The e ect of helpers on reproductive success in the laughing kookaburra

Ecology 2000, 69, The e ect of helpers on reproductive success in the laughing kookaburra SARAH LEGGE Evolutionary Ecology Group, Botany and Zoology, Australian National University, Anu, Canberra, ACT 0200, Australia Summary 1. In co-operatively breeding birds and mammals that live in family groups, helpers may gain indirect tness bene ts by increasing the number of breeding attempts in a season, by increasing the success of each nesting attempt, and by increasing the survivorship of related breeders. 2. The e ects of helpers were examined in an analysis of reproductive success in the laughing kookaburra (Dacelo novaeguineae), where a monogamously breeding pair is assisted by o spring of both sexes. 3. Helpers could not increase the number of attempts in a season, because kookaburras are single-brooded. 4. In a paired analysis controlling for pair and territory quality, increases in group size were not matched by increases in the number or weight of edglings in each breeding attempt, even though brood reduction was the major source of productivity loss. This concurs with previous results showing that group size does not a ect overall provisioning levels to broods, because all group members reduced their provisioning e ort as group size increased. 5. Because kookaburras prefer to reduce workloads during breeding rather than raise larger broods, the costs of caring for young must be high. Thus breeders with helpers may enjoy energetic savings and enhanced survivorship. 6. Helpers had di erent e ects on nest success depending on their sex. Whereas male helpers had a neutral e ect on edging success and edgling weight, additional female helpers reduced edging success. Females are relatively poor helpers, and other group members may not properly compensate for the female's `inadequate' help. 7. High-quality kookaburra pairs, or pairs on good territories, were more likely to be in larger groups because they produced more potential recruits for their groups than relatively `poor' performers. This was demonstrated by comparing the results of the paired comparisons with those of a correlational analysis that showed that in groups of two to six birds, group size was positively correlated with edging success and edgling weight (which a ects the probability of juvenile survival). 8. However, unusually large groups of seven or more birds su ered drastically reduced nest success because eggs were damaged during incubation. Key-words: co-operative breeding, costs of helping, king sher, kin selection, philopatry. Ecology (2000) 69, Introduction The search for the adaptive basis to co-operative breeding has generated a long list of potential bene- Correspondence: Sarah Legge (fax (Australia) 26249 5573; e-mail Sarah.Legge@anu.edu.au). ts to both breeders and helpers (review by Cockburn 1998). Most co-operatively breeding birds live in groups of relatives (Brown 1987) and this has strongly encouraged the idea that kin selection (Hamilton 1964) is a major selective force in the evolution of co-operative breeding (Emlen 1997). Helpers could boost the production of non-descen-

715 S. Legge dent kin by increasing the breeder's productivity in each breeding attempt (Brown et al. 1982; Mumme 1992) or by increasing the number of breeding attempts in a season (Brown & Brown 1981; Innes & Johnston 1996). In addition, they could increase the survivorship and reproductive life span of breeders by reducing the latter's workload in each breeding attempt (Mumme, Koenig & Ratnieks 1989; Crick 1992). However, the evidence that helpers increase group productivity is inconsistent. Some correlational studies have shown strong positive e ects (Lewis 1981; Emlen & Wrege 1991; Heinsohn 1992) but others have shown that helpers have no discernible e ect on productivity (Gaston 1978a; Monadjem, Owen- Smith & Kemp 1995). More importantly, positive correlations may be driven by the confounding e ects of pair and territory quality (Brown 1987; Cockburn 1998). Enhanced reproductive performance by high quality, older or experienced birds, or pairs on premium territories, may result in larger groups. Continued high performance can then be wrongly attributed to the e ects of group size. Confounding e ects have been controlled for by monitoring productivity after experimental removal of helpers (Brown et al. 1982; Leonard, Horn & Eden 1989; Mumme 1992). Less disruptive approaches include manipulating the postulated source of mortality that helpers alleviate (Austad & Rabenold 1985; Boland, Heinsohn & Cockburn 1997) or comparing the productivity of the same groups at di erent group sizes (Green et al. 1995), or groups of the same size but with di erent levels of help (Magrath & Yezerinac 1997). In some of these studies helpers did indeed increase productivity through alloparental care (Mumme 1992; Boland et al. 1997), but in others they had no apparent e ect (Leonard et al. 1989; Magrath & Yezerinac 1997). Group living can have advantages for the reproduction or survival of individuals (Ekman, Bylin & TegelstroÈ m 1999) irrespective of whether group members are related or not. For example, groups may perform better in territorial contests (Craig 1984), be more e ective in predator detection (Woolfenden & Fitzpatrick 1990) or cope more e ectively with ecological constraints, such as thermoregulation in cold environments (Sydeman, Guntert & Balda 1988; Ligon & Ligon 1990). Groups may also be good forums for nding mates (Reyer, Dittami & Hall 1986; Sherley 1989). Young birds in particular may reap the rewards from a `supportive' environment, allowing them to learn foraging and breeding skills under the wing of extended parental care (Brown 1987; Heinsohn 1991). Even if they are secondarily evolved to some extent, parental facilitation and other bene ts of group living could be powerful selective forces maintaining co-operative breeding. The e ects of helpers on reproductive success are examined here in the laughing kookaburra (Dacelo novaeguineae, Hermann), a large terrestrial king- sher endemic to eastern Australia. Laughing kookaburras live in co-operatively breeding groups of two to eight birds. A socially monogamous and dominant pair is assisted by helpers of either sex that are o spring from previous breeding attempts (Legge & Cockburn 2000). All helpers participate with all aspects of clutch, nestling and edgling care, even developing brood patches (Legge, 2000a). They also vigorously help to defend the group territory, which is maintained all year (Parry 1973; Reyer & Schmidl 1988; S. Legge, unpublished data). Territorial inheritance, a common feature of other co-operative breeders, is rare or absent, and helpers gain extremely few direct mating bene ts (Legge & Cockburn 2000). Thus it is pertinent to ask whether helpers gain indirect tness bene ts by in uencing the production of non-descendent kin. As kookaburras are single-brooded, helpers cannot increase the number of nesting attempts in a season. However, brood reduction is common, therefore additional helpers may increase the success of each breeding attempt. In addition, their presence and contribution during breeding may enhance the survival and lifetime reproductive success of related breeders. Methods The study was carried out between December 1994 and February 1998 in 20 km 2 of eucalypt woodland near Canberra, south-east Australia (Mt Majura, Mt Ainslie, O'Connor ridge, Aranda Bushland and Black Mountain, including the Australian National Botanic Gardens). This period included years with very di erent weather conditions. The annual mean rainfall for 1988±98 was 655 mm (Bureau of Meteorology, Canberra, Australia). 1994 was characterized by the coldest winter in 150 years, followed by a severe drought during the breeding season. 1997 was also a dry year (411 mm), but 1995 was a particularly wet year (801 mm). About 130 birds in 35 groups lived in the study area each year. Group territorial displays marked the position of each group's territory boundaries (Parry 1973; Reyer & Schmidl 1988). Birds were identi ed from coloured aluminium bands, patagial tags, or distinctive plumage markings that were retained between moults. Group composition and the relationships between birds in this population were known in detail from long-term monitoring and DNA ngerprinting (Legge & Cockburn 2000). Although both sexes help, male helpers predominate (68±82% in any one year) because females disperse at an earlier age. Helpers disperse to breeding positions, but never to helping positions in other groups. Dominant birds retain their status for many years, so in most groups all helpers were the o spring of

716 Reproductive success in kookaburras the dominant pair. However, step-families form through death and replacement of one of the breeders, usually the female, making it possible to compare the productivity of groups where the relatedness of helpers to the brood varies. The tenure of individuals within groups was monitored with weekly nest watches during nesting, then fortnightly censuses for 2 months after edging, and opportunistically throughout the rest of the year. Kookaburras were normally found sitting in old gum trees. The exact age of some dominant birds was known because they had been banded as edglings, but most dominants were assigned a minimum age based on the number of o spring helpers (known from DNA ngerprinting) in their group when rst encountered. Birds that immigrated into a dominant position during the study could be aged dichotomously as 1 year old or 2, based on tail feather shape. Birds were sexed using a molecular technique (Gri ths et al. 1998). The few uncaught birds (less than 15%) were assigned a sex based on behaviour (e.g. the mate of a known-sex bird) or plumage characteristics (Higgins 1999). Clutches were laid in the austral spring (September±December) in unprepared hollows 4±12 m from the ground, mainly in eucalypts (Eucalyptus rossii, R.T. Baker & H.G. Smith, E. mannifera, Mudie, E. blakleyii, Maiden, E. melliodora, A. Cunn ex Schauer). Hollows form as a result of termite action, branch fall and wind shear. Kookaburras normally lay three eggs (range 1±4) at 1±4 day intervals. Incubation begins before the clutch is complete and lasts 24±26 days. Eggs hatch asynchronously at 2- to 72-h intervals. Nestlings edge at about 35 days, and forage independently after a further 8 weeks (Parry 1973; S. Legge, unpublished data). Nests were usually found by ushing incubating birds, and reached using single-rope techniques. The laying date was estimated by candling eggs. Nests were visited every 1 or 2 days as required near the estimated hatching date until all eggs had hatched. Thereafter the progress of nests was checked weekly. Unhatched or missing eggs were noted, as were nestling disappearances. Fledging success and edging weight were taken as the number and weight of nestlings present on the nal nest visit at 32 days, when nestlings were also banded. In 16 group-years no clutch was found; although it was suspected that eggs were never laid, a clutch may have been initiated but quickly lost. These data were omitted from the analysis of clutch sizes and hatching success, but