Patterns of reproductive effort and success in birds: path analyses of long-term data from European ducks


  • Peter Blums,

    Corresponding author
    1. * Institute of Biology, University of Latvia, Miera 3, LV–2169, Salaspils, Latvia; andCanadian Wildlife Service, Prairie & Northern Wildlife Research Centre, 115 Perimeter Road, Saskatoon, Saskatchewan S7N 0X4, Canada
      ‡Present address and correspondence: Peter Blums. Gaylord Memorial Laboratory, School of Natural Resources, University of Missouri-Columbia, Puxico, MO 63960–9686, USA. E-mail:
    Search for more papers by this author
  • * Robert G. Clark,

    1. * Institute of Biology, University of Latvia, Miera 3, LV–2169, Salaspils, Latvia; andCanadian Wildlife Service, Prairie & Northern Wildlife Research Centre, 115 Perimeter Road, Saskatoon, Saskatchewan S7N 0X4, Canada
    Search for more papers by this author
  • and * Aivars Mednis

    1. * Institute of Biology, University of Latvia, Miera 3, LV–2169, Salaspils, Latvia; andCanadian Wildlife Service, Prairie & Northern Wildlife Research Centre, 115 Perimeter Road, Saskatoon, Saskatchewan S7N 0X4, Canada
    Search for more papers by this author

‡Present address and correspondence: Peter Blums. Gaylord Memorial Laboratory, School of Natural Resources, University of Missouri-Columbia, Puxico, MO 63960–9686, USA. E-mail:


  • 1We tested ecological hypotheses about timing of breeding and reproductive effort in birds, by analysing > 15-year data sets for individually marked females in three species of Latvian ducks (northern shoveler, tufted duck, common pochard).
  • 2Duckling survival and recruitment declined with advancing hatch date in pochard and tufted duck, after controlling for effects of female age and other factors with path analysis, a novel finding which indicates that fitness advantages associated with early hatching extended beyond the prefledging period. Logistic regression analysis suggested further that individual duckling prefledging survival was moderate in the earliest phase of the breeding season, greatest in mid-season and lowest later on.
  • 3However, selection acting against early hatched ducklings was surpassed by strong directional selection favouring recruitment of the earliest hatching females. The absolute and relative numbers of female recruits produced by a breeding female declined sharply with advancing hatch date in all species.
  • 4Unlike previous studies, an hypothesized intraspecific trade-off between duckling mass and brood size was detected, being very robust in two of three species.
  • 5Unexpectedly, female age effects on recruitment were manifested only indirectly by several pathways, the most important being the earlier hatching dates of older females. Size-adjusted body mass (i.e. condition index) was positively related to reproductive success, and was 2–8-fold more influential than female size (indexed by wing length).
  • 6Overall, fecundity-independent variables (e.g. hatching date, weather, indices of duckling production and habitat quality) generally had 2–10 times greater influence on recruitment rates than did fecundity-dependent variables such as female size or condition, duckling mass and brood size, suggesting a critical role for external environmental factors vs. individual female-specific traits in the recruitment process.


Ecologists are challenged to explain the remarkably diverse ways in which organisms allocate time and energy to reproduction, and to unravel fitness consequences of these complex investment patterns. Because individuals typically have limited time or nutrients available for reproduction, advantages of raising more offspring may be offset by costs of lowered parental survival or progeny quality, as well as diminished future reproductive output; trade-offs such as these form the foundation of life-history theory (Roff 1992; Stearns 1992). Studies of optimal timing of breeding indicate that offspring production and recruitment are generally greater for early breeding individuals in numerous taxa [e.g. plants (Kelly & Levin 2000); fish (Schultz 1993); mammals (see Dobson, Risch & Murie 1999); birds (Perrins 1970; Daan, Dijkstra & Tinbergen 1990)], but forces countering early nesting have been neglected in most avian studies (see Nilsson 1994; Svensson 1997; Brown & Brown 1999; for notable exceptions). Here, we focus on these and related processes by analysing patterns of reproduction and breeding success in large samples of individually marked birds of three species, principally with path analysis.

Path analysis begins with the construction of a diagram(s) portraying possible causal and non-causal relationships among response and explanatory variables in a system based on a priori knowledge, and the method is being used increasingly for testing complex evolutionary and ecological hypotheses (Kingsolver & Schemske 1991; Mitchell 1992; Wootton 1994; Sinervo & DeNardo 1996; Grace & Pugesek 1997). With large sample sizes, path analysis may reveal critical relationships among variables correctly or pinpoint where key experimental manipulations are warranted. Our principal objective was to evaluate sources of variation in offspring survival and recruitment, while testing general ecological hypotheses about natural selection on timing of breeding, brood size and female body size and condition. Previous studies have investigated thoroughly some of these hypotheses but, to our knowledge, this is among the first to simultaneously evaluate effects of multiple, interacting factors on recruitment processes.


We generated a general diagram about interrelationships among variables that could affect offspring prefledging survival and subsequent recruitment (Fig. 1); other obvious paths not shown may also be retained during analysis, but the main hypotheses and assumptions are explained below. In general, factors positioned to the right are causally affected by those on the left, and all directly or indirectly (i.e. effects are channelled through intermediate variables) influence one or both response variables, fledging success and recruitment (see Methods for variable definitions and further explanation).

Figure 1.

Hypothetical path diagram showing hypothesized interrelationships among factors affecting duckling survival to fledging (FLEDGE) and female recruitment (RECRUIT) in three species of European ducks. Known relationships are also indicated (in parentheses). Arrows indicate direction of causality assumed in the model. Plus or minus signs indicate positive or negative effects. See Methods for definitions of abbreviations of variables.

Brood size. There is little evidence that brood size affects reproductive success in waterfowl (Rohwer 1992; Johnson, Nichols & Schwartz 1992; Milonoff, Pöysä & Virtanen 1995), but larger broods could contribute more fledglings or recruits simply by virtue of initially having more ducklings; therefore, other things being equal, there is a greater likelihood that more ducklings will survive to be recruited. Females may also devote greater maternal care to larger broods because they perceive them to be more valuable (e.g. Eadie, Kehoe & Nudds 1988), possibly enhancing duckling survival in larger broods.

