutionary ‘fitness’ is a relative measure, a greater ratio of increase in poor compared to good conditions is important, even if there is no incremental difference. The overall benefit to breeding in groups would depend on how frequently birds breed in different conditions.
The conclusion from this comparison of group-size effects within species is consistent with that from Hatchwell’s (1999) comparison among species. Hatchwell found that species in which brood reduction is more common (among individuals breeding in pairs) are more likely to show a higher reproductive success in groups compared to pairs. This result suggests that group size has a greater effect on reproductive success in species in which food is more likely to be scarce. However, Hatchwell emphasizes that his study was designed to look at risk of starvation and feeding rates, not at reproductive consequences, and the effect on reproductive success could be an artefact of using brood reduction in pairs as an estimate of environmental conditions. It is also unclear whether the differences in brood reduction and the group-size effects represent values typical of species or simply the conditions prevailing in specific studies.
Variability in group-size effects within species has broad implications for the study of cooperative breeding in birds. It is clearly inappropriate to accept or reject hypotheses about the evolution of cooperative breeding on the basis that there was or was not an effect of group size on reproductive success in a particular study. For example, helping to raise collateral kin might be important in the evolution or maintenance of cooperative breeding even if a particular study found ‘no effect’ of group size on reproductive success. The lack of effect might be specific to the conditions under which the study was conducted, rather than being typical of the species. Conversely, a positive effect of group size in a particular study does not necessarily mean that this is typical of the species. Both points are illustrated by Reyer’s (1990) study of pied kingfishers, which found a strong effect of group size at one site, but none at another. A geographically limited study would at best have gained a limited insight into cooperative breeding in that species. Comparisons within populations at single sites can also give insight into variability in group-size effects, and may be more practicable. Variation among years and among breeders of different age reveal the same pattern of a greater group-size effect in poorer conditions for breeding.
Variation in group-size effects within species also raises the issue of helper flexibility and how individuals assess fitness benefits and costs of helping. Individuals could estimate the benefits of helping most accurately if they were able to assess their potential effect on the group’s reproductive success in specific circumstances, in addition to their relatedness to breeders. A likely problem in interpreting the causes of helper flexibility is that the benefits and costs of helping could covary with conditions for breeding (Heinsohn & Legge 1999). For example, poor environmental conditions could raise both the costs and benefits of helping. Focusing on breeder age may help isolate variation in the benefits of helping, rather than the costs, since contributing a constant effort can have larger effects for some breeders than others.
The work presented here would not have been possible without the help and collaboration of many people. Beth Bobroff, Janet Gardner, Tony Giannasca, Ashley Leedman, Anjeli Nathan, James Nicholls and Linda Whittingham made substantial contributions to fieldwork in more than one year, and Camille Crowley, Megan MacKenzie, Helen Osmond, Amy Rogers, Derek Smith, Lynda Sharpe, Kate Trumper and Stephen Yezerinac also made valuable contributions. Sam Portelli and Belinda Mitterdorfer helped with data entry. The paper was largely written while on sabbatical with Jamie Smith, and I thank him and all of those in the Department of Zoology at UBC who provided a stimulating environment in which to work. Andrew Cockburn, Rob Heinsohn, Elsie Krebs, David Green, Jamie Smith, Liana Zanette and two anonymous referees provided insightful comments on earlier versions of the manuscript. This research was supported by grants from the Australian Research Council, and was carried out under permits from the Australian National University ethics committee, the Australian Bird and Bat Banding Scheme, the Australian National Botanic Gardens and Environment ACT.
The paper is dedicated to the memory of Anjeli Nathan, who worked on the scrubwrens during two field seasons and died in a car accident in South Africa in 1999, aged 24. In addition to the measurable contribution of gathering data, she contributed immeasurably to the project through her enthusiasm and dedication.
Received 3 June 2000; revision received 3 October 2000