Last but not least: nestling growth and survival in asynchronously hatching crimson rosellas

Authors


Elizabeth Krebs, Division of Botany and Zoology, The Australian National University, Canberra 0200, A.C.T., Australia

Abstract

1. Many species of altricial birds hatch their young asynchronously within broods. Although there are many potential benefits to parents, hatching asynchrony reduces the growth and often the survival of last-hatched nestlings. The consequences of hatching asynchrony on the growth, size at fledging and survival of male and female nestlings of an Australian parrot, the crimson rosella (Platycercus elegans) were examined.

2. Crimson rosella broods hatched over 1·5–7 days, creating mass hierarchies where first-hatched chicks were up to seven times larger than last-hatched chicks. Hatching asynchrony and mass hierarchies increased over the breeding season, but were not strongly correlated with brood size.

3. Male chicks grew faster and were larger at fledging than females. Growth rates did not differ between hatching ranks. Chicks of all ranks were of equal sizes at fledging, but last-hatched male chicks had lower fledging mass in pairwise analyses. Female mass at fledging did not decrease with hatching rank. Chick growth rates and size or mass at fledging were not related to hatching asynchrony, mass or size hierarchies in broods, brood size, laying date or year in mixed-model analyses.

4. Last-hatched chicks had the same post-fledging survival as other chicks, however, they were more likely to die during the nestling period. Increased mortality of last-hatched nestlings occurred only at hatching and chicks had equal mortality rates over the remainder of the nestling period. Early brood reduction was not associated with brood size or hatching asynchrony but increased over the breeding season, and in broods with high hatching success.

5. Hatching asynchrony in rosellas, unlike in most previous studies, did not lead to poor growth and subsequent survival of last-hatched chicks. This suggests that the costs of hatching asynchrony are low in this species and that selective feeding by parents may increase the growth and survival of last-hatched chicks.

6. Reduced growth and survival of later hatched chicks is not an inevitable consequence of asynchronous hatching; however, the costs to parents of overcoming competitive interactions between chicks may be higher than the benefits in most species.

Introduction

Young hatch sequentially in many species of birds, producing an age and size hierarchy within the brood. Hatching asynchrony within a brood varies from several hours in many passerines (Clark & Wilson 1981), to several weeks in some owls and parrots (Saunders 1982; Wilson, Wilson & Durkin 1986; Waltman & Beissinger 1992). The function of hatching asynchrony has been the subject of many studies and continues to be hotly debated (reviews by Magrath 1990; Amundsen & Slagsvold 1991; Stoleson & Beissinger 1995). Producing a size and age hierarchy within the brood via asynchronous hatching could promote parental feeding efficiency (e.g. Hussell 1972; Hahn 1981), facilitate efficient brood reduction when resources are limited (Lack 1947, 1968), or be the non-selected consequence of constraints on incubation patterns (Clark & Wilson 1981).

Despite the variety of explanations for asynchronous hatching, the consequences are remarkably similar across species. Hatching asynchrony is associated with reduced growth and survival of last-hatched nestlings (e.g. Bryant & Tatner 1990; Veiga 1990; Wiebe 1996; Ostreiher 1997), even in species with small hatching spreads (Greig-Smith 1985; Stouffer & Power 1990). Hatching patterns are clearly controlled by parents through the timing of the onset of incubation (Bryant 1978; Magrath 1992). However, size hierarchies within a brood are often reinforced by processes controlling the distribution of food after hatching. In most species access to food is determined by a dominance hierarchy among chicks, or through scramble competition in the nest, both of which favour larger chicks (e.g. Bengtsson & Ryden 1983; Poole 1989; Kacelnik et al. 1995). Thus, reduced growth of last-hatched chicks is probably a consequence of their poor competitive ability in interactions over food.

Whether poor growth and survival of last-hatched chicks is congruent with parental interests is less clear. Increased mortality of last-hatched nestlings may benefit parents and surviving chicks when food is limited. On the other hand, if reduced nestling growth results in low fledging weights, post-fledging survival of chicks can be reduced (Perrins 1965; Nur 1984; Magrath 1991). Reduced growth and survival of last-hatched chicks may occur in asynchronous broods regardless of food availability (e.g. Bryant 1978; Werschkul 1979; Amundsen & Stokland 1988; Wiebe & Bortolotti 1995; Nilsson & Svensson 1996; Stoleson & Beissinger 1997); for example, in little blue herons (Egretta caerulea), broods of even-aged nestlings grew at the same rate as early hatched nestlings in asynchronous broods, suggesting that parental ability to distribute food was more important than the amount of food delivered (Werschkul 1979). Similarly, last-hatched chicks in house martin (Delichon urbica) broods with large size hierarchies were 10 times more likely to die than those in broods with subtle hierarchies, even when food was abundant (Bryant 1978). Consequently, parents may be forced to trade-off the benefits of asynchronous hatching against reduced ability to control the distribution of food within the broods.

The majority of studies of hatching asynchrony have focused on passerines, which have relatively low levels of hatching asynchrony. In contrast, nonpasserines show extreme variability in hatching patterns (Ricklefs 1993; Stoleson & Beissinger 1995); for example, parrots (Aves: Psittaciformes) have predominantly asynchronous hatching patterns (Stoleson & Beissinger 1995), and considerable variation in the degree of hatching asynchrony both within and between species. Some parrots initiate incubation on the first egg; examples include green-rumped parrotlets (Forpus passerinus;Waltman & Beissinger 1992) and white-tailed black cockatoos (Calyptorhynchus funereus;Saunders 1982). In others, incubation can begin on the last egg (e.g. Major Mitchell's cockatoo Cacatua leadbeaterii;Rowley & Chapman 1991). Detailed studies on groups such as parrots are necessary to determine whether poor growth and subsequent survival of last-hatched chicks is a general cost of hatching asynchrony, or if species with extreme hatching asynchrony have evolved mechanisms to reduce the costs of a large size hierarchy.

In this study, variation in hatching asynchrony was quantified in a medium-sized parrot, the crimson rosella (Platycercus elegans Gmelin). The consequences of hatching asynchrony for chick growth, size at fledging and survival of chicks, both in the nest and after fledging were examined. Male rosellas are larger than female rosellas and molecular sexing of nestlings was used to evaluate growth and size at fledging for each sex independently. If hatching asynchrony disadvantages last-hatched chicks, it may be more costly for last-hatched males and an assessment was made of whether the costs of hatching asynchrony differed between the sexes.

Methods

Study area and methods

Crimson rosellas breeding in nestboxes in Black Mountain Nature Reserve, Canberra, Australia were studied between August 1993 and January 1997. Fifty nestboxes were placed in a grid 50 m apart in dry sclerophyll Eucalypt forest (Eucalyptus rossii, E. mannifera and E. macrorhyncha) during August 1993. Up to three pairs of rosellas attempted to breed in a particular nestbox over a breeding season, but high levels of egg destruction and hatching failure resulted in many pairs failing or having very poor breeding success. In total the breeding success of 64 pairs incubating clutches was monitored.

