Nestbox availability affects extra-pair fertilizations and conspecific nest parasitism in eastern bluebirds, Sialia sialis

Anim. Behav., 1991, 41, 661-675

Nestbox availability affects extra-pair fertilizations and conspecific nest parasitism in eastern bluebirds, Sialia sialis P A T R I C I A A D A I R G O W A T Y * & W I L L I A M C. B R I D G E S t *Department of Biological Sciences, Clemson University, Clemson, SC 29634, U.S.A. t Department of Experimental Statistics, Clemson University, Clemson, SC 29634, U.S.A.

(Received27 October1989; initialacceptance30 January 1990; final acceptance 14 September1990; MS. number:3480)

Abstract. Eastern bluebird pairs are usually monogamous, cooperating m caring for their offspring.

However, alternative reproductive tactics including copulations with more than one partner and the laying of eggs in nests of conspecifics also occur. These alternative reproductive tactics could be adaptive when nest site availability varies, and different predictions have been made of who should employ these tactics, and when. Two experiments involving manipulation of nestboxes discriminated between these possibilities. The first experiment involved placement of nestboxes so that breeding pairs of bluebirds were relatively close together, a standard distance apart or relatively far apart. Nestlings non-descendant (measured using isozyme exclusion analysis) from one or both of their putative parents, the male and female care-givers, were least likely when nesting territories were relatively far apart and scarce, and increased significantly in experimental areas where nesting pairs were relatively close together and abundant, allowing the inference that neighbours were the source of non-directly descendant offspring. This is the first experimental evidence for birds demonstrating that individuals engage in genetically effective alternative reproductive tactics as a function of their opportunity to do so. The second experiment involved removal of nestboxes so that only two-thirds of known breeders had nesting sites. The rate of conspecific nest parasitism increased after removal providing the first experimental evidence for alternative reproduction by females when nest sites were not available. These two experiments were performed sequentially over two years. During these years extra-pair fertilization was significantly more frequent than conspecific nest parasitism.

Extra-pair fertilization and conspecific nest parasitism both occur in bird species with biparental care (Gowaty & Karlin 1984; Gavin & Bollinger 1985; Emlen & Wrege 1986; Gowaty & Davies 1986; Westneat 1987a; Sherman& Morton 1988; Brown& Brown 1989; Price et al. 1989). Two hypotheses about the identities of individuals that engage in extra-pair fertilization and conspecific nest parasitism exist. The first states that female and male breeders should behave opportunistically with regard to reproduction. In other words, males should be selected to care for the offspring of primary mates and to seek copulations with females for whose offspring they do not care (Trivers 1972), and females should be selected to reduce the costs of their own reproduction by laying eggs both in their own nests and in the nests of conspecifics (Gowaty 1985; Davies 1988). The other hypothesis states that extrapair fertilization and conspecific nest parasitism are reproductive tactics of 'floater' individuals, those 0003-3472/91/040661+15 $03.00/0

without access to breeding territories or resources necessary for reproduction, such as nesting sites or mates, An assumption of this hypothesis is that floating is facultative rather than obligate such that individuals with access to breeding resources will behave as territorial breeders. These two hypotheses, the opportunistic breeder hypothesis and the facultative floater hypothesis, make alternative predictions. The facultative floater hypothesis predicts that extra-pair fertilization and conspecific nest parasitism will be most common when resources for reproduction are rare relative to the population of potentially breeding individuals. The opportunistic breeder hypothesis predicts that they will be most common when resources are close together and abundant. These two hypotheses are not mutually exclusive, because opportunistic breeders and facultative floaters may both exist is some species. However, for species in which it is difficult to identify clearly individuals responsible

9 1991 The Association for the Study of Animal Behaviour 661

662

Animal Behaviour, 41, 4

for extra-pair fertilization and conspecific nest parasitism, variation in the frequencies of nondescendant offspring should shed light on the identities (or on their relative proportions within a population) of individuals engaging in extra-pair fertilization or conspecific nest parasitism. For example, should extra-pair fertilization and/or conspecific nest parasitism be most common when resources necessary for reproduction are sparsely distributed, rather than close together and common, the conclusion that facultative floaters are primarily responsible would be supported. Should extra-pair fertilization and conspecific nest parasitism be most common when resources necessary for reproduction are close together and common, the conclusion that opportunistic breeders are engaging in mixed reproductive tactics would be supported. Extra-pair fertilization and conspecific nest parasitism occur among eastern bluebirds, Sialia sialis (Gowaty & Karlin 1984; P. A. Gowaty, unpublished data). On our study sites eastern bluebirds nest primarily in nestboxes provided for them. Thus, we were able to manipulate experimentally the availability of a resource essential to reproduction in this species. If extra-pair fertilization and conspecific nest parasitism are by territorial bluebirds that are employing a mixed reproductive strategy, nestlings non-descendant from their caregivers (their putative parents) should be a function of opportunity and should be most common where breeding territories are relatively close together and abundant relative to the density of potential breeders. If extra-pair fertilization and conspecific nest parasitism are by bluebirds without access to breeding resources, nestlings genetically unrelated to their care-givers should be most common where breeding territories are relatively far apart and scarce. These predictions assume that bluebirds without access to nestboxes are capable of reproduction and that bluebirds behave facultatively, 'floating' only when nestboxes are unavailable to them. Here we report the first experimental manipulations to test these predictions directly. We used allozymic variation in blood proteins as evidence of parentage to determine frequencies of genetic mismatches (also called 'exclusions') between caregivers and nestlings, and adjusted these raw frequencies by probabilities of detection. Our study is similar to the more common survey studies of uncertain parentage (Gowaty & Karlin 1984; Westneat 1987b; Wrege & Emlen 1987; Sherman

