Food-supplementation does not override the e ect of egg mass on tness-related traits of nestling house wrens

Transcription

1 Ecology 2000, 69, Food-supplementation does not override the e ect of egg mass on tness-related traits of nestling house wrens JOHN D. STYRSKY*, ROBERT C. DOBBS and CHARLES F. THOMPSON Behavior, Ecology, Evolution, and Systematics Section, Department of Biological Sciences, Illinois State University, Normal, IL, 61790±4120, USA Summary 1. Cross-fostering experiments have shown that egg mass per se positively a ects post-hatching growth and development in passerine birds. In most studies, however, the initial in uence of egg mass was not sustained, leading to questions about the importance of egg mass relative to the cumulative e ects of environmental factors during the period of post-hatching care. 2. In house wrens, Troglodytes aedon Vieillot, evidence from cross-fostering experiments suggests that food availability mitigates the in uence of egg mass on posthatching growth: early in the season, when food was putatively abundant, the initial e ect of egg mass was lost as nestlings aged; however, later, when food was putatively scarcer, the e ect of egg mass persisted until nestlings left the nest. 3. We tested the hypothesis that abundant food for provisioning nestlings can decouple the relationship between tness-related traits of nestlings and the mass of the eggs from which they hatch (`egg-mass override hypothesis') by providing daily mealworm supplements to broods of house wrens. 4. The food supplements did not increase nal nestling mass, nal nestling tarsus length, growth rate, or survival until nest-leaving in late-season broods, when the hypothesis of food limitation predicts that there should be such an e ect. Instead, the food supplements signi cantly, but only slightly, increased nal nestling mass in early-season broods, when food is putatively abundant. 5. The e ect of the food supplements on nal nestling mass was independent of egg mass. Nestling survival in early-season broods, and nal nestling mass and tarsus length in both early and late-season broods signi cantly increased with increasing egg mass in both experimental and control broods. 6. We propose that food was not limiting during the course of the experiments and that nal nestling mass and tarsus length remained positively related to egg mass because nestling growth and development in house wrens is `predetermined' by some component or property of eggs that covaries with egg mass. Troglo- Key-words: egg mass, food availability, maternal e ects, nestling growth, dytes aedon. Ecology (2000) 69, Introduction Intraspeci c variation in egg mass is common in wild birds (Boag & van Noordwijk 1987; Perrins *Present address and correspondence: J.D. Styrsky, Department of Biological Sciences, 331 Funchess Hall, Auburn University, Auburn, AL 36849±5414, USA. Tel.: (334) 844±4850, Fax: (334) 844±9235, styrsjd@mallard.duc.auburn.edu 1996), and the amount of energy and resources invested in individual eggs varies with egg mass (reviewed in Williams 1994). There is, however, limited support for the common assumption that tness-related traits of nestlings are related to the size (mass or volume) of the eggs from which they hatch (Ricklefs 1984; Magrath 1992; Williams 1994). This is because the assumption can be tested only by using cross-fostering techniques to decouple experimentally variance due to pre-hatching attributes of

2 691 J.D. Styrsky, R.C. Dobbs & C.F. Thompson egg size (e.g. genetic factors or maternal e ects such as nutritional or hormonal factors [sensu Mather & Jinks 1971: 293; see also Price 1998]) from variance caused by environmental factors (e.g. territory quality or quality of post-hatching care) (Bernardo 1996). The results of cross-fostering experiments are mixed, although many have demonstrated that chick growth (Amundsen & Stokland 1990; Magrath 1992; Smith, Ohlsson & Wettermark 1995; Amundsen, Lorentsen & Tveraa 1996; Hipfner & Gaston 1999; Reed, Turner & Sotherland 1999; Styrsky, Eckerle & Thompson 1999; but see Ricklefs & Peters 1981; Ricklefs 1984) and, to a lesser extent, chick survival (Bolton 1991; Blomqvist, Johansson &GoÈ tmark 1997; but see Magrath 1992; Smith et al. 1995; Amundsen et al. 1996; Reed et al. 1999; Styrsky et al. 1999) are related to egg mass per se in avian taxa across the altricial±precocial spectrum of nestling development. Whether an investment in larger eggs increases female tness through the e ects of egg mass on nestling growth and survival, however, remains unclear because e ects typically do not persist throughout the nesting cycle. In most cross-fostering experiments, the relationship between tness-related traits of nestlings and egg mass was detected only early in post-hatching development (Amundsen & Stokland 1990; Reid & Boersma 1990; Magrath 1992; Smith et al. 1995; Amundsen et al. 1996; Reed et al. 1999), although Styrsky et al. (1999) found that this varied between broods within a breeding season. The absence of a sustained in uence of egg mass on tness-related traits of nestlings suggests that variance caused by environmental factors eventually swamps variance caused by egg mass per se, thereby making it undetectable (Magrath 1992; Smith et al. 1995; Styrsky et al. 1999). This may be particularly true of altricial passerines, in which parental investment in post-hatching care is much greater than the initial investment in eggs (Walsberg 1983; O'Connor 1984). Environmental factors that in uence the quality of parental care may therefore exert a greater in uence on post-hatching development in altricial birds than do pre-hatching attributes of eggs (Ricklefs 1984; Magrath 1992; see also Smith 1993; Smith & Wettermark 1995). Food availability during the period of brood care is generally considered to be the most important aspect of the environment a ecting nestling growth and development (reviewed in Martin 1987; Gebhardt-Henrich & Richner 1998). Experimental studies have shown that not only does limited availability of food during the nestling period constrain several components of