In this study, females rosellas initiated breeding in late September or early October, laying a mean ± se of 5·3 ± 0·1 eggs (range = 3–8; n = 64 broods) at 2–4 day intervals. Rosellas hatched 4·0 ± 0·2 chicks (range = 1–7; n = 64), and fledged 2·7 ± 0·2 chicks (range = 0–7; n = 64). Hatching success in some broods was very low, averaging 75 ± 3% (range = 13–100% n = 64) for the entire study.

Nestboxes were inspected at least weekly during the laying period. In most cases, boxes were visited daily during the predicted hatching period until all nestlings had hatched. Where the order of hatching was observed directly, a hatching order was assigned to each chick as well as a hatching time to the nearest half day, based on the dryness of a chick's down and whether it had been fed (chicks are not fed until 6 h after hatching). Where more than one chick hatched on the same day, hatching time was assigned based on a combination of wing chord length and dryness of the down. If chicks were so similar in size that it was unclear which chick had hatched first, tied hatching ranks were assigned. In broods where several chicks had already hatched, each chick was assigned an age retrospectively using a regression of wing chord on age for all chicks of known hatching times. These broods were never more than 4 days old, and rank reversals were never noted early in the nestling period.

All newly hatched chicks were marked by trimming the down on different parts of the body, and 7-day-old chicks were marked with a single colour band to allow individual recognition. Nestlings were weighed to the nearest 0·1 g at hatching and subsequently to the nearest 1 g, using Pesola spring balances. Wing chord length was measured to the nearest 0·5 mm using a stainless steel ruler. Tibia length (outer edges of joints) was measured to the nearest 0·1 mm using dial callipers. All chicks in a brood were measured each time a new chick hatched to obtain a measure of the size hierarchy at the completion of hatching. After hatching was complete, the brood was measured at least weekly until the oldest chick was 30–33 days of age. Broods were not measured after this age, because disturbance was likely to cause the premature fledging of older chicks. Consequently, last-hatched chicks in broods with large hatching spreads may have been measured for a final time when only 23 days old. All chicks were banded with a numbered Australian Bird and Bat Banding Scheme stainless steel band and a unique combination of anodized aluminium colour bands prior to fledging.

Sexing individuals

Adult male crimson rosellas are 10 – 20% larger than females, and differ in bill morphology. Nestlings, however, can be difficult to sex because size differences are small at hatching and because weight and size measurements overlap between the sexes. Therefore, a sex-specific genetic marker was used to determine the sex of all nestlings,. A blood sample (100 μL) was taken from each chick at least 15 days after hatching for use in sexing and other genetic work. Two polymerase chain reaction (PCR) primers (P2 and P8; Griffiths et al. 1998) were used to amplify a gene (CHD) on the W-chromosome. The heterogametic sex, females, were identified by the presence of two distinct bands on an agarose gel, and males by the presence of only one band (see Griffiths et al. 1998 for details of the method).

Hatching hierarchies

The hatching hierarchy within a brood was quantified in two ways: (i) the total hatching asynchrony (first-hatched to last-hatched) within a brood to the nearest 0·5 day, as outlined above; and (ii) the relative difference in nestling size for chick mass (‘mass hierarchy’), and for tibia length (‘size hierarchy’). The relative difference in nestling size was calculated as: [(size of first-hatched chick) – (size of last-hatched chick)]/ (mean size of all nestlings) (Bryant & Tatner 1990; Wiebe & Bortolotti 1994a). Because measures of the size or mass hierarchy within a brood could be biased by the age at which the last-hatched chick was first measured, it has only been calculated for broods where the last-hatched chick was measured within 24 h of hatching.

All broods where more than two eggs or chicks failed to hatch or died were excluded. In total, the growth of 130 individual chicks in brood sizes ranging from three to seven was examined. In addition, 13 broods from 1993 which were depredated during the nestling period, were included in order to maximize the number of broods with known hatching asynchrony and size or mass hierarchies.

Growth

Growth curves for tibia length and mass were calculated for each chick. Tibia length provides a measure of overall skeletal size. Chick mass is a more variable measure because of differences in crop and stomach contents; however, it provides a composite measure of chick size and condition. Growth curves were not fitted to wing length because primaries continued to grow after fledging.

The logistic form of a growth curve (Ricklefs 1971) described changes in chick mass and tibia length well. Variance in chick mass increased with age, therefore the data were transformed and loge(mass) was fitted to a loge(logistic curve). Three parameters describe the shape of a logistic curve: the asymptote, the rate constant (k) and the point of inflection. The estimate of asymptote from the curve was used as a measure of chick mass or size at fledging. However, the point of inflection and the rate constant (k), in logistic curves, are both dependent on the value of the asymptote (Ricklefs 1971). This makes these measures difficult to interpret for chicks of differing sizes and therefore of limited biological meaning. Estimates of inflection points or rate constants have therefore not been included in analyses. Instead, the average growth rate over the linear portion of a growth curve was calculated, a measure not dependent on the final mass or size of a chick.

Two difficulties were encountered fitting logistic growth curves to mass. First, many rosella nestlings peaked in mass about 10 days prior to fledging and subsequently decreased in mass, with large fluctuations until fledging. Thus estimates of asymptotic mass varied between chicks depending on the distribution of measurements during the plateau of the curve. Second, in some broods, chicks hatching last were only measured once during the plateau of the curve, making estimates of the asymptote less accurate. In general, the logistic curve appeared to produce asymptotic estimates of mass which appeared unrealistically low for last-hatched chicks. A second estimate of mass was therefore calculated at fledging by using the mean mass for each chick for measures taken after the growth curve reached a plateau at 21 days post-hatch. Measures of mean maximum mass were highly correlated to predicted asymptotic mass (r = 0·87, P < 0·001, n = 123), especially for first-hatched chicks (r = 0·92, P < 0·001, n = 38) but, as suspected, less strongly for last-hatched chicks (r = 0·83, P < 0·001, n = 29).

Tibia length reached a plateau earlier in the nestling period than mass and varied less between measurements. Consequently, estimates of asymptotic tibia length from the logistic growth curves were assumed to be unbiased and accurate for all individuals. It was not possible to calculate adequate measures of linear growth rates for tibia length because the duration of the linear phase was short, lasting only from 1 to 10 days after hatching, during which too few measures were taken to give reliable estimates. The tibia of many chicks was only measured twice during this period, and the slope of some individuals was unrealistic, particularly if the measures were taken 1 or 2 days apart. Linear growth rates calculated from wing chord measurements were therefore used. Wing chord growth is much slower than tibia length and remains approximately linear between 8 and 28 days after hatching.

If less than five measurements were taken for a chick, a growth curve was not calculated. Because mass, tibia and wing were not measured on all occasions, sample sizes vary depending on the measure used.

