• clutch size;
  • individual optimization;
  • laying date;
  • reproductive success;
  • seasonal decline


  1. Top of page
  2. Summary
  3. Journal of Animal Ecology (2000), 69, 000–000
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

1. Even though feeding conditions typically improve over time during the laying period, clutch size decreases over the course of the nesting period in most bird species. We examined whether seasonal decrease in offspring value could explain the seasonal decline in clutch size in arctic-nesting greater snow geese (Anser caerulescens atlanticus L.).

2. Nesting was synchronized within a year, with more than 90% of the nests being initiated within about 8 days. Despite this high synchrony, there was a steep seasonal decline in clutch size in each of the 7 years of the study, from about five eggs in early clutches to three in late ones (−0·20 egg day−1).

3. Late parents performed more poorly than early parents in most components of reproductive success. The relationship between laying date and nesting success was curvilinear, early and late nests having a higher failure rate than those initiated near the median. Prefledging survival decreased by about 50% over the season, although the earliest hatched goslings also tended to have a reduced survival. The postfledging survival showed the strongest seasonal decline, as survival probability of late-hatched birds was about five times lower than in early-hatched ones.

4. Overall reproductive success showed a very steep seasonal decline as the number of young surviving to the first winter was about eight times lower in late-nesting birds than in early-nesting ones. Reproductive success declined slightly in the earliest-nesting birds, suggesting a cost to nesting too early.

5. The observed clutch size generally matched the clutch size that yielded the highest reproductive success for each laying date, except in earliest-nesting birds which should have done better by slightly delaying nesting. Our data suggest that trading off additional eggs for earlier nesting to increase reproductive success is an option in geese. Consequently, the seasonal decline in clutch size may be an adaptive response to seasonally declining survival prospects of offspring.


  1. Top of page
  2. Summary
  3. Journal of Animal Ecology (2000), 69, 000–000
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

Many organisms live in seasonal environments where fitness costs and benefits associated with survival and reproduction vary over time. Parents should therefore time their reproductive effort to maximize the number of viable offspring produced (Perrins 1970). When to start laying (i.e. laying date) and how many eggs to lay (i.e. clutch size) are two of the most critical decisions that breeding birds face (Perrins & McCleery 1989; Daan, Dijkstra & Tinbergen 1990).

Two paradoxes emerge from most studies of seasonal variation in reproductive success. The first paradox is that the laying date which maximizes reproductive success is typically earlier than the laying date of most birds in the population (Lack 1968; Perrins 1970; Perrins & Moss 1975). A second paradox is the seasonal decline in clutch size. As a rule, early-nesting birds lay a larger clutch size than late-nesting ones (Klomp 1970). Because most bird species of temperate areas nest in spring when food resources are increasing, late-nesting birds should have access to higher food availability at the time of laying and should thus be able to lay more eggs than early-nesting ones (Perrins 1970; Daan et al. 1988). It therefore appears that late-nesting birds are behaving suboptimally, both in terms of their decisions about laying date and clutch size.

One hypothesis to explain these paradoxes is that the seasonal decline in clutch size and reproductive success is simply an epiphenomenon resulting from a direct effect of body condition on both laying date and reproductive success. (i) If individuals in good body condition both breed earlier and have higher reproductive success, and (ii) if body condition is not heritable, then a negative correlation between reproductive success and laying date will persist even though laying date is heritable (Price, Kirkpatrick & Arnold 1988; a similar argument can be applied to clutch size, Price & Liou 1989). Results of food addition experiments during prelaying have been taken as evidence for a causal effect of body condition on both laying date and reproductive success. Although food-provisioning almost always advances laying date, many experiments failed to find an effect on clutch size (Daan et al. 1988; Meijer, Daan & Hall 1990; but see Korpimäki & Wiehn 1998). In the tree swallow (Tachycineta bicolor Vieillot), Winkler & Allen (1996) rejected the hypothesis that variations in condition alone could explain the seasonal decline in clutch size. In waterfowl, there is good evidence that body condition influences clutch size (Ankney & MacInnes 1978; Raveling 1979; Ankney, Afton & Alisauskas 1991; Mann & Sedinger 1993) but evidence for an effect on number of offspring produced is weak (Teunissen, Spaans & Drent 1985; Ebbinge & Spaans 1995).

An alternative explanation to these paradoxes is that reproductive decisions are optimized at the individual level. Högstedt (1980) and Pettifor, Perrins & McCleery (1988) provided such evidence for clutch size. Drent & Daan (1980) and Daan et al. (1990) further proposed that optimal clutch size will covary with laying date and that both should therefore be optimized, an idea that was recently formalized into a model by Rowe, Ludwig & Schluter (1994). This hypothesis assumes that improvement in body condition of individuals takes time and that condition will vary among individuals. Because of the fitness costs associated with delayed nesting and the time required to gather enough nutrients for egg production, their model predicts an optimal combination of clutch size and timing for every individual based on its ability to acquire nutrients from the environment (Daan et al. 1990; see also McNamara & Houston 1996 for additional details on state-dependent approaches). At any given date, the relative costs (lower offspring value) and benefits (higher clutch size) of delaying nesting will depend on the time required to accumulate more nutrients (given a specific rate of accumulation for each individual) and will determine whether a female should start laying now or wait to lay a larger clutch later. Seasonal change in clutch size is here viewed as a strategic adjustment in response to temporal change in body condition of the female and survival prospect of offspring (Rowe et al. 1994). Because of the difficulties in obtaining the appropriate data, few studies have been able to provide a test of this model (Perrins & McCleery 1989; Daan et al. 1990; Smith 1993; Winkler & Allen 1996; Korpimäki & Wiehn 1998).