reproductive success (e.g. nestling mass, survival, and edgling production) (Simons & Martin 1990; Richner 1992; Soler & Soler 1996; Wiehn & KorpimaÈ ki 1997; SiikamaÈ ki 1998), but also that seasonally limited food resources account for intra-seasonal variation in components of reproductive success, such as chick survival and edgling production (Lindholm, Gauthier & Desrochers 1994; Wiggins, Part & Gustafsson 1994; Brinkhof & Cave 1997; SiikamaÈ ki 1998). If food resources are particularly abundant (e.g. in optimal habitats or during a portion of the breeding season), the cumulative e ect of nestling provisioning may override any initial e ects of egg mass as nestlings age (Smith & Bruun 1998; Styrsky et al. 1999). Indirect evidence that food availability overrides initial e ects of egg mass is provided by a nonexperimental study of the e ect of egg mass and habitat on nestling growth and survival in European starlings (Sturnus vulgaris Linnaeus) (Smith & Bruun 1998) and by a cross-fostering experiment in house wrens (Troglodytes aedon Vieillot) (Styrsky et al. 1999). In the former, nestling survival, mediated through an e ect of egg mass on nestling mass, was related to egg mass in areas with a low proportion of pasture (a high-food habitat preferred by foraging starling parents) but not in areas with a high proportion of pasture (Smith & Bruun 1998). Styrsky et al. (1999) demonstrated that nestling mass in house wrens was positively related to egg mass per se, but that the persistence of the relationship between nestling mass and egg mass depended on the time of the season. Whereas nestling mass was signi cantly correlated with egg mass only in the rst half of the nestling period early in the breeding season, nestling mass was signi cantly correlated with egg mass until nestlings achieved asymptotic mass late in the breeding season, at a time when food resources may be limiting (see Styrsky et al. 1999). The results of both studies suggest that egg mass per se has a more pronounced e ect on tness-related traits of nestlings when food availability is low than when food is abundant, which we term the egg-mass override hypothesis. We are not aware of any experiment with birds in which food availability has been manipulated to investigate speci cally the in uence of egg mass on nestling growth and survival. In this paper, we present the results from a food-supplement experiment designed to test the egg-mass override hypothesis in nestling house wrens. The egg-mass override hypothesis predicts that when food resources are limiting, food supplementation should decouple any positive association of egg mass with nestling mass, tarsus length, growth rate, and survival by disproportionately enhancing these traits in nestlings that hatch from smaller eggs relative to nestlings that hatch from larger eggs. Increased food should more greatly bene t nestlings from small-egg broods because such nestlings are presumably farther from the typical asymptotic values of these traits than are nestlings from large-egg broods. Speci c predictions for early and late-season broods di er based on evidence that food resources are abundant early but

3 692 Egg mass and food in nestling wrens then decline as the breeding season progresses (see Styrsky et al. 1999). In early-season broods, the eggmass override hypothesis predicts that nestling mass, tarsus length, growth rate, and survival until nest-leaving should not be in uenced either by egg mass or by food supplementation (because food is abundant and should override the e ect of egg mass) (Fig. 1a). In late-season broods, these tnessrelated traits should be in uenced by egg mass in control broods, but not in food-supplemented broods in which the additional food should override the e ect of egg mass (Fig. 1b). Therefore, in addition to testing the egg-mass override hypothesis, the food-supplement experiments also tested the assumption that food availability declines seasonally on our study area (see gure legend, Fig. 1). Methods STUDY AREA AND SPECIES We conducted the food-supplement experiments between 1 May and 31 August 1998 with house wrens that bred in nestboxes on the forested 108-ha Mackinaw Study Area in McLean County, Illinois, USA ( N, W). Nestboxes were mounted on 15-m greased steel poles to deter predators and were distributed 30 m apart along N±S lines separated by 60 m (see Drilling & Thompson 1988). House wrens are insectivorous, short-distance migrant passerines and are typically double-brooded on the Mackinaw Study Area, with two distinct peaks of clutch production (early season, mid-may; late season, late June±early July) (Finke, Milinkovich & Thompson 1987). Early and late-season clutches were designated as such based on whether clutch egg-1 dates (the date on which the rst egg is laid) fell before, or on or after, respectively, the median egg-1 date for For this study, we sampled early-season clutches with egg-1 dates between 10 and 26 May and late-season clutches with egg-1 dates between 21 June and 14 July. House wrens in this population exhibit a marked seasonal pattern of reproductive investment with a decrease in clutch size (early-season mode, 7 eggs; late-season mode, 6 eggs) and a concomitant increase in per-clutch mean egg mass as the breeding season progresses (Finke et al. 1987; Styrsky et al. 1999, respectively). GENERAL FIELD PROCEDURES We inspected nestboxes at least twice weekly throughout the breeding season to determine when nest-building began. Nestboxes with nests that were near completion were inspected at least every other day to determine when clutches were initiated. We weighed eggs to the nearest 0001 g on a portable electronic balance (Acculab PocketPro 2060D) Fig. 1. Predicted results under the egg-mass override hypothesis for the in uence of food availability on the relationship between egg mass and several tness-related traits of nestling house wrens in (a) early-season, and (b) lateseason broods. The light grey line represents the hypothetical relationship between tness-related traits and egg mass if food is limiting. The expected relationship between