Mortality

Chicks that died during the nestling period either died soon after hatching without being fed, or later in the nestling period as a result of predation, physical defects or unknown causes (E.A. Krebs, unpublished data). Mortality was categorized as occurring early in the nestling period, late in the nestling period or post-fledging for all nests that hatched at least three chicks (allowing a first-, middle- and last-hatched category, see below). The number of banded fledglings resighted at least 3 months after leaving the nest was used as a measure of survival after fledging.

Analyses

The effect on chick growth of six different explanatory variables was examined: year, laying date, brood size, total hatching asynchrony (in days) or the brood size or mass hierarchy, sex and hatching rank. As hatching asynchrony and brood size and mass hierarchies are highly correlated and describe similar biological variables, each was entered separately as explanatory variables in alternate models in order to assess their independent impact on chick growth. Hatching ranks were assigned to chicks by grouping them as first-hatched, middle-hatched or last-hatched within the brood. This resulted in large broods having more than one middle-hatched chick, however, it allowed comparisons based on hatching order for broods of different sizes. If hatching ranks were tied for first or last chicks within a brood (see above), this resulted in the brood having more than one chick categorized as first or last. However, this only occurred in a few large broods, and never eliminated the middle-hatched category.

Analysing differences between chicks of different hatching ranks across broods requires a repeated-measured design. However, least-squares models do not allow for unbalanced data within a repeated-measures design. Therefore, to analyse differences in growth across differing brood sizes, a mixed model incorporating random and fixed effects was used (REML, restricted maximum likelihood; GENSTAT 5 GENSTAT, 5 Committee 1993). Five variables describing chick growth and size at fledging were each used as response variables in separate models: asymptotic mass, mean maximum mass, the linear rate of mass increase, asymptotic tibia length and linear rate of wing growth. Nestbox was included as a random term in all mixed models. A model was fitted by initially including all explanatory variables and two-way interactions. Terms were dropped from a model in a stepwise procedure by assessing the change in deviance between the full model and the submodel. Any significant terms were included in subsequent model fitting. A final model was selected by sequentially dropping nonsignificant interactions and then nonsignificant main effects, until only significant terms remained. Any confounding effects resulting from the order of deletion were avoided by adding and dropping any term that was close to significant (P < 0·20) from the final model. The results of model fitting for chick growth rates are presented for all main effects and four biologically important interactions (sex*rank, sex*hatching asynchrony, rank*hatching asynchrony and brood size*rank). For simplicity, interaction terms were not presented if nonsignificant.

For analyses where each brood was represented only once, least-squares models were used for continuous response variables and logistic regression for dichotomous variables (SPSS software; Norusis 1994).

For all fitted models, the normality of the data was verified using residual and normal probability plots. Log transformations were used to produce a linear relationship between hatching asynchrony and laying date. Means and standard errors are reported throughout, except where otherwise indicated.

Results

Hatching asynchrony

Crimson rosella broods hatched over an interval of 1·5–7 days (mean ± sd = 3·6 ± 1·6, n = 48). Hatching asynchrony increased with laying date (Fig. 1) and brood size (Fig. 2a). However, total hatching asynchrony increased most between three-chick broods and all other brood sizes (Fig. 2a), and when only clutches with no hatching failure were considered did not increase with brood size (ancova controlling for laying date: F4,17 = 0·48, P = 0·49).

Figure 1.

The relationship between total hatching asynchrony and laying date in crimson rosellas. Hatching interval increased significantly over the breeding season and with increasing brood sizes (ancova&mdash;laying date covariate: F4,31 = 7·2, P = 0·01; brood size effect: F4,31 = 3·6, P = 0·02). The predicted lines are shown for each brood size.

Mass and size hierarchies

Mass hierarchies at the end of hatching ranged from 0·43 to 1·42 (mean ± sd = 0·96 ± 0·28, n = 34). Mass hierarchies larger than mean brood mass were common and first-hatched chicks could be 40 g or 700% heavier than last-hatched chicks. Mass hierarchies increased with hatching asynchrony (Fig. 3a) and laying date (Fig. 4), but not with brood size (ancova, controlling for hatching asynchrony and laying date: F4,33 = 0·3, P = 0·80; Fig. 2b).

Figure 4.

The residuals of mass hierarchy regressed against hatching asynchrony (HA) plotted against laying date for rosella broods. The residuals were positively correlated to laying date (residuals = –1·02 + 0·003 * laying date; regression analysis, controlling for hatching asynchrony, t = 2·1, P = 0·04, n = (34).

Mass and size hierarchies within a brood were highly intercorrelated (r = 0·78, P = 0·0001, n = 29). Overall, size hierarchies at hatching were smaller, and ranged from 0·14 to 0·70 (mean ± sd = 0·41 ± 0·16). Size hierarchies increased with hatching asynchrony (Fig. 3b), but did not increase with laying date or brood size (ancova, controlling for hatching asynchrony: laying date covariate, t = –0·7, P = 0·56; brood size effect, F4,30 = –0·7, P = 0·62).

Nestling growth and size at fledging

Male rosella chicks grew faster and had higher asymptotic masses and tibia lengths at fledging than female chicks (Tables 1 and 2; Fig. 5). No other variable consistently explained variation in chick growth or size at fledging (Tables 1 and 2; Fig. 5).

Table 1.  Summary of REML model fitting for the following response variables: (a) asymptotic mass (g), (b) mean maximum mass (g), (c) linear growth rate of mass (g), (d) asymptotic tibia length (mm) and (e) linear growth rate of wing chord (mm). Asymptotic estimates were derived by fitting a logistic curve to individual chick measurements (n = 130; see Methods for further details). The order in which the variables are presented in the table does not reflect the order in which they were dropped from the model (see Methods). The change in deviance is reported when each variable is removed from the model. All possible two-way interactions were tested, but results are presented from those of biological interest. No other interactions were found to be significant. All P-values less than or equal to 0·10 are reported, any greater than 0·10 are termed nonsignificant (NS)
Explanatory variableModel variable removed fromChange in deviance (χ2) d.f.P
  1. BS, brood size; HA, hatching asynchrony (in days); laydate, laying date; rank, hatching rank (first, middle, last).

