Brood parasitic European starlings do not lay high-quality eggs

Behavioral Ecology doi:10.1093/beheco/ari017 Advance Access publication 19 January 2005 Brood parasitic European starlings do not lay high-quality eggs Kevin M. Pilz, a Henrik G. Smith, b and Malte Andersson c a Department of Neurobiology and Behavior, Cornell University, Ithaca, NY 14853, USA, b Department of Animal Ecology, Lund University, SE-223 62 Lund, Sweden, and c Department of Zoology, University of Gothenburg, 405 30 Göteborg, Sweden Chicks of obligate brood parasites employ a variety of morphological and behavioral strategies to outcompete nest mates. Elevated competitiveness is favored by natural selection because parasitic chicks are not related to their host parents or nest mates. When chicks of conspecific brood parasites (CBPs) are unrelated to their hosts, they and their parents would also benefit from traits that enhance competitiveness. However, these traits must be inducible tactics in CBPs, since conspecific brood parasitism (CBP) is facultative. Such tactics could be induced by resources passed to offspring through the egg. Thus, females engaging in CBP should allocate to their eggs resources that will enhance offspring competitiveness. We tested this prediction in a population of European starlings (Sturnus vulgaris) breeding in southern Sweden. Previous research showed that almost all CBPs in this population are floater females that have yet to breed in the current season. We identified putative brood parasitic eggs through monitoring egg laying and verified parasitism using protein fingerprinting. We then determined whether parasitic eggs were larger, larger-yolked, or had higher concentrations of yolk testosterone or androstenedione than control eggs. The 14 brood parasitic eggs laid in active nests (those with clutches of at least two eggs that were eventually incubated) did not differ from controls in any of these characteristics. Ten dumped eggs, laid in nonactive nest-boxes or on the ground, were smaller and smaller-yolked than control eggs but did not differ in yolk androgen concentrations. The failure of our prediction could be the result of high costs of investing in eggs, lack of competition-based benefits for chicks, or physiological constraints on egg manipulation. Key words: conspecific brood parasitism, egg size, European starlings, nestling competition, Sturnus vulgaris, yolk androgens. [Behav Ecol 16:507 513 (2005)] Brood parasitic birds lay eggs in the nests of host birds that raise the parasitic offspring to independence. About 1% of bird species reproduce solely via brood parasitism, laying their eggs in the nests of other species (obligate interspecific brood parasitism, IBP; Payne, 1977). In another 185þ species of birds, individuals sometimes parasitize conspecifics (conspecific brood parasitism, CBP; Eadie et al., 1998). IBP species have evolved various characteristics to enhance the success of offspring, including egg and chick mimicry to impede host discrimination (Davies, 2000; Lichtenstein, 2001) and strong eggshells to deter egg damage by hosts or during rushed laying (Davies, 2000; Rahin et al., 1988). In species where CBP is a persistent strategy that produces some successful offspring (such as in the European starling, see below), selection should also favor characteristics among CBP offspring that will enhance their success. Characteristics most beneficial to altricial CBP offspring would probably be those that enhance competitive success of chicks. A brood parasitic chick has an a priori competitive advantage over its nest mates: lack of relatedness. If a brood parasite chick is not related either to its competitors in the nest or to the adults feeding it, it does not lose inclusive fitness when its nest mates receive less food or when the parents expend more feeding effort (Hamilton, 1964). An unrelated parasitic chick pays no fitness cost for selfishness, whereas host chicks pay inclusive fitness costs for nest mate or parental fitness losses. This situation extends to the parents: parasitic Address correspondence to K.M. Pilz, who is now at Department of Evolutionary Ecology, Museo Nacional de Ciencias Naturales, Jose Gutierrez Abascal, 2, 28006 Madrid, Spain. E-mail: kevin.pilz@ cornell.edu. Received 18 December 2003; revised 27 September 2004; accepted 3 November 2004. parents lose no inclusive fitness from enhanced offspring competitiveness, but host parents do lose fitness if their own offspring behave selfishly toward siblings (Godfray, 1995; Trivers, 1974). Moreover, relatedness between brood parasitic females and hosts does not change the predictions, since parasitic chicks are still less related to their nest mates than host offspring are (in the European starling, brood parasites and hosts are probably not closely related due to low recruitment to the natal colony; Cabe, 1999; H.G. Smith, unpublished data). Because of these differences in relatedness, natural selection should always maintain more vigorous competition among parasitic chicks than among host chicks (i.e., even if host chicks respond evolutionarily to parasitism with increased competitiveness; Johnstone and Grafen, 1993; see also Godfray, 1995). Brood parasitic chicks can outcompete host chicks through tactics such as large size and exaggerated begging (Cotton et al., 1999; Kilner and Johnstone, 1997). Traits such as these have become fixed in IBP species (Dearborn, 1998; Lichtenstein, 2001; Lichtenstein and Sealy, 1998), presumably due to natural selection favoring these traits. While CBP offspring should also behave more competitively than host offspring, in this case selection cannot fix elevated competitiveness at the species level because host young would then elevate competition in step with parasites. Instead, selection should favor an inducible tactic where only parasitic young show increased competitiveness. This conditional tactic could be triggered by either the parasitic offspring or the parasitic mother. As there is little evidence that birds are capable of discriminating genetic relatedness per se (Davies, 2000; Kempenaers and Sheldon, 1996; but see Petrie et al., 1999), parasitic chicks are unlikely to discern their foreign origin. However, mothers can discriminate whether or not they are laying parasitic eggs. Thus, mothers are more likely to Behavioral Ecology Ó International Society for Behavioral Ecology; all rights reserved.