retained in the analysis of edging success. Failure to lay eggs is as relevant as egg or nestling loss in determining the number of young edged in a season, and the decision to lay or not may be in uenced by the presence of helpers. In 126 groupyears, ve groups had a second breeding attempt in a single season after the rst attempt failed, making 131 attempts available for analysis of edging success. Data on clutch size and hatching success were fewer (89 attempts in 84 group-years) because most nests from 1994 were found after hatching, some nests in later years were inaccessible, and because the 16 group-years with `no clutch' were excluded. Breeding in 43 of the 126 group-years (341%) was by unassisted pairs. The remainder involved groups with one to ve helpers. The median group size was three. ANALYSIS Using Genstat 5 (release 4.1; Genstat Committee 1993), a statistical modelling approach was used to consider environmental as well as demographic variables that might a ect reproductive success. The analysis was framed as three questions. First, what a ects hatching and edging success? Second, as an alternative to numbers of young, what a ects edgling weight? and third, what in uences edgling survival to 2 months, when it is nutritionally independent? In all these questions, the e ect of helpers was of particular interest, but other social as well as environmental variables were considered in the models. Generally, there was potential for dependency within the data, as groups were represented one to four times, and nests contained one to three edglings. To account for repeated sampling with an unbalanced design, mixed models were used initially, incorporating random (`group' and `nest' where appropriate) as well as xed e ects. Variance components were estimated using the residual maximum likelihood (REML) procedure in Genstat, and xed e ects from weighted least squares. However, models were simpli ed by omitting random e ects if there was no evidence of dependency. This was assessed by examining the change in deviance, which approximates to a chi-square distribution, when the random e ect was dropped from a full model. Speci- c details on the analysis of nest success, edgling weight and edgling survival are given separately below. Hatching and edging success There was no evidence of intragroup dependence between years, as dropping the random e ect `group' from the initial linear mixed model resulted in a non-signi cant change in deviance. Consequently, because the responses for hatching and edging success were discrete with a limited number of values (0, 1, 2, 3), they were treated as ordinal categorical responses in a proportional odds regression (McCullagh & Nelder 1989; Genstat Committee 1993). These models are extensions of generalized linear models using a logit link function, and describe the cumulative probability of the response

717 S. Legge being in each ordinal category given the signi cant covariates. The statistical signi cance of terms was assessed using the change in deviance. Fledgling weight Kookaburras exhibit reverse sexual dimorphism, which is evident even at edging (65%; S. Legge, unpublished data). Each edgling's weight was divided by the average weight for its sex and this corrected `weight ratio' was used as the response variable in a linear mixed model, with `group' and `nest' speci ed as the random e ects. To assess the signi cance of xed e ects, the deviance explained by a full model was contrasted with the deviance of a submodel with the same variance components but excluding the variable of interest. Fledgling survival A generalized linear mixed model with a binomial distribution was used to model the probability that a edgling would survive to independence. The variance components for `group' and `nest' were again estimated with the REML procedure; the signi cance of xed e ects was assessed using the Wald statistic when the e ect of interest was last in the model (Genstat Committee 1993). VARIABLES OF INTEREST The following variables were considered in all models: year, hatch date, family type, pair duration, the age of the breeding female, group size, and hatching sex ratio (% male). Groups were de ned as full families if all helpers were full o spring of the breeding pair, and step-families if one (or more) helper(s) was unrelated to one of the breeders; 14/ 126 group-years involved step-families. Pair duration had two levels, `novice' or `experienced', depending whether the pair had prior breeding experience together. Because the birds in a breeding pair were usually of similar age, only the age of the female was used in analysis. Group size was considered as a dichotomous factor (pair or group) and as a multilevel categorical variable because the e ect of additional helpers may be not be additive in a simple way. The models for edgling weight and edgling survival also considered the edging success of the nest, the edging sex ratio, and the sex of the individual edgling. Finally, the weight ratio of each edgling was considered in the model for edgling survival. Interactions were tested where pertinent. For example, helpers may only have a positive e ect in years when conditions are poor, and the weight ratios of male and female nestlings may be a ected di erently by variation in group size. PAIRED COMPARISONS The modelling analysis indicated that helpers had a positive e ect on edging success and edgling weight. To check whether this resulted from the confounding e ects of pair and/or territory quality, a restricted subset of the data was used for withingroup paired comparisons (Wilcoxon's paired-sample