Duckling body mass. A classic reproductive trade-off hypothesis states that individuals producing more offspring do so at the expense of progeny size or quality. This hypothesis is supported by diverse empirical studies [e.g. arthropods (Fox & Czesak 2000); reptiles (Rowe 1994); mammals (Dobson et al. 1999); birds (Nager, Monaghan & Houston 2000; Williams 2001)], but Bernardo (1996) argued convincingly that the failure to find this trade-off in many studies may be due to confounding effects of factors that shape reproductive allocation regimes (e.g. state-dependent reproductive effort: Schwarzkopf 1992; McNamara & Houston 1996). Thus, in addition to maternal fitness, optimal nutrient allocation to individual eggs and clutches may depend critically on both the expected future value of neonates as well as resources available to invest in offspring (e.g. Eldridge & Krapu 1988; Glazier 2000). This trade-off is expected to be most pronounced in avian species which make large nutrient investments in clutches and have limited post-hatch parental care (Lack 1967; Lack 1968), yet there is disagreement about the existence let alone strength of this trade-off in waterfowl, a group which more than satisfies both of these prerequisites (Blackburn 1991; Rohwer 1991).

Because intraspecific analyses may not account fully for effects of other interacting variables (e.g. Bernardo 1996), we tested specifically for an inverse relationship between mean (per brood) duckling mass and brood size, while controlling effects of other variables. Duckling body mass was placed in the path diagram to the left of brood size because inheritance estimates for egg size (i.e. duckling mass) generally are greater than those for brood size (Lessells, Cooke & Rockwell 1989; Larsson & Forslund 1992; Potti 1999); females are expected to balance energetic investment in clutches by lowering egg number, not by compromising egg size (Smith & Fretwell 1974). Indeed, heavier ducklings may have higher survival than lighter ones because they are better able to endure inclement weather and periods of food shortage, capture prey or evade predators (reviewed by Williams 1994).

Female body mass and wing length. The importance of female size and condition to reproductive investment and success in birds is debated. Some workers report a positive correlation between female size and both egg and clutch sizes (Barbraud et al. 1999), but others have not found these relationships (Winkler & Allen 1996). Nutrient storage and use may be influenced by body size, but the interplay of body size and condition in reproductive effort and success is uncertain (Alisauskas & Ankney 1990; Cooke, Davies & Rockwell 1990; Congdon et al. 1999), and has not been evaluated fully in birds (Barbraud et al. 1999). Larger individuals may be heavier, and size adjustments are performed typically in many animal species to derive an index of ‘condition’. Larger birds, and those with better condition, may have greater breeding success than smaller, light-weight individuals because they may be able to retain larger nutrient reserves after hatching or because larger size confers an advantage when brooding ducklings or in acquiring food (i.e. via social dominance). Therefore, we placed wing length (an index of size) before female mass in the path diagram.

Hatching date and annual effects. The influence of hatching time on offspring survival and recruitment, favouring early hatched young, is a ubiquitous pattern in birds (Perrins 1970; Daan et al. 1988), including some ducks (e.g. Dzus & Clark 1998), but it is unclear if this pattern results from higher fledging success, enhanced first-year survival rates combined with greater dominance (Spear & Nur 1994) or both. Therefore, we tested first whether duckling survival and recruitment declined with hatching date in ducks and then also evaluated the hypothesis that early hatched young have an additional advantage that carries over beyond the fledging stage to recruitment.

If early hatched offspring have higher survival and recruitment than late-hatched ones, what counteracts very early breeding (Nilsson 1994)? Hypotheses invoked to explain delays in early nesting in waterfowl centre primarily on female quality or nutrient reserve acquisition (reviewed by Rohwer 1992). We determined whether there was a cost of very early breeding, by explicitly considering curvilinear relationships between hatching date and survival or recruitment using logistic regression analysis (details below). We also looked for yearly fluctuations in production as part of this analysis.

Female age. Typically, avian reproductive rates and success are age-dependent (Sæther 1990; Stearns 1992), and ducks are no exception (Blums, Hepp & Mednis 1997a). Older birds generally breed earlier than younger ones, they may be larger or have better body condition, invest more heavily in broods, and therefore have higher breeding success, but interactions among these variables are not well quantified. Thus, we gauged effects of age on duckling survival and recruitment by exploring direct and indirect paths to all variables on the right side of the path diagram.

Environmental effects. Adverse conditions during the first few days post-hatching can cause high duckling mortality (e.g. Mendenhall & Milne 1985; Korschgen et al. 1996), so we looked specifically for effects of weather on fledging success. The relative roles of abiotic vs. biotic factors on duckling survival are virtually unknown in most duck species, nor have mediating effects of female age (i.e. behaviour) been estimated. Therefore, we included indirect paths from female age and hatching date to duckling survival via weather conditions.

Duckling production, harvest index and breeding density. Total duckling production in the year of hatching may affect recruitment negatively the following year if ducklings compete for food and space, and survival decreases (Savard, Smith & Smith 1991). Similarly, if a large proportion of fledged ducklings is harvested by hunters, recruitment might increase because of compensatory effects on overwinter survival or reduced competition for breeding space the next spring. Alternatively, excessive harvest of young females might lower recruitment if potential recruits are removed from the population.

If females compete for nesting space or other resource then the number of recruits could be related inversely to breeding density [total number of recorded nests (nesting attempts); see below]; alternatively, if the latter estimate reflects habitat suitability rather than population density (an unknown number of breeding females may have been present but did not nest) then we would predict a positive relationship between recruitment and breeding density (Fig. 1). Female age, hatching date and female or duckling size could mitigate effects of duckling production, harvest or breeding density on recruitment, so we considered indirect paths to female recruitment via these variables.



Northern shoveler (Anas clypeata L., hereafter shoveler), common pochard (Aythya ferina[L.], hereafter pochard), and tufted duck (Aythya fuligula[L.]) were captured and banded from 1960 to 1995 on the 35-km2 Engure Marsh (57°15′N, 23°07′E), Latvia, Eastern Europe; both Aythya species are diving ducks. The marsh is an isolated, shallow, permanently flooded palustrine wetland dominated by common reed (Phragmites australis[Cav.] Trin. ex Steud.) and cattail (Typha spp.). About 2000 pairs of 13 duck species nested on the marsh during the study period, with roughly 60% consisting of pochards, tufted ducks and shovelers; of these, 99% of shovelers, 42% of tufted ducks and 23% of pochards nested within permanent sampling areas (details below). Permanent sampling areas included natural and artificial islands and persistent emergent marshes with a total maximum area of about 140 ha (excluding open water). Most of the marsh is open to waterfowl hunting from early August to early November. Detailed information on the study site, field methods, predator removal and breeding populations is provided elsewhere (Blums et al. 1996; Blums et al. 1997a; Viksne 1997).