& Morton 1988; Price et al. 1989), but is unique in the use of wide-scale manipulations of habitats and resources to test experimentally a priori predictions. We stress that our primary interest at this time is in differences between treatments in the frequencies of nestlings non-directly descendant from their care-givers. However, the number of passerine species in which both conspecific nest parasitism and extra-pair fertilization are known to occur is still small. Therefore, we also report estimates, based on novel combinations of techniques, of the relative frequencies of extra-pair fertilization and conspecific nest parasitism in eastern bluebirds.

METHODS

Subjects Eastern bluebirds are sexually dichromatic thrushes (Musicapidae) that obligately breed in cavities that they do not excavate themselves (secondary cavity nesters). Widespread populations of bluebirds are now found most reliably where nestboxes are provided for them, because their access to natural cavities has been reduced. This reduction has been caused by intensive management practices that eliminate snags and dead branches, and by competition with starlings, Sturnus vulgaris, and house sparrows, Passer domesticus, both species introduced to North America. Eastern bluebirds consort primarily in monogamous pairs, but polygynous and polyandrous trios are known to occur (Verner & Willson 1969). Despite the fact that both males and females feed nestlings and fledglings (Pinkowski 1977), care of nestlings by males is not essential to the reproductive success of females (Gowaty 1983; P. A. Gowaty, R. R Robertson, G. Ball & A. Dufty, unpublished data), who alone incubate eggs and brood nestlings. Incubation lasts for 14 days and nestlings fledge when they are between 15 and 22 days old. In north-west South Carolina the breeding season lasts from March to August and eastern bluebirds attempt as many as three and four broods per season.

Nest-site Density Experiment To test the effect of nearest-neighbour distances and nest-site abundances on the frequency of nestlings non-descendant from their care-givers, we placed new nestboxes designed for use by eastern

Gowaty & Bridges: Bluebird breeding strategies bluebirds along fence rows in high, standard and low nest-site densities. We placed the boxes in areas known to be used traditionally by bluebirds on privately owned and on Clemson University properties during the winter of 1985. In the high nest-site density category, we placed 51 nestboxes an average _ SE of 32"7_+ 0"924 m apart, in the standard category we placed 51 nestboxes at least 150m apart, and in the low category, we placed 51 nestboxes at least 1 km apart. The majority of the nestboxes in each density treatment were segregated into different areas. Fifteen air miles separated the nearest boundaries of the low density and high density areas; 6 air miles separated the low and standard treatment areas; and 4 air miles separated the standard and high density areas. In early March we began censusing of each nestbox to determine the onset of nesting activity, the identities of uniquely colour-banded individuals, the numbers of eggs laid and hatched and of nestlings fledged. In many cases already-marked birds returned to the study sites to breed. In some cases we captured unmarked adults on the day the clutches they cared for hatched. We also captured and marked females during late incubation; however, we never handled unmarked females during nest building or egg laying as such handling occasionally caused abandonment of nests during earlier studies (1977-1982; P. A. Gowaty, personal observation). We visited nest sites daily after the nest cups were completed in order to determine the date the first egg of each clutch was laid. We then visited each nest every 24 h to measure and mark each egg that was laid. We assumed a female was the mother of the brood if she had been at the nest during egg laying or if she incubated the eggs. We assumed paternity if males were seen to copulate with females, guard females (Gowaty et al. 1989), or attend territories during incubation. We eliminated only one nest from our sample because of adoption of nestlings by neighbouring breeding adults (Plissner & Gowaty 1988). We censused 239 nesting attempts. Nestlings fledged from 62% of 37 attempts in the high category, 64% of 101 attempts in the low category and 63 % of l01 attempts in the standard category. The mean_+ SEnearest-neighbour distances were 69.5 _ 82.9 m in the high, about 150 m in the standard, and about 1 km in the low density category. Fifty-seven per cent of available boxes were unoccupied all season in the high category, but only 27% in each of the other categories remained unoccupied all season. We systematically collected, labelled,