by rank test). Fledging success and the mean edgling weight ratio for the brood were compared for the same group (i.e. same breeding pair on the same territory) when the group size had increased by the addition of one, two or three helpers. Each group was only used in one comparison. To separate the e ects of male and female helpers, the success of groups was compared over two breeding attempts when the number of helpers of one sex had increased, the number of the other sex having remained constant. Again, this included increases of one, two or three males (or females). Comparisons went forwards and backwards in time, thus removing the potentially confounding e ects of year (e.g. one group might be compared in 1996 vs. 1995, another group in 1994 vs. 1997, etc.). Results Kookaburras in the study area occupied large territories (mean 69 ha, range 16±224 ha, 30 territories from 1997). Territory size was positively correlated with group size (F 1,29 ˆ 204, P <0001, r 2 ˆ 08; Fig. 1). Hatching began between the 11 and 16 October each year, and 75% of clutches had hatched within 37, 31 and 17 days of the rst clutch hatching in 1995, 1996 and 1997, respectively. The modal clutch size was three eggs (752% of 89 clutches); 165% of the 254 eggs did not hatch, mainly because of infertility or developmental failure (23 eggs in 20 clutches). Other causes of egg loss included egg breakage (eight in four clutches) and egg tossing by unknown individuals ( ve in three clutches were found on the ground under the nest). Loss of eggs to predators was rare: six eggs (23%) in two clutches from the same group suddenly disappeared, implying predation; there were three birds in this group. Just over half of all eggs resulted in a edgling (Table 1). The largest loss in potential productivity occurred during the nestling phase: although 165% (42/254) of eggs failed to hatch, 321% (68/212) of hatchlings failed to edge. Loss of young also occurred between the time young edged until nutritional independence at 2 months: 20% (24/125) of edglings died in this period. Nestling losses were attributed to predation (healthy nestling suddenly disappeared, 3/68, 44%), disease (nestling sickly on previous visit, 6/68, 88%), accident (nest caved in, branch broke, 7/68,

718 Reproductive success in kookaburras Fig. 1. Territory size (in hectares) against group size measured in September 1997. 103%), abandonment (brood found dead in nest, 15/68, 221%) or brood reduction (nestling underweight or injured from sibling aggression on previous visit, 37/68, 544%). In a subset of 72 nests where at least two eggs hatched, intensive observations revealed that 30 nests (417%) experienced brood reduction due either to sibling aggression (21/ 72 nests, 292%) or starvation (12/63 nests, 19%; S. Legge 2000b). Young nestlings were extremely aggressive to each other, often continuing to ght after being brought down to the ground temporarily to be weighed. A few nests experienced both early siblicide and late starvation. Fig. 2. Summary of the relationship between group size and (a) clutch size, (b) hatching success and (c) edging success. Bars represent means and SE of data, sample sizes are given above each bar. HATCHING AND FLEDGING SUCCESS Proportional odds regression showed that hatching success was signi cantly a ected by group size (89 nests; w 52 ˆ 121, P <005; Fig. 2b). Groups of two to six birds were most likely to hatch three nestlings and very unlikely to hatch none. However, there was a sharp reduction in hatching success for groups of seven or more, which were much more likely to experience complete hatching failure, and very unlikely to hatch three nestlings (Fig. 3a). This was not explained by di erences in clutch size, which varied little with group size (Fig. 2a). The sample of groups of seven was small (four groups), but hatch failure in each case was caused by damage to eggs. Of 10 eggs hatched to groups of seven or more, six were dented or broken. In contrast, in groups of six or less, only two eggs out of 244 failed to hatch because of damage (contingency w 12 ˆ 344, P < 0001). Year, hatch date, family type, pair duration, Table 1. Distribution of edging and hatching success for 89 clutches, for which the clutch size, hatching success and edging success were all known Distribution of clutch, nestling and edging sizes 4 3 2 1 0 Total eggs, nestlings, edglings % surviving from eggs Clutch size 3 67 16 1 254 Number at hatching 2 52 20 8 7 212 835 Number at edging 0 26 24 18 21 144 567 679 % surviving from nestlings

719 S. Legge Fig. 3. Model predictions for (a) the probability of hatching 0, 1, 2 or 3 nestlings, and (b) the probability of edging 0, 1, 2 or 3 young. Hatching and edging probabilities are given for groups of di erent size. age of the dominant female, the hatching sex ratio and interactions were all non-signi cant. Fledging success was also signi cantly a ected by group size (131 nests; w 52 ˆ 117, P <005; Fig. 2c). As group size increased from two to six birds, groups were less likely to experience complete edging failure, and more likely to edge three young (Fig. 3b). Once again, the largest groups of seven or more birds were least likely to edge young, re ecting their poor hatching success. Year, hatch date, family type, pair duration, age of the dominant female, the hatching sex ratio and interactions were non-signi cant. FLEDGLING WEIGHT The weight ratio of a edgling decreased with the number of siblings that edged together (155 edglings from 76 nests; w 22 ˆ 85, P <0025), indicating a trade-o between numbers of young and their mass (Fig. 4a). The weight ratio of edglings from unassisted pairs was lower than for groups of three or more birds (w 12 ˆ 171, P <0001). There was a non-signi cant trend for this e ect to be more pronounced for female edglings (interaction between edgling sex and group type: w 12 ˆ 27, P <01; Fig. 4b). Year, hatch date, family type, pair duration, age of the female breeder, hatching and edging sex ratios and interactions were non-signi cant. FLEDGLING SURVIVAL Heavier edglings had a higher chance of surviving to independence at 2 months (111 edglings; logistic regression w 12 ˆ 81, P<0005). The model predicted that 83% of edglings of average weight for their