Each year two to three complete searches for duck nests were conducted on permanent sampling sites within the marsh from mid-May to late-June. All breeding habitats within permanent sampling areas were searched systematically to locate nests by walking parallel transects. Effectiveness of nest searches was probably very high and few females/nests were missed (capture probability estimates ranged from 0·73 to 0·81; P. Blums, unpublished data).

Nests were found during egg-laying (23%) or after incubation had commenced (77%). Hatching date was confirmed when ducklings were banded at the nest. Conspecific brood parasitism was common in diving ducks, but very rare in shoveler. Factors validated by DNA fingerprinting (Dugger 1996), used singly or in combination, to indicate nest parasitism were: eggs had widely different incubation stages; eggs differed in colouration, size and shape; laying rate of > 1 egg/day; and clutch sizes of ≥ 14 for shoveler and pochard, and of ≥ 15 for tufted duck. We used non-parasitized broods in all analyses.

Females were captured on nests during the last week of incubation using drop-door nest traps (Blums et al. 1983) or dip nets. Unmarked females were banded with conventional leg bands. About 66 300 day-old ducklings were individually marked using plasticine-filled leg bands (Blums, Mednis & Nichols 1994; Blums et al. 1999); subsequent recaptures of these birds as breeding females enabled us to assign exact ages. Unmarked females were aged as yearlings (SY, 1 year) or adults (ASY, > 1 year) using wing feather characteristics (Blums et al. 1996). Previously marked females were identified by band number. Body mass of females was measured with a 1000-g Pesola scale (nearest 10 g). Flattened wing chord, our measure of female size, was measured from the wrist to the end of the longest primary (nearest 1 mm). Newly hatched ducklings were weighed with a 100-g Pesola scale (nearest 1 g).


Data were obtained from the Latvian National Meteorological Center for the closest weather station, located 10 km from permanent study plots. Weather data were assigned to each brood for the periods of 5, 10 and 15 days after hatching because most mortality occurs during the first 2 weeks after hatching, when the thermoregulatory ability of ducklings is incomplete. Variables (except rain) referred to conditions during daylight hours (recorded every 3 h from 06 : 00 to 21 : 00 h local time) when broods were active and usually not brooded by a female. Preliminary analyses revealed that best model performance and strongest associations were obtained when using weather variables during the first 10 days after hatching. Similarly, initial screening identified three influential variables: mean daily temperature, mean daily wind speed and total accumulated precipitation. Therefore, subsequent analyses used these variables, averaged over the 10-day period immediately following hatch.


Structural equation modelling and path analysis. We evaluated duckling survival and recruitment on a per brood basis (i.e. to correct for possible non-independence of fates of sibs) by developing structural equation models which link response and explanatory variables (PROC CALIS, SAS Institute 1989). The CALIS procedure (similar to LISREL, Johnson, Huggins & DeNoyelles 1991) tests goodness-of-fit (GOF) of models directly to data for different combinations of variables as well as estimate direct, indirect and total effects of different variables. We calculated indirect effects of one factor on another as the compound path between two variables, the strength of a compound path being calculated by multiplying all values of the constituent paths. Summary effect coefficients were calculated by summing the values of compound paths.

GOF of our main hypothetical model (Fig. 1) and alternatives was evaluated with indices provided by CALIS, including chi-square test, Bentler’s comparative fit index (CFI), and normalized residuals (see Hatcher 1996 for details). Chi-square test P-values should exceed 0·05; the closer to 1, the better the model fit. Use of CFI was justified because we could not normalize the distribution of some variables fully (other GOF indices yielded very similar estimates, so we only report CFI results); CFI indices > 0·9 (range = 0–1·0) are indicative of good model fit (Hatcher 1996). Sample sizes should be 5–20 times the number of parameters being estimated (SAS Institute 1989), so ours were optimal (diving ducks) or nearly so (shoveler). Data outliers were identified by the procedure and excluded from analyses.

The CALIS procedure also provides standardized path coefficients, significance tests (t) for path coefficients, residual terms (E), and partial coefficients of multiple determination (R2). Path coefficients represent the size of the effect (in standard deviations) that a given explanatory (antecedent) variable has on a dependent variable, while holding statistically constant all other explanatory variables. Descriptions of the methodology are available (Bollen 1989; Loehlin 1992; Brown & Weis 1995; Pugesek & Tomer 1996).

Recruitment (RECRUIT) was our main response variable. Recapture in any subsequent year of nesting females (range = 0–3/brood) banded as day-old ducklings provided recruitment data. Because only females incubated clutches and could be captured on nests, recruitment analyses deal exclusively with females.

Fledging success (FLEDGE) estimates were available for diving ducks only. We used band recoveries (range = 0–4/brood) of both sexes banded at hatch and killed by hunters during the early post-fledging period (August–early September) as a relative index of prefledging survival. The average age of shot juveniles used for duckling survival analyses was 65 and 73 days for tufted duck and pochard, respectively. Most recoveries were obtained during regular hunter bag checks at all six hunting stations on Engure Marsh during the first month of hunting but before the onset of fall migration for these species. Ducklings that died before fledging provided no subsequent information for path analysis. Surviving juveniles were exposed to hunting over the entire marsh, so these recoveries should provide a good index of prefledging survival for ducklings of both sexes. Shoveler band recoveries could not be used to index duckling survival because autumn migration began in late July–early August and few locally hatched birds were harvested during the early post-fledging period.