663

stored and retrieved blood samples using standardized protocols to reduce the likelihood of identification errors. These protocols were developed jointly by David G. Heckel and P. A. Gowaty in 1985. The most critical aspect of these protocols included the use of data sheets and labels preprinted with the number on the United States Department of the Interior, Fish and Wildlife aluminium band (used under Master banding permit number 21815; Scientific Collecting permit, PRT-71833305; and a South Carolina Wildlife and Marine Resources Scientific and Research Permit all to P. A. Gowaty), which was placed on each individual in our study and used to identify a given individual. We sampled 50 txl of blood from the distal portion of the tibiotarsus. We pricked the skin and blood vessel using a lancet and collected blood into heparinized microhaematocrit capillary tubes using sterile techniques. We collected blood from all 10-day-old nestlings and from all adults at each nest, so that samples from 23 adults and 80 nestlings were available from high category nests, 78 adults and 281 nestlings from low category nests, and 56 adults and 212 nestlings from standard category nests. We stored blood samples on ice in the field, placing them in a freezer at - 8 0 ~ within 2--4 h of collection. The samples were then submitted to isozyme analysis by B. May using electrophoresis during July 1986 (1985 samples) and June 1987 (1986 samples). P. A. Gowaty and co-workers organized data sheets, which were preprinted and provided to B. May for scoring of individuals. We pre-sorted samples by individuals and families and assigned individuals to particular lanes in particular runs before they were analysed. This allowed all offspring associated with a consorting pair of adults (sometimes including up to three broods) to be run together on gels. B. May used standard horizontal starch gel electrophoresis techniques including modification of buffers (May et al. 1979; Selander et al. 1971), scoring gels blind with respect to age and kinship of individuals as well as to experimental treatments. Photographs ofalt runs are archived by P. A. Gowaty. Initially 40 eastern bluebird blood samples were screened for enzyme activity at 38 loci (a list of these results is available from B. May). Six of these exhibited readable variability, and were thus chosen for our analyses. We considered the electromorphic variants genotypes and compared those of care-givers (the presumea parents) with the genotypes of the nestlings for which they cared.

664

Animal Behaviour, 41, 4

Although our design is balanced for the number of boxes in each treatment, we did not control for the differences in the overall area effect, i.e. for variation in sampling area associated with the relatively large amount of land necessary for treatments in which nestboxes were 1 km apart relative to the small amount of land for treatments in which nestboxes were 25 m apart. More birds nested in the standard and low treatment boxes than in the high treatment. This resulted in variation in the number of nestlings available for testing from each nest-site density category. There were many more nesting attempts and resulting nestlings available for comparison from the standard and low category boxes than from the high category boxes. To correct for sample size variation in the three treatments, we randomly sampled from among the nestlings from the standard and low category boxes, so that we had similar numbers of tested nestlings in each of the three categories. We report results from both the full sample and the subsample below.

Nestbox Removal Experiment To determine whether individuals without access to nesting resources could breed successfully by alternative tactics, we created experimental 'floaters' by removing a resource essential for reproduction, nestboxes. The facultative floater hypothesis predicted that mismatches of electromorphic genotypes between care-givers and nestlings would increase after box removal in experimental areas compared to control areas. We observed regular nesting activities during the spring broods of 1986 on 12 study sites, using methods described above. Between 9 May and 16 May, we removed 65 of 99 available nestboxes on six experimental study sites, so that only 64% ( N = 53) of pairs breeding in these areas during the spring would have access to nest cavities during the intermediate and summer brood periods. We used a random scheme to determine which nestboxes to remove. We waited to remove boxes until all nestlings from earlier broods were fledged; none of the boxes we removed contained eggs or nestlings on the removal date. Bluebirds have two and often three broods per year on our study sites. Thirty-six per cent of spring breeding pairs thus had no access to nestboxes on their original nesting areas during later brood periods. Throughout the remainder of the breeding season, we documented all nesting activities at each of these remaining nestboxes and at the 34 nestboxes on

the six control study sites, from which we removed no nestboxes. We again sampled and had blood electrophoresed from all 10-day-old nestlings and each of their care-givers.

Relationship of the Two Experiments We performed the nest-site density experiment during 1985 and the nestbox removal experiment during 1986. We assumed that the responses of the birds were independent between years. We modified the distributions and abundances of available nestboxes between years. The 1985 experiments necessarily required more space than the 1986 experiments; thus we used study sites on and off Clemson University properties during 1985. The 1986 experiment took place primarily on the Clemson University, Simpson Station study sites.

Initial Statistical Analyses We examined all loci for evidence of sex-linkage. We found female heterozygotes for all six loci, and, thus, we assumed that all loci were autosomal. These six loci were esterase one (ESTI), esterase two (EST2), mannose phosphate isomerase (MPI), glutathione reductase (GR), glucose phosphate isomerase (GPI) and isocitrate dehydrogenase (IDH). We tested alleles at each locus for conformance to Hardy-Weinberg expected frequencies. G R and EST2 in 1985 samples and MPI in 1986 samples were the only loci to deviate from Hardy-Weinberg expected frequencies. We further examined the allele frequencies at these loci for associations with our experimental manipulations. In all three cases, deviations in observed and expected allele frequencies were associated clearly with our experimental treatments. Two points should be made concerning the tests for Hardy-Weinberg equilibrium. First, the tests of deviation from equilibrium were significant even when we adjusted the significance level for simultaneous tests (Holland & Copenhaver 1987). Second, the Hardy-Weinberg equilibrium at each locus was not a necessary statistical condition for the subsequent exclusion analysis. However, it was necessary for the biological interpretations of the exclusion results, because the probability of detection that we use (see below) assumes that nestling genotypes occur at random with respect to the mating types in our population. We use H a r d y Weinberg equilibrium between parent generation and offspring generation as evidence that nestling genotypes occur at random with respect to mating types.