508 Behavioral Ecology be the initiators of competition-enhancing mechanisms in parasitic offspring than the offspring themselves. Because parasitic mothers do not directly interact with their offspring as nestlings, they can only initiate these mechanisms through the egg. The European starling (Sturnus vulgaris; hereafter starling ) provides an excellent opportunity to test these predictions. CBP has been extensively studied in starlings (Evans, 1988; Feare, 1991; Lombardo et al., 1989; Pinxten et al., 1991a,b; Sandell and Diemer, 1999; Stouffer and Power, 1991; Stouffer et al., 1987; Yom-Tov et al., 1973) and occurs in all studied populations, with 0 to 37% of nests receiving parasitic eggs in a given year (Evans, 1988; Pinxten et al., 1991a). CBP is a widespread and common tactic in starlings, implying that brood parasitism is not a recently derived aspect of starling reproductive behavior and that natural selection has therefore had opportunity to favor traits that benefit CBP success in this species. One difficulty of studying maternal manipulation of eggs as a means of enhancing success of CBP young is knowing the reproductive strategy of the brood parasite. In starlings, as in most altricial species where CBP occurs, CBPs are typically disrupted breeders that lost their nests or floater females that have not yet established nests (Pinxten et al., 1991a; Sandell and Diemer, 1999; Stouffer and Power, 1991). Conspecific brood parasites (CBPs) could theoretically also be high-fitness females that lay parasitic eggs after completing their own clutches to enhance reproductive fitness, although this has not been reported in starlings. These distinctions are important because each egg yolk takes several days to form (Challenger et al., 2001; Etches, 1996), and thus temporal constraints may prevent females from tailoring egg characters to a parasitic strategy when resorting to CBP directly before or after laying in their own nest. Our predictions are, therefore, best tested when CBPs are nonnesting floaters that did not have previous nests of their own. We studied CBP in one of the few populations where this is known to be the case, a population of starlings nesting in the Revinge area of southern Sweden. Brood parasitism has been consistently observed in this population for more than three decades (Karlsson, 1983; Smith et al., 1994; this study). Sandell and Diemer (1999) studied the identity of CBP females in this population, in many of the same nest-box colonies we study. They caught 17 CBP females using nest-box traps with artificial nests. Only one of the females was known to have an active nest before being caught (which she had abandoned 2 days before being captured as a brood parasite), and at least 15 (possibly all 16) of the other females were floaters that had not yet settled in a nest-box. This population provides an excellent opportunity to test our predictions because we know that brood parasites rarely have concurrent or previous nests of their own (probably because they lack a nest and mate) and can therefore tailor eggs for CBP. We examine two mechanisms by which brood parasitic starling females could increase the competitiveness of their offspring via materials deposited in the egg. The first mechanism is to lay larger eggs or larger-yolked eggs. Chicks hatching from large eggs are larger at hatching and have a survival and growth advantage early in the nestling period over chicks hatching from small eggs, especially when food shortage occurs during that time (Smith and Bruun, 1998; see also Amundsen et al., 1996; Styrsky et al., 1999; Williams, 1994). Large-yolked eggs carry more lipid energy and may result in larger hatchlings or hatchlings with larger residual yolk reserves (Carey, 1996; Finkler et al., 1998). The second mechanism is to allocate high levels of androgens to eggs. Androgenic hormones (primarily testosterone [T], androstenedione [A4], and dihydrotestosterone) occur in avian egg yolk and are known to increase the growth of chicks (Eising et al., 2001; Schwabl, 1996b; but see Sockman and Schwabl, 2000), including starlings (Pilz, 2003; Pilz et al., 2004). The competition hypothesis of yolk androgen allocation proposes that females differentially allocate yolk androgens to eggs within a clutch to create asymmetries in chick competitive ability, as benefits the female (Pilz et al., 2003; Schwabl, 1993, 1996b; Schwabl et al., 1997). This hypothesis predicts that CBPs should allocate high levels of androgens to their eggs, so that their offspring outcompete unrelated nest mates. Thus, we test whether brood parasitic females laid larger eggs, larger-yolked eggs, or eggs with higher concentrations of T or A4. In addition to manipulating own-egg characteristics, CBPs may attempt to enhance offspring success by targeting particular hosts. We therefore compare egg characteristics of parasitized and nonparasitized nests to examine whether CBPs target nests with high- or low-quality eggs; the former may ensure high-quality host parents, whereas the latter may improve the competitive success of parasitic young. To our knowledge, this is the first study to examine differences in egg size or yolk hormone characteristics between parasitic and nonparasitic eggs of an avian species engaging in CBP. METHODS Field methods We studied CBP in European starlings (S. vulgaris) breeding in the Revinge area of southern Sweden. These starlings are single brooded and the vast majority of females begin laying within a short, highly synchronized period of about 1 week (Smith, 2004). We have previously examined the effects of egg androgens on chick growth and begging behavior, as well as relationships of maternal breeding density and maternal quality with yolk androgen allocation, in these same colonies (Pilz and Smith, 2004; Pilz et al., 2003, 2004). We checked for new eggs daily during the synchronous laying period at 170 nest-boxes in 1999 and 84 nest-boxes in 2000. Nest checks were usually begun at 1100 h, since starlings usually lay between 0800 and 1000 h (Feare et al., 1982) and finished by 1800 h. Eggs were marked with nontoxic permanent ink as they were laid. There were 120 active nests in 1999 and 61 nests in 2000. We defined active nests as nests with a final clutch size of more than two eggs; all clutches with more than two eggs were incubated, whereas no clutch with only one or two eggs was incubated. We assumed a nest had been parasitized when two new eggs were found in a nest on a given day (since birds do not lay more than one egg per day) or when a single egg differed markedly in appearance from the other eggs in a clutch. We collected these clutches as soon as the potential parasitism was observed and, as they were laid, collected the subsequent eggs laid by the host female of the parasitized nest. For control eggs, in each year we collected complete clutches from randomly selected nest-boxes. Most eggs from these clutches were collected on the day they were laid, and all were collected before clutch completion. Collected eggs were replaced with fresh starling eggs or mimetic plastic eggs to prevent abandonment by host females. We collected 14 putatively parasitized clutches in 1999 and 2 in 2000, as well as 17 control clutches in 1999 and 5 control clutches in 2000. Two clutches designated as controls in 1999 were subsequently parasitized and thus shifted to the parasitism group. Another two clutches from 1999 were collected due to the presence of an atypically large egg, which turned out to be double yolked; these two clutches were shifted from the parasitism group to the control group, and double-yolked eggs were excluded from analysis. Finally, we collected all dumped eggs,

Pilz et al. Brood parasitic starlings do not lay high-quality eggs 509 defined as any eggs in nest-boxes that did not have active nests and any eggs found unbroken on the ground (Feare, 1991). We also conducted analyses of dumped eggs excluding eggs found on the ground, since these may represent eggs removed from nest-boxes by hosts or parasites (Lombardo et al., 1989). We found eight dumped eggs in 1999 (one on the ground) and four dumped eggs in 2000 (two on the ground). Collected eggs were kept cool and brought to a laboratory in the evening for processing. We first weighed the whole egg and then cracked it and took several samples of egg albumen. We then separated the yolk from the albumen by absorbing the albumen with filter paper. We weighed the yolk, homogenized it, and diluted several samples in distilled water. Yolk and albumen samples were frozen at ÿ80 C until analysis. Yolk mass was not recorded for a few eggs because of yolk rupture during handling. We did not assay yolk androgen concentrations in the two dumped eggs found on the ground in 2000 due to space constraints in the androgen assay. Androgen extraction and assay We assayed yolk androgen concentrations using competitive binding radioimmunoassay of yolk extract. Methods for yolk androgen extraction and radioimmunoassay followed those of Schwabl (1993), as described in detail by Pilz et al. (2003). Radioimmunoassays were conducted for A4, dihydrotestosterone, and T. As we have done previously (Pilz and Smith, 2004; Pilz et al., 2003), here we only consider the androgens A4 and T, which are the most ecologically relevant in this system (see Pilz et al., 2003). Extraction recoveries averaged 53.3 6 6.4% (mean 6 SD) for A4 and 61.5 6 5.7% for T, intra-assay variation was 9.2% for A4 and 9.8% for T, and interassay variation was 12.1% for both A4 and T. Since all eggs could not be extracted and radioimmunoassayed simultaneously, eggs were divided into three assays (two for eggs collected in 1999 and one for 2000). Clutches were randomly distributed across the assays, but all eggs within a clutch were run side by side within the same assay. Thus, intra- and interassay variability should not create any bias in our data. Maternity analysis We identified brood parasitic eggs in putatively parasitized clutches using isoelectric focussing electrophoresis (IEF) of crude egg albumen, as described by Andersson and Åhlund (2001). In this method, albumen proteins are run to their isoelectric point on polyacrylamide gels with immobilized ph gradients. The suite of maternal proteins in the crude albumen creates complex banding patterns, with high probability of differences between females. Full clutches were run together on gel types A and D (sensu Andersson and Åhlund, 2001). Eggs whose banding patterns were dissimilar from all other eggs in the same nest were considered brood parasitic eggs (Figure 1). When results from a gel were ambiguous, the gel was run again. In two clutches, one egg was identified as probably parasitic, but this conclusion was considered ambiguous due to relatively little difference in the banding patterns of the putative parasitic egg and the rest of the clutch. These two clutches were also run on gel type B, which confirmed that the ambiguous egg was indeed parasitic. Two authors (K.M.P., M.A.) scored all gels independently, blind to the identity of eggs within clutches, and came to the same final conclusions about which eggs were brood parasitic in all cases. False positives, where a nonparasitic egg is classified as a parasite, would be problematic for our study due to the sample size. However, Figure 1 Representative isoelectric focussing electrophoresis gel of type D (as described in Andersson and Åhlund, 2001), showing two clutches of six eggs (including the parasitic egg). Inferred parasitic eggs are indicated with *. In clutch A, eggs in lanes 5 and 6 were laid on the same day. In clutch B, eggs in lanes 2 and 3 were laid on the same day. IEF analysis of 69 eggs that were not expected to be parasitic (based on laying patterns and size) across 17 clutches revealed only 1 parasitic egg. If this egg was a false positive (unlikely, since it was different from other eggs in its clutch in both the A and D gel type in several bands), this would correspond to less than a 1.5% probability of false positives. False negatives (where the parasitic and host females share protein fingerprints) are also rare; for example, whenever two eggs were laid on the same day in the same clutch, one was always identified as parasitic (N ¼ 10, excluding two eggs on the first day of laying see Results). Statistical analysis Yolk androgen concentrations were log transformed [log (concentration þ 1)] to normalize the data. We analyzed the difference between parasitic or dumped eggs and controls using paired t tests. We used two controls for parasitic eggs. We first compared each parasitic egg to the host egg laid in the same nest on the same day or to the egg laid on the previous day if none was laid on the same day (hereafter referred to as paired egg controls ). We then compared values for parasitic eggs to the mean value for all eggs laid on the same day as the parasitic egg ( date controls ). Dumped eggs were compared only to date controls. Date controls were useful since yolk androgens vary with laying order (later eggs within clutches having more androgens; Pilz et al., 2003). We could not control for laying order itself because we could not determine the laying-order position of parasitized eggs. However, since the synchronous laying season is so short in this population (e.g., all eggs were laid over 10 days in 1999), lay date and laying order are highly correlated (analysis including 176 eggs from 36 clutches, b ¼ 0.559, r 2 ¼.429; for both the regression and the correlation, p,.0001). Date is also a strong predictor of yolk androgen concentration (e.g., for A4: F 1,55.6 ¼ 26.32; p,.0001). Thus, lay date is a good surrogate for laying order. Sample sizes for comparisons involving parasitic eggs were typically 14. For this sample size, paired t tests have 80% power to detect a mean difference of 0.75 3 SD. Sample sizes for comparisons involving dumped