In this study, we examined whether the seasonal decrease in offspring value could explain the seasonal decline in clutch size in the greater snow goose (Anser caerulescens atlanticus L.), a precocial species nesting in the High Arctic, where seasonal change in environmental conditions is very marked. In geese, Ryder (1970) originally proposed the nutrient re-allocation hypothesis to explain the seasonal decline in clutch size. According to this hypothesis, birds delaying nesting had to deplete their nutrient reserves accumulated on spring staging areas to survive, and hence had less body reserves left to invest in eggs. This depletion occurred because the short arctic summer presumably forced geese to initiate laying so early in spring that food was not available during egg-laying. Since then, however, many studies, including in greater snow geese, have shown that laying geese actually feed on the breeding grounds, and that these nutrients could contribute to the formation of the eggs (Budeau, Ratti & Ely 1991; Bromley & Jarvis 1993; Choinière & Gauthier 1995; Ganter & Cooke 1996). Moreover, birds in poor condition can also delay arrival on the breeding grounds to improve their condition at staging areas further south. Therefore, the nutrient re-allocation hypothesis, as originally stated, cannot explain why late-nesting geese lay fewer eggs when these birds have opportunities to improve their body condition while waiting to initiate laying (see also Hamann & Cooke 1989).

The individual optimization hypothesis predicts that observed clutch sizes are optimum for each laying date and that parents could not do better by laying smaller clutches earlier or larger clutches later (Daan et al. 1990; Rowe et al. 1994). To test this prediction, we analysed the seasonal variation in several components of reproductive success in greater snow geese, from clutch size to postfledging survival, and examined whether the individual combinations of clutch size and laying date observed in the population were the most productive in terms of number of offspring surviving to the first winter.


  1. Top of page
  2. Summary
  3. Journal of Animal Ecology (2000), 69, 000–000
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

Field methods

The study was conducted on Bylot Island, Nunavut, Canada (73°08′N, 80°00′W). From 1991 to 1997 extensive nest searches were conducted during the laying and incubation periods in June. Nests were revisited in the first half of incubation, during the hatching period and after the goslings had left in order to determine the outcome. Eggs were marked individually and signs of nest predation were noted on each visit. At hatch, we marked goslings before they left the nest with numbered web-tags (see Lepage, Gauthier & Reed 1996 for details of the area and methods).

Parents with their young were captured about 5 weeks after hatching, shortly before fledging, in mass-banding drives. At that time, moulting adults are unable to fly because of the loss of their flight feathers. Goslings were checked for the presence of web-tags and all captured birds were marked with US Fish & Wildlife metal bands. A second recapture event for web-tagged birds occurred in the autumn when tags from birds killed by hunters were returned to us by outfitters or when tagged birds were captured in cannon-nets during banding operations on the autumn staging area in southern Quebec (Lesage & Gauthier 1997). To estimate postfledging survival, we used bands returned by hunters to the Bird Banding Laboratory (see below).

Laying date, clutch size and survival estimates during nesting

Laying date (date of the first egg laid; see Lepage, Desrochers & Gauthier 1999 for details) was transformed as the deviation from the median laying date in each year (relative laying date). For web-tagged goslings, hatching date was defined as the date at which at least half the brood was hatched. Hatching was synchronized within a brood, most eggs usually hatching within 24 h. Again, hatching date was transformed as deviation from the median hatching date in each year (relative hatching date). Total clutch laid (TCL) is the number of eggs laid in a nest and was defined as the maximum number of eggs found in a nest. Nests with a TCL of only one egg accounted for a small proportion of nests found (4%) and they were removed from the sample because many of them result from partial predation (Gauthier, personal observation). Clutch size at hatch (CSH) is the number of eggs in nests where at least one egg hatched. Finally, GLN is the number of goslings leaving the nest.

Following Rockwell et al. (1993), we calculated three transition probabilities at the nesting stage: nest success (NS) is the proportion of nests where at least one egg hatched successfully, and was calculated using the Mayfield method (Johnson 1979); egg survival (P1 = CSH/TCL) is the proportion of eggs surviving to hatch in successful nests; hatchability (P2 = GLN/CSH) is the proportion of surviving eggs that hatched. The transition probabilities NS, P1 and P2 were estimated for each date over the range of laying dates using the maximum sample available for each component.

Because TCL is related to laying date, and because the lowest bounds of P1 and P2 are not independent of TCL (e.g. the lowest value possible for P1 is 0·5 when TCL = 2, and 0·25 when TCL = 4), we could not test directly the effect of laying date on P1 and P2. We thus used the method of Rockwell et al. (1993) to evaluate seasonal effects on egg survival (P1) and hatchability (P2). We calculated DevP1 and DevP2 as deviations from the expected value of CSH and GLN under the null hypothesis that P1 and P2 did not depend on laying date but only on year. We calculated expected clutch size at hatch, E(CSH), and DevP1 for each parent i in year y as:

  • image

Similar equations were used to estimate E(GLN) and DevP2. We used general linear models (SAS Institute Inc. 1996) to test the effect of relative laying date (d and d2) and year on individual values of DevP1, DevP2 and TCL and weighted stepwise regression on average values of NS, using the inverse of the standard error as a weighting factor.

Pre-fledging survival estimates

We used the surge program (version 5) to obtain estimates of prefledging survival using the capture–recapture data of web-tagged goslings (Lebreton et al. 1992). Each year, we had one marking event (web-tagging at hatching) and two recapture events (banding period and autumn hunting in Quebec), making four possible capture–recapture histories each year. We tested the effect of year (y), hatching date (h; divided into seven groups according to the relative hatching date of individuals: −6 to −4, −3 and −2, −1, 0, +1, +2 and +3, and from +4 to +6), and time (t; pre- and postfledging periods) on the probabilities of survival (φ) and capture (p).

A potential problem with this approach is that early-hatched goslings recaptured during banding had a longer exposure period than late-hatched ones. This longer exposure period will negatively bias survival in the former group. However, this bias should be quite small because the first two weeks after hatching is a period of much higher mortality for goslings than the subsequent three weeks (Williams et al. 1993). Therefore, despite the difference in length of exposure, the difference in mortality risk between early- and late-hatched goslings is actually much smaller because both groups had an equal exposure to the period of highest mortality risk. Another potential problem is that young recaptured at banding also received a metal leg band which may have increased reporting rate by hunters in the autumn compared to young with only web-tags. Although we minimized this bias by using only data from outfitters that were trained by us to look for the presence of both markers, this heterogeneity may have negatively biased survival. However, this bias should be independent of hatching date.