Ten predictor variables were incorporated into path analyses (see Hypotheses, above): female age (FAGE), wing length (FWING) and adjusted body mass (FMASS; see below); standardized duckling hatching date (HDATE; see below) and average duckling body mass per brood (DMASS); brood size (BROOD); weather conditions (WEATHER), a latent variable (derived with principal components analysis, PROC PRINCOMP, SAS Institute 1989) which integrated temperature, wind speed and precipitation data; duckling production (DUCKLPROD), the total number of day-old ducklings (shoveler, range = 145–489; tufted duck, 618–1251; pochard, 586–2119) hatched from nests on permanent study areas in the year the recruit was banded (newly hatched ducklings later found dead at or near the nest were excluded); annual harvest index (HARVEST) was the proportion (shoveler, range = 0·02–0·15; tufted duck, 0·01–0·08; pochard, 0·02–0·09) of banded ducklings shot during the first year of life either on or outside the native marsh, including migration or wintering periods; breeding density (NESTS [t + 1]) was indexed as the total number of nest attempts (shoveler, range = 22–57; tufted duck, 97–176; pochard, 111–440) recorded annually on permanent study areas and likely reflects breeding habitat suitability [because ducklings that hatched in year t were recruited predominantly into the local population (70–77% across species) in the year following hatching we indexed breeding density as the total number of nest attempts in year t + 1].

To maximize sample sizes, we used only two age categories (SY and ASY females), a decision which is also justified by analyses which showed that the principal break-point for age-specific reproductive attributes occurs between 1 and 2 years old (Blums et al. 1997a). Body mass adjusted for day of incubation (Blums et al. 1997a:739) was used in all analyses. Hatching date was also standardized to control for annual variation by expressing it as positive or negative deviations from hatching dates of the first 10% (tufted duck and pochard) or 30% of nests (shoveler) each year (Blums, Mednis & Clark 1997b). Brood size was the number of ducklings that hatched and left the nest. Each brood appears no more than once in path analyses. Because of specific requirements for data selection (see above), the average number of broods produced per female and used in path analysis was: tufted duck 2·0 ± 0·06 (range = 1–9), pochard 1·6 ± 0·03 (1–9), shoveler 1·7 ± 0·08 (1–7).

We tested and progressively simplified alternative path models using combinations of modification and model performance indices (above). To further validate results, we performed 100 iterations of final models with random samples of 50% (tufted duck, pochard) or 63% (shoveler) of broods. GOF and number of times each path was nonsignificant (Wald test) were determined, enabling us to better evaluate the relative importance of each path (see Results).

Logistic regression. Logistic regression (PROC CATMOD, SAS Institute 1989) was used to evaluate the effect of year and hatching date on duckling survival (i.e. whether or not a banded duckling of either sex was shot and reported by hunters) and recruitment (i.e. whether or not a banded female was recaptured on a nest in years subsequent to marking), using individual ducklings as the measurement unit. Variables were checked for normality, and transformed if necessary. Second- and third-order polynomial terms involving hatching date were included to evaluate non-linear patterns. We only considered two-way interactions among continuous variables because higher order interactions and those involving year (class variable) were difficult to interpret or could not be resolved. A full model was run and then simpler models, reflecting hypothesized relationships among the variables, were explored. Non-significant two-way interactions and main effects were removed sequentially from models until only significant predictors (P < 0·05) remained.



Because hunting had no effect on female recruitment in tufted duck we removed HARVEST index and all paths associated with this variable. The final model (Fig. 2a) retained nine explanatory variables and had a good fit with the data (Table 1; all standardized residuals < 1·9). WEATHER, a latent variable, consisted of positive coefficients for temperature and precipitation (0·80 and 0·04, respectively), and a negative coefficient for wind speed (–0·81), and represented a gradient from warm, calm weather to cool, windy conditions; rainfall was relatively unimportant.

Figure 2.

Path diagram summarizing effects of maternal attributes, annual reproductive and harvest indices and weather conditions during the early post-hatching period on duckling survival and recruitment in (a) tufted duck (b) common pochard and (c) northern shoveler. All arrows represent significant pathways and indicate positive or negative direct effects of explanatory variables on response variables. Solid lines indicate strong and moderate paths, dashed lines indicate weak paths. Numbers beside arrows are standardized path coefficients. E is unexplained (residual) variance for each explanatory variable, R2 is partial coefficient of multiple determination and respective values are given for all explanatory variables. See Methods for definitions of abbreviations of variables. *P < 0·05; **P < 0·01; ***P < 0·001.

Table 1.  Goodness-of-fit indicators for final path models (using broods as sampling units), and 100 randomly selected data subsets (resampling; see Methods), evaluating patterns of fledging success and recruitment in three species of Latvian ducks. Model performance is based on Bentler’s comparative fit index (CFI), and chi-square (Chisq) test probability (P) levels. Mean CFI and chi-square P-values, with 95% confidence intervals (CI), are also given for resampling results
GOF indicatorsTufted duckCommon pochardNorthern shoveler
Final model
 Chisq (P)0·0750·1480·237
 95% CI0·988–1·00·992–1·00·973–1·0
 Chisq (P)0·3020·3400·314
 95% CI0·016–0·8400·012–0·8780·021–0·768

Duckling survival declined (standardized direct path coefficient =−0·36) as the season progressed, and was greater during post-hatching periods of warm, calm weather (Fig. 2a). The negative effect of hatching date was mitigated by a slight positive effect (0·04) of an indirect path via weather, suggesting that adverse effects of date were dampened when weather conditions were good. Total duckling production affected prefledging survival negatively, fewer ducklings fledging when local production was large. There was some evidence that heavier ducklings survived better (0·07 summary effect coefficient of direct and indirect paths); however, resampling indicated that two contributing paths (DMASS–FLEDGE, DMASS–DUCKLPROD) were weak (83 and 72 non-significant occurrences, respectively, in Wald tests). Female age was related only indirectly to duckling survival, mainly via earlier hatching of nests of older females.