Gowaty & Bridges: Bluebird breeding strategies We tested all combinations of loci for association. We use 'association' instead of 'linkage' intentionally. To establish linkage requires test crosses. Association can be established with chisquared tests. An association can be due to linkage or to other factors. The cause of the association is not an issue in this analysis. We are concerned only whether a strong relationship exists in this data set, because our further analysis (see below) requires independent loci. We detected association between loci in one out of 15 pairwise chi-squared tests (MPI*IDH) in 1985. Three pairs of loci were related in 1986 samples (EST1 and GPI, EST2 and GR, MPI and IDH). On re-analysis for association of data pooled over 2 years, only MPI and IDH, and EST2 and G R were related. Again, these tests were significant even with the correction for simultaneous tests. Our later analyses required independent loci, because we used multiplicative laws of probability to combine probabilities of detection from each locus (see below); therefore, we deleted one locus of each associated pair, basing our choice on the number of individuals scored for the locus and variability at the locus. We used GPI, EST1, MPI and G R in our exclusion analyses.

Isozyme ExclusionAnalysis For ease of calculation we included in our initial analyses only nestlings for which we had sampled both male and female care-givers, the, so-called, putative parents. In addition and for the same reason, we also limited our analyses only to nestlings for which all loci were scored and for which all loci for each of its putative parents were scored as in one previous report of these data (given at the XXth International Ethological Congress, Madison, 1987). Later, we relaxed these limiting criteria and included all nestlings scored for at least one locus and for whom at least one parent was also scored. It is these later results we report here. To determine whether alternative reproductive tactics had been successful, we compared offspring genotypes for mismatches ('exclusions') with their putative parents' genotypes. Exclusions at one locus can be of four types: (1) unambiguous maternal exclusions in which neither allele in the tested nestling occurred in the putative mother; (2) unambiguous paternal exclusions in which neither allele in the tested nestling occurred in the putative father; (3) exclusion of both putative parents, in which neither allele in the tested nestling occurred in either of the putative

665

parents; and (4) ambiguous exclusions in which one allele in the tested nestling could not be accounted for by reference to the alleles in its putative parents (see Gowaty 1985; Westneat et al. 1987 for a discussion of these exclusion types). F o r the purposes of the experimental evaluations described here we did not differentiate between exclusion types and let any exclusion type contribute to our estimates of the frequency of nestlings non-directly descendant from their care-givers, because our primary purpose with these experiments was to determine the frequency of non-directly descendant nestlings associated with each experimental treatment. Excluded nestlings were used only once in our exclusion sums even when excluded at more than one locus, so that our estimates of non-directly descendant nestlings remain on the conservative side, uninflated by counting some individuals with exclusions at more than one locus more than once. Raw mismatch (detection) frequencies are reported in Table I and II, along with the full sample of nestlings tested in each experimental category. We did not test for differences between experimental groups in raw exclusion frequencies, because studies of parentage based on allozymes are conservative (the majority of loci have common alleles, shared by many individuals in the population, so that many cases of caring for nondescendant nestlings go undetected). Any conclusions one makes about detection frequencies between groups must be tempered by the possibility that variation in raw detections is due to (even minor) variation in allele frequencies. Our method (described below) corrects estimates of non-directly descendant nestlings to eliminate the potential bias in the raw frequency scores. We corrected raw frequencies of nestlings whose putative parent(s) were excluded from genetic parentage by the probability of detection of an exclusion. We calculated these probabilities using a technique (Gowaty & Davies 1986; Wrege & Emlen 1987) based on the actual observed mating type frequencies (combinations of putative parents' genotypes) and on the observed frequencies of offspring genotypes. In their simplest form, probabilities of detection are observed frequencies of nestling genotypes capable of rejecting particular mating types (combinations of care-giver genotypes) from genetic parentage; these we indicate as Pmt" F o r example, at a two-allele locus, if the mating type was, say, F F (maternal genotype) x F F (paternal genotype), the nestling genotypes that

Animal Behaviour, 41, 4

666

Table I. Number and percentage of non-directly descendant nestlings (NDDN) in the nest-site density experiment (1985) using the full sample of all available data and the subsampled data

Nestbox density

No. of nestlings tested (T)

No. of observed detectionst (D)