510 Behavioral Ecology Table 1 Mean 6 SE of egg mass, yolk mass, and yolk androgen concentrations (log transformed) of parasitic eggs, paired eggs, and date controls Variable Parasitic egg Paired egg Date control N SD pop d 80 Egg mass (g) 7.128 6 0.108 7.273 6 0.144 7.157 6 0.029 14 0.570 0.254 Yolk mass (g) 1.156 6 0.035 1.180 6 0.023 1.162 6 0.007 13 0.098 0.046 log (pg A4 mg ÿ1 þ 1) 1.272 6 0.089 1.246 6 0.107 1.273 6 0.081 14 0.324 0.144 log (pg T mg ÿ1 þ 1) 0.287 6 0.023 0.284 6 0.034 0.283 6 0.015 14 0.118 0.053 The paired egg was laid in the same nest on the same day (or on the previous day when no egg was laid the same day) as the parasitic egg. The date control is the mean value for eggs laid on the day the parasitic egg was laid. N shows the sample size (which is one less for yolk mass because a yolk ruptured before weighing). SD pop shows the standard deviation for that trait in the population, based on a sample of 180 eggs collected in 1999. d 80 is the minimum detectable mean difference between samples (in absolute units), given the sample size and 80% power (Zar, 1996). eggs were typically 10, for which paired t tests have 80% power to detect a mean difference of 0.89 3 SD. We consider this power reasonable since parasitic chicks must outcompete unknown host chicks, whose eggs could have any level of size or hormone content. For each paired t test, we present the minimum difference (in absolute units) that our test could detect with 80% power (d 80 ). For these calculations, we used estimates of the population standard deviation (SD pop ) based on 180 nonparasitic eggs collected in 1999 (Zar, 1996). We analyzed differences in egg characteristics between parasitized (host) clutches and nonparasitized clutches using repeated measures mixed models in the SAS System, Version 8.0. Eggs within clutches were the repeated measures. When marginally significant (p,.1), laying order, clutch size, and laying start date were included in the models as fixed effect covariates. Included covariates are described parenthetically with the relevant results below. For further details on the mixed models methods see Pilz et al. (2003). RESULTS Brood parasitism Of the 16 putatively parasitized clutches, we found 3 clutches that contained no parasitic eggs. These three clutches were suspected of parasitism because two eggs were found on the first day of laying in the nest. The explanation of this phenomenon may be that starlings lay the first egg of the clutch later in the day than their second egg (Feare et al., 1982). We identified 14 parasitic eggs in 13 clutches, including one that we did not suspect of parasitism. Characteristics of parasitic eggs Means and standard errors of egg parameters from parasitic eggs, paired eggs, and date controls are given in Table 1, and mean differences between parasitic eggs and controls are shown in Figure 2. Parasitic eggs did not differ in mass from paired eggs (t 13 ¼ÿ0.86, p ¼.41) or from date controls (t 13 ¼ ÿ0.25, p ¼.81). Yolk mass of parasitic eggs also did not differ significantly from that of paired eggs (t 12 ¼ÿ0.61, p ¼.56) or date controls (t 12 ¼ÿ0.15, p ¼.88). Examining residuals of yolk mass regressed on egg mass indicated that parasitic eggs have relatively smaller yolk mass for their egg size than controls, although not significantly so (parasitic versus paired eggs: t 10 ¼ÿ0.98, p ¼.35; parasitic eggs versus date controls: t 12 ¼ÿ0.38, p ¼.71). Concentrations of yolk A4 in parasitic eggs did not differ from those in paired eggs (t 13 ¼ 0.22, p ¼.83) or date controls (t 13 ¼ 0.01, p ¼.99), neither did concentrations of yolk T in parasitic eggs differ from those in paired eggs (t 13 ¼ 0.07, p ¼.94) or date controls (t 13 ¼ 0.22, p ¼.83). Characteristics of dumped eggs Comparisons of egg characteristics from dumped eggs and date controls are given in Table 2, and relative differences are shown in Figure 3. Dumped eggs were significantly smaller than date controls in both egg mass (t 10 ¼ÿ3.39, p ¼.0069) and yolk mass (t 9 ¼ÿ2.29, p ¼.047). Dumped eggs did not differ from date controls in either concentration of yolk A4 (t 9 ¼ÿ0.37, p ¼.72) or yolk T (t 9 ¼ 0.59, p ¼.57). Excluding eggs found on the ground from these analyses did not change the results, except that yolk mass was then only marginally smaller in dumped eggs (t 8 ¼ÿ1.85, p ¼.10). Differences between parasitized and nonparasitized clutches Repeated measures ANOVAs revealed that parasitized (host) and nonparasitized clutches did not differ in egg mass (F 1,28.2 ¼ 0.53, p ¼.47), yolk mass (F 1,28.2 ¼ 0.64, p ¼.43; laying start date: F 1,28.7 ¼ 8.91, p ¼.0057), concentration of yolk A4 (F 1,28.0 ¼ 0.02, p ¼.90; laying order: F 1,123 ¼ 66.28, p,.0001), or concentration of yolk T (F 1,27.2 ¼ 0.36, p ¼.55; laying order: F 1,123 ¼ 27.26, p,.0001; laying start date: F 1,27.5 ¼ 4.19, p ¼.05). Thus, brood parasites did not preferentially parasitize clutches with particular egg characteristics (e.g., they did not target clutches with small or androgen-poor eggs). Figure 2 Bars represent absolute mean differences (6SEs) between parasite eggs and paired eggs or date controls (see Methods). Represented are egg mass, yolk mass, log androstenedione (A4) concentration [log(pg A4 mg ÿ1 þ 1)], and log testosterone (T) concentration [log(pg T mg ÿ1 þ 1)]. For each characteristic, we subtracted from the value for each parasitic egg the value for its paired egg or date control, and then calculated the mean and SE of those differences. In no case do parasitic eggs differ significantly from paired eggs or date controls.