tness-related traits and egg mass in the control treatments (based on Styrsky et al. 1999) is indicated by the solid black lines. The arrows indicate the predicted direction of the e ect of additional food in the food-supplement treatments (shown as the dotted black lines). The predictions for early and late-season broods di er based on evidence that food resources are abundant early but then decline as the breeding season progresses (see Introduction). In earlyseason broods, abundant food in both control and foodsupplemented treatments should decouple the positive relationship between egg mass and tness-related traits of nestlings. In contrast, in late-season broods, a marked disparity in food availability between the two treatments should result in a positive e ect of egg mass in the control treatment but no e ect of egg mass in the food-supplement treatment (and therefore a statistical interaction between e ect of egg mass and food availability). If the assumption of a seasonal decline in food availability does not hold true and, instead, food availability increases seasonally, we predicted there would be an interaction between the e ects of egg mass and food availability in early-season broods but not in late-season broods. If food availability is always limiting, there should be an interaction between the e ects of egg mass and food availability in both early and late-season broods. Finally, if food availability is never limiting, egg mass should not in uence tness-related traits of nestlings in either the control or food-supplement treatments (i.e. no interaction) in either early or late-season broods. before noon on the day they were laid and numbered each egg with a permanent marker. A few eggs were weighed 6±48 h after they were laid, but

4 693 J.D. Styrsky, R.C. Dobbs & C.F. Thompson prior to initiation of incubation. Mass loss of unincubated eggs during the laying period is negligible (00005 g day 1 ) (J.D. Styrsky, unpublished data). As the expected date of hatching approached, we visited nests daily to determine when the rst egg of the clutch hatched (brood-day 0). Individual nestlings were weighed to the nearest 01 g on a portable electronic balance (Acculab PocketPro 150 or 250) on brood-days 0, 2, 4, 6, 8, 10, and 12. We occasionally missed rst-hatched nestlings on the day they hatched because they hatched after we had visited the nest. Brood-day 0 was determined for those nests using the criteria of Harper, Juliano & Thompson (1992). On brood-day 12, nestling tarsometatarsus length was measured to the nearest 01 mm with dial calipers and each nestling was marked with a numbered, aluminium US Fish and Wildlife Service (USFWS) leg ring. Nests were visited daily after brood-day 13 to determine the day of nest-leaving (typically brood-day 15±17) and the number of nestlings that survived to leave the nest. We captured adults at their nests by closing a sliding metal trap door mounted on each nestbox or by placing a mist-net near the entrance of the nestbox. We marked each adult with a numbered, aluminium USFWS leg ring. Adult males were also marked with a unique combination of three coloured-plastic rings (max. two rings per leg) to allow identi cation without recapture. FOOD-SUPPLEMENT EXPERIMENT After egg-laying had begun, we randomly assigned nests to three treatments (Table 1). In the experimental treatment, supplemental food was placed daily during the nestling period in lm canisters pinned inside nestboxes. In the sham-control treatment, lm canisters were pinned inside nestboxes but no supplemental food was added. In the unmanipulated control treatment, neither food nor lm canisters were added. In both experimental and sham-control nestboxes, one lm canister was pinned in each of the two corners immediately adjacent to the nestbox's entrance. Film canisters were pinned to the walls of the nestboxes in such a way that the rim of each canister was ush with the bottom edge of the entrance hole. We added lm canisters to nestboxes 3±4 days before the expected date of hatching in order to give the adults time to acclimate to them before nestlings were present. A sham-control treatment was not used in lateseason broods because there was no di erence in mean brood-day 12 nestling mass (two-sample equal variance t-test; t 27 ˆ 018, P ˆ 086) or tarsus length (two-sample equal variance t-test; t 26 ˆ 106, P ˆ 030) between the sham-control and unmanipulated control treatments in early-season broods. Therefore, the data from the two control treatments were pooled for all subsequent analyses (see Table 1). There was no di erence in mean egg mass between control and experimental treatments in either earlyseason (two-sample equal variance t-test; t 60 ˆ 056, P ˆ 058) or late-season (t 38 ˆ 072, P ˆ 048) broods. Nests were generally assigned to treatments before the identity of the parents could be determined, so some adults were included in both early and lateseason experiments. It is possible that food supplementation of early-season broods a ected parents' ability to provision nestlings in late-season broods. Therefore, data from late-season nests of females that raised food-supplemented broods in the early season (n late-season experimental nests ˆ 5; n control nests ˆ 3) were excluded from all subsequent analyses. Supplemental food consisted of commercially purchased mealworms (Tenebrio molitor Linnaeus). Each day (between and h) during the nestling period, 15 g of live mealworms were placed in each of the two lm canisters (30 g total) in the experimental nestboxes. This amount of food was chosen based on an estimate of the mean peak daily metabolizable energy (DME) requirement for nest- Table 1. Mean egg mass of house wren clutches for each treatment in early-season and late-season food supplement experiments, 1998 Brood Treatment n clutches Egg mass (mean SD) (g) Early-season Experimental Sham control Unmanipulated control Pooled control* Late-season Experimental Unmanipulated control *Data from the sham- and unmanipulated-control treatments in early-season broods were pooled for all statistical analyses (see Methods section).