Hatching date and total duckling production had strong residual negative effects on recruitment (Fig. 2a), after also being estimated as critical factors affecting duckling prefledging survival (above). However, the effect of hatching date on recruitment was weaker than on duckling survival (–0·11 vs. −0·36), whereas the effect of duckling production on recruitment was substantially stronger than on duckling survival (–0·22 vs. −0·15). The effect of hatching date was further strengthened by a summary effect (–0·15) of other indirect paths to recruitment. The strong negative effect (–0·22) of duckling production on recruitment was diminished to −0·06 by the summary effect of two indirect paths [via NESTS (t + 1) and FLEDGE]. A strong positive path between breeding density [NESTS (t + 1)] and recruitment suggests that more birds were recruited in years when nesting conditions were favourable and more females attempted to nest. Heavier females had a higher likelihood of producing a recruit. Although BROOD and FLEDGE were positively related to recruitment, resampling indicated that both paths were weak (73 and 64 occurrences in Wald test). Female age affected recruitment only indirectly, via hatching date (0·04) and many other variables, yielding a positive summary effect coefficient (0·08) of all compound paths.

There was a very strong (0·84) positive path between duckling production and breeding density, suggesting that more females attempted to breed after years of good duckling production. SY females nested later than ASY females. ASY females had longer wings and therefore also tended to be heavier, produced slightly larger broods (one direct and six indirect paths yielded a coefficient of 0·20) and heavier ducklings (a relationship which involved one direct and three indirect paths, yielding a summary effect coefficient of 0·20). The influence of female body mass on recruitment was about eight times stronger (0·097) than that of female wing length (0·012). As expected, brood size declined seasonally. After controlling for other significant effects, there was a strong negative association (–0·15) between mean (per brood) duckling body mass and brood size, larger broods being composed of lighter ducklings. However, the relationship was marginally significant (r = −0·06, P = 0·04) when evaluated with simple correlation.


The final model for pochard was similar to that of tufted duck (Fig. 2b), and model fit was excellent (Table 1). The main difference was that the pochard model retained HARVEST instead of WEATHER, suggesting that hunting impacts rather than post-hatching weather conditions were more important for pochard.

Hatching date had a strong direct (–0·26) and indirect (–0·02, via brood size) negative effect, with survival declining as the season progressed. Brood size and female mass had moderate positive effects on duckling survival, suggesting that larger broods produced more fledglings, and heavier females had a higher likelihood of producing a fledged juvenile. The direct positive effect (0·07) of female mass was strengthened by the indirect path (0·01) via brood size, but resampling revealed that the path FMASS–FLEDGE was weak (non-significant path in 60% of resampling occasions). Female age affected duckling survival only indirectly, mainly via hatching date (0·10) and several other indirect paths (combined = 0·02).

Hatching date and duckling production had strong direct negative effects on recruitment (Fig. 2b). Hatching date effects remained unchanged after five indirect paths from this variable to recruitment were considered. The strong negative effect (–0·24) of duckling production on recruitment was diminished to −0·07 by moderating effects of two positive indirect paths via breeding density and harvest index. Unlike other species, the direct path from the index of breeding density to recruitment was very weak (76 occurrences in Wald test). Brood size had a strong residual positive effect on recruitment after also being estimated as moderately affecting duckling survival (above).

Several indirect paths between female age and recruitment were channeled through duckling mass and production, the overall effect of female age on recruitment being positive (0·16–0·03 = 0·13) and translated mainly via hatching date (0·12), duckling production (0·05) and, to a lesser extent, brood size (0·01). Several indirect paths from female mass to recruitment yielded a weak positive summary coefficient (0·02), suggesting that heavier females tended to be more successful in producing recruits. Body mass was about 3·5 times more influential than body size (wing length) to recruitment, although both had weak total effects (0·014 vs. 0·004) compared with other species. Several other associations among variables in the model (including female age, wing length, body mass and breeding density, as well as duckling production, duckling mass, and brood size) were similar to those for tufted duck (above).

The negative path between duckling mass and brood size was the weakest (–0·06, P < 0·05) of all three species and this was confirmed also by resampling (69 in Wald test). This path was stronger (–0·07, P < 0·01) when we substituted brood size with clutch size (most trade-off studies use clutch, not brood size). The relationship was nonsignificant when simple correlation was employed (brood and clutch size, r = 0·03, Ps > 0·21).


We investigated only recruitment in shovelers (see Methods). The final model (Fig. 2c) had a good fit with the data (Table 1). Resampling also indicated that the final model was robust, with only two paths (FAGE–DMASS and HDATE–DMASS) occurring > 50 times in Wald tests.

Hatching date and total duckling production had strong direct negative effects on recruitment, whereas a positive association was found with the breeding density of females in the year after recruits were banded. The direct negative effect of duckling production on recruitment was moderated by an indirect positive path via breeding density; nevertheless, the resulting effect of duckling production remained negative (0·23–0·37 = −0·14). No other direct paths to recruitment were retained, but the influence of female age, wing length and body mass was reflected in several indirect paths. Older, perhaps more experienced females, produced more recruits than younger females [indirect paths via hatching date and wing length yielded a moderate (0·05) positive effect], and heavier females, regardless of age, tended to produce more recruits [two indirect paths from female mass to recruit via duckling production and breeding density yielded a rather trivial positive effect (0·02), roughly twice that of wing length (0·01)].

Although many other strong paths remained in the model, none were related directly to recruitment. There was a strong (0·81) positive association between duckling production and breeding density a year later. SY females nested later than ASY females. ASY females had longer wings and also tended to be heavier, and produced larger broods. Brood size declined seasonally with slightly increasing mean duckling body mass. Older females, especially larger, heavier ones, also produced heavier ducklings (a relationship which involved one direct and two indirect paths, yielding a reasonably large summary effect coefficient of 0·20), and larger broods (five indirect paths yielded a coefficient of 0·13). Finally, larger broods were composed of lighter ducklings and associations between brood or clutch size and duckling mass held negative (P < 0·05) when simple correlation was employed.


For both diving ducks, there was a strong curvilinear relationship between hatching date and duckling prefledging survival (Table 2), with recovery rates increasing from moderate to high during the first 15–18 days of the hatching period and then declining steeply (Fig. 3). Recovery rate of ducklings hatching near the end of the breeding season was very low despite high production of hatched ducklings.

Table 2.  Temporal (year and hatching date) relationships for duckling survival and recruitment indices in three species of European ducks banded at hatch on Engure Marsh, Latvia. Results of logistic regression analyses (per duckling basis) include model goodness-of-fit (GOF), logistic regression coefficients (estimate) and significance (P)
Species (n)PeriodOverall model GOFPredictorEstimateχ2P
  1. To maximize sample sizes, only the main predictors hatching date and year (and their interaction) were used in the models. Standardized hatching date was square-root transformed. Because hatch date3 was non-significant, it was removed from all models. Number of parameters in model. §Coefficient estimates for each year are not shown. When we replaced the quadratic term with a linear (negative) term, the latter performed nearly as well.