Pilz et al. Brood parasitic starlings do not lay high-quality eggs 511 Table 2 Mean 6 SE of egg mass, yolk mass, and yolk androgen concentrations (log transformed) of dumped eggs and date controls Variable Dumped egg Date control N SD pop d 80 Egg mass (g) 6.506 6 0.176 7.202 6 0.047 11 0.570 0.286 Yolk mass (g) 1.078 6 0.033 1.203 6 0.024 10 0.098 0.053 log (pg A4 mg ÿ1 þ 1) 1.221 6 0.201 1.277 6 0.127 10 0.324 0.171 log (pg T mg ÿ1 þ 1) 0.306 6 0.043 0.281 6 0.018 10 0.118 0.062 See legend of Table 1 for explanation of table presentation. A total of 12 dumped eggs were found. One egg was excluded from the egg and yolk mass analyses because its shell was cracked when found; one egg was excluded from the yolk mass analysis due to yolk rupture before weighing, and two eggs were not analyzed for yolk steroid content due to assay size constraints. DISCUSSION Since brood parasitic chicks should compete more vigorously with nest mates than nonparasitic chicks, we predicted that CBP eggs would be larger, larger-yolked, and/or have higher yolk androgen concentrations than nonparasitic eggs. Despite the fact that we tested these predictions in a species where CBP is common and widespread, and in a population where almost all CBPs are engaging exclusively in CBP, we failed to confirm our predictions. Moreover, dumped eggs were significantly smaller in size than normal eggs and did not differ in androgen concentrations. Our failure to find that CBP eggs are larger or have larger yolks is unlikely to be a problem of sample size. Parasitic eggs were slightly (nonsignificantly) smaller than controls, in direct opposition to our prediction, so increased sample size is unlikely to resolve this problem. Furthermore, we had sufficient power to detect that dumped eggs were significantly smaller than normal eggs, despite an even smaller sample size than that for the parasitic comparisons. Failure of CBPs to lay larger eggs or larger-yolked eggs could result from costs of investment in these traits. There is Figure 3 Absolute mean differences (6SEs) between dumped eggs and date controls for egg mass, yolk mass, log androstenedione (A4) concentration [log(pg A4 mg ÿ1 þ 1)], and log testosterone (T) concentration [log(pg T mg ÿ1 þ 1)]. For each characteristic, we subtracted from the value for each dumped egg the value for its date control, and then calculated the mean and SE of those differences. Asterisks indicate significant differences between dumped and control eggs based on paired t tests: **p,.01, *p,.05. substantial evidence that female investment in egg size is costly (Carey, 1996; Monaghan and Nager, 1997), for example, egg laying depletes energy and protein reserves in many bird species (Drent and Daan, 1980). Starlings are income breeders and use daily food intake to provide the nutrients for eggs (Meijer and Drent, 1999), so laying extra eggs does not deplete macronutrient reserves (Christians, 2000). However, a starling female s ability to produce larger eggs might be constrained by her daily food intake (Källander and Karlsson, 1993; but see Meijer and Langer, 1995). Furthermore, while hatching from larger eggs is beneficial to chicks (see Introduction), enhancing growth and survival, these benefits may not be mediated by competition with nest mates. Large eggs could merely improve chick quality, allowing chicks to survive and grow better without actually outcompeting nest mates. In this case, CBPs would not benefit more than nesting females from laying larger or larger-yolked eggs. Thus, the fact that CBPs do not lay larger eggs than nesting females could imply that the costs of egg investment outweigh the benefits for these females. This is especially likely considering that brood parasites in this population appear to be low-quality females (Sandell and Diemer, 1999). Because dumped eggs may be laid by parasites of particularly low quality, for example, that are unable to find or successfully parasitize active nests, our finding that dumped eggs are significantly smaller than control eggs is consistent with this interpretation. The competition hypothesis of yolk androgen allocation proposes that females differentially allocate yolk androgens to eggs within a clutch to create asymmetries in chick competitive ability. This hypothesis does not require that yolk androgen allocation is costly to females. Previous support for the hypothesis comes from Eising et al. (2001), who found that untreated black-headed gull (Larius ridibundus) chicks grow more slowly when their broodmates had hatched from eggs treated with androgens, and Schwabl (1996b), who found that canary (Serinus canaria) chicks hatched from T- treated eggs beg more than control chicks. This hypothesis predicts that CBPs will allocate higher levels of androgens to their eggs than nesting females, but we did not find such differences. In accordance with this failure of the competition hypothesis in starlings, we recently found that experimental increase of yolk T in starling eggs enhances chick growth and survival but does not increase begging behavior (Pilz et al., 2004). Although our sample size was small, lack of statistical power probably does not explain our failure to find higher concentrations of yolk androgens in CBP eggs (see Tables 1 and 2). Parasitic chicks must compete with chicks hatching from eggs from a range of sizes and yolk androgen concentrations, and parasitic mothers should always favor greater competitive ability in offspring than nesting females. Thus, if amplifying yolk androgens enhance offspring competitiveness and mothers can manipulate these characters without substantial cost, we expect parasitic eggs to surpass all