Full sample High Standard Low

69 222 232

Subsample* High Standard Low

69 45 44

P:~

No. of NDDNw

% NDDNw __+confidence interval

6 10 6

0.1962 0.2622 0.3113

31.0 38.0 19.0

44.3 __+11.7 17.1 +4.9 8.3+3.5

6 4 0

0.1962 0.3780 0.3780

31.0 11.0 0.0

44.9 ___11.7 23.6 ___12.4 0.000+0

*See text for the rationale for subsampling. tThree broods from the high density treatment, five broods from the standard, and six broods from the low density treatment had exclusions. :~Combined probability of detection over all loci; see text for computational details. w of NDDN=D/P and % N D D N = N o . of NDDN/T. The large-sample confidence interval for binomial proportions is given. Table II. Number and percentage of non-directly descendant nestlings (NDDN) in treatments of the nestbox removal

experiment (1986)

Box manipulation

Time in relation to 9 May

No. of nestlings tested (T)

No. of observed detections (D)

P*

No. of NDDNt

% NDDNt + confidence interval

Controls Not removed Not removed

Before After

19 90

7 7

0.569 0.314

12 22

64.68+21.49 24.76+ 8.92

Experimental boxes Removed Removed

Before After

146 90

12 7

0.274 0.306

44 23

29.98__+ 7.43 25.41 _+ 8.99

*Combined probability of detection over all loci; see text for computational details. tNo. of NDDN = D/P; % NDDN = No. of NDDN/T. The large-sample confidence interval for binomial proportions is given.

would indicate that one or more of the care-givers was not the genetic parent(s) (capable of rejecting this mating type) would be FS, the heterozygote, and SS, the alternative homozygote. If the frequencies of these nestling genotypes equalled 0.3 and 0.3, respectively, Pro, would equal 0"6. We used the following formula to estimate the number of non-directly descendant nestlings ( N D D N ; those detected and undetected) in each year and level of experimental manipulation over all loci N D D N = (D/P)

where P is the overall probability of detection, and D is the actual number of detected (raw) exclusions. We calculated P over all four loci (Tables I and II) as follows: (1) for each locus we calculated a probability (P~: Table III) which was the sum of the PintS; and (2) we combined the PlS using the additive law of probability. Several conditions and assumptions concerning this calculation of non-directly descendant nestlings are necessary. (1) Theoretically this method gives large weights to detections that are unlikely and smaller weights to detections that are likely to occur.

Gowaty &Bridges: Bluebirdbreedingstrategies

667

Table IlL Probabilities of detection of non-directly descendant nestlings at each electrophoretic locus in experimentaltreatments Electrophoretic loci Experimental treatments

EST1

GR

Nest site density experiment (1985) Full sample* High 0.0610 Standard 0.0099 Low 0.0289 Subsample* High 0-0734 Standard 0.0212 Low 0.0413 Nestbox removal experiment (1986) Not removed,'{"beforew 0 . 1 0 5 3 Not removed, afterw 0.0000 Removed,$ before 0.0058 Removed, after 0.0000

GPI

MPI

0.1222 0.0000 0.0248 0.1584 0.0653 0.0528 0.2256 0.0620 0.0237 0.1169 0.0000 0.0297 0.2405 0.0830 0.0864 0.2887 0.0878 0.0000 0.2105 0.1770 0.1576 0.1667

0.3906 0.1242 0.0888 0.0964

0.0000 0-0484 0.0489 0.0783

*'Full sample' indicates values computed over the entire sample of 408 nestlings.'Subsample' indicatesvaluescomputed for a subsampleof 152 nestlings. See text for calculation of probabilities of detection and rationale for subsampling. tControl areas from which boxeswere not removed. SExperimentalareas from which boxes were removed. w before and after, nestbox removal.

(2) The reliability of these corrections depends on the assumption that excluded nestling genotypes occur at random with respect to combinations of care-giver genotypes (i.e. that allele frequencies are in Hardy-Weinberg equilibrium). Recall that at its most basic the probability of detection is the sum of the frequencies of nestling genotypes capable of rejecting a particular mating type from genetic parentage. If the Hardy-Weinberg assumption is not met, our estimates would be conservative underestimates of the true frequencies of non-directly descendant nestlings. (3) The calculation of P assumed independent assortment of loci. (4) When an individual was excluded at more than one locus, we counted these exclusions as only one detection. (5) We corrected raw frequencies from each experimental manipulation from each year by the unique probability of detection associated with a given treatment. The percentage of non-directly descendant nestlings was calculated as the number of nondirectly descendant nestlings (NDDN) divided by the total number of tested nestlings (T) in each level of experimental manipulation. (We developed a computer program for the computation of these

values from observations that include only observed offspring genotypes and observed (putative) paternal and maternal genotypes for each individual offspring. Additional information about this program is available from us.) Note that this technique differs from that used by some other workers (e.g. Westneat et al. 1987) in which probabilities of exclusion types were calculated from theoretically derived genotype frequencies (from Hardy-Weinberg frequencies of alleles) rather than observed frequencies of genotypes. The purpose of their technique was to determine whether extra-pair fertilizationor conspecificnest parasitism alone could account for non-directly descendant nestlings. Our purposes were to obtain the best estimates of non-directly descendant nestlings and to determine whether our experimental manipulations changed the frequency of non-directly descendant nestlings regardless of whether the source was conspecific nest parasitism or extra-pair fertilization. So we used probabilities of exclusion to correct our raw exclusions (D/P = NDDN) and then we calculated a frequency within each treatment (% N D D N = N D D N / T ) , where T = the total number of nestlings

668

Anirnal B e h a v i o u r , 41, 4

tested in each experimental treatment. In subsequent analyses we discuss possible apportionments of N D D N to conspecific nest parasitism or extra-pair fertilization. We used categorical data analysis (Grizzle et al. 1969) for all tests among our experimental treatment groups for all years. Categorical data analysis techniques involve linear models, generalized least squares and chi-squared tests. We used the SAS procedure CATMOD (SAS) for all calculations. We calculated confidence intervals for proportions using large sample binomial approximations.