5 694 Egg mass and food in nestling wrens ling house wrens. Peak DME was estimated using the allometric equation peak DME kj day 1 ˆ501 m 0782 ; eqn 1 where m ˆ individual nestling mass (g) at nest-leaving (sensu Weathers 1992). For this experiment, we used mean brood-day 12 nestling mass (mean SD ˆ g) from 77 broods weighed on the Mackinaw Study Area in Brood-day 12 mass represents nal (asymptotic) nestling mass in this species, a slight overestimate of mass at nest-leaving in our study population (C.F. Thompson, unpublished data). Our estimate of mean peak DME (mean SD ˆ kj day 1 ) is very similar to the value empirically determined by Dykstra & Karasov (1993) for brood-day 12±15 nestling house wrens in a Wisconsin population (mean peak DME SDˆ kj day 1 ). Assuming that the energy content of mealworms is 1159 kj g 1 (calculated from Bell 1990) and that assimilation e ciency of mealworms by house wrens is 065 (Kacelnik 1984), an individual nestling would need to consume 41 g day 1 of mealworms to satisfy its peak DME requirement. Therefore, a daily food supplement of 30 g of mealworms could satisfy the mean peak DME requirement of a brood of seven nestlings (the modal clutch size of early-season broods) throughout the nestling period. We added mealworms to experimental nests from brood-days 1±11. On brood-days 2, 4, 6, 8 and 10, we weighed nestlings in experimental broods prior to adding mealworms. Although food-supplemented nests were visited more frequently than control nests, the e ects of the disturbance on brooding females should have been negligible. Brooding females always returned to their nests within 5 min of delivering mealworms (personal observation). Any mealworms remaining from the previous day were discarded in the vegetation near the nestbox. Mealworm sizes were scaled to nestling age to ensure that nestlings would be able to consume them. Hence, the food supplement consisted of `mini' mealworms (length: < 63 mm) from brooddays 1±4, `small' mealworms (63±127 mm) from brood-days 5±8, and `medium' mealworms (128±191 mm) from brood-days 9±11. We recorded the proportion of mealworms that remained in each lm canister each day to estimate utilization of the food supplement. Mean per-day mealworm utilization ranged from 709% to 909% (mean SD ˆ %) in early-season broods and ranged from 753% to 929% (mean SD ˆ %) in late-season broods (twosample equal variance t-test: t 53 ˆ 044, P ˆ 066). One author (J.D.S.) veri ed nestling ingestion of the food-supplement by analysing nestling faecal samples. For each experimental nest, faecal sacs were collected from as many nestlings as possible 2±3 h after providing the daily supplement of mealworms on one randomly chosen day between brood-days 7 and 11. Similarly, faecal sacs were collected from nestlings in 10 control nests. Faecal samples from every experimental nest contained easily identi able remnants of mealworms. Analysis of the faecal samples revealed that parents also provisioned experimental nestlings with other insect taxa besides the mealworm supplement, including lepidopteran larvae and adults, orthopterans, dipterans, and arachnids. None of the faecal sacs collected from the control nestlings contained remnants of mealworms. STATISTICAL ANALYSES Mean egg mass was 1432 g (n clutches ˆ 121, SE ˆ 0010) in early-season clutches and 1464 g (n clutches ˆ 86, SE ˆ 0012) in late-season clutches (two-sample equal variance t-test: t 205 ˆ 217, P ˆ 003). Mean clutch size was 69 eggs (n clutches ˆ 121, SE ˆ 007) in early-season clutches and 56 eggs (n clutches ˆ 86, SE ˆ 007) in late-season clutches (two-sample equal variance t-test: t 205 ˆ 1274, P < 00001). Because of the signi cant di erences in mean egg mass and mean clutch size between early and late-season broods and because of evidence for a seasonal decline in food availability, we analysed data from early and late-season broods separately. The egg-mass override hypothesis predicts an interaction between treatment e ects and egg-mass e ects when food availability is limiting (see Fig. 1). Therefore, we used analyses of covariance (GLM procedure; SAS Institute 1990) with mean egg mass as a covariate to test for e ects of food supplementation on mean nestling mass, tarsus length, growth rate, and survival. Mean values for each brood were used because individual nests represent the unit of replication. In the analyses of nestling mass and tarsus length, we focused speci cally on brood-day 12 (asymptotic) mass and tarsus length because broodday 12 mass is positively correlated with juvenile survival and recruitment in this population of house wrens (C.F. Thompson, unpublished data), as it is in many other passerines (e.g. Tinbergen & Boerlijst 1990; Hochachka & Smith 1991; Magrath 1991). Nestling mass and tarsus length may be a ected by factors other than food-supplementation or egg mass, thereby confounding detection of treatment or egg-mass e ects. Therefore, analyses of mean brood-day 12 mass were performed on residuals from multiple regressions (REG procedure, SAS Institute 1990) with hatching date, brood size, and hour of weighing as independent variables. Similarly, analyses of mean brood-day 12 tarsus length were performed on residuals from multiple regressions with hatching date and brood size as independent variables. Residuals from the regression analyses as well as the ANCOVA met the assumptions of normality and homoscedasticity.