(a) Asymptotic mass
LaydateSex + rank + BS + HA + year + laydate 0·241NS
YearSex + rank + BS + HA + year 1·603NS
HASex + rank + BS + HA 1·481NS
BSSex + rank + BS 3·364NS
RankSex + rank 6·242< 0·05
SexSex + rank63·91< 0·001
Sex*rankSex + rank + sex*rank 3·422NS
Sex*HASex + rank + HA + sex*HA 2·871< 0·10
Rank*HASex + rank + HA + rank*HA 0·792NS
Rank*BSSex + rank + BS + rank*BS 8·668NS
(b) Mean maximum mass
LaydateSex + rank + BS + HA + year + laydate 0·051NS
YearSex + rank + BS + HA + year 3·593NS
HASex + rank + BS + HA 0·481NS
BSSex + rank + BS 2·304NS
RankSex + rank 3·632NS
SexSex + rank55·131< 0·0001
Sex*rankSex + rank + sex*rank 1·042NS
Sex*HASex + HA + sex*HA 3·231< 0·10
Rank*HASex + rank + HA + rank*HA 0·552NS
Rank*BSSex + rank + BS + rank*BS 6·948NS
(c) Linear growth rate of mass
LaydateSex + rank + BS + HA + year + laydate 0·281NS
YearSex + rank + BS + HA + year 1·113NS
HASex + rank + BS + HA 0·001NS
BSSex + rank + BS 7·384NS
RankSex + rank 1·052NS
SexSex + rank17·241< 0·0001
Sex*rankSex + rank + sex*rank 4·342< 0·10
Sex*HASex + HA + sex*HA 0·691NS
Rank*HASex + rank + HA + rank*HA 2·262NS
Rank*BSSex + rank + BS + rank*BS 6·228NS
(d) Asymptotic tibia length
LaydateSex + rank + BS + HA + year + laydate 0·051NS
YearSex + rank + BS + HA + year14·863< 0·01
HASex + rank + BS + HA + year 0·001NS
BSSex + rank + BS + year 6·914NS
RankSex + rank + year 0·172NS
SexSex + year46·111< 0·0001
Sex*rankSex + rank + year + sex*rank 3·192NS
Sex*HASex + HA + year + sex*HA 0·261NS
Rank*HASex + rank + HA + year + rank*HA 0·662NS
Rank*BSSex + rank + BS + year + rank*BS 5·178NS
(e) Linear growth rate of wing chord
LaydateSex + rank + BS + HA + year + laydate 0·281NS
YearSex + rank + BS + HA + year 0·763NS
HASex + rank + BS + HA 0·741NS
BSSex + rank + BS 7·484NS
RankSex + rank 0·142NS
SexSex + rank 1·491NS
Sex*rankSex + rank + sex*rank 0·622NS
Sex*HASex + HA + sex*HA 3·811< 0·10
Rank*HARank + HA + rank*HA 1·422NS
Rank*BSRank + BS + rank*BS10·127NS
Table 2.  Comparison of asymptotic mass, growth rates and asymptotic tibia length for crimson rosella chicks grouped by sex and hatching order (first-, middle- and last-hatched). Hatch order did not significantly affect any measure except asymptotic mass (Table 1)
First hatchMid hatchLast hatchOverall
Growth
variable
(x¯ ± SE) n(x¯ ± SE) n(x¯ ± SE) n(x¯ ± SE) n
(a) Males
Asymptotic mass138·2 ± 3·015139·2 ± 1·828136·8 ± 4·010138·5 ± 1·553
Mean mass139·4 ± 2·613142·9 ± 1·827137·6 ± 2·910140·9 ± 1·350
Linear growth rate (mass) 7·94 ± 0·3815 7·49 ± 0·2025 7·33 ± 0·278 7·61 ± 0·1748
Tibia length 56·2 ± 0·415 56·7 ± 0·3627 56·3 ± 0·58 56·5 ± 0·2450
Linear growth rate (wing) 5·75 ± 0·0915 5·54 ± 0·0925 5·56 ± 0·138 5·61 ± 0·0648
(b) Females
Asymptotic mass130·2 ± 2·925124·7 ± 2·031128·7 + 2·620127·6 ± 1·476
Mean mass132·1 ± 2·225127·2 ± 2·029127·8 ± 2·319129·0 ± 1·373
Linear growth rate (mass) 6·86 ± 0·1521 7·20 ± 0·2728 7·36 ± 0·2520 7·15 ± 0·1469
Tibia length 55·0 ± 0·325 54·2 ± 0·331 54·4 ± 0·519 54·5 ± 0·275
Linear growth rate (wing) 5·56 ± 0·1019 5·51 ± 0·0730 5·63 ± 0·1418 5·55 ± 0·0667
Figure 5.

Predicted logistic growth curves for the mass of crimson rosella chicks presented separately for (a) males (first-hatched n = 15, middle-hatched n = 28, last-hatched n = 10), and (b) females (first-hatched n = 25, middle-hatched n = 31, last-hatched n = 20). Logistic curves were calculated using the mean parameter values from individually fitted curves for chick categorized as first-, middle- and last-hatched.

Although hatching rank did not generally explain differences in chick size or growth in mixed models, last-hatched chicks had lower asymptotic masses than first-hatched chicks (Tables 1 and 2). Mean maximum masses were not significantly smaller for last-hatched chicks, further suggesting that logistic growth curves may have underestimated the mass of last-hatched nestlings at fledging (see Methods). To further assess any impact of chick rank on growth or size at fledging, a pair-wise comparison of the differences for each sex was undertaken. To maximize the number of useable broods, chicks of either sex were assigned as ‘first-hatched’ if they hatched first or second, and chicks of either sex as ‘last-hatched’ if they were penultimate or last-hatched in a brood. Pairwise comparisons for each sex were only considered where the chicks differed by at least two ranks. These criteria were met in only a subsample of broods, particularly for males, as last-hatched males are much less common than last-hatched females. Analysis on this subset of data suggested that first-hatched male chicks were heavier, but not larger at fledging than last-hatched male chicks (Table 3). There were no pairwise differences in linear rates of mass gain or linear rates of increase of wing growth. In contrast, no measure of female chick growth differed in pairwise comparisons (Table 3). Despite the fact hatching rank potentially affected male mass at fledging, no hatching rank*sex interactions were detected in the mixed models (Table 1), suggesting that the subsample of data used in pairwise comparisons may have been biased

Table 3.  Summary of pairwise comparisons between first and last-hatched rosella chicks (see Results for definition) summarized separately for (a) male and (b) female nestlings for the five measures of nestling growth analysed: asymptotic mass (g); mean maximum mass (g); asymptotic tibia length (mm); linear rate of mass gain (g day–1); and linear rate of increase in wing length (mm day–1). For each variable, mean paired differences (first–last hatched), standard deviations and the number of broods (in brackets) in each comparison are presented. The results of paired t-tests and probability levels are included for each variable
Growth variableMean paired difference ± SD Paired t-statistic Two-tailed significance
  • *

    Note for this comparison, last-hatched males were significantly larger than first-hatched males, as indicated by the minus sign

(a) Male nestlings
Asymptotic mass 6·6 ± 6·4 (9) 3·10·02
Mean maximum mass 7·5 ± 7·6 (8) 2·80·03
Asymptotic tibia length–0·95 ± 0·94 (7)*–2·70·04
Linear rate of mass gain 0·45 ± 1·1 (9) 0·870·26
Linear rate of wing growth 0·14 ± 0·47 (9) 0·920·39
(b) Female nestlings
Asymptotic mass 1·1 ± 7·0 (19) 0·710·48
Mean maximum mass 2·5 ± 7·3 (19) 1·510·15
Asymptotic tibia length 0·74 ± 2·4 (18) 1·310·21
Linear rate of mass gain 0·22 ± 1·1 (18) 0·870·40
Linear rate of wing growth–0·04 ± 0·64 (18)–0·290·78

Patterns of annual and seasonal variation did not consistently affect nestling mass, or growth (Table 1). However, asymptotic tibia length varied between years, primarily as result of shorter tibias in 1993 [predicted asymptotic values (in mm): males 1993 = 55·0 ± 0·4 1994 = 56·3 ± 0·5 1995 = 57·3 ± 0·4 1996 = 57·2 ± 0·8; Females 1993 = 53·8 ± 0·4 1994 = 53·5 ± 0·3 1995 = 55·6 ± 0·3 1996 = 54·5 ± 0·5]. Annual differences in tibia length may reflect differing resource levels as the longest tibias were recorded in the wettest year; however, there were no interannual differences in mass or wing linear growth rates to support this view (Table 1).