Two-eggs-in-l-day Analyses Because the statistical analysis of genetic mismatch data from the nest removal experiment was not as straightforward as we would have hoped (see below), we also used a direct measure of the frequency of conspecific nest parasitism to test for an effect of nestbox removal. The appearance of two eggs in a nest within one 24-h period is often used as a measure of the frequency ofconspecific nest parasitism (Yom-Tov 1980). The use of this measure of parasitism is based on the assumption that birds lay only one egg a day. On the day they were laid we marked and measured from all study sites 888 eggs from 211 clutches in 1985, 642 eggs from 154 clutches in 1986, and 981 eggs from 241 clutches in 1987. Bluebirds do not remove eggs of parasites experimentally placed in their nests, whether these appear before, during, or after their own egg laying has commenced (P. A. Gowaty, unpublished observations). Therefore, we corrected the raw frequency of observed 'two-eggs-in- 1-day' by a probability of detection (Pc; Fredericks & Shields 1986). In this case Pc is a function of variation in clutch size (rather than of variation in genotypes), and is based on the fact that parasitism on the day before a host begins to lay and the day after a host ends her laying is not detectable, while parasitism during any other day of the egg-laying period is detectable. We weighted Po for a given clutch size with the frequency of that clutch size to get an overall probability (Pe) for each year before and after 9 May, the date of our experimental manipulation in 1986 (Table IV). To determine the estimated number of parasitic eggs, one divides the number of observed two-eggs-in-l-day by Pe. We predicted that there would be significant differences between the periods after 9 May for the year of nestbox removal and for the same periods in the year before and the

year after the experimental year. We again used categorical data analysis (Grizzle et al. 1969) on adjusted frequencies of estimates ofconspecific nest parasitism within and between years.

Extra-pair Fertilization and Nest Parasitism The distribution of exclusion types is significantly different from both an all-conspecific-nest-parasitism or all-extra-pair-fertilization distribution (unpublished data). However, exclusion type frequencies are difficult, sometimes impossible, to partition into frequencies due conclusively to extra-pair fertilization and conspecific nest parasitism. However, because we had two independent measures of non-directly descendant nestlings, namely from allozyme variants that may indicate both extra-pair fertilization and conspecific nest parasitism, and from two-eggs-in-l-day data that indicate conspecific nest parasitism only, we are able to report the relative contributions to our estimates of non-directly descendant nestlings from conspecific nest parasitism and from extra-pair fertilization. To estimate the frequency of non-directly descendant nestlings resulting from extra-pair fertilization, we subtracted our estimates of the frequency of parasitic eggs from estimates of the total non-directly descendant nestlings in a sample (Table V). The validity of the conclusions from this calculation relies on the assumptions that (1) the frequency of parasitic eggs equals the frequency of parasitic nestlings (e.g. that the likelihood of survival of descendant and non-descendant nestlings is the same); and (2) parasitic nestlings are distributed randomly with respect to host genotypes (which will not be the case if quasi-parasitism (Emlen & Wrege 1986) is an important source of non-directly descendant nestlings, or if parasitic females lay eggs in relatives' nests, or if females preferentially copulate with other males whose genotypes are like that of their territorial consorts); (3) parasitic females do not remove host eggs; and (4) parasitic females are equally likely to lay on any day of the host's laying schedule or on the day before a host begins laying Or on the day after a host completes her clutch. We used categorical data analysis (Grizzle et al. 1969) for comparisons of the relative rates of non-directly descendant nestlings due to extra-pair fertilization and conspecific nest parasitism.

Gowaty & Bridges: Bluebird breeding strategies

669

Table IV. Distributions (N) of clutch sizes, the frequencies of clutch sizes during each period [f(N)], probabilities of detection associated with each clutch size (PC), and overall probabilities of detection (P~) in bold face from two-eggs-in-1day data before and after 9 May for 1985, 1986, and 1987 Before 9 May Clutch size

After 9 May

Pc

N

fiN)

fiN) P~

N

fiN)