Because our sample size did not allow estimation of parameters in the most general model with all three effects simultaneously included (this model φy*h*tpy*h*t, where both survival and capture probabilities are specific to year, hatching date and time, had 196 parameters when considering all possible interactions), our two most general models had a maximum of 28 parameters (i.e. φh*tpy*t and φy*tph*t). We tested additional models where φ or p were constrained equally among hatching dates or years to test the significance of these variables following the procedure outlined by Lebreton et al. (1992). We also tested whether φ of the first time interval (t1) could be related through either a linear or a polynomial regression (second or third degree) to hatching date using linear constraints in surge.

Post-fledging survival estimates

We used the recovery rate (R) of goslings marked with US Fish & Wildlife metal bands near fledging to obtain an index of their postfledging survival, because all goslings recovered through hunting in their first winter were known to have survived at least the autumn migration. Although R underestimates true survival, it can nonetheless be used to test the effect of hatching date of postfledging survival if the probability of shooting a bird in winter is independent of its hatching date (Menu 1998). Even though we cannot test this assumption, we have evidence that hunting vulnerability is not related to body condition in this species (Morez 1997).

We used logistic models (SAS Institute Inc. 1996) to test the effect of relative hatching date on the probability of being recovered. Hatching date of all banded young was calculated from the estimated age at banding using a year-specific linear relationship between age of web-tagged goslings recaptured and length of their ninth primary feather (Lepage, Gauthier & Reed 1998).

Composite measures of reproductive success

We calculated composite measures of reproductive success following the method used by Rockwell et al. (1993). We used the product of average TCL, the transition probabilities during the nesting stage (NS, P1 and P2), prefledging survival (P3 as estimated by φ) and postfledging survival (P4 as estimated by R) to provide a final index of individual reproductive success (RS = the number of goslings surviving to first winter) for each relative laying date d:

  • image

Because P3 and P4 were measured with respect to hatching dates, we had to find the relative hatching date (d′) corresponding to each clutch size laid on a given relative laying date (d). This was achieved using the equation:

  • image

This correction is based on the median clutch size which is four eggs. Each egg added or removed from this value should result in a delay or an advance of hatching date by 1 day, respectively. The 1-day value was estimated by the difference (rounded to the nearest integer) between observed median relative laying and hatching dates for each clutch size. For instance, whereas a clutch of four eggs initiated on relative day 0 would have hatched on day 0, a clutch of six initiated on the same day would have hatched on day +2.

We generated 95% confidence intervals for each RS estimate with 10 000 Monte-Carlo simulations (see appendix 2 in Rockwell et al. 1993), and we converted them to standard error by dividing the confidence interval by 1·96. The effect of relative laying date on E(RS) was evaluated with an ordered series of weighted polynomial regressions, using the AIC to select the best model, and using the inverse of the standard error as weighting factor for each RS estimate.

In order to evaluate the consequences of individual egg-laying decisions (laying date and number of eggs laid) in terms of reproductive success, we estimated the expected probability of individual eggs laid at different dates producing viable offspring at two stages. We first calculated the expected probability of producing a viable offspring leaving the nest (OS), as the product of the average successive survival probabilities, again calculated for each relative laying date d:

  • image

When combining OS with the pre- and postfledging survival probabilities (P3 and P4), we could estimate the final reproductive success (i.e. the number of offspring surviving to winter) of individual parents laying a hypothetical clutch size C initiated at date d as follows:

  • image

In this equation, we used the point estimates of OS, P3 and P4 obtained from the relationships with relative laying date rather than actual average values. The relationship between OS and laying date was obtained with an ordered series of weighted polynomial regressions as described for RS. Again, because P3 and P4 were measured with respect to hatching dates and not laying dates, we used:

  • image


  1. Top of page
  2. Summary
  3. Journal of Animal Ecology (2000), 69, 000–000
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

Nesting components

The number of nests monitored each year ranged from 143 in 1992 to 755 in 1994. Median laying dates varied from 6 to 20 June between 1991 and 1997, whereas median hatching dates were between 3 and 15 July. Laying was earliest in 1993, late in 1992 and 1996, and average in the four other years (Fig. 1). Nesting was very synchronized within a year, with usually more than 90% of the nests being initiated in about 8 days. No female was known to have attempted laying a second brood following a nest failure.


Figure 1. Distribution of laying date of the first egg (hatched bars) and hatching date (solid bars) in greater snow geese at Bylot Island from 1991 to 1997. Stippled line is the median for each variable.

Download figure to PowerPoint

Mean clutch size (TCL) was 3·91 ± 0·02 (SE) (n = 2253). Late-nesting females laid fewer eggs than early-nesting females (Fig. 2a). The seasonal decline varied from −0·11 to −0·26 egg per day of delay and was significant in all years (Table 1), although the rate of decline differed among years as shown by the significant interaction between year and laying date.


Figure 2. Relationship between (a) total clutch laid, (b) egg survival, and (c) hatchability (means with standard errors), and laying date, all years combined (1991–97). Egg survival and hatchability are illustrated as deviation (DevP1 and DevP2, respectively) from null expected values for a given year (see text for more details). Laying date is standardized relative to the median (0 = median laying date in each year). Sample size is shown above each point.

Download figure to PowerPoint

Table 1. . General linear models testing the effect of year and laying date on total clutch laid (TCL), egg survival (DevP1) and hatchability (DevP2). Slopes were calculated with the estimate statement of sas proc glm (SAS Institute Inc. 1996)
  1. F-values. ***P ≤ 0·001; **P ≤ 0·01; *P ≤ 0·05; NS P > 0·05.