Pre-fledging survival
 Tufted duck (25 593)1961–94976690·75Hatch date 2·4 22·9< 0·0001
     Hatch date2−0·3 42·9< 0·0001
     Year§  51·1  0·02
     Date × year  63·2  0·001
 Common pochard (28 185)1972–94728470·98Hatch date 4·1 70·6< 0·0001
     Hatch date2−0·5104·1< 0·0001
     Year§ 133·2< 0·0001
     Date × year 161·3< 0·0001
 Tufted duck (16 957)1972–93669230·76Hatch date2−0·1 92·3< 0·0001
     Year§ 137·8< 0·0001
 Common pochard (20 823)1972–93738450·45Hatch date 1·0  3·7  0·054
     Hatch date2−0·2 14·6  0·0001
     Year§  35·7  0·02
     Date × year  36·3  0·02
 Northern shoveler (4625)1972–93281240·95Hatch date 1·4  4·4  0·04
     Hatch date2−0·2  8·8  0·003
     Year§  56·0< 0·0001
Figure 3.

Relationship between fledging success and relative timing of hatching in tufted duck (n = 24 425) and common pochard (n = 27 964) ducklings. Fledging success is the proportion of juveniles (both sexes) banded as day-old ducklings and shot later in the same year by hunters on Engure Marsh before the onset of fall migration. Vertical arrows denote median hatching date on a standardized scale for each species. Numbers at symbols represent sample sizes (number of day-old ducklings banded during each 3-day period). The first 3-day interval included ducklings hatched before ‘–16’ standardized date, and the last interval included ducklings hatched after ‘+13’ standardized date for each species.

Relative recruitment rates declined exponentially with hatching dates in all species and, unlike duckling survival, there was no evidence that very early hatched diving duck ducklings were recruited at lower rates (Fig. 4, Table 2). Broods which hatched during the last week of the nesting season provided almost no recruits. A seasonal decline in recruitment could stem from differential emigration by late-hatched individuals, but this did not account for the pattern. In all species, proportions of ducks shot by hunters away from Engure Marsh (i.e. dispersing birds) did not increase in relation to hatching date (Fig. 5). Indeed, patterns resembled those found for recruitment (Fig. 4), but date-dependency was weaker.

Figure 4.

Relationship between recruitment rate and timing of hatching in common pochard [n = 633/13 191 (total recruited/banded)], tufted duck (n = 484/12 416) and northern shoveler (n = 273/3514). Data were grouped by 3-day intervals. The first interval included ducklings hatched before ‘–16’ standardized date (diving ducks) or before ‘–13’ standardized date (shoveler). The last interval included ducklings hatched after ‘+16’ standardized date (all species). Vertical arrows denote median hatching date on standardized scale for each species. Models of best fit: pochard, y = 0·22 · x−0·25; tufted duck, y = 0·51 · x−0·26; shoveler, y = 0·27 · x−0·17.

Figure 5.

Relationship between emigration index and timing of hatching in ducks. Emigration index is proportion of ducks banded as day-old ducklings and recovered by hunters outside Engure Marsh after more than (i) 1 year in non-migrational directions up to 4055 km (both sexes); (ii) 2 years > 80 km outside Engure Marsh regardless of the direction of movement (females and individuals of unknown sex); these birds were never recorded returning to Engure Marsh. Models of best fit: common pochard, y = 0·003 · x−0·09, n = 108/26 652 (total emigrated/banded); tufted duck, y = 0·002 · x−0·06, n = 50/19 208; northern shoveler, y = 0·001 · x−0·23, n = 37/6276.


Indices of duckling survival and recruitment varied by approximately an order of magnitude across years in diving ducks, with as few as 0–5 recruits produced in some years vs. 30–60 in others. There was a positive association between indices of female recruitment and fledging success in diving ducks [pochard: r = 0·88 (Pearson), P = 0·0001, n = 22 years; tufted duck: r = 0·43, P = 0·047, n = 22 years], with no indication of nonlinear relationships. The median number of juvenile shovelers shot during the first year of life (i.e. from fledging to the next breeding season) was 18 (range = 3–41 annually, n = 33 years), whereas median number of SY females recruited was 5 (range = 0–20 annually, n = 33 years).



The strongest patterns emerging from our analyses involved effects of hatching date, and featured counteracting processes acting at different stages of an individual’s early life. All species exhibited strong directional selection for early nesting when we examined offspring recruitment. Early nesting females did not necessarily produce the most fledged ducklings (below), but early nesting attempts resulted in remarkable payoffs in terms of recruitment (Fig. 4). These patterns were probably not produced by greater permanent dispersal by late-hatching individuals (Fig. 5), as reported in some species (van Balen & Hage 1989; Verhulst, Perrins & Riddington 1997). Ducklings banded ≥ 1 day before the mean hatch date (shoveler, 54% of ducklings; tufted duck and pochard, 48%) accounted for > 73% (shoveler, 77%; tufted duck, 74%; pochard, 78%) of all recruits, and these were produced by < 27% (shoveler, 26%; tufted duck and pochard, 16%) of individual females that hatched ducklings. Recruitment advantages of early breeding are well documented in temperate-breeding birds and other organisms, and may be as common a phenomenon as seasonal brood (clutch, litter) size decline. Path coefficients linking hatching date with duckling survival and recruitment consistently rivaled or surpassed the magnitudes of those between hatching date and brood size (Fig. 2).

We obtained evidence of greater prefledging mortality of the earliest and late-hatching diving duck ducklings, being most pronounced in pochard (Fig. 3), but this selective force was insufficient to supersede the advantage of early nesting (above). Hypotheses advanced to explain delays in nesting after spring arrival by migratory, temperate-nesting waterfowl focus on female nutrient reserve acquisition requirements (reviewed by Rohwer 1992), and largely ignore the possibility that females optimize the onset of nesting to enhance duckling production (Dawson & Clark 1996). We suspect that the earliest-hatching broods more probably encountered adverse weather in some years, lower food availability, higher per capita predation rates or some combination of these factors. Nutrition is important for the onset of breeding in birds and reproductive success (below), but our result underscores the need to consider multiple constraints when looking at selective processes. For instance, females could obtain greater recruitment return on energy invested if they delay nesting slightly (early broods, being largest, require greater investment; Fig. 2), a hypothesis we investigate elsewhere.