~N) P~

1985 I 2 3 4 5 6

0.000 0.333 0.500 0.600 0.667 0.714

1 2 5 24 44 4 80

0.013 0.025 0.063 0'300 0.550 0.050

0.0000 0.0083 0.0315 0.1800 0.3669 0.0357 0"6224

5 1 17 70 39 0 131

0.038 0.008 0.130 0.534 0.290 0.000

0'0000 0.0027 0.0650 0.3204 0.1934 0.0000 0"5815

1986 1 2 3 4 5 6

0.000 0.333 0.500 0.600 0-667 0.714

3 1 4 21 42 I 72

0.042 0.014 0.056 0.292 0.583 0.014

0.0000 0.0047 0.0280 0.1752 0.3889 0.0100 0"6068

5 4 5 44 23 1 82

0.061 0.049 0.061 0.537 0.280 0.012

0.0000 0.0163 0.0305 0.3222 0.1868 0.0086 0"5644

1987 I 2 3 4 5 6

0.000 0.333 0.500 0-600 0.667 0.714

i 2 6 39 50 1 99

0.010 0.020 0.061 0.394 0.505 0.010

0.0000 0.0067 0.0305 0.2364 0.3368 0.0071 0"6175

6 3 27 77 29 0 142

0.042 0.021 0.190 0.542 0.204 0.000

0.0000 0.0070 0.0950 0.3252 0.1361 0.0000 0"5633

Table V. Computation of non-directly descendant nestlings (NDDN) due to extra-pair fertilizations using estimates of non-directly descendant nestlings (from both conspecific nest parasitism and extra-pair fertilization) from allozyme data and estimates of parasite nestlings from two-eggs-in-1-daydata

Year

Time in relation to 9 May

No. of nestlings tested (T)

No. of observed detections

1985 1985

Before After

215 308

1986 1986

Before After

254 91

P*

% NDDNt __+confidence interval

% Parasites:~

% EPFw _ confidence interval

7 15

0.218 0.335

14.9+4.76 14.5 _ 3.93

0.89 0.49

14.01 +__4.64 14-01 _ 3.88

26 7

0.300 0-330

33.8 __+5.82 23.2___7.08

0-52 4.36

32-28 _+5.75 18.84___8.03

*Combined probability of detection over all loci; see text for computational details. tCalculated from allozyme data. ~ ~:An estimate of the number of nestlings from conspecific nest parasitism calculated from two-eggs-in-1-daydata. w is the difference in % non-directly descendant nestlings and the % parasites estimated from conspecific nest parasitism. The large-sample confidence interval for binomial proportions is given.

Animal Behaviour, 41, 4

670

RESULTS AND DISCUSSION

Nest-site Density Experiment In both the full samples and the subsample analyses, non-directly descendant nestlings were significantly more common in nests of the high density treatment, those with relatively close neighbours, than in the standard and low treatments where nearest neighbours were further apart. In fact the likelihood of non-directly descendant nestlings from both samples increased as the nearest neighbour distance decreased. Furthermore, the percentage of non-directly descendant nestlings was greatest in areas with the highest number of unused nestboxes, suggesting that floaters were unlikely to be the source of non-directly descendant nestlings. For the full samples the percentage of non-directly descendant nestlings was significantly associated with the density of nest sites (Table I: ~2=41.91, dr=2, P=0"0001). Each density treatment differed from each of the others (contrast analyses: high-standard Z2=20.07, P=0.0001; high-low ~z=40.97, P=0.0001; low standard Z2= 7.49, P = 0"0056). For the subsample the percentage of non-directly descendant nestlings was also significantly associated with the density of nest sites (Table I: X2= 12.41, dr=2, P=0.002). Again each density treatment differed from each of the others (contrast analyses: high-standard ~2 = 4.79, P=0.0285; high-low ~2=8.74, P=0.0031; lowstandard ~2 = 5.22, P = 0-0223). Thus, we conclude that non-directly descendant nestlings are most likely to be the genetic offspring of opportunistic neighbours rather than nonterritorial individuals without access to essential resources. This result is consistent with inferences of extra-pair copulations and conspecific nest parasitism by neighbours in other species (e.g. extra-pair fertilization by neighbouring indigo buntings, Passerina cyanea (Westneat 1987a, b); conspecific nest parasitism by neighbouring whitefronted bee-eaters, Merops bullockoides (Emlen & Wrege 1986); by neighbouring wood ducks, Aix sponsa (Semel & Sherman 1986, 1988); and by neighbouring house wrens, Troglodytes aedon (Price et al. 1989). Furthermore, these data are consistent with the predictions that individuals should be selected to cooperate with a primary mate and to take alternative opportunities for additional reproductive success with individuals with whom they do not cooperate in caring for offspring (Trivers 1972).