720 Reproductive success in kookaburras Fig. 4. Fledgling weight ratio (weight of edgling divided by the mean weight for its sex) (a) against number of edglings in the brood and (b) of males and females in unassisted pairs vs. groups. sex (i.e. weight ratio ˆ 1) would survive to independence. A 10% reduction in weight ratio reduced survival probability to 72%, and a 20% weight ratio reduction reduced survival probability to 58%. In contrast, a 20% increase in weight ratio improved the probability of survival by 11%. The data suggested that females were more prone to mortality than males: of 61 male edglings with reliable survival data, eight died (131%). In contrast, 16/64 female edglings died (25%). (Note this sample is slightly larger than that used in the model of weight ratio because it includes edglings for which no weight ratio was available). However, neither edgling sex nor its interaction with weight ratio were signi cant in the model. Year, hatch date, family type, pair experience, age of the female breeder, hatching and edging sex ratios and interactions were all non-signi cant. PAIRED ANALYSIS ± DO HELPERS REALLY HELP? Paired comparisons showed that within groups, group size did not a ect edging success (Fig. 5a; Wilcoxon's paired sample by rank test T ˆ 255, P <018, n ˆ 24 pairs; 13/24 forward in time) nor the mean weight ratio of edglings (T ˆ 5, P <082, n ˆ 16 pairs; 9/16 forward in time). In comparisons Fig. 5. Di erences in the number of young edged when (a) group size increases by one to three helpers, (b) male helpers increase by one to three (number of female helpers unchanged), and (c) female helpers increase by one to three (number of male helpers unchanged). Symmetrical distributions around the dotted line indicate no consistent change in the number of young edged. within groups where the number of female helpers remained constant but male helpers varied, an increase in male helpers also had no e ect on edging success (Fig. 5b; T ˆ 15, P <089, n ˆ 16 pairs; 8/16 forward in time) nor mean weight ratio of edglings (T ˆ 115, P <02, n ˆ 9 pairs; 5/9 forward in time). In contrast, if the number of male helpers in a group remained constant but the number of female helpers increased, edging success signi cantly decreased (Fig. 5c; T ˆ 105, P <003, n ˆ 7 pairs; 3/7 forward in time). Although the available sample for this last test was small, edging success decreased for ve of the seven groups when the number of females increased, and the remaining two groups showed no change. Note that groups of seven birds, which showed markedly reduced reproductive success, were not included in any of the paired comparisons. The negative e ect of female helpers appeared to operate during the nestling

721 S. Legge phase rather than at the egg stage, as an increase in female helpers did not have any e ect on the hatching success of a group (T ˆ 3, P <05, n ˆ 6 pairs; 3/6 forward in time). Data were insu cient to test whether female helpers also had a detrimental e ect on edgling weight ratios. Discussion The data presented here reveal four interesting results. First, the paired analysis showed that between the usual group sizes of two to six birds, additional helpers did not improve the success of nesting attempts. Secondly, helpers had di erent e ects on nest success depending on their sex. Male helpers had a neutral e ect, whereas female helpers actually reduced edging success. Thirdly, unusually large kookaburra groups su ered drastically reduced nest success. Finally, the quality of the breeding pair and/or their territory clearly had a strong e ect on the reproductive success of their group, generating a non-causal correlation between group size and edging success, edging weight and, by extension, edgling survival. These four points are discussed in turn. GROUP SIZE DOES NOT AFFECT NEST SUCCESS The productivity of kookaburra nests is strongly in uenced by food availability. Nearly half of all nests experienced brood reduction, losing at least one nestling. Sometimes nestlings were lost sequentially, youngest to eldest, until the entire attempt failed. Young that edged below average weight were more likely to die before reaching independence, and edgling weight decreased with the number of siblings, indicating that kookaburras have trouble raising a full brood. Also, the extra cost of raising a female (the larger sex) is su ered disproportionately by `poorer' groups (pairs). Besides alleviating brood reduction, kookaburra helpers lack other avenues to increase nest success. Helpers in some co-operatively breeding birds reduce predation of eggs and nestlings (Austad & Rabenold 1985; Woolfenden & Fitzpatrick 1990), but like many hole-nesters (Emlen & Wrege 1991) nest-predation is rare in kookaburras. The prevalence of brood reduction in kookaburras suggested that extra help feeding young would translate into higher success. Yet additional helpers were not associated with greater nest success. This result