512 Behavioral Ecology eggs laid by nesting females in yolk androgens, due to the much higher potential fitness benefits. With our sample size, we could detect with 80% power a mean difference of 0.75 3 SD between parasitic and control eggs. Thus, we were able to detect relatively moderate differences in yolk androgen concentrations. Mean differences between parasitic and control eggs in yolk androgen concentrations never exceeded 2.1%, far from the dramatic differences predicted. Parasitism explains approximately 0.3% of the variation in yolk A4 concentration and 0.9% of the variation in yolk T concentration. Such minor differences are unlikely to be biologically important. Our data may be better explained by the investment hypothesis of yolk androgen allocation, which proposes that yolk androgens represent a form of reproductive investment by females (Gil et al., 1999; Pilz et al., 2003). This hypothesis implies that yolk androgens are costly for females to invest but enhance mean offspring fitness. Costs for females are unknown but could include reduced immunity (De Ridder et al., 2002) or inhibition of gonadotropin/gnrh release (Norris, 1997). Benefits to chicks would probably be mediated by a noncompetitive mechanism, such as increased growth, immunity, or survival, since increased competitive vigor of all chicks in a clutch is unlikely to increase mean chick fitness. Our previous research in starlings has supported predictions of the investment hypothesis from both the maternal and offspring perspectives (Pilz et al., 2003, 2004). The investment hypothesis has also received support from Eising et al. (2003), who found that black-headed gull chicks hatching from androgen-treated eggs do not expend more energy than controls even though they grow faster, and from Gil et al. (1999, 2004), who found that avian females invest more androgen in eggs when mated/exposed to apparently higher quality males. The investment hypothesis does not predict higher levels of androgens in the eggs of CBPs than in the eggs of nesting females, since the predicted benefits of high yolk androgen levels for offspring are not competition mediated (i.e., would not provide greater benefits to CBP mothers than to nonparasitic mothers). Furthermore, since parasitic females in our population appear to be low quality (Sandell and Diemer, 1999), they may be unable to afford such costs of investment (van Noordwijk and de Jong, 1986; Williams, 1966). Thus, although this study does not explicitly test the investment hypothesis, the results are consistent with its predictions. The third hypothesis of yolk androgen allocation, the physiological epiphenomenon hypothesis, proposes that the occurrence of androgens in egg yolk is an incidental byproduct of ovarian androgen synthesis related to maternal physiology (Pilz et al., 2003; Schwabl, 1993). Under this hypothesis, one might expect yolk androgen levels in parasite eggs to reflect the social environment of parasites, for example, whether they engage in more or less aggressive social interactions than breeding females. However, this information is not known for starlings or other CBPs. This hypothesis makes no a priori predictions regarding differences in yolk androgen allocation between CBPs and nesting females. If females are physiologically constrained in terms of the amount of hormone they allocate to eggs or in the size of the eggs they lay, CBPs may be unable to respond to selection pressure for laying larger eggs or eggs with more androgen. Within females, yolk androgens vary across laying order and across clutches (Gil, 2003; starlings: Pilz et al., 2003). Also, huge variation in yolk androgens is typical among females: androgen concentrations can vary 10-fold across eggs (Pilz et al., 2003; Reed and Vleck, 2001; Schwabl, 1993). Yolk androgen levels respond to photoperiod (Schwabl, 1996a) and to perceived mate quality (Gil et al., 1999, 2004), indicating that avian females are able to manipulate yolk androgen levels. Egg size varies relatively little within females in most species of birds (Christians, 2002), including starlings (Christians and Williams, 2001; Smith et al., 1993), and anatomical or physiological traits of the female may set important constraints (Christians, 2002). However, a critical study by Cunningham and Russell (2000) has shown that female mallards can flexibly manipulate egg size in response to mate quality, and food manipulation has been reported to affect egg size in eight species, including the starling (Christians, 2002; Källander and Karlsson, 1993). Thus, egg size is malleable in at least some avian species possibly including the starling. In conclusion, both egg size and yolk androgen levels appear manipulable, although more evidence is needed. Failure to find support for our predictions regarding characteristics of CBP eggs could result from high costs of investing in eggs, lack of competition-based benefits for chicks, or physiological constraints on egg manipulation. To