The hatching date pattern in duckling prefledging survival was not merely an artefact of the departure of early hatched ducklings before hunting began because only a small percentage of birds had left the marsh before our cut-off date (9 September) and these birds were also included in the analysis. When we used progressively earlier cut-off dates (i.e. 31 or 25 August) for hunter band returns, the patterns held. In any year, broods averaged 60 ± 1 (range = 50–69) days old for pochard and 50 ± 1 (range = 40–58) for tufted duck when hunting began; because ducklings fledge at about 50–60 days in these species, the earliest and latest hatching ducklings typically had only recently attained flight when hunters were actively harvesting young ducks. Finally, given similar fledging periods (above), this hypothesis would not explain why the pattern existed in tufted duck (late nesting species), and not just in pochard (early nesting).

By explicitly connecting hatching date with both duckling survival and recruitment in path analyses, we were able to test whether there was an added advantage of early nesting that extended beyond the duckling stage to recruitment in diving ducks. To our knowledge, this is one of the clearest demonstrations of this effect in birds (contraDzus & Clark 1998). Early hatching individuals may gain advantages of social dominance, more access to higher quality food and greater first-year survival (Spear & Nur 1994), mechanism(s) we were unable to evaluate.

There may be other costs associated with early or late nesting that we did not explore, such as reduced egg hatching success stemming from very cold (early season freezes) or hot (late season, Arnold, Rohwer & Armstrong 1987) temperatures during egg laying. By contrast, breeding females that begin nesting early also have more time either to lay replacement clutches or to replenish body reserves before migration. Survival and recruitment advantages of early breeding were detected while explicitly incorporating measures of female quality (age, wing length, mass); therefore, we believe that these separate, additive effects of hatching date probably reflected environmental variation (e.g. food, predation) or seasonal differences in female behaviour such as maternal care (e.g. brooding, vigilance).

Empirical and experimental studies (e.g. Brinkhof et al. 1993; Nilsson 1994; Moreno et al. 1997; Aparicio 1998; Nilsson 1999) imply the existence of an individual optimization process about laying date decisions, and suggest further that timing of breeding should be viewed as a trade-off between present and future reproductive success (Daan et al. 1990; Svensson & Nilsson 1995; Nilsson 1999). Parental investment should become more restrictive later in the breeding season (Clutton-Brock 1991), as females perceive their own survival and future reproductive prospects as being increasingly important. Studies of temperate waterfowl (Talent, Jarvis & Krapu 1983; Rushforth Guinn & Batt 1985) have reported lower parental care for late season than early season broods. Late nests of our study species have a higher probability of abandonment (Blums et al. 1997b) and the same may be true for late-hatched broods.


Path analyses were not expected to capture fully the relationship between fledging success and recruitment because brood was the unit of measurement, and the loss of most brood mates to hunting would necessarily weaken the relationship; however, paths between FLEDGE and RECRUIT were positive for both species (indirect in pochard). When we looked at year-specific fledging success and subsequent recruitment from yearly cohorts, the positive relationship was much stronger in pochard (r = 0·88) than tufted duck r = 0·43). Tufted ducks showed very conservative nesting patterns, nested almost exclusively in association with gulls and waders, and long-term average apparent nest success (0·81) was consistently very high for this species. We believe the carrying capacity of tufted duck nesting habitats became progressively saturated, and the number of breeding females was limited by suitable nesting sites. In contrast, pochards showed very flexible nesting patterns, were not constrained to gull colonies and bred in a wide variety of habitats, including islands and emergent marshes. Thus, habitat saturation was probably not a limiting factor for population growth in pochard. These results suggest that annual fledging success may be a good index of year-specific recruitment potential (see Weatherhead & Dufour 2000), particularly when nesting habitats are not fully saturated and competition for available nest sites is non-existent or low.


Contrary to initial expectations (Fig. 1), female age had no direct effect on duckling survival and recruitment. None the less, age was highly influential, because ASY females nested earlier (all species) and, in diving ducks, were larger and heavier, and produced bigger broods composed of heavier ducklings than SY females. Thus, ASY female diving ducks produced more offspring than SY breeders. We found a positive relationship between duckling mass and survival only in tufted duck. However, we believe that this is an artefact of using mean (per brood) duckling mass. For instance, when we analysed these relationships using individual ducklings as the unit of measurement, heavier ducklings survived better than light-weight ones in all species (logistic regression, P < 0·03, P. Blums, unpublished data).


Theory predicts a trade-off between size and number of reproductive units but an inverse relationship between egg size and number has eluded avian (e.g. Rohwer 1988) and other ecologists (Bernardo 1996), except perhaps at the interspecific level (Christians 2000). If egg production is limited by nutrient availability, ducks seem to be ideally suited for testing this relationship because they produce large clutches of large eggs, and offspring are self-feeding, thus lowering the need for parental care. However, intraspecific analyses have also failed to reveal this trade-off (e.g. Rohwer 1988; Lessells et al. 1989; Rohwer & Eisenhauer 1989; Guyn & Clark 2000), even after controlling factors that might influence egg and brood size variation.

We detected a negative relationship between brood size or clutch size and duckling size in all three species and this relationship was clearly method-dependent (see Blackburn 1991; Rohwer 1991). When simple correlation was employed the relationship was significant in only shoveler (brood and clutch size) and tufted duck (brood size only) but not in pochard. In diving ducks, especially in pochard, consistent negative relationships (P-values from 0·02 to 0·0001) were not detected until hatching date and all female characteristics were included in covariance and path analyses. The critical need to control female characteristics (e.g. age, size, condition) and breeding date statistically when looking for an egg size–clutch size trade-off is reminiscent of the argument that allocation patterns may be state-dependent (McNamara & Houston 1996).