  2. ‡Slopes in egg day−1. ***P ≤ 0·001; **P ≤ 0·01; *P ≤ 0·05; otherwise: P > 0·05.

n 225311541143
R2 0·2830·0150·013
Analysis of variance tablemodel laying date67·99*** 413·81***1·34 NS 1·61 NS1·15 NS 1·65 NS
year47·72***0·10 NS0·14 NS
laying date*year5·46***2·58*2·22*
Estimates of slopes1991−0·230***−0·101*0·097**
all years−0·203***−0·0150·012

Mean egg survival (P1) was 89·7 ± 0·9% (SE) (n = 1154). Overall, laying date had no significant effect on DevP1 (Fig. 2b, Table 1). However, the interaction between year and laying date was significant, a consequence of slight negative effect of date in 1991 and 1994 (0·10 and 0·05 higher egg loss per day of delay, respectively; Table 1).

Mean hatchability (P2) was 92·8 ± 0·8% (SE) (n = 1143). Overall, DevP2 was not related to laying date (Fig. 2c, Table 1). Again, the interaction term between year and laying date was significant as laying date had a slight but significant effect in 1991 (a seasonal increase in hatchability; Table 1). Thus, effects of laying date on egg survival and hatchability were absent in most years, and small and inconsistent in the few years where present.

Overall nesting success (NS) averaged 63·3 ± 2·1% (SE) (n = 2068) although it was quite variable among years. There was a significantly higher failure rate in early and late nests compared to average ones in 1992, 1995 and 1997, whereas nest success declined linearly in 1993 and 1996 (Fig. 3). The relationship between NS averaged over the 7 years and relative laying date also showed a significantly higher failure rate for early and late nests. The earliest (from relative date −6 to −4) and the latest (from date +4 to +6) nests were about 10% less successful on average than nests initiated closer to the median date.


Figure 3. Relationship between nesting success (mean ± SE) and laying date from 1991 to 1997, and for all years combined (n = 2068). Laying date is standardized relative to the median (0 = median laying date in each year). In all years, we grouped early and late nests together because sample sizes were too low. The regression curves (solid lines) were obtained with weighted least-squares (see Methods).

Download figure to PowerPoint

Pre- and postfledging survival

The number of goslings web-tagged at hatch each year was 315 in 1991, 289 in 1992, and between 1099 and 1987 afterward. Numbers recaptured each year at banding ranged from 47 to 102 (except in 1991 where n = 17), and from 4 to 27 during the autumn.

Mean survival rate of goslings from hatch to fledging was 39%. Among models without linear constraint, the model φh*tpy*t best described the capture–recapture data of web-tagged goslings, as indicated by its AIC (Table 2). This suggests that survival was specific to hatching date, and that capture rate was specific to the year. Removal of the hatching date effect on the survival estimates (model φtpy*t compared to model φh*tpy*t) caused a significant increase of the deviance (P < 0·001), indicating that survival differed among goslings hatched at different dates. The hypothesis that prefledging survival was a simple linear function of hatching date was rejected, but the hypothesis of a curvilinear relationship with hatching date was accepted as the model where survival during the first interval (t1) was constrained by a third level polynomial regression [cnstr (h3 + h) φh*tpy*t] was the most parsimonious (Table 2). Gosling prefledging survival was maximal for birds hatched on days −3 and −2 (50%), and minimal for goslings hatched after the peak (24–30%; Fig. 4).

Table 2. . Model selection for gosling prefledging survival with surge. The Akaike Information Criterion (AIC = deviance + 2 × number of parameters estimated) with the lowest value indicates the best models (shown in bold for models with and without linear constraints). n = 8939 goslings web-tagged at hatch
  • †Indices for the parameters φ (survival) and p (capture probability) are as follows. h: hatching date (divided into seven groups; see Methods); y: year (1991–97); t: time period (pre- and postfledging period). Cnst (h) φht indicates that the survival parameters of the first time interval (t1) were tested for a linear regression (or polynomial regression for h2 and h3) with hatching date.

  • NP = number of identifiable parameters.

cnst (h3 + h2 + h) φh*tpy*t2344224468
cnst (h2 + h) φh*tpy*t2444314479
cnst (h) φh*tpy*t2244314475

Figure 4. Relationship between prefledging survival of gosling (φ ± SE) and hatching date (h). The regression curve (solid line) and standard errors (dotted lines) correspond to the equation given and were obtained using a linear constraint of hatching date on survival with surge (see text). Hatching date is standardized relative to the median (0 = median hatching date in each year). We grouped early- and late-hatched birds together because sample sizes were too low.

Download figure to PowerPoint

The number of goslings banded at fledging each year ranged from 950 to 2308 and numbers recovered in the autumn each year ranged from 90 to 263. Both hatching date and year had a significant effect on postfledging survival (Table 3). The logistic analyses indicated that the model with a quadratic term (h2) provided a better fit than a first-order regression. Postfledging survival of goslings was high until about day −2 but showed a very strong seasonal decline thereafter (Fig. 5). Inclusion of an interaction term between year and hatching date improved the fit of the model, indicating that the seasonal effects differed among years (Table 3). Although year-specific relationships differed in shape, they all showed a strong negative relationship between postfledging survival and hatching date (Menu 1998).

Table 3. . Model selection for gosling postfledging survival. The Akaike index criterion (AIC = deviance + 2 × number of parameters estimated) with the lowest value indicates the best model (shown in bold). n = 12 289 goslings banded at fledging
ModelAIC−2 Log L
  • h: hatching date, y: year.

h + h2 + y + y*h + y*h275907548
h + h2 + y + y*h75927562
h + h2 + y76437625
h + h282078201
h + y + h*y76547626
h + y76887672
Intercept only88518849

Figure 5. Relationship between recovery probability (mean ± SE), an index of postfledging gosling survival and hatching date (h). The regression curve (solid line) and standard errors (dotted lines) correspond to the equation given and were obtained from the logistic model (see text). Laying date is standardized relative to the median (0 = median laying date in each year).