concurs with an analysis of feeding contributions presented elsewhere, showing that group size did not a ect the overall amount of food delivered to the nest. In larger groups, breeders reduced their own contribution to compensate completely for the e orts of their helpers (Legge 2000a). Why do kookaburras miss this opportunity to boost productivity? There are two possibilities. Helpers may be ine ectual at preventing early siblicide, which accounts for three-quarters of all the nestlings lost through brood reduction. Early siblicide occurs soon after hatching, well before food demands peak, and is heavily in uenced by, for example, the degree of hatch asynchrony and nestling sex (Legge 2000b). Thus competitive interactions between nestlings may be more important than group size in determining nestling loss (Innes & Johnston 1996). However, hatching asynchrony and other factors a ecting the outcome of sibling rivalry, including group size, are probably modi ed according to the group's ability to rear its young (Legge 2000b). In addition, some nestlings die from starvation, rather than sibling aggression, later in the nestling period. Alternatively, if provisioning young is costly, then the bene ts of reducing workloads as group size increases may out-weigh the bene ts of edging more young for both breeders and helpers (Boland et al. 1997; Hatchwell 1999; Heinsohn & Legge 1999). The corollary to this is that breeders with helpers should enjoy greater survivorship. Kookaburras are long-lived birds (15 years; Higgins 1999) and data on survival from this 4-year study are weak: only 10 breeders disappeared during the entire period, and only four of these were con rmed as dead (from recovered carcasses). Moreover, three out of these four birds died from collisions with cars. FEMALE HELPERS REDUCE NEST SUCCESS Female helpers may reduce nest success because they disrupt breeding attempts. Other group members, especially the breeding female, may perceive female helpers as reproductive competition and spend time chasing them away (Gaston 1978b; Stallcup & Woolfenden 1978; Mumme, Koenig & Pitelka 1983; McGowan & Woolfenden 1990). However, antagonistic behaviour around nests was never observed. As well as their di ering e ects on nest success, male and female helpers also di er in their helping e ort: males incubate eggs and provision nestlings at a higher rate than females (Legge 2000a). If other group members use group size to gauge their appropriate level of feeding, they may overvalue female contributions to the detriment of the corporate e ort. TOO MANY HELPERS SPOIL NEST SUCCESS Although the available sample was small, the largest group sizes of seven birds experienced drastically reduced nest success because eggs were much more likely to be damaged (Figs 2b, 3a). Large groups tend to have more female helpers than small groups, but it seems unlikely that egg damage was caused by female competition (Mumme et al. 1983). In within-

722 Reproductive success in kookaburras group paired comparisons, an increase in the number of female helpers did not reduce hatching success, although it did reduce edging success. Also, DNA ngerprinting showed that female helpers never gain maternity in nests (Legge & Cockburn 2000). It seems more likely that breakages occur more frequently when incubation is co-ordinated between many birds (Koenig 1982; Komdeur 1994). intruders (McComb, Packer & Pusey 1994). Parry (1973) noted that changes in group size were associated with concomitant changes in territory size (see also Gaston 1978c; Brooker & Rowley 1995). During this study, territory boundaries did not appear to shift conspicuously, but territory expansions may require neighbouring territories to retract or `die'. PAIR AND TERRITORY QUALITY Group size had no e ect on either edging success or edgling weight in paired comparisons that controlled for the quality of the breeding pair and their territory. Yet the results of the mixed model indicated that group size was associated with higher edging success and edgling weight. This highlights the problems inherent in correlational studies of the e ect of helpers on reproductive success, and demonstrates that extreme caution is required when interpreting the results of such studies. If territory quality is important in kookaburras, it seems strange that other ecological variables tested in the mixed model were non-signi cant. There were no interannual di erences in edging success or edgling weight despite the fact that the study spanned years of extremely low and high rainfall. Similarly, hatch date was not signi cant, although seasonal e ects are known to be strong in other species. However, most nests hatched within a very short period (e.g. in 1997, 50% of all clutches hatched within 7 days), possibly dampening variability in conditions experienced between nests with di erent hatch dates. In addition, the kookaburra's diet is extremely broad (reviewed in Higgins 1999), and breeding birds may be able to switch prey types in response to short-term uctuations in availability. The earliest hatch date occurred between 11 and 16 October in every year of the study, despite the variation in environmental conditions between years, again suggesting a compensatory response to any short-falls. The consistently high nest success of some breeding pairs may be explained by the size of their territory, as larger groups live on larger territories (Parry 1973; Reyer & Schmidl 1988; this study). Although kookaburras hunt a wide variety of prey, presumably the absolute biomass is greater on larger territories. Helpers may be important