6 695 J.D. Styrsky, R.C. Dobbs & C.F. Thompson We used nonlinear least-squares estimation (NLIN procedure; SAS Institute 1990) to t mean nestling mass on successive 2-day intervals to the generalized logistic growth equation W t ˆ W 1 W 0 e gt W 1 W 0 W 0 e gt eqn 2 where W t ˆ body mass at age t, W 1ˆ asymptotic body mass, W o ˆ initial body mass, and g ˆ an exponential rate constant of growth (sensu Peters 1983). Mean nestling growth for each brood is estimated by g, which describes the rate (day 1 ) at which asymptotic mass is achieved. We included only broods in which nestlings were weighed at least ve times during the nestling period. Analyses of mean nestling growth were performed on residuals from multiple regressions with hatching date and mean brood size (between brood-days 2±12) as independent variables. Brood size was included in these analyses because brood size varied and individual nestlings occasionally died and were not replaced. Residuals from the regression analyses as well as the ANCOVA met the assumptions of normality and homoscedasticity. We considered nestling survival for each brood as the proportion of nestlings that survived to leave the nest; therefore, we analysed nestling survival using generalized linear models with a binomial distribution and a cloglog-link function (GENMOD procedure; Allison 1995: 211 ). We speci ed likelihood ratio chi-square statistics for all hypothesis tests because Allison (1995: 219) suggests that likelihood ratio statistics more closely approximate a true chisquare distribution than do Wald chi-square statistics in small to moderate-sized samples. We included hatching date and mean brood size as independent variables in these analyses for the same reason as described above. As both egg mass and clutch size di er between early and late-season clutches, we used two-sample t-tests to test for seasonal di erences in mean residual brood-day 12 nestling mass and tarsus length (controlling for mean egg mass and brood size), and mean residual nestling growth and survival (controlling for mean egg mass and mean brood size [see above]) in control broods. Statistical tests for the predicted positive e ects of food-supplementation and egg mass on mean brood-day 12 nestling mass and tarsus length, growth rate, and survival are onetailed. In all analyses we consider P-values R 005 as signi cant. Finally, we investigated the statistical power of the analyses to detect an interaction between e ects of food availability and egg mass on each of the four tness-related traits of nestlings measured in this study following Cohen's (1988: 273 ) procedures to estimate the power of ANCOVA for large, Fig. 2. Mean nestling mass (mean 2 SE) in relation to brood day for control nests (W) and for food-supplemented nests (*) in (a) early and (b) late-season house wren broods Brood-day 0 is the day the rst egg of a clutch hatched. medium, and small e ect sizes, as de ned by Cohen (1988). Results NESTLING MASS In early season broods, nestlings in the control treatment consistently lagged behind food-supplemented nestlings in mean mass (Fig. 2a). In contrast, in lateseason broods, food-supplemented nestlings initially lagged behind control nestlings in mean mass (but not signi cantly), and then caught up by brood-day 8 (Fig. 2b). There was no interaction between the e ects of food availability and mean egg mass on mean brood-day 12 nestling mass in either early or late-season broods (Table 2a). However, food-supplementation signi cantly enhanced mean broodday 12 mass in early-season broods but not in lateseason broods (Table 2a, Fig. 3a). Further, mean brood-day 12 mass increased signi cantly with increasing mean egg mass in both early (slope [b] SE ˆ ) and late-season broods (b SE ˆ ) (Table 2a). Finally, mean residual

7 696 Egg mass and food in nestling wrens brood-day 12 mass was signi cantly greater (t 46 ˆ 174, P ˆ 004) in early-season control broods (unadjusted mean SE ˆ g) than in late-season control broods (unadjusted mean SE ˆ g). Food supplementation could have enhanced nestling mass disproportionately within individual broods (e.g. smaller siblings could have bene ted more than larger siblings), an e ect that may not have been detected when considering mass averaged over all siblings within broods. Therefore, we tested for an e ect of food supplementation on the coe cient of variation of brood-day 12 nestling mass. Food supplementation did not signi cantly decrease the coe cient of variation in brood-day 12 mass in either early (Wilcoxon two-sample test; z ˆ 143, P ˆ 015) or late-season (z ˆ 032, P ˆ 075) broods. The power of the analyses of covariance to detect an interaction between e ects of food availability and egg mass on mean brood-day 12 nestling mass in early-season broods was 84% for a large e ect size ( f ˆ 040; see Cohen 1988), 46% for a medium e ect size ( f ˆ 025), and 10% for a small e ect size ( f ˆ 010). The power of the analyses of covariance to detect an interaction in late-season broods was 66%, 31%, and 8% for large, medium, and small e ect sizes, respectively. NESTLING TARSUS LENGTH There was no interaction between the e ects of food availability and mean egg mass on mean brood-day 12 nestling tarsus length in either early or late-season broods (Table 2b). Food supplementation did not signi cantly enhance mean brood-day 12 tarsus length in either brood, although the e ect approached signi cance in late-season broods (Table 2b, Fig. 3b). However, mean brood-day 12 tarsus length increased signi cantly with increasing mean egg mass in both early (b SEˆ ) and late-season broods (b SEˆ ) (Table 2b). There was no di erence in mean residual brood-day 12 tarsus length between earlyseason (unadjusted mean SE ˆ mm) and late-season (unadjusted mean SE ˆ mm) control broods (t 45 ˆ 087, P ˆ 020). Finally, food supplementation did not signi cantly decrease the coe cient of variation of mean broodday 12 tarsus length in either early (Wilcoxon twosample test; z ˆ 063, P ˆ 053) or late-season (z ˆ 111, P ˆ 027) broods. The power of the analyses of covariance to detect an interaction between e ects of food availability and egg mass on mean brood-day 12 nestling tarsus length in early-season broods was 83%, 45%, and 10% for large, medium, and small e ect sizes, respectively, and in late-season broods was 66%, Table 2. ANCOVA F-tests for e ects of treatment on (a) mean brood-day 12 nestling mass, (b) mean brood-day 12 tarsus length, and (c) mean nestling growth, with mean egg mass as a covariate, in early and late-season house wren broods 1998 Early-season broods Late-season broods Source F d.f. P { F d.f. P { (a) Mean brood-day 12 mass Full model 398 3, , Mean egg mass treatment 141 1, , Reduced model* 523 2, , Treatment 575 1, , Mean egg mass 400 1, , (b) Mean brood-day 12 tarsus Full model 168 3, , Mean egg mass treatment 161 1, , Reduced model* 170 2, , Treatment 008 1, , Mean egg mass 340 1, , (c) Mean nestling growth ( g) Full model 073 3, , Mean egg mass treatment 022 1, , Reduced model* 100 2, , Treatment 013 1, , Mean egg mass 183 1, 47 ± { 002 1, 33 ± { *ANCOVA assumes the slopes of the two treatments (control and experimental) are equal. This assumption is tested by incorporating in the model a term describing the interaction between the treatment and covariate. If it is not signi cant, the interaction term is dropped from the model and treatment and covariate e ects are tested separately. {P-values for treatment and covariate e ects in reduced models are one-tailed. {P-value omitted because e ect of mean egg mass opposite to that predicted.