Although few explanatory variables related to chick growth, growth rates were not constant between nestboxes. For each response variable tested, significant variation occurred between nestboxes even after the final model was fitted (change in deviance for model with and without nestbox as a random factor: asymptotic mass, χ21 = 84·9, P < 0·0001; mean maximum mass, χ21 = 42·7, P < 0·0001; linear mass growth rate, χ21 = 6·0, P < 0·02; tibia, χ21 = 10·2, P < 0·005; linear wing growth rate, χ21 = 11·1, P < 0·001). Thus other unmeasured variables, such as parental age, parental condition, or genetic factors, also affect nestling growth.

Nestling mortality

Twenty-two percent of all chicks hatched (43/198), died during the nestling period. Overall mortality during the nestling period tended to be higher for chicks hatching last [percent dying: first-hatched = 18% (8/44); middle-hatched = 17% (19/110); last-hatched = 34% (15/44); χ22 = 5·7, P < 0·10]. Increased mortality of last-hatched chicks was entirely a result of mortality immediately after hatching. Early mortality occurred in 10 cases (23% of all chick mortality) and always involved last-hatched, or in one case, both penultimate and last-hatched chicks [percent dying immediately after hatching: first- and middle-hatched = 1% (1/154); last-hatched = 20% (9/44); χ22 = 28·4, P < 0·0001]. In contrast, nestling mortality over the remainder of the nestling period did not differ between hatching ranks [first-hatched = 18% (8/44); middle-hatched = 17% (18/109); last-hatched = 17% (6/35); χ22 = 0·16, P > 0·97].

The probability of early brood reduction increased over the breeding season (Fig. 6). However, brood reduction did not increase with hatching asynchrony (logistic regression, model improvement statistic; χ21 = 0·7, P = 0·39) or brood size (χ21 = 0·4, P = 0·52). Brood reduction might be more likely when the interval between the penultimate and last-hatched chick was large. However, there was no difference in the hatching intervals between the penultimate and last-hatched chicks for broods that experienced brood reduction and those that did not (final hatching interval: brood reduced = 1·7 ± 0·3 days, n = 10; non-brood reduced = 1·6 ± 0·4 days, n = 11). Broods that hatched a higher proportion of their eggs tended to have a higher probability of brood reduction (logistic regression, χ21 = 3·59, P = 0·06; proportion hatched: reduced broods = 0·89; non-reduced broods = 0·73, n = 28).

Figure 6.

The probability of brood reduction in rosella broods vs. laying date for all years combined (logistic regression, χ21 = 12·3, P < 0·001). The fitted line and the proportion of broods experiencing brood reduction over each 20-day interval are shown. The number of broods sampled is indicated above each dot.

Post-fledging survival

A total of 11% (18/168) of rosellas banded as nestlings were resighted as juveniles in or adjacent to the study area. Most individuals were sighted more than once (mean = 2·8; range = 1–12), and several were observed paired and defending hollows, suggesting that they were resident in the area. Male juvenile rosellas were more likely to be sighted than females [25% (15/61) of males vs. 4% (3/80) of females resighted; χ21 = 10·3, P < 0·0001]. If post-fledging survival of males and females is equal, this suggests that female crimson rosellas move further from their natal area before breeding. It was assumed that any juvenile male in the area was equally likely to be resighted, however, because females may have survived but moved beyond the vicinity of the study area, they were not included in survival analyses. Survival of male rosella chicks did not differ between chicks hatched first, middle or last in a brood [first-hatched = 25% (3/12), middle-hatched = 20% (7/35), last-hatched = 44% (4/9); χ22 = 0·92, P > 0·9]. Surviving male chicks were not heavier at fledging, nor larger at fledging and did not fledge earlier in the season than chicks who did not survive (Table 4).

Table 4.  A comparison of chick size and mass at fledging for male crimson rosella chicks which survived to at least 3 months post fledging and chicks that did not. Two measures of chick mass at fledging (asymptotic mass and the mean maximum mass; see Methods) asymptotic tibia length and laying date were compared between individuals who survived and those who did not. Sample sizes are presented in brackets. Test statistics for t-tests and P-values are reported for each variable. There were also no significant differences between the characteristics of surviving and non-surviving chicks detected by a logistic regression
VariableMales survivingMales not survivingt-statistic Two-tailed significance
Asymptotic mass139·6 ± 3·3 (13)139·0 ± 1·6 (43)–0·20·88
Mean maximum mass141·5 ± 3·7 (11)141·2 ± 1·5 (42)–0·10·94
Asymptotic tibia length 56·6 ± 0·3 (13) 56·5 ± 0·3 (40)–0·00·97
Laying date298·9 ± 6·6 (14)300·5 ± 4·3 (40) 0·20·84

Discussion

The surprising result of this study is that high levels of hatching asynchrony in crimson rosella broods did not reduce the growth of later-hatched chicks, and only affected survival immediately after hatching. Overall, neither hatching asynchrony nor hatching order affected the growth rates, size at fledging, or post-fledging survival of rosella chicks, but may have reduced mass at fledging for last-hatched male chicks. In contrast, hatching asynchrony in most species leads to poor growth and sometimes to lower post-fledging survival of last-hatched chicks (Table 5).