for maintaining ownership of territories, because they contribute conspicuously to territory defence. Intruders are aggressively repelled, the borders with neighbours are patrolled during ritualized displays, and groups regularly chorus from roost sites at dawn and dusk (Parry 1973; Reyer & Schmidl 1988; S. Legge, unpublished data). The duration and loudness of choruses is positively related to group size (Reyer & Schmidl 1988), and chorusing probably conveys information about numbers to neighbours and OTHER REASONS TO HELP Helping kin may occur for reasons other than gaining indirect tness bene ts. Helping may be an unselected response to stimuli normally presented during parenthood (Jamieson 1991). This seems unlikely in kookaburras, because the feeding contribution of helpers is plastic, modi ed according to their group size, breeding status and sex (Legge 2000a). Alternatively, helping may provide valuable breeding experience (Komdeur 1996). However, kookaburras breeding together for the rst time did not fare worse than experienced pairs, nor did the age of breeders a ect any aspect of reproduction, suggesting breeding success does not improve with experience. The measure of experience (pair duration) used in the analysis may be too crude, as `novice' pairs may nevertheless have prior breeding experience with other partners, or prior helping experience; unfortunately only one pair was known of during the study that formed from completely inexperienced birds. Conclusions This study aimed to nd whether kookaburras gained indirect tness bene ts from helping to raise siblings, as other bene ts of helping such as territorial inheritance or mating opportunities are not available to them (Legge & Cockburn 2000). Helpers were unable to increase the number of nesting attempts in a season, because kookaburras are single-brooded. Surprisingly, despite the high frequency of brood reduction, helpers did not increase the number or weight of young in each nesting attempt. Indeed, very large groups su ered reduced hatching success and female helpers exacted a toll on edging success. Although it was not possible to test whether helper contributions increased the survivorship of breeders, the alacrity of all kookaburras to reduce their workload rather than alleviate brood reduction suggests that raising young is costly. It therefore seems likely that breeders gain an energetic advantage from having helpers. This study is unusual because the e ect of helpers on nest success was examined after controlling for pair and territory quality. As well as revealing the absence of a group size e ect, when this controlled analysis was compared to a more extensive correlational analysis, pair and territory quality were shown to have a con-

723 S. Legge siderable impact on nest success. The importance of such quality e ects for the reproductive success of co-operative breeders is often discussed, but rarely demonstrated. Acknowledgements I thank A. Cockburn, E. Doerr, V. Doerr, two anonymous referees and especially R. Heinsohn for comments on the manuscript, and R. Cunningham and C. Donnelly for statistical advice. Canberra Nature Park, the Australian National Botanic Gardens and the Australian Bird and Bat Banding Scheme (ABBBS) were always helpful. I was supported by Overseas Postgraduate Research and ANU Scholarships. The work was carried out under permit from the ANU Animal Experimentation Ethics Committee, ABBBS and ACT Parks and Conservation. References Austad, S.N. & Rabenold, K.N. (1985) Reproductive enhancement by helpers and an experimental inquiry into its mechanism in the bicolored wren. Behavioral Ecology and Sociobiology, 17, 19±27. Boland, C.R.J., Heinsohn, R. & Cockburn, A. (1997) Experimental manipulation of brood reduction and parental care in co-operatively breeding white-winged choughs. Ecology, 66, 683±691. Brooker, M. & Rowley, I. (1995) The signi cance of territory size and quality in the mating strategy of the splendid fairy-wren. Ecology, 64, 614±627. Brown, J.L. (1987) Helping and Communal Breeding in Birds: Ecology and Evolution. Princeton University Press, Princeton, NJ. Brown, J.L. & Brown, E.R. (1981) Kin selection and individual selection in babblers. Natural Selection and Social Behaviour (eds R.D. Alexander & D.W. Tinkle). Chiron Press, New York, NY. Brown, J.L., Brown, E.R., Brown, S.D. & Dow, D.D. (1982) Helpers: e ects of experimental removal on reproductive success. Science, 215, 421±422. Cockburn, A. (1998) Evolution of helping behaviour in cooperatively breeding birds. Annual Review of Ecology and Systematics, 29, 141±177. Craig, J.L. (1984) Are communal pukeko caught in the prisoner's dilemma? Behavioral Ecology and Sociobiology, 14, 147±150. Crick, H.Q.P. (1992) Load-lightening in cooperatively breeding birds and the cost of reproduction. Ibis, 134, 56±61. Ekman, J., Bylin, A. & TegelstroÈ m, H. (1999) Increased lifetime reproductive success for Siberian jay (Perisoreus infaustus) males with delayed dispersal. Proceedings of the Royal Society of London, Series B, Biological Sciences, 266, 911±915. Emlen, S.T. (1997) Predicting family dynamics in social vertebrates. Behavioural Ecology: An Evolutionary Approach, 4th edn (eds J.R. Krebs & N.B. Davies), pp. 228±253. Blackwell Science, Oxford, UK. Emlen, S.T. & Wrege, P. (1991) Breeding biology of whitefronted bee-eaters at Nakuru: the in uence of helpers on breeder tness. Ecology, 60, 309± 326. Gaston, A.J. (1978a) Demography of the jungle babbler Turdoides striatus. Ecology, 47, 845± 870. Gaston, A.J. (1978b) Ecology of the common babbler Turdoides caudatus. Ibis, 120, 