8 697 J.D. Styrsky, R.C. Dobbs & C.F. Thompson respectively, and in late-season broods was 66%, 31% and 8% for large, medium, and small e ect sizes, respectively. NESTLING SURVIVAL Mean nestling survival was high in both early-season (909%) and late-season (911%) control broods (t 47 ˆ 013, P ˆ 045). The mean number of nestlings that survived to leave the nest was signi cantly greater (two-sample equal variance t-test; t 47 ˆ 173, P ˆ 0045) in early-season control broods (mean SE ˆ 53 03) than in late-season control broods (mean SE ˆ 46 03), but this di erence is due primarily to the di erence in mean clutch size between early and late-season broods (see Methods section). There was no interaction between the e ects of food availability and mean egg mass on mean nestling survival in either early (w 2 ˆ 029, d.f. ˆ 1, n ˆ 53, P ˆ 059) or late-season (w 2 ˆ 178, d.f. ˆ 1, n ˆ 39, P ˆ 018) broods. In addition, food sup- Fig. 3. Mean ( 2 SE) residual brood-day 12 (a) nestling mass and (b) nestling tarsus length in early and late-season 1998, control (W) and food-supplemented (*) house wren broods. Presented are least-squares means derived from ANCOVA described in the text. In panel (a), the vertical axis represents the mean deviation in mass (g) from that expected based on hatching date, brood size, and hour of weighing. In panel (b), the vertical axis represents the mean deviation in tarsus length (mm) from that expected based on hatching date and brood size. 31%, and 8% for large, medium, and small e ect sizes, respectively. NESTLING GROWTH There was no interaction between the e ects of food availability and mean egg mass on mean nestling growth ( g) in either early or late-season broods, although the interaction approached signi cance in late-season broods (Table 2c). Mean nestling growth was not signi cantly a ected by either food-supplementation (Fig. 4a) or mean egg mass in either early or late-season broods (Table 2c). However, mean nestling growth was signi cantly faster (t 41 ˆ 200, P ˆ 003) in early-season broods (unadjusted mean SEˆ ) than in late-season broods (unadjusted mean SE ˆ ). The power of the analyses of covariance to detect an interaction between e ects of food availability and egg mass on mean brood-day 12 nestling growth in early-season broods was 78%, 40% and 10% for large, medium, and small e ect sizes, Fig. 4. (a) Mean ( 2 SE) residual rate of nestling growth, and (b) mean ( 2 SE) proportion of brood surviving in early and late-season 1998, control (W) and food-supplemented (*) house wren broods. Presented in panel (a) are least-squares means derived from ANCOVA described in the text. Presented in panel (b) are least-squares means derived from regressions of residual nestling survival on mean egg mass. The vertical axis represents the mean deviation in (a) growth rate, or (b) proportion of brood surviving from that expected based on hatching date and mean brood size (see text).

9 698 Egg mass and food in nestling wrens plementation did not enhance mean nestling survival in either early (w 2 ˆ 025, d.f. ˆ 1, n ˆ 53, P ˆ 031) or late-season (w 2 ˆ 172, d.f. ˆ 1, n ˆ 39, P ˆ 009) broods (Fig. 4b). However, mean nestling survival did increase signi cantly with mean egg mass in early-season broods (w 2 ˆ 453, d.f. ˆ 1, n ˆ 53, P ˆ 002), but not in late-season broods (w 2 ˆ 027, d.f. ˆ 1, n ˆ 39, P ˆ 030). The power of the analyses of covariance to detect an interaction between e ects of food availability and egg mass on mean brood-day 12 nestling survival in early-season broods was 82%, 43% and 10% for large, medium, and small e ect sizes, respectively, and in late-season broods was 68%, 33% and 9% for large, medium, and small e ect sizes, respectively. Discussion RESULTS INCONSISTENT WITH EGG-MASS OVERRIDE HYPOTHESIS The results of the food-supplement experiments failed to support the egg-mass override hypothesis that abundant food resources decouple the relationship between mean egg mass and tness-related traits of nestling house wrens. Inconsistent with our predictions based on the assumption that food availability declines seasonally (see Fig. 1), experimentally increased food availability did not disproportionately enhance mean brood-day 12 nestling mass or tarsus length, or mean growth rate or survival of nestlings that hatched from smaller eggs relative to nestlings that hatched from larger eggs in late-season broods. Also inconsistent with our predictions based on the assumption that food availability declines seasonally, the food supplements had no detectable e ect in late-season broods but, instead, had a limited e ect in early-season broods. Finally, although we predicted that tnessrelated traits of nestlings would be positively associated with mean egg mass in late-season control broods only (see Fig. 1), mean nestling survival increased signi cantly with mean egg mass in both treatments in early-season broods, and mean broodday 12 mass and tarsus length increased signi cantly with mean egg mass in both treatments in both early and late-season broods. The results of the food-supplement experiments are also inconsistent with our predictions based on the other assumptions that (i) food availability increases seasonally, (ii) food resources are always limiting, and (iii) food resources are never limiting (see Fig. 1). In a similar study, Smith & Bruun (1998) also failed to detect a signi cant interaction between egg mass and availability of pasture (an estimate of food available to provision nestlings) on tness-related traits of nestling European starlings; however, early nestling survival was positively correlated with egg mass in areas where availability of pasture was low but not in areas where availability of pasture was high. Although Smith & Bruun's (1998) results suggest that egg mass had a more pronounced e ect on nestling survival when food availability was low, they did not experimentally manipulate food availability and thereby did not directly test the egg-mass override hypothesis. INTERPRETATION I: FOOD LIMITING BUT SUPPLEMENTS INEFFECTIVE The results of our study are surprising because they are contrary to any expectation related to food availability. It is highly unlikely that any other house wrens or individuals of other species removed the mealworms as house wrens vigorously defend nestboxes from others. Although some mealworms may have escaped over the lip of the lm canisters, only a very small proportion (< 5%) would have been able to do so because the mealworms could not climb the canisters' polished plastic walls (personal observation). One possibility is that food availability was indeed limiting during the nestling period in both early and late-season broods, but that food supplements had no appreciable e ect because they were ine ectively used by parents to provision nestlings. Although we do not know how much of the food supplement was actually fed to nestlings, a substantial fraction of the mealworm supplements (> 70%) was utilized on a daily basis and the analyses of nestling faecal samples indicate that experimental nestlings were always provisioned with mealworms within 2±3 h of our supplying the daily supplement. Moreover, that a small percentage of mealworms commonly remained in the lm canisters 24 h later indicates that mealworms were constantly available for adults to provision nestlings. Even if the parents consumed the majority of the mealworm supplements, this should have a orded them with additional time and energy to provision nestlings with the food they would have fed themselves. Of the four tness-related traits of nestlings that we measured, experimentally increased food resources signi cantly enhanced only mean broodday 12 mass and did so only in early-season broods. Despite the statistical signi cance of the e ect, the increase in nestling mass was slight, approximately 02 g, which represents only a 2% increase in mass for a 10-g bird. Likewise, although the e ect of the food supplements on mean brood-day 12 nestling tarsus length in late-season broods approached signi cance, mean tarsus length of food-supplemented broods was approximately 02 mm greater than that of control broods, only a 1% increase.