Several clear patterns emerged from analyses involving indices of duckling production and breeding density, but this was not true for post-hatching weather conditions or harvest index. In all species, path coefficients between duckling production and recruitment were negative (range =−0·22 to −0·37), effects that were moderated to some extent by positive associations among duckling production, breeding density and recruitment. None the less, the strongest summary negative association between production and recruitment was obtained with shoveler (–0·14) and was weaker for tufted duck (–0·06) and pochard (–0·07). Female age had virtually no influence on these relationships in diving ducks, and none in shoveler. Overall, an inverse relationship between production and duckling survival/recruitment could be indicative of competition during brood-rearing or for breeding space in subsequent years. Shoveler are territorial relative to the diving ducks studied (Seymour 1974; Titman & Seymour 1981), and have a 2–4 times greater duckling survival rate [0·12 vs. 0·06 (pochard) and 0·03 (tufted duck), Blums et al. 1996] and a lower proportion of SY breeders relative to tufted duck (0·70 vs. 0·89, Blums et al. 1996), conditions that could favour collectively the development of competitive interactions. Resource limitation and inverse density-dependent effects have not been determined clearly in ducks (Savard et al. 1991), although there has been considerable speculation. Anderson, Emery & Arnold (1997) were unable to demonstrate an inverse relationship between duckling production and survival in an experimentally enlarged population of canvasbacks [Aythya valisineria (Wilson)]. Given that predators were removed on the Engure Marsh and nesting success was generally high (Blums et al. 1996), initial duckling population levels were probably larger than normally encountered. Thus, potential for competition may have been amplified by these management activities.

The number of nests found of each species was related directly to duckling production the previous year, and this index of breeding density was related positively to recruitment. We suspect that the number of nests reflected breeding habitat quality, so the larger number of recruits detected in years with more nests was a consequence of higher breeding propensity, greater probability of renesting, and lower nest abandonment (i.e. greater maternal investment and risk-taking). When we subtracted the number of recruits from number of nests found each year (increasing independence of the two variables) and re-ran path analyses, coefficients between NESTS (t + 1) and RECRUIT remained significant.


Our novel path analytical results have helped to unravel several controversial aspects of (avian) evolutionary ecology. Specifically, we obtained unequivocal evidence of hatching date effects that extended beyond duckling survival to recruitment, amplifying the importance of early breeding for recruitment much more than seasonally declining duckling survival estimates alone would do. This effect also illustrates clearly the very low value of late-hatched offspring, and may therefore help to explain greater late-season nest (Blums et al. 1997b) and brood abandonment and seasonal clutch size decline (Toft, Trauger & Murdy 1984).

We obtained the first compelling evidence for an hypothesized intraspecific trade-off between egg and clutch size. We encourage others to (re)consider effects of breeding date and female attributes, including age and nutritional state when looking for this relationship, an essential step for detecting this pattern in diving duck species, particularly pochard. Similarly, understanding the nutrient content of eggs (and neonates) in relation to egg and clutch size variation seems critical to improving our understanding of allocation tactics because important chemical properties may be overlooked only by measuring egg or neonate size.

Female condition was more important than size for reproductive success. This conclusion is consistent with our previous work and that of others which showed that poor-condition females, regardless of size, were more prone to nest failure (Blums et al. 1997b; Kellett & Alisauskas 2000). We do not believe that this was merely an age-related artefact because age was considered explicitly when these interrelationships were evaluated. Overall, female mass was > 2 times more influential than size.

Our results are also relevant to the question of whether or not avian breeding time is shaped by fecundity-dependent or fecundity-independent selection (Svensson 1997). Duckling mass (three species) and brood size (diving ducks) were related directly to female size-adjusted mass and age, and indirectly (shoveler brood size) to female age. Furthermore, female shovelers and close relatives of tufted duck and pochard often use nutrient reserves for egg formation [shoveler, Ankney & Afton 1988; lesser scaup, Aythya affinis (Eyton), Afton & Ankney 1991; canvasback, Barzen & Serie 1990] or to complete incubation successfully (Blums et al. 1997b). We summed direct and indirect path coefficients (Fig. 2) linking recruitment to: (i) hatching date, weather, duckling production and breeding density, these representing fecundity-independent variables; and (ii) duckling mass and brood size (fecundity-dependent traits). Summed coefficients were generally low for duckling mass (range = 0−0·02), weather (tufted duck only, 0·02), and brood size (range = 0−0·11). By contrast, coefficients were much larger for duckling production (range =−0·37 to −0·06), breeding density (range = 0·13−0·29) and hatching date (range =−0·29 to −0·14). Effects of hatching date were strongest in pochard (−0·29) and shoveler (–0·28), species which nest earlier than tufted duck (Fig. 3). Thus, fecundity-independent variables generally were 2–10 times more influential than fecundity-dependent traits. This strong contrast between individual vs. external forces may also help to explain why egg and clutch sizes are so variable, and inheritance estimates for these traits are << 1·0.

Our results also raise questions about interspecific variation in the form and magnitude of relationships among variables, and their impact on productivity. Although these three species nested and raised ducklings in the same wetland (i.e. common environment), differences in nest sites, predators, diet, brood-rearing behaviour and wintering sites may account for species-specific patterns revealed by path analysis. Analyses of survival and lifetime reproductive success of individual females may also demonstrate important interspecific differences and parallels, patterns which may shed further insight on hypotheses tested in this study.


We thank the more than one hundred people who assisted with field work or contributed to the maintenance of the database, in particular J. Baltvilks, I. Bauga, A. Celmins, A. Graubica, G. Graubics, M. Janaus, J. Kats, M. Kazubierne, J. Kazubiernis, P. Leja, G. Lejins (deceased), J. Lipsbergs, H. Mihelsons (deceased), A. Petrins, V. Pilats, V. Reders, J. Viksne and A. Stipniece. D. Spals and V. Klimpins provided invaluable technical support throughout the study. Ken Norris, Kevin Dufour and two anonymous referees gave helpful comments on previous versions of this manuscript. Funding for the fieldwork and data computerization of this long-term research project was provided by the Institute of Biology, University of Latvia (formerly Latvian Academy of Sciences). A grant from the National Science Foundation, USA (DEB–9974077), supported P. Blums during data analysis and manuscript preparation. This is Missouri Agricultural Experiment Station Project 183, Journal Series 13194.