415±432. Gaston, A.J. (1978c) The evolution of group territorial behaviour and cooperative breeding. American Naturalist, 112, 1091±1100. Genstat Committee (1993) Genstat 5 Release 3 Reference Manual. Clarendon Press, Oxford, UK. Green, D.J., Cockburn, A., Hall, M.L., Osmond, H.L. & Dunn, P.O. (1995) Increased opportunities for cuckoldry may be why dominant male fairy-wrens tolerate helpers. Proceedings of the Royal Society of London, Series B, Biological Sciences, 262, 297±303. Gri ths, R., Double, M.C., Orr, K. & Dawson, R.J.G. (1998) A simple DNA test to sex most birds. Molecular Ecology, 7, 1071±1075. Hamilton, W.D. (1964) The genetical evolution of social behaviour. II. Journal of Theoretical Biology, 7, 17±52. Hatchwell, B.J. (1999) Investment strategies of breeders in avian cooperative breeding systems. American Naturalist, 154, 205±219. Heinsohn, R.G. (1991) Slow learning of foraging skills and extended parental care in cooperatively breeding whitewinged choughs. American Naturalist, 137, 864±881. Heinsohn, R.G. (1992) Cooperative enhancement of reproductive success in white-winged choughs. Evolutionary Ecology, 6, 97±114. Heinsohn, R. & Legge, S. (1999) The cost of helping. Trends in Ecology and Evolution, 14, 53±57. Higgins, P.J. (1999) Laughing Kookaburra. Handbook of Australian, New Zealand, and Antarctic Birds. Oxford University Press, Melbourne, Australia. Innes, K.E. & Johnston, R.E. (1996) Cooperative breeding in the white-throated magpie-jay. How do auxiliaries in uence nesting success? Animal Behaviour, 51, 519± 533. Jamieson, I.G. (1991) The unselected hypothesis for the evolution of helping behavior: too much or too little emphasis on natural selection? American Naturalist, 138, 271±282. Koenig, W.D. (1982) Ecological and social factors a ecting hatchability of eggs. Auk, 99, 526±536. Komdeur, J. (1994) Experimental evidence for helping and hindering by previous o spring in the cooperativebreeding Seychelles warbler Acrocephalus sechellensis. Behavioral Ecology and Sociobiology, 34, 175±186. Komdeur, J. (1996) In uence of helping and breeding experience on reproductive performance in the Seychelles warbler: a translocation experiment. Behavioral Ecology, 7, 326±333. Legge, S. (2000a) Helper contributions in the laughing kookaburra: feeding young is no laughing matter. Animal Behaviour, in press. Legge, S. (2000b) Siblicide in the cooperatively breeding laughing kookaburra. Behavioural Ecology and Sociobiology, in press. Legge, S. & Cockburn, A. (2000) The social and mating system of the cooperatively breeding laughing kookaburra. Behavioral Ecology and Sociobiology, 47, 220± 229. Leonard, M.L., Horn, A.G. & Eden, S.F. (1989) Does juvenile helping enhance breeder reproductive success? A removal experiment on moorhens. Behavioral Ecology and Sociobiology, 25, 357±362. Lewis, D.M. (1981) Determinants of reproductive success of the white-browed sparrow weaver, Plocepasser mahali. Behavioral Ecology and Sociobiology, 9, 83±93.

724 Reproductive success in kookaburras Ligon, J.D. & Ligon, S.H. (1990) Green woodhoopoes: life history traits and sociality. Cooperative Breeding in Birds (eds P.B. Stacey & W.D. Koenig), pp. 33±65. Cambridge University Press, Cambridge, UK. McComb, K., Packer, C. & Pusey, A. (1994) Roaring and numerical assessment in contests between groups of female lions, Panthera leo. Animal Behaviour, 47, 379± 387. McCullagh, P. & Nelder, J.A. (1989) Generalized Linear Models. Cambridge University Press, Cambridge, UK. McGowan, K.J. & Woolfenden, G.E. (1990) Contributions to edgling feeding in the Florida scrub jay. Journal of Animal Ecology, 59, 691±707. Magrath, D. & Yezerinac, S.M. (1997) Facultative helping does not in uence reproductive success or survival in cooperatively breeding white-browed scrubwrens. Journal of Animal Ecology, 66, 658±670. Monadjem, A., Owen-Smith, N. & Kemp, A.C. (1995) Aspects of the breeding biology of the arrowmarked babbler Turdoides jardineii in South Africa. Ibis, 137, 515±518. Mumme, R.L. (1992) Do helpers increase reproductive success ± an experimental analysis in the Florida scrub jay. Behavioral Ecology and Sociobiology, 31, 319±328. Mumme, R.L., Koenig, W.D. & Pitelka, F.A. (1983) Reproductive competition in the acorn woodpecker: sisters destroy each other's eggs. Nature, 306, 583±584. Mumme, R.L., Koenig, W.D. & Ratnieks, F.L.W. (1989) Helping behaviour, reproductive value, and the future component of direct tness. Animal Behaviour, 38, 331±343. Parry, V. (1973) The auxiliary social system and its e ect on territory and breeding in kookaburras. Emu, 73, 81±100. Reyer, H.-U., Dittami, J.P. & Hall, M.R. (1986) Avian helpers at the nest: are they psychologically castrated? Ethology, 71, 216±228. Reyer, H.-U. & Schmidl, D. (1988) Helpers have little to laugh about: group structure and vocalisation in the laughing kookaburra Dacelo novaeguineae. Emu, 88, 150±160. Sherley, G.H. (1989) Bene ts of courtship-feeding for ri eman (Acanthisitta chloris) parents. Behaviour, 109, 309±318. Stallcup, J.A. & Woolfenden, G.E. (1978) Family status and contributions to breeding by Florida scrub jays. Animal Behaviour, 26, 1144±1156. Sydeman, W.J., Guntert, M. & Balda, R.P. (1988) Annual reproductive yield in the cooperative pygmy nuthatch (Sitta pygmaea). Auk, 105, 70±77. Woolfenden, G.E. & Fitzpatrick, J.W. (1990) Florida scrub-jays: a synopsis after 18 years study. Cooperative Breeding in Birds: Long-Term Studies of Ecology and Behaviour (eds P.B. Stacey & W.D. Koenig), pp. 239± 266. Cambridge University Press, Cambridge, UK. Received 13 September 1999; revision received 7 February 2000