10 699 J.D. Styrsky, R.C. Dobbs & C.F. Thompson The apparent lack of a marked e ect of supplemental food in this study is surprising considering that food-supplement experiments in many other bird species have demonstrated that food availability during the nestling period not only limits reproductive success (e.g. Arcese & Smith 1988; Simons & Martin 1990; Richner 1992), but also that a seasonal decline in food abundance or quality accounts for a seasonal decline in nestling growth and survival (e.g. Lindholm et al. 1994; Brinkhof & Cave 1997; SiikamaÈ ki 1998). As is common in many bird species (e.g. Perrins 1970; Daan et al. 1988; Brinkhof & Cave 1997), house wrens in our study population also exhibit a seasonal decline in reproductive success (Drilling & Thompson 1991). Females produce signi cantly fewer eggs in late-season clutches than in early-season clutches (this study; Finke et al. 1987; Styrsky et al. 1999). In addition, nestlings in the smaller late-season broods grow signi cantly more slowly and weigh signi cantly less on broodday 12 than nestlings in early-season broods (this study; Styrsky et al. 1999), despite a signi cantly greater per capita rate of nestling provisioning by the parents in late-season broods (K.P. Eckerle, unpublished data). Moreover, a 38-year data set on seasonal changes in arthropod abundance in forested areas of central Illinois that are very similar in habitat structure and within a 100-km radius of the Mackinaw Study Area indicates that the arthropod taxa most commonly fed to nestling house wrens (Lepidoptera, Orthoptera, and Arachnida [Morton 1984]) decline in abundance seasonally (Kendeigh 1979). Clearly, declining prey availability during the production of second broods coincides with a decrease each year in house wren reproductive success. It is di cult to understand why, then, experimentally increased food resources in uenced nestling growth in early but not in late-season broods in INTERPRETATION II: FITNESS-RELATED TRAITS CONSTRAINED BY EGG MASS REGARDLESS OF FOOD AVAILABILITY Perhaps there was no interaction between the e ects of food availability and egg mass on tness-related traits of nestlings because food availability was not limiting during the nestling period in either early or late-season broods; thus, food supplements had no appreciable e ect. Previous brood-size manipulation experiments in our study population of house wrens demonstrated that brood-day 12 nestling mass and nestling survival were usually not adversely a ected if nestlings were raised in enlarged broods. Further, females that raised enlarged rst broods were not less likely than other females to attempt a second brood, did not have a longer interbrood interval, and did not produce smaller second clutches than did females that raised rst broods of a normal size (Finke et al. 1987; Harper et al. 1992). These results suggest that food availability is not limiting on our study area, at least in most years. If we interpret the results to indicate that food availability was not limiting on our study area during the food-supplement experiments, the fact that mean brood-day 12 nestling mass and tarsus length, and mean nestling survival were positively associated with mean egg mass, even in the experimental treatments (in which food supplements signi cantly, albeit slightly, enhanced mean brood-day 12 nestling mass), suggests that tness-related traits of nestling house wrens may be `predetermined' by some prehatching attribute of the egg (i.e. maternal e ect) that covaries with egg mass. For example, embryonic development is limited by the functional capacity of the yolk sac for nutrient absorption and of the chorioallantois for gas exchange. The functional capacity of these two extraembryonic membranes is in turn limited, in part, by the size of the yolk and the size of the egg, respectively (Starck 1998). Embryonic development may also be constrained simply by the amount of protein and lipids or essential vitamins and minerals present in the egg (Royle et al. 1999), or by genetic factors that in uence embryonic metabolic e ciency (Magrath 1992). Whole yolk mass, as well as lean dry and lipid components of yolk and albumen increase with egg mass in some species of birds (e.g. Ward 1995; Kennamer, Alsum & Colwell 1997), including house wrens (S.S. Soukup, unpublished data). Magrath (1992) suggested that the same gene that confers metabolic e ciency to o spring development may confer metabolic e ciency in the production of yolk by the female; thus, both egg mass and nestling development may covary because of the independent expression of a shared gene (Magrath 1992). Once set on a speci c developmental track by some property of the egg, any increase in food availability above a minimum threshold necessary for normal post-hatching development may not a ect nestling growth because of pre-set limits to intrinsic growth capacities of tissues (Ricklefs, Starck & Konarzewski 1998). Some empirical support for this comes from food-restriction experiments conducted on domestic fowl, turkeys, and Japanese quail (Coturnix japonica Temminck & Schlegel) (reviewed in Schew & Ricklefs 1998). In most studies, the response in chick growth to realimentation after varying periods of undernutrition was parallel, rather than accelerated, relative to that of chicks that were fed ad libitum, suggesting that compensatory (accelerated) growth in birds may be unlikely because of strong genetic control over growth and development. Thus, in house wrens, supplemental food may enhance tness-related traits of nestlings when food is limiting, but still not decouple the relationship between such traits and egg mass. Alterna-