Table 5.  Summary of patterns of growth and survival for last-hatched nestlings in asynchronously hatching species. Only studies on wild birds are included
Order*/Species Nestling growth reduced?Post-fledging survival reduced?Reference
  • *

    After Sibley et al. 1990)

Passeriformes
Arabian babblersYesUnknownOstreiher 1997
(Turdoides squamiceps)
Blue titsYesYes/noNur 1984; Slagsvold et al. 1995
(Parus caeruleus)
European blackbirdsYesYesMagrath 1989, 1991
(Turdus merula)
Marsh titsYesUnknownNilsson & Svensson 1996
(Parus palustris)
StarlingsYesYesStouffer & Power 1990
(Sturnus vulgaris)
Pied flycatcherYesUnknownSlagsvold 1986
(Ficedula hypoleuca)
Tree swallowsYesUnknownZach 1982
(Tachycineta bicolor)
White-winged choughsYesNoHeinsohn 1995; Boland et al. 1997
(Corcorax melanorhamphos)
Psittaciformes
Crimson rosellasNoNoThis study
(Platycercus elegans)
Green-rumped parrotletsYesNoStoleson & Beissinger 1997
(Forpus passerinus)
Major Mitchell's cockatooYesNoRowley & Chapman 1991
(Cacatua leadbeateri)
Long-billed corellaYesNoSmith 1991
(Cacatua pastinator)
Coraciiformes
Blue-throated bee-eatersYesUnknownBryant & Tatner 1990
(Merops viridis)
European bee-eatersYesNoLessells & Avery 1989
(Merops apiaster)
White-fronted bee-eatersYesUnknownEmlen et al. 1991
(Merops bullockoides)
Ciconiiformes
American kestrelsYesUnknownWiebe & Bortolotti 1995
(Falco sparverius)
Black kitesYesUnknownVeiga & Hiraldo 1990
(Milvus migrans)
Magellanic penguinsYesUnknownBlanco et al. 1996
(Spheniscus magellanicus)
Maguari storksNoUnknownThomas 1984
(Ciconiia maguari)
Red kitesYesYesVeiga & Hiraldo 1990
(Milvus milvus)
ShagNoUnknownAmundsen & Stokland 1988
(Phalacrocorax aristotelis)
White-fronted bee-eatersYesUnknownEmlen et al. 1991
(Merops bullockoides)

Hatching asynchrony and size hierarchies in broods

Hatching asynchrony in rosella broods averaged 3·6 days, longer than that observed in 85% of the bird species summarized by Clark & Wilson (1981). Hatching asynchrony in broods never exceeded 1 week, although laying could take up to 2 week, suggesting that females initiated incubation midway through egg-laying. Hatching asynchrony and mass hierarchies observed in crimson rosella broods were highly variable and not strongly related to brood size. This contrasts with most asynchronous species studied, where hatching spreads and size hierarchies increase with brood size [fieldfares (Turdus pilaris) Wiklund 1985; European blackbirds (Turdus merula) Magrath 1992; green-rumped parrotlets (Forpus passerinus) Waltman & Beissinger 1992; white-winged choughs (Corcorax melanorhamphos) Heinsohn 1995; American kestrels (Falco sparverius) Wiebe & Bortolotti 1994a], suggesting that females initiate incubation relative to the first-laid egg, leading to longer hatching spreads as clutch size increases. In crimson rosellas, the weak relationship between hatching asynchrony and clutch size and the increase in hatching asynchrony over the breeding season suggests that females varied when they initiated incubation in response to factors other than clutch size.

Some of the variation in hatching patterns observed in rosellas was due to hatching failure. Hatching failure is common in rosellas, averaging 34% of all eggs laid in this study (Krebs 1998a). Relatively high rates of hatching failure appear to be characteristic of parrots and similar levels have been described in several species [33%, long-billed corellas (Cacatua pastinator pastinator) Smith 1991; 32% (mean for all areas), white-tailed black cockatoos (Calyptorhynchus funereus) Saunders 1982; 30%, galahs (Cacatua roseicapillus) Rowley 1990; 32%, Puerto Rican parrots (Amazona vittata vittata) Snyder, Wiley & Kepler 1987]. In this study, the probability of an egg hatching did not vary with laying sequence (E.A. Krebs, unpublished data), so failure to hatch was not simply a result of neglect of last-laid eggs. Regardless of whether hatching failure occurred or not, there was substantial variation in hatching asynchrony at all brood sizes. In large rosella broods, up to four chicks could hatch on 1 day, or chicks could hatch at 1-day intervals.

Hatching asynchrony and size hierarchies increased with laying date, suggesting that more synchronous broods may be advantageous early in the breeding season. However, because large clutch sizes only occurred early in the breeding season, it was difficult to assess whether hatching asynchrony varied because of ecological factors associated with laying date, or because of differences between females. Either way, the variability of hatching patterns in rosella broods suggests that the onset of incubation varied between individual females. Mass hierarchies also increased with laying date, independent of hatching asynchrony, suggesting that female rosellas use two mechanisms to set the size hierarchy within the brood, first by varying the onset of incubation and second by increasing or decreasing the feeding rates to early-hatched chicks.

Females in some species can alter incubation patterns in response to food availability, but whether this is a result of energetic constraints during egg-laying or facultative manipulation of hatching intervals is less clear. Supplementary food in marsh tits (Parus palustris) lead to females initiating incubation earlier than control females, suggesting that the timing of incubation was energetically constrained (Nilsson 1993). Female marsh tits therefore appear to trade-off between foraging and incubation, regardless of the resulting hatching patterns. In contrast, in American kestrels, supplementary food lead females to delay incubation and reduce hatching asynchrony in broods (Wiebe & Bortolotti 1994a). Because the advantages of differing hatching patterns depended on food availability in American kestrels, females altered incubation patterns to produce levels of hatching asynchrony best suited to predicted levels of food (Wiebe & Bortolotti 1994b).

Energetic constraints did not appear to explain the incubation patterns observed in crimson rosellas. Rosellas competed intensely for nesting hollows and one or both members of the pair defended the hollow. Females spent long periods of time in the nestbox during egg-laying and presumably could begin incubation early in laying, although in some clutches they did not initiate incubation until late in the laying sequence. Female rosellas are also fed by their mate from before egg-laying until the brood is 1 week old, suggesting that they do not have to trade-off between foraging and initiating incubation. If energetic constraints determined the incubation patterns observed in rosellas, a female who was poorly fed by her mate may be forced to increase her foraging time during egg-laying and delay the onset of incubation. However, the observed seasonal decreases in clutch size and increases in hatching asynchrony in rosellas do not support this hypothesis. Pairs breeding early in the season are generally older, more experienced, or in better condition than pairs breeding late in a season [e.g. sparrowhawks (Accipter nisus) Newton 1986; European kestrels (Falco tinnunculus) Village 1990; kittiwakes (Rissa tridactyla) Thomas & Coulson 1988; female great tits (Parus major) Perrins & McCleery 1989; pied flycatcher (Ficedula hypoleuca) Harvey, Stenning & Campbell 1988]. If incubation patterns in rosellas were the result of energetic constraints, pairs breeding early would be predicted to have higher hatching asynchrony than pairs breeding later in the season, opposite to the pattern observed in this study.