Download figure to PowerPoint

Reproductive consequences of individual egg-laying decisions

There was a nearly constant decline of reproductive success (RS) during the season, the earliest-laying birds producing about eight times as many offspring surviving to the first winter as the latest birds (Fig. 6). The fourth-order regression explained 99·6% of the variation in the data and was the most parsimonious model based on its AIC. The observed RS was slightly lower on average for birds laying before day −4, but the error associated with this estimate was large because of the small sample size. We could not estimate RS at laying dates −6 and +6 because the corresponding hatching date fell outside the period for which we could estimate P3.


Figure 6. (Top) Composite measure of reproductive success (mean ± SE number of young surviving to the first winter) calculated for each laying date. Standard errors were estimated with Monte-Carlo simulations and the regression curve (solid line) corresponds to the equation given (see Methods). Dotted lines illustrate 95% confidence limits for the regression. (Bottom) Expected reproductive success (RS) according to laying date. Each curve represents RS for females laying a different clutch size (TCL) over the range of possible clutch size. Dots associated with the heavy line show the RS for the median clutch size observed in the population at each laying date (see text for details). Laying date is standardized relative to the median (0 = median laying date in each year).

Download figure to PowerPoint

The expected RS associated with each combination of clutch size and laying date is given in Fig. 6. According to this model, the median clutch size laid by geese yielded the highest possible RS for all laying dates after day −4. In fact, after day −4, the seasonal decline in RS was so steep that even if a bird had a body condition sufficient to lay an additional egg at the onset of laying, the 1-day delay incurred at hatching was sufficient to eliminate the potential gain in RS from this additional egg. One should note, however, that the discriminatory power of the model becomes low for laying dates > +2 because all RS curves converged.

There were, however, some discrepancies between the predictions of the model and observed clutch size. First, birds initiating laying on days −5 and −4 would have a higher RS by laying an additional egg (6 vs. 5 eggs). However, this assumes that birds already had a condition sufficient to lay an extra egg on those dates. If not, then the observed clutch size could still be the one yielding the highest RS, providing that the time required to accumulate the nutrients to lay an additional egg would delay laying by more than 2 days (for birds that laid on day −5) or 1 day (for those on day −4). Second, some birds may have had a higher RS by laying on a different date without changes in body condition. For instance, birds that laid five eggs on day −5 would obtain a higher RS by laying the same clutch size one day later (day −4). Likewise, birds that started laying on day 0 or later could obtain a higher RS by laying fewer eggs earlier.


  1. Top of page
  2. Summary
  3. Journal of Animal Ecology (2000), 69, 000–000
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

Seasonal variations in reproductive success

Although there was some variation among years, late parents consistently performed more poorly than early parents in most components of reproductive success. Clutch size and pre- and postfledging survival showed the strongest seasonal effects. The resulting value of reproductive success showed a continuous seasonal decline after day −4. Our data also suggest that the earliest birds did not achieve the maximum reproductive success, indicating a possible cost associated with nesting too early.

The curvilinear relationship between nesting success and date (i.e. lower success of earliest and latest nests) has been reported before in lesser snow geese A.c. caerulescens L. (Findlay & Cooke 1982) and other species (Brinkhof et al. 1993; Norris 1993; Nilsson 1994). Differential susceptibility to predation of nests initiated at different dates may explain this seasonal pattern. This is especially true for colonial species where success may depend on a predator swamping effect.

Predation may also explain in part the observed seasonal variation in prefledging survival, the earliest-hatched goslings being more susceptible to predation because of lack of a swamping effect. In addition, because late-hatched goslings have access to less food (Lepage, Gauthier & Reed 1998), and are more exposed to harsher (e.g. colder) environmental conditions, they are potentially more susceptible to predation, diseases and exposure than early goslings (Lindholm, Gauthier & Desrochers 1994; Renaud 1999). Late-nesting parents can also be less efficient against predators if they have to devote more time to feeding at the expense of vigilance (Sedinger, Eichholz & Flint 1995). In a manipulation experiment where we exchanged clutches between early and late parents, we did not find any effect of parental quality on either offspring growth rate or prefledging survival (Lepage, Gauthier & Reed 1999), suggesting that environmental conditions were largely responsible for the observed decline.

The strong seasonal decline of postfledging survival that we observed is likely to result from differences in body size and condition of goslings at the time of fledging (Francis et al. 1992; Sedinger, Flint & Lindberg 1995). Because late-hatched birds have less time to complete their body development and because food is less abundant for them, they fledge with a lower body condition and at a later date than early-hatched birds (Lesage & Gauthier 1997, 1998), and are thus less likely to survive until their first winter (Menu 1998).

The seasonal patterns of variation in reproductive success observed in this study are thus similar to those reported in other species (e.g. Perrins & McCleery 1989; Daan et al. 1990; Brinkhof et al. 1993; Smith 1993). What is perhaps unusual in this study is the steepness of the seasonal decline which occurs over a very short period of time, about 12 days. This is a consequence of the constraints faced by a species with a relatively long breeding cycle reproducing in an environment (i.e. the Arctic) where the summer season is very short. The optimal time window to breed is therefore small and a short delay in nesting can result in a high cost in terms of reproductive success. Although the steepness of the decline may facilitate the study of seasonal variations and the analysis of the trade-off between laying date and clutch size, we do not believe that this limits the relevance of our conclusions for species inhabiting environments where seasonal variations are not as rapid. For instance, Daan et al. (1990) found a similar trade-off in Eurasian kestrels (Falco tinnunculus L.), a species inhabiting temperate areas where the breeding season is much longer and the steepness of the decline much shallower.

Is there a trade-off between laying date and clutch size?

Our results suggested that trading off an additional egg for an earlier laying date, and thereby obtaining an equal or even a higher reproductive success, is an option in greater snow geese. There was, however, some discrepancy between the predictions of the model based on expected reproductive success and observed clutch size. First, when reproductive success increases early in the season before declining, optimal laying dates will always fall on the declining portion of the curve because in this case there is no cost in delaying nesting (Rowe et al. 1994). Therefore, the observed clutch size of five eggs laid by birds on day −5 and earlier was suboptimal, as delaying until day −4 would have increased reproductive success even without adding an extra egg (Fig. 6). This discrepancy could result from unknown biases in our estimate of reproductive success, especially very early in the season when sample sizes are small. Alternatively, this may represent birds that made a wrong decision in their choice of laying date and/or clutch size. Variable nesting conditions among years (e.g. Lepage et al. 1996) could make the choice of the earliest-nesting birds suboptimal in some years but not in others. One should note, however, that birds nesting on day −5 or earlier account for a very small fraction of the breeding population (< 4%).