11 700 Egg mass and food in nestling wrens tively, when food is not limiting, supplemental food may, instead, be allocated to tissue maturation (cell di erentiation) rather than tissue growth (increase in undi erentiated cells) (sensu Ricklefs, Shea & Choi 1994), in which case the positive relationship between tness-related traits of nestlings and egg mass would also be maintained. If food-supplemented nestlings in this study did mature at a greater rate than control nestlings, we could predict that food-supplemented nestlings should have left the nest earlier. Although food-supplemented nestlings did leave the nest approximately 05 days earlier than did control nestlings, the reduction in time was not statistically signi cant (J.D. Styrsky, unpublished data). Conclusions Previous cross-fostering studies designed to partition genetic and environmental e ects have shown that egg mass per se can a ect post-hatching nestling growth in passerine birds (Magrath 1992; Smith et al. 1995; Reed et al. 1999), including house wrens (Styrsky et al. 1999). In each of these studies, however, the initial in uence of egg mass was not sustained as nestlings aged, except in late-season broods of house wrens (Styrsky et al. 1999), leading to questions about the importance of egg size relative to the cumulative e ects of environmental factors, such as food availability, during the period of post-hatching care (Magrath 1992; Smith et al. 1995; Smith & Bruun 1998; Styrsky et al. 1999). Results from the food-supplement experiments presented here fail to support the egg-mass override hypothesis that abundant food resources decouple the relationship between mean egg mass and tnessrelated traits of nestlings, and instead indicate a sustained e ect of egg mass on nestling mass and tarsus length, even when food supplements enhanced nestling mass. These results suggest that some component or property of eggs that covaries with egg mass constrains nestling growth and development in birds. In contrast to the conclusions drawn by Styrsky et al. (1999), we conclude that if environmental variance does swamp variance due to egg mass as nestlings age, a source of environmental variance other than food availability per se is responsible for the loss of the initial e ect of egg mass on nestling house wrens. The tness-related consequences of egg mass in relation to food availability, particularly for altricial passerines, clearly merit further investigation. Acknowledgements We thank D.E. Davis, Mr and Mrs D. Sears, and the ParkLands Foundation for allowing us to use their properties. This research would not have been possible without the able eld assistance of Kevin P. Eckerle, Lisa A. Ellis, Alexandra Latham, and Eric Sgariglia. This study greatly bene ted from thoughtful comments by Sabine S. Loew, Steven A. Juliano and Scott K. Sakaluk. We especially thank Karen M. Thompson for a priority delivery of mealworms. This eld study was conducted with support from the Illinois State University Department of Biological Sciences and the Beta Lambda Chapter of the Phi Sigma National Biological Honor Society. References Allison, P.D. (1995). Survival Analysis Using the SAS System: A Practical Guide. SAS Institute Inc., Cary, North Carolina. Amundsen, T., Lorentsen, S.-H. & Tveraa, T. (1996) E ects of egg size and parental quality on early nestling growth: an experiment with the Antarctic petrel. Ecology, 65, 545±555. Amundsen, T. & Stokland, J.N. (1990) Egg size and parental quality in uence nestling growth in the shag. Auk, 107, 410±413. Arcese, P. & Smith, J.N.M. (1988) E ects of population density and supplemental food on reproduction in song sparrows. Ecology, 57, 119± 136. Bell, G.P. (1990) Birds and mammals on an insect diet: a primer on diet composition analysis in relation to ecological energetics. Studies in Avian Biology, 13, 416± 422. Bernardo, J. (1996) Maternal e ects in animal ecology. American Zoologist, 36, 83±105. Blomqvist, D., Johansson, O.C. & GoÈ tmark, F. (1997) Parental quality and egg size a ect chick survival in a precocial bird, the lapwing Vanellus vanellus. Oecologia, 110, 18±24. Boag, P.T. & van Noordwijk, A.J. (1987) Quantitative genetics. Avian Genetics (eds F. Cooke & P.A. Buckley), pp. 45±78. Academic Press, London. Bolton, M. (1991) Determinants of chick survival in the lesser black-backed gull: relative contributions of egg size and parental quality. Ecology, 60, 949±960. Brinkhof, M.W.G. & CaveÂ, A.J. (1997) Food supply and seasonal variation in breeding success: an experiment in the European coot. Proceedings of the Royal Society of London Series B, 264, 291±296. Cohen, J. (1988) Statistical Power Analysis for the Behavioral Sciences, 2nd edn. Lawrence Erlbaum Associates, New Jersey. Daan, S., Dijkstra, C., Drent, R. & Meijer, T. (1988) Food supply and the annual timing of avian reproduction. Acta XIX Congressus Internationalis Ornithologici, 1, 392±407. Drilling, N.E. & Thompson, C.F. (1988) Natal and breeding dispersal in house wrens (Troglodytes aedon). Auk, 105, 480±491. Drilling, N.E. & Thompson, C.F. (1991) Mate switching in multibrooded house wrens. Auk, 108, 60±70. Dykstra, C.R. & Karasov, W.H. (1993) Daily energy expenditure by nestling house wrens. Condor, 95, 1028±1030.