discriminate between these possibilities, the direct consequences of variation in egg traits, in terms of costs and benefits, needs further attention. The evidence available favors the hypothesis that parasitic females cannot afford to invest costly resources in eggs and that parasitic females are of lower than average quality. Whether eggs of parasitic and nesting females might differ in other components remains to be determined. We thank H. Schwabl for use of his laboratory for steroid extraction and assay, A.P. Bretscher, D.L. Deitcher and I. Vilinsky for materials and support for maternity analyses, M. Åhlund for advice on isoelectric focussing, and M. Granbom and M. Bruun for assistance with fieldwork. We thank E. Adkins-Regan, D.W. Winkler, S.T. Emlen, and A.H. Bass and two anonymous reviewers for helpful comments on the manuscript. H.G.S. was supported by the Swedish Agricultural and Forestry Research Council (SJFR), M.A. by the Swedish Science Research Council (VR) and K.M.P. by a Howard Hughes Medical Institute Predoctoral Research Fellowship. Research funding was provided to K.M.P. by the Chapman Fund of the American Museum of Natural History, a Bleitz Award and an AOU Supplement Award from the American Ornithologists Union, the Fuertes Award from the Wilson Ornithological Society, an NSF Doctoral Dissertation Improvement Grant (IBN-0104907), and by NSF grant IBN-9514088 awarded to E. Adkins-Regan. The methods adhere to the Guide for the Care and Use of Laboratory Animals (NIH Publication 80-23) and were approved by the Malmö/Lund Animal Care Committee. REFERENCES Amundsen T, Lorentsen S, Tveraa T, 1996. Effects of egg size and parental quality on early nestling growth: an experiment with Antarctic petrel. J Anim Ecol 65:545 555. Andersson M, Åhlund M, 2001. Protein fingerprinting: a new technique reveals extensive conspecific brood parasitism. Ecology 82:1433 1442. Cabe PR, 1999. Dispersal and population structure in the European starling. Condor 101:451 454. Carey C, 1996. Female reproductive energetics. In: Avian energetics and nutritional ecology (Carey C, ed). New York: International Thomson Publishing; 324 369. Challenger WO, Williams TD, Christians JK, Vezina F, 2001. Follicular development and plasma yolk precursor dynamics through the laying cycle in the European starling (Sturnus vulgaris). Physiol Biochem Zool 74:356 365. Christians JK, 2000. Producing extra eggs does not deplete macronutrient reserves in European starlings Sturnus vulgaris. J Avian Biol 31:312 318. Christians JK, 2002. Avian egg size: variation within species and inflexibility within individuals. Biol Rev 77:1 26.

Pilz et al. Brood parasitic starlings do not lay high-quality eggs 513 Christians JK, Williams TD, 2001. Intraspecific variation in reproductive physiology and egg quality in the European starling Sturnus vulgaris. J Avian Biol 32:31 37. Cotton PA, Wright J, Kacelnik A, 1999. Chick begging strategies in relation to brood hierarchies and hatching asynchrony. Am Nat 153:412 420. Cunningham EJA, Russell AF, 2000. Egg investment is influenced by male attractiveness in the mallard. Nature 404:74 76. Davies NB, 2000. Cuckoos, cowbirds and other cheats. London: Poyser. Dearborn DC, 1998. Begging behavior and food acquisition by brownheaded cowbird nestlings. Behav Ecol Sociobiol 43:259 270. De Ridder E, Pinxten R, Mees V, Eens M, 2002. Short- and long-term effects of male-like concentrations of testosterone on female European starlings (Sturnus vulgaris). Auk 119:487 497. Drent RH, Daan S, 1980. The prudent parent: energetic adjustment in avian breeding. Ardea 68:225 252. Eadie J, Sherman P, Semel B, 1998. Conspecific brood parasitism, population dynamics, and the conservation of cavity-nesting birds. In: Behavioral ecology and conservation biology (Caro T, ed). Oxford: Oxford University Press; 306 340. Eising CM, Eikernaar C, Schwabl H, Groothuis TGG, 2001. Maternal androgens in black-headed gull (Larius ridibundus) eggs: consequences for chick development. Proc R Soc Lond B 268:839 846. Eising CM, Visser GH, Muller W, Groothuis TG, 2003. Steroids for free? No metabolic costs of elevated maternal androgen levels in the black-headed gull. J Exp Biol 206:3211 3218. Etches RJ, 1996. Reproduction in poultry. Wallingford, UK: CAB International. Evans PGH, 1988. Intraspecific nest parasitism in the European starling Sturnus vulgaris. Anim Behav 36:1282 1294. Feare CJ, 1991. Intraspecific nest parasitism in starlings Sturnus vulgaris: effects of disturbance on laying females. Ibis 133:75 79. Feare CJ, Spencer PL, Constantine DAT, 1982. Time of egg-laying of starlings Sturnus vulgaris. Ibis 124:174 178. Finkler MS, Van Orman JB, Sotherland PR, 1998. Experimental manipulation of egg quality in chickens: influence of albumen and yolk on the size and body composition of near-term embryos in a precocial bird. J Comp Physiol [B] 168:17 24. Gil D, 2003. Golden eggs: maternal manipulation of offspring phenotype by egg androgen in birds. Ardeola 50:281 294. Gil D, Graves J, Hazon N, Wells A, 1999. Male attractiveness and differential testosterone investment in zebra finch eggs. Science 286:126 128. Gil D, Leboucher G, Lacroix A, Cue R, Kreutzer M, 2004. Female canaries produce eggs with greater amounts of testosterone when exposed to preferred male song. Horm Behav 45:64 70. Godfray