A second discrepancy between the predictions of the model and observed results is that the seasonal decline in clutch size should have been steeper than that we observed. Indeed, birds laying four eggs on day 0 or +1 would have done better by laying only three eggs earlier, and birds laying three eggs from day +2 to +6 would have done better by laying only two eggs earlier, assuming that they had the condition to do so at that time (Fig. 6). Why then did these birds not lay a smaller clutch earlier? One possible explanation is that there may be additional benefits to laying a large clutch size in geese that are not taken into account in this model. It has been shown that social dominance, and thus access to food resources, is positively related to family size in geese (Prop, van Eerden & Drent 1984; Gregoire & Ankney 1990). An experimental manipulation of brood size in this species suggested that this effect was a direct result of the number of young per se and not a correlate of parental quality (Lepage, Gauthier & Desrochers 1998). Because juveniles from large families grow faster and thus recruit at higher rate, laying a large clutch may compensate for some of the cost of the delay.

After day +1, the expected reproductive success for birds laying different clutch sizes all converged. This means that a female that had accumulated enough nutrients to lay three eggs on day +2, for instance, would achieve about the same reproductive success by reabsorbing one egg in order to advance hatching. If there is some physiological cost associated with laying ‘unnecessary’ eggs, then females should always opt for the smallest clutch size in this situation. This is not what we observed, as most birds laying after day +1 had clutches of three eggs although they could potentially have achieved a similar reproductive success by laying two eggs.

The model of Rowe et al. (1994) assumes that eggs are costly to produce in terms of nutrients and that clutch size depends on body condition acquired during the prelaying period. In greater snow geese, much of the fat and protein required for the production of eggs is acquired during the prelaying and laying periods in the Arctic (Choinière & Gauthier 1995). The rate of nutrient accumulation will obviously influence the trade-off between laying date and clutch size at the individual level. The rates of nutrient accumulation in tissues (both somatic and reproductive) is on average 2·3 g day−1 for fat and 7·4 g day−1 for protein at that time. Because one egg contains about 14·5 g of fat and 17·5 g of protein (Gauthier, unpublished data), the time required to acquire the nutrients for one egg is about 6·3 days for fat and 2·4 days for proteins. Peak rate of nutrient accumulation can be higher as it was estimated at 14 g day−1 of fat and 12 g day−1 of protein around the onset of laying (Choinière & Gauthier 1995), a period when food becomes more readily accessible (Gauthier 1993). Our analysis nonetheless suggests that delaying nesting to acquire more nutrients and hence lay more eggs could not compensate for the cost of the delay over the range of laying dates observed in snow geese.

Some studies have found evidence that a trade-off between laying date and clutch size is involved in reproductive decisions. In lesser snow geese, the analysis by Cooke, Rockwell & Lank (1995) suggested that the earliest-laying females could have an equal reproductive success by delaying nesting to lay an additional egg, but the seasonal decline in reproductive success was much less steep than in this study. In the Eurasian kestrel, such a trade-off explained relatively well the seasonal decline in clutch size (Daan et al. 1990). However, in great tits (Parus major L.), Perrins & McCleery (1989) concluded that the seasonal decline in reproductive success was not sufficient to account fully for the seasonal decline in clutch size observed in this species.

A weakness of our analyses is that they are limited to annual reproductive success. Therefore, we do not account for possible costs associated with reproduction; individuals may reduce their reproductive effort in one year in order to maximize their lifetime reproductive output (Williams 1966; Charnov & Krebs 1974). Rowe et al. (1994) explored the consequence of including a cost of reproduction in their model. They concluded that in the presence of a significant cost, females should both lay earlier and have smaller clutches. This is contrary to the trend observed in this study where, for some laying dates (especially late ones), geese tended to lay later, larger clutches than predicted by the model. Moreover, the few studies that examined this hypothesis in geese found little evidence for a cost of reproduction (Lessells 1986; Williams, Loonen & Cooke 1994; Tombre & Erikstad 1996), although measuring such costs is very difficult in wild populations.

Individual optimization of clutch size and laying date

Optimization of reproductive decisions at the individual level has been found in some studies. In the great tit, Pettifor et al. (1988) showed that parents were producing the optimal number of offspring that they were able to fledge; experimentally adding or removing young both resulted in a lower number of viable offspring produced. The costs and benefits associated with timing of reproduction may also vary among individuals according to their ability to raise young or to accumulate nutrients for egg formation. However, only Daan et al. (1990) provided good evidence that birds can optimize both clutch size and laying date at the individual level. The use of composite measures of reproductive success in our analyses precluded the analysis of reproductive trade-off at the individual level.

Our results nonetheless provided evidence that the seasonal decline in clutch size in birds is partly adaptive, that is when reproductive success declines seasonally, late-nesting birds can increase their reproductive success by laying fewer eggs than early-nesting ones. However, in greater snow geese this strategic adjustment in clutch size was not sufficient to compensate the seasonal decline in offspring survival, such that late-nesting individuals always experienced a lower reproductive success than early-nesting ones. This suggests that some proximal constraint acts on laying date and prevents earlier laying in many individuals. Rate of gain in body condition in spring is likely to be a major proximal constraint on laying date (Rowe et al. 1994). In contrast, both constraint (e.g. body condition of the female) and restraint on the part of the female (i.e. adjustment of clutch size in response to survival prospect of offspring) may act on clutch size determination.