Consequences of asynchronous hatching

Effects on chick size and post-fledging survival

Despite the substantial hatching asynchrony observed in crimson rosellas, variation in chick sizes at fledging and chick growth rates were unrelated to differences in hatching asynchrony and mass/size hierarchies. Rosella broods in this study never hatched synchronously, so potentially all last-hatched chicks were at a competitive disadvantage. Last-hatched chicks did not grow more slowly than first-hatched chicks. Linear growth rates of chick mass and wing length did not differ between hatching ranks in multivariate analyses and pairwise tests. In addition, last-hatched chicks did not have smaller tibias, a measure of skeletal size, than first-hatched chicks in any analysis, in fact last-hatched male chicks had significantly larger tibias (Table 3). However, asymptotic mass decreased with hatching rank in multivariate analysis, although asymptotic and mean maximum masses were lighter only for last-hatched male chicks in pairwise comparisons. Thus, last-hatched male and female rosella chicks grew at the same rate as first-hatched chicks over most of the nestling period and reached the same size at fledging. Last-hatched males however, were estimated to be lighter than their earlier hatched male siblings and potentially in poorer condition at fledging.

Differences in male mass at fledging were small and averaged 6 g, or 5% of male weight at fledging. Several lines of evidence suggest that the measured differences are not biologically meaningful. First, the high degree of hatching asynchrony in many rosella broods may result in underestimates for the mass of last-hatched chicks at fledging (see Methods). Second, fledging ages in parrots are variable, and underweight nestlings may be able to increase their condition by remaining in the nest for a few extra days (Stamps et al. 1985; Rowley 1990; Rowley & Chapman 1991). Third, post-fledging survival of last-hatched chicks was not reduced, and males with low fledging weights were not less likely to survive in this study.

It is not clear why last-hatched male and not female chicks should have lower masses at fledging. The larger sex in dimorphic species may have a competitive advantage in interactions over food (Teather 1992; Anderson et al. 1993; Price, Harvey & Ydenberg 1996). First-hatched male rosellas, particularly in broods with large hatching asynchronies, may have a competitive advantage relative to last-hatched chicks because size differences will be largest about halfway through the nestling period. Any competitive disadvantage would affect later-hatched male chicks more than female chicks because male energy requirements should increase more rapidly over the nestling period. Alternately, if male size affects subsequent breeding success, it may be advantageous for parents preferentially to feed first-hatched male, but not first-hatched female chicks. Interestingly, most last-hatched chicks in rosella broods are female (74%, 23/31 broods). Thus female rosellas may avoid producing males late in the laying sequence to minimize the costs of competitive disparities for last-hatched chicks.

Equal growth rates have rarely been described in asynchronously hatching species (Table 5). Thomas (1984) found no differences in the growth of nestlings hatched 4 days apart in maguari storks (Ciconiia maguari), although she measured only one brood. Despite hatching at 1-day intervals, nestling shag (Phalacrocorax aristotelis) had similar weights at fledging during a year of high food availability (Amundsen & Stokland 1988). In the majority of species, asynchronous hatching leads to decreased growth and increased mortality of later-hatched nestlings (Table 5). Last-hatched chicks in asynchronous broods may have reduced growth and survival because parents are not able to provide sufficient food for the whole brood, and older chicks have first access to food (e.g. Ploger & Mock 1986; Forbes & Ankney 1987; Osorno & Drummond 1995; Blanco, Yorio & Boersma 1996). Even when food is abundant, later-hatched chicks may grow poorly; for example, in green-rumped parrotlets (Forpus passerinus), supplemental feeding in large broods only slightly improved the low survival rates of last-hatched chicks and did not increase the survival of penultimate chicks (Stoleson & Beissinger 1997). Stoleson & Beissinger (1997) argued that this was a result of the inability of parents to distribute food to later-hatched chicks because of the large competitive asymmetries within the brood. Last-hatched chicks in large highly asynchronous broods may also be vulnerable to being crushed or smothered in the nest by their much larger siblings (Stoleson & Beissinger 1997).

If the size differences between chicks in asynchronous broods reflect differences in competitive abilities, equal growth rates suggest that food distribution is under parental rather than chick control. Selective feeding of chicks, not determined by competitive interactions, has been shown in several species [budgerigars (Melopsittacus undulatus) Stamps et al. 1985; pied flycatchers (Ficedula hypoleuca) Gottlander 1987; tree swallows (Tachycineta bicolor) Leonard & Horn 1996; American kestrels (Falco sparverius) Anderson et al. 1993; white-winged choughs (Corcorax melanorhamphos) Boland, Heinsohn & Cockburn 1997]. In most studies, selective feeding was only observed when food was abundant. Small and underweight chicks were selectively fed by female captive budgerigars with food ad lib (Stamps et al. 1985). Female pied flycatchers selectively fed small chicks in brood, but this preference disappeared when the brood hunger was experimentally increased (Gottlander 1987). In addition, selective feeding of last-hatched chicks, may not be sufficient to overcome their overall competitive inferiority in highly asynchronous species. In white-winged choughs, groups supplemented with food preferentially fed last-hatched chicks. Last-hatched chicks were more likely to survive in these broods, however, they still grew more slowly and had lower fledging weights than first-hatched chicks (Boland et al. 1997).

Brood reduction

Although the present study could find no effect of asynchronous hatching on the growth or survival of crimson rosella chicks for most of the period of parental care, last-hatched chicks were more likely to die shortly after hatching than other chicks. Three lines of evidence suggest that this early mortality resulted more from parental refusal to feed recently hatched chicks than competition between chicks. First, even in broods with long hatching intervals, the largest chicks in a brood would not be coordinated enough to dominate parental feeds during the first few days of a last-hatched chick's life. Chicks younger than 1-week-old rarely begged when their parents entered the nestbox and most feedings began when parents approached and offered food to inert chicks (E.A. Krebs, unpublished data). Second, newly hatched chicks that had died had empty crops. Third, videotapes showed that parents in many broods actively sought out small chicks to feed, unconstrained by chick behaviours (Krebs 1998b).

The ability to reduce brood size may be important in rosellas, not because of unpredictable food supplies, as originally suggested by Lack (1947), but because of high levels of hatching failure. Rosellas forage primarily on seeds (Forshaw 1981), and although food availability varies between seasons, mainly as a result of rainfall patterns, overall levels of food availability do not change unpredictably over the breeding season. However, hatching failure was common in rosella broods. In species with obigate siblicide, last-laid eggs appear to function mostly as replacement chicks in the event of random mortality or hatching failure (Mock & Parker 1986; Anderson 1990; Evans 1996). Last-hatched young in species with passive brood reduction can also function as replacement chicks, especially when hatching failure is high (Wiebe 1996). Forbes et al. (1997) have shown that last-hatched nestlings were more likely to survive in broods with hatching failure in red-winged blackbirds (Agelaius phoeniceus). Because high levels of hatching failure in rosellas result in initial clutch sizes not reliably correlating to brood sizes, a female may lay a larger clutch than desired and reduce the brood size through brood reduction if too many eggs hatch. Consistent with this hypothesis, brood reduction in rosellas tended to increase with the proportion of eggs successfully hatched. The probability of early brood reduction also increased over the breeding season, as would be predicted if large broods were less successful or more costly to raise late in the breeding season.