HCJ, 1995. Evolutionary theory of parent-offspring conflict. Nature 376:133 138. Hamilton WD, 1964. The genetical evolution of social behavior. J Theor Biol 7:1 52. Johnstone RA, Grafen A, 1993. Dishonesty and the handicap principle. Anim Behav 46:759 764. Källander H, Karlsson J, 1993. Supplemental food and laying order in the European starling. Condor 95:1031 1034. Karlsson J, 1983. Breeding in the starling (Sturnus vulgaris) (PhD thesis). Lund, Sweden: University of Lund. Kempenaers B, Sheldon BC, 1996. Why do male birds not discriminate between their own and extra-pair offspring? Anim Behav 51:1165 1173. Kilner R, Johnstone RA, 1997. Begging the question: are offspring solicitation behaviours signals of need? Trends Ecol Evol 12:11 15. Lichtenstein G, 2001. Selfish begging by screaming cowbirds, a mimetic brood parasite of the bay-winged cowbird. Anim Behav 61:1151 1158. Lichtenstein G, Sealy SG, 1998. Nestling competition, rather than supernormal stimulus, explains the success of parasitic brownheaded cowbird chicks in yellow warbler nests. Proc R Soc Lond B 265:249 254. Lombardo MP, Power HW, Stouffer PC, Romagnano LC, Hoffenberg AS, 1989. Egg removal and intraspecific brood parasitism in the European starling (Sturnus vulgaris). Behav Ecol Sociobiol 24: 217 223. Meijer T, Drent R, 1999. Re-examination of the capital and income dichotomy in breeding birds. Ibis 141:399 414. Meijer T, Langer U, 1995. Food availability and egg-laying of captive European starlings. Condor 97:718 728. Monaghan P, Nager RG, 1997. Why don t birds lay more eggs? Trends Ecol Evol 12:270 274. Norris DO, 1997. Vertebrate endocrinology, 3rd ed. San Diego: Academic Press. Payne RB, 1977. The ecology of brood parasitism in birds. Ann Rev Ecol Syst 8:1 28. Petrie M, Krupa A, Burke T, 1999. Peacocks lek with relatives even in the absence of social and environmental cues. Nature 40:155 157. Pilz KM, 2003. Egg yolk androgens in the European starling (Sturnus vulgaris): maternal allocation and offspring effects (PhD dissertation). Ithaca, NY: Cornell University. Pilz KM, Quiroga M, Schwabl H, Adkins-Regan E, 2004. European starling chicks benefit from high yolk testosterone levels during a drought year. Horm Behav 46:179 192. Pilz KM, Smith HG, 2004. Egg yolk androgen levels increase with breeding density in the European starling, Sturnus vulgaris. Funct Ecol 18:58 66. Pilz KM, Smith HG, Sandell M, Schwabl H, 2003. Inter-female variation in egg yolk androgen allocation in the European starling: do high quality females invest more? Anim Behav 65:841 850. Pinxten R, Eens M, Verheyen RF, 1991a. Conspecific nest parasitism in the European starling. Ardea 79:15 30. Pinxten R, Eens M, Verheyen RF, 1991b. Responses of male starlings to experimental intraspecific brood parasitism. Anim Behav 42: 1028 1030. Rahin H, Curran-Everett L, Booth DT, 1988. Eggshell differences between parasitic and nonparasitic icteridae. Condor 90:962 964. Reed WL, Vleck CM, 2001. Functional significance of variation in eggyolk androgens in the American coot. Oecologia 128:164 171. Sandell MI, Diemer M, 1999. Intraspecific brood parasitism: a strategy for floating females in the European starling. Anim Behav 57: 197 202. Schwabl H, 1993. Yolk is a source of maternal testosterone for developing birds. Proc Natl Acad Sci USA 90:11446 11450. Schwabl H, 1996a. Environment modifies the testosterone levels of a female bird and its eggs. J Exp Zool 276:157 163. Schwabl H, 1996b. Maternal testosterone in the avian egg enhances postnatal growth. Comp Biochem Physiol A 114:271 276. Schwabl H, Mock DW, Gieg JA, 1997. A harmonal mechanism for parental favouritism. Nature 386:231. Smith HG, 2004. Selection for synchronous breeding in the European starling. Oikos 105:301 311. Smith HG, Bruun M, 1998. The effect of egg size and habitat on starling nestling growth and survival. Oecologia 115:59 63. Smith HG, Ottosson U, Ohlsson T, 1993. Interclutch variation in egg mass among starlings Sturnus vulgaris reflects female condition. Ornis Scand 24:311 316. Smith HG, Ottosson U, Sandell MI, 1994. Intrasexual competition among polygynously mated female starlings (Sturnus vulgaris). Behav Ecol 5:57 63. Sockman KW, Schwabl H, 2000. Yolk androgens reduce offspring survival. Proc R Soc Lond B 267:1451 1456. Stouffer PC, Kennedy ED, Power HW, 1987. Recognition and removal of intraspecific parasite eggs by starlings. Anim Behav 35:1583 1584. Stouffer PC, Power HW, 1991. Brood parasitism by starlings experimentally forced to desert their nests. Anim Behav 41:537 539. Styrsky JD, Eckerle KP, Thompson CF, 1999. Fitness-related consequences of egg mass in nestling house wrens. Proc R Soc Lond B 266:1253 1258. Trivers RL, 1974. Parent-offspring conflict. Am Zool 14:249 264. van Noordwijk AJ, de Jong G, 1986. Acquisition and allocation of resources: their influence on variation in life history tactics. Am Nat 128:137 142. Williams GC, 1966. Natural selection, the costs of reproduction, and a refinement of Lack s principle. Am Nat 100:687 690. Williams TD, 1994. Intraspecific variations in egg size and egg composition in birds: effects on offspring fitness. Biol Rev 68:35 59. Yom-Tov Y, Dunnet GM, Anderson A, 1973. Intraspecific nest parasitism in the starling Sturnus vulgaris. Ibis 116:87 90. Zar JH, 1996. Biostatistical analysis. New Jersey: Prentice-Hall.