  1. Top of page
  2. Summary
  3. Journal of Animal Ecology (2000), 69, 000–000
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

Funding was provided by a Natural Sciences and Engineering Research Council of Canada (NSERC) grant to G. Gauthier, Ducks Unlimited Canada, the Arctic Goose Joint Venture (Environment Canada), the Fonds pour la Formation de Chercheurs et l’Aide à la Recherche (FCAR, Ministère de l’Éducation du Québec) and the Department of Indian and Northern Affairs Canada. The NSERC, FCAR and Centre d’Études Nordiques provided financial assistance to D. Lepage. Logistic support was generously provided by the Polar Continental Shelf Project (Natural Resources Canada). Thanks to all the people who participated in the field work, especially J. Bêty, R.J. Hughes, S. Lapointe, A. Lindholm, D. Leclerc, T. Pewatoaluk, G. Picard, Y. Poirier, C. Poussart, A. Reed, M. Salathé, J.-P. Tremblay and F. Villeneuve. We also thank A. Desrochers, C. Barette, R. Rockwell, E. Cooch, R. Pradel, J.-D. Lebreton, M. Lambrechts, R.A. Pettifor and L. Underhill for their comments on the manuscript or the analyses, and the Hunters and Trappers Association of Pond Inlet, North-west Territories, for assistance and support.