Nevertheless, it seems unlikely that asynchronous hatching in rosellas has evolved simply as an adaptation to allow brood reduction. First, brood reduction was uncommon and its occurrence did not differ between years apparently differing in food availability. Second, the probability of early brood reduction in rosellas was not positively correlated with the degree of hatching asynchrony, suggesting that large hatching intervals were not necessary to allow rapid early mortality of nestlings. Several studies have noted that hatching asynchrony is not necessary to create competitive asymmetries between chicks, and strong dominance hierarchies for food were observed even in experimentally synchronized broods (Bengtsson & Ryden 1983; Drummond, Gonzalez & Osorno 1986; Slagsvold 1986; Mock & Ploger 1987; Wiebe & Bortolotti 1995). Thus, some degree of hatching asynchrony may minimize the costs of brood reduction, but it is not clear that increasing levels of hatching asynchrony will further decrease costs.

Why do crimson rosellas hatch asynchronously?

It is difficult to identify a single factor explaining asynchronous hatching in rosellas. Breeding in rosellas, as in parrots in general, is characterized by 2-day (or longer) egg-laying intervals, relatively long incubation and nestling periods, and asynchronous hatching (Saunders, Smith & Campbell 1984; Ricklefs 1993; Krebs 1998a). Ricklefs (1993) has argued that this suite of life history characteristics is commonly associated and may relate to longevity benefits derived from longer embryonic and nestling development times and a reduction of sibling competition between developing embryos. In addition, the costs of asynchronous hatching in rosellas may be relatively low in comparison to synchronous hatching, particularly if low predation rates on incubating females and young in nesting hollows minimizes the costs of longer incubation and nestling periods. Thus, the costs of asynchronous hatching in rosellas may be mainly in maintaining an equitable food distribution to the brood, whereas synchronous hatching may lead to additional costs such as reduced egg viability (Arnold, Rohwer & Armstrong 1987), increased levels of sibling competition (Hahn 1981), or increased food demands by the brood (Hussell 1972).

Regardless of the ultimate function of hatching asynchrony in rosellas, the variation in hatching intervals and size hierarchies observed in this study suggests that the costs and benefits associated with a particular degree of hatching asynchrony vary between individuals. Optimal levels of hatching asynchrony might vary between individuals because of the age or experience of a female, her condition or the quality of her mate. Several other studies have proposed that optimal breeding behaviour may vary between individuals; for example, female magpies (Pica pica) adjusted their clutch size to match territory quality (Hogstedt 1980) and optimal clutch sizes for female great tits varied between individuals (Pettifor et al. 1988). Few studies of hatching asynchrony have focused on individual variation within populations. Wiebe & Bortolotti (1994b) showed that females could alter their levels of hatching asynchrony in response to male provisioning during egg-laying. However, the only study which has specifically tested whether individual optimization explained the variation in hatching asynchrony observed within a population, found that individual house wrens (Troglodytes aedon) did no better or worse when hatching asynchrony in their nest was experimentally altered (Harper, Juliano & Thompson 1994). In rosellas, seasonal increases in hatching asynchrony suggest that high quality females, or individuals breeding early in the season, benefit from increased synchrony of hatching. Early rosella broods, were often large, relatively synchronous and appeared extremely healthy throughout the nestling period, suggesting that relative synchrony was advantageous for some individuals. Experimental manipulations of hatching asynchrony in crimson rosellas would be useful to determine the relative costs and benefits of hatching asynchrony in this species.

Why are equal growth rates rare?

This study has shown that reduced growth and survival of last-hatched nestlings is not an unavoidable cost of hatching asynchrony, as suggested by several authors (Clark & Wilson 1981; Slagsvold 1985; Stoleson & Beissinger 1995; Nilsson & Svensson 1996). However, in most species, the costs of producing equal growth rates in asynchronous broods may be higher than the benefits. The major costs to parents of maintaining an equitable distribution of food are likely to be overriding competitive interactions between chicks (Amundsen & Stokland 1988; Stoleson & Beissinger 1995). Mock (1985) proposed that levels of sibling competition within broods are determined by the extent to which food deliveries can be monopolized by a single chick. Crimson rosellas may have lower levels of sibling competition within broods because parents distribute food through many small regurgitations over a single visit, making it difficult and perhaps costly for one chick to monopolize food. In addition, feeding visits to the brood were infrequent and load sizes were large (up to 25% of adult body weight; Krebs 1998b), which should satiate chicks of high competitive ability, and produce a more equitable distribution of food (Forbes 1993).

Another way parents can directly overcome differences in the competitive abilities of chicks is by selectively feeding small or underweight individuals. This behaviour has been observed in another species of parrot, the budgerigar, where selective feeding produced equal growth rates of nestlings in captivity (Stamps et al. 1985). Selective feeding however, reduced the efficiency with which food was distributed and so may be costly to parents (Stamps et al. 1985). Although reduced feeding efficiency may restrict the extent of selective feeding in species with high delivery rates, the costs should be relatively low in rosellas because of infrequent food deliveries. Thus rosellas may be able to use selective feeding of last-hatched nestlings to maintain equal growth. If there are competitive differences between male and female nestlings in broods, the costs and benefits of selective feeding to parents may also differ between the sexes within a brood.

Equal growth rates are not characteristic of parrots as a group (Table 5), but several aspects of their breeding biology may allow the evolution of more complex food allocation within the brood. Parrots have long nestling periods for their body size, twice that of Columbiformes, and 60% higher than Falconiformes (Saunders et al. 1984), probably because of reduced predation pressure during the nestling stage as hole nesters (Lack 1968). Chick ages at fledging are highly variable in rosellas and other species of parrots, and fledging can be very asynchronous (E.A. Krebs, unpublished data; Rowley 1990; Rowley & Chapman 1991; Waltman & Beissinger 1992). In addition, post-fledging survival in parrots does not appear to be strongly dependent on size or condition of nestlings at fledging (Rowley & Chapman 1991; Smith 1991; Stoleson & Beissinger 1997; this study). These traits may allow parents to engage in provisioning behaviours that potentially reduce the growth of all members of the brood but maximize the number of nestlings produced.

Acknowledgements

I thank Rob Magrath for his excellent theoretical and editorial advice throughout the project. Andrew Cockburn, David Green, Charley Krebs, Sarah Legge and Rob Magrath made many helpful comments on earlier versions of this manuscript. Thanks also to David ‘sherpa’ Green, for his exceptional help with all aspects of the fieldwork. Benj Whitworth and Kate Trumper also provided stoical and invaluable help in the field. This project would not have been possible without the excellent technical support from Bruce Barrie, Alan Muir and Frank Train, who built and perfected the design of my nestboxes. I was supported during this study by an ANU/OPRS postgraduate scholarship and the project was partially financed by the M.A. Ingram Trust, the Australian Federation of University Women and a Cayley Memorial Scholarship.

Received 17 February 1998; revision received 16 June 1998

Ancillary