  1. Top of page
  2. Summary
  3. Journal of Animal Ecology (2000), 69, 000–000
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References
  • Ankney, C.D. & MacInnes, C.D. (1978) Nutrient reserves and reproductive performance of female Lesser Snow Geese. Auk, 95, 459 471.
  • Ankney, C.D., Afton, A., Alisauskas, R.T. (1991) The role of nutrient reserves in limiting waterfowl reproduction. Condor, 93, 1029 1032.
  • Brinkhof, M.W.G., Cavé, A.J., Hage, F.J., Verhulst, S. (1993) Timing of reproduction and fledging success in the Coot Fulica atra: evidence for causal relationship. Journal of Animal Ecology, 62, 577 587.
  • Bromley, R.G.H. & Jarvis, R.L. (1993) The energetics of migration and reproduction of Dusky Canada Geese (Branta canadensis occidentalis). Condor, 95, 193 210.
  • Budeau, D.A., Ratti, J., Ely, C.R. (1991) Energy dynamics, foraging ecology, and behaviour of pre-nesting Greater White-fronted Geese. Journal of Wildlife Management, 55, 556 563.
  • Charnov, E.L. & Krebs, J.R. (1974) On clutch size and fitness. Ibis, 116, 217 219.
  • Choinière, L. & Gauthier, G. (1995) Energetics of reproduction in female and male Greater Snow Geese. Oecologia, 103, 379 389.
  • Cooke, F., Rockwell, R.F., Lank, D.B. (1995) The Snow Geese of la Pérouse Bay: Natural Selection in the Wild. Oxford University Press, Oxford.
  • Daan, S., Dijkstra, C., Drent, R.H., Meijer, T. (1988) Food supply and the annual timing of avian reproduction. Acta XIX Congressus Internationalis Ornithologici Ottawa, Ontario, 1986 (ed.H.Ouellet), pp. 392 407. Museum of Natural Science, Ottawa.
  • Daan, S., Dijkstra, C., Tinbergen, J.M. (1990) Family planning in the Kestrel (Falco tinnunculus): the ultimate control of covariation of laying date and clutch size. Behaviour, 114, 83 116.
  • Drent, R.H. & Daan, S. (1980) The prudent parent: energetic adjustments in avian breeding. Ardea, 68, 225 252.
  • Ebbinge, B.S. & Spaans, B. (1995) The importance of body reserves accumulated in spring staging areas in the temperate zone for breeding in dark-bellied brent geese Branta b. bernicla in the High Arctic. Journal of Avian Biology, 26, 105 113.
  • Findlay, C.S. & Cooke, F. (1982) Synchrony in the Lesser Snow Goose (Anser caerulescens caerulescens). II. The adaptative value of reproductive synchrony. Evolution, 36, 786 799.
  • Francis, C.M., Richards, M.H., Cooke, F., Rockwell, R.F. (1992) Long-term changes in survival rates of Lesser Snow Geese. Ecology, 73, 1346 1362.
  • Ganter, B. & Cooke, F. (1996) Pre-incubation feeding activities and energy budget of Snow Geese: can food on the breeding ground influence fecundity. Oecologia, 106, 153 165.
  • Gauthier, G. (1993) Feeding ecology of nesting Greater Snow Geese. Journal of Wildlife Management, 57, 216 223.
  • Gregoire, P.E. & Ankney, C.D. (1990) Agonistic behavior and dominance relationships among lesser snow geese during winter and spring migration. Auk, 107, 550 560.
  • Hamann, J. & Cooke, F. (1989) Intra-seasonal decline of clutch size in lesser snow geese. Oecologia, 79, 83 90.
  • Högstedt, G. (1980) Evolution of clutch size in birds: adaptative variation in relation to territory quality. Science, 210, 1148 1150.
  • Johnson, D.H. (1979) Estimating nest success: the Mayfield method and an alternative. Auk, 96, 651 661.
  • Klomp, H. (1970) The determination of clutch size in birds. A review. Ardea, 58, 1 124.
  • Korpimäki, E. & Wiehn, J. (1998) Clutch size of kestrels: seasonal decline and experimental evidence for food limitation under fluctuating food conditions. Oikos, 83, 259 272.
  • Lack, D. (1968). Ecological Adaptations for Breeding in Birds. Methuen, London.
  • Lebreton, J.D., Burnham, K.P., Clobert, J., Anderson, D.R. (1992) Modeling survival and testing biological hypotheses using marked animals: a unified approach with case studies. Ecological Monographs, 62, 67 118.
  • Lepage, D., Gauthier, G., Reed, A. (1996) Breeding site infidelity in Greater Snow Geese: a consequence of constraints on laying date? Canadian Journal of Zoology, 74, 1866 1875.
  • Lepage, D., Gauthier, G., Reed, A. (1998) Seasonal variation in growth of greater snow goose goslings: the role of food supply. Oecologia, 114, 226 235.
  • Lepage, D., Gauthier, G., Desrochers, A. (1998) Larger offspring increases fledging success and offspring quality in a precocial species. Journal of Animal Ecology, 67, 210 216.
  • Lepage, D., Desrochers, A., Gauthier, G. (1999) Seasonal decline of growth and fledging success in Snow Geese Anser caerulescens: an effect of date or parental quality? Journal of Avian Biology, 30, 72 78.
  • Lesage, L. & Gauthier, G. (1997) Growth and organ development in greater snow goose goslings. Auk, 114, 229 241.
  • Lesage, L. & Gauthier, G. (1998) Effect of hatching date on body and organ development in greater snow goose goslings. Condor, 100, 316 325.
  • Lessells, C.M. (1986) Brood size in Canada Geese: a manipulation experiment. Journal of Animal Ecology, 55, 669 689.
  • Lindholm, A., Gauthier, G., Desrochers, A. (1994) Effects of hatch date and food supply on gosling growth in arctic-nesting Greater Snow Geese. Condor, 96, 898 908.
  • McNamara, J.M. & Houston, A.I. (1996) State-dependent life histories. Nature, 380, 215 221.
  • Mann, F.E. & Sedinger, J.S. (1993) Nutrient-reserve dynamics and control of clutch size in northern pintails breeding in Alaska. Auk, 110, 264 278.
  • Meijer, T., Daan, S., Hall, M. (1990) Family planning in the Kestrel (Falco tinnunculus): the proximate control of covariation of laying date and clutch size. Behaviour, 114, 117 136.
  • Menu, S. (1998) Survie de la grande oie des neiges: aspects méthodologiques et implications dans la dynamique de population. PhD Thesis, Université Laval.
  • Morez, V. (1997) L’effet de la condition physique de la grande oie blanche sur sa vulnérabilitéà la chasse. Msc Thesis, Université Laval.
  • Nilsson, J.Å. (1994) Energetic bottle-necks during breeding and the reproductive costs of being too early. Journal of Animal Ecology, 63, 200 208.
  • Norris, K. (1993) Seasonal variation in the reproductive success of Blue Tits: an experimental study. Journal of Animal Ecology, 62, 287 294.
  • Perrins, C.M. (1970) The timing of birds’ breeding seasons. Ibis, 112, 242 255.
  • Perrins, C.M. & McCleery, R.H. (1989) Laying dates and clutch size in the Great Tit. Wilson Bulletin, 101, 236 253.
  • Perrins, C.M. & Moss, D. (1975) Reproductive rates in the Great Tit. Journal of Animal Ecology, 44, 695 706.
  • Pettifor, R.A., Perrins, C.M., McCleery, R.H. (1988) Individual optimization of clutch size in Great Tit. Nature, 336, 160 162.
  • Price, T. & Liou, L. (1989) Selection on clutch size in birds. American Naturalist, 134, 950 959.
  • Price, T., Kirkpatrick, M., Arnold, S.J. (1988) Directional selection and the evolution of breeding date in birds. Science, 240, 798 799.
  • Prop, J., Van Eerden, M.R., Drent, R.H. (1984) Reproductive success of the Barnacle Goose Branta leucopsis in relation to food exploitation on the breeding grounds, western Spitsbergen. Norsk Polarinstitutt Skrifter, 181, 87 117.
  • Raveling, D.G. (1979) The annual cycle of body composition of Canada geese with special reference to control of reproduction. Auk, 96, 234 252.
  • Renaud, M. (1999) Coûts énergétiques de la thermorégulation chez les jeunes de la grande oie des neiges en milieu naturel. MSc Thesis, University Laval.
  • Rockwell, R. F., Cooch, E.G., Thompson, C.B., Cooke, F. (1993) Age and reproductive success in female Lesser Snow Geese—experience, senescence and the cost of philopatry. Journal of Animal Ecology, 62, 323 333.
  • Rowe, L., Ludwig, D., Schluter, D. (1994) Time, condition, and the seasonal decline of avian clutch size. American Naturalist, 143, 698 722.
  • Ryder, J.P. (1970) A possible factor in the evolution of clutch size in Ross’ Goose. Wilson Bulletin, 82, 5 13.
  • SAS Institute Inc. (1996) SAS/STAT User's Guide. Version 6. SAS Institute, Cary, North Carolina.
  • Sedinger, J.S., Eichholz, M.W., Flint, P.L. (1995) Variation in brood behavior of Black Brant. Condor, 97, 107 115.
  • Sedinger, J.S., Flint, P.L., Lindberg, M.S. (1995) Environmental influence on life-history traits: growth, survival, and fecundity in Black Brant (Branta bernicla). Ecology, 76, 2404 2414.
  • Smith, H.G. (1993) Seasonal decline in clutch size of the Marsh Tit (Parus palustris) in relation to date-specific survival of offspring. Auk, 110, 889 899.
  • Teunissen, W., Spaans, B., Drent, R.H. (1985) Breeding success in brent in relation to individual feeding opportunities during spring staging in the Wadden sea. Ardea, 73, 109 119.
  • Tombre, I.M. & Erikstad, K.E. (1996) An experimental study of incubation effort in high-Arctic barnacle geese. Journal of Animal Ecology, 65, 325 331.
  • Williams, C.G. (1966) Natural selection, the cost of reproduction, and a refinement of Lack's principle. American Naturalist, 100, 687 690.
  • Williams, T.D., Cooch, E.G., Jefferies, R.L., Cooke, F. (1993) Environmental degradation, food limitation and reproductive output: juvenile survival in lesser snow geese. Journal of Animal Ecology, 62, 766 777.
  • Williams, T.D., Loonen, M.J.J.E., Cooke, F. (1994) Fitness consequences of parental behavior in relation to offspring number in a precocial species: the lesser snow goose. Auk, 111, 563 572.
  • Winkler, D.W. & Allen, P.E. (1996) The seasonal decline in Tree Swallow clutch size: physiological constraints or strategic adjustment? Ecology, 77, 922 932.

Received 8 March 1999;revisionreceived 1 October 1999