We sought additional corroboration of the hypothesis that neighbours were the genetic parents of non-directly descendant nestlings by examining the genotypes of adult males and females whose territories were within three territories of those with observed exclusions. In the high density nests, neigbouring females carried a matching allele in four of the six exclusions. Two of the six neighbouring males carried a matching allele. In the standard density nests, neighbouring females had matching alleles in nine of 10 cases, and neighbouring males had matching alleles in three of the 10 cases. In only one case was there no near neighbour with a matching allele. It was notable that in three cases there were at least 10 neighbouring adults that carried the excluding allele. In the low density nests, four of six exclusions could not be explained by even distant neighbours, two of six nearest neighbour males had matching alleles, and in one case a neighbouring female had the matching allele. These observations indicate that alleles matching the 'stray' alleles in mismatched nestlings are most often represented in nearby individuals within the local population. However, for two reasons, these comparisons of neighbours' alleles to 'stray' alleles in the nestlings do not allow general conclusions about the source of the excluding alleles. First, it is unlikely that we sampled the entire population, and, second, in many cases the excluding allele is one that occurs frequently in the population. These conclusions are logically satisfying: neighbouring bluebirds have access to information critical to the success of alternative reproductive tactics. Neighbouring males can relatively easily observe that females are fertilizable, using either the behaviour of females or the behaviour of territorial males, who stay closer to their mates and follow them more often when they are fertilizable than when they are not (Gowaty et al. 1989). Neighbouring females can theoretically observe that nests are suitable for parasitism simply by observing the marked change in foraging behaviour (P. A. Gowaty, unpublished data) of females who have just initiated incubation.

Nestbox Removal Experiment Over the experimental and control study sites combined the percentage of non-directly descendant nestlings before box removal was 33.94%, and 25% after box removal (Table II: Z2= 9.79, df= 1, P=0.0018), a significant difference, and one

Gowaty & Bridges: Bluebirdbreeding strategies

671

80

g

.r

60

g

-o

40

"6 o~ i

g

z

2O

Before

After

Before After Experimen~'ol

Control

Figure 1. Results of contrasts of the differences in the percentage of non-directly descendant nestlings (% NDDN) before and after 9 May 1986 for the Control and Experimental areas. This analysis computed the expected values as the product of the row and column totals divided by the grand total, where the rows were the estimated number of nondirectly descendant nestlings and directly descendant nestlings (DDN) in a given sample. The probability values for the contrasts are based on chi-squared tests from categorical data analyses. II, % NDDN; D, expected. opposite to the prediction that the frequency of non-directly descendant nestlings would increase, rather than decrease, when individuals did not have access to nestboxes. There was also a significant difference between experimental (boxes removed) and control study sites in that 31.2% of nestlings from control areas were non-directly descendant from their care-givers, but only 28.4% from experimental areas were non-descendant (;(2 = 5.22, df= 1, P = 0-0223). This result is explained by an area and time interaction (;(2= 5.75, dr= 1, P = 0.0165), which arose because of the relatively high rate of non-directly descendant nestlings from control areas before nestbox removal (Table II, Fig. 1). This very high estimate may be due to the small number of tested nestlings in this class (i.e. it may be an artefact of small sample size), or it may be due to some biological factor(s) that we are at a loss to explain. Notably, contrasts analysis (Fig. 1) failed to show differences between study sites after nestbox removal (;(2 = 0-01, df-- 1, P = 0.9204) or a time effect within the experimental areas (;(2 = 0.54, df= 1, P=0.4619). The experimental removal of nestboxes failed to increase the frequency of nondirectly descendant nestlings in the population, and suggested that individuals without nestboxes may not attempt to breed using alternative reproductive behaviour patterns. Because we are unable to account for the very high estimated percentage of

B

2O

-G

~_ Io

0

Before After Before After 1985

1986

Before After 1987

Figure 2. The percentage of total clutches parasitized before and after 9 May for 1985, 1986 and 1987. A and B indicate significantly different contrasts between years. non-directly descendant nestlings in the control areas before nestbox removal, we necessarily consider these overall results somewhat questionable. We realized that, because bluebirds often fly great distances even in the breeding season (P. A. Gowaty, personal observation), the 'control' areas might have been subject to extra-pair fertilization and conspecific nest parasitism by individuals from experimental areas, something that may have affected our comparisons between control a n d ' experimental areas after nestbox removals. In further attempts to understand these results, we

672

Animal Behaviour, 41, 4

speculated that the control area results before the time of nestbox removal might be associated with seasonal variation in extra-pair fertilization and/or conspecific nest parasitism, a thought that led us to examine another independent source of data germane to alternative reproductive tactics. Seasonal Variation If seasonal variation in the occurrence of conspecific nest parasitism is regularly greater when laying synchrony is most pronounced (i.e. during earlier nesting attempts: Gowaty 1980), conspecific nest parasitism could provide a partial explanation for the high rate of non-directly descendant nestlings in control areas before nestbox removal. To test this idea we examined laying data for evidence of two-eggs-in-l-day. In 1986 we observed twoeggs-in-l-day once at one nest before 9 May, the first date of nestbox removal, and nine times after 9 May. The frequency of eggs from conspecific parasites adjusted for their probabilities of detection (Pc; Table IV) indicated that 2.29___3-5% (% ___ confidence interval) of clutches before 9 May and 19.45__ 8-6% after were parasitized, a significant difference (22=3-8764, dr=l, P<0.05). To test whether this result was associated with our experimental manipulations or with an unanticipated seasonal effect, we computed Pe and estimated the frequencies of clutches containing parasitic eggs before and after 9 May over three years, 1985, 1986 and 1987. We observed two-eggs-in-l-day twice each before and after 9 May 1985. In 1987 we never observed two-eggs-in-l-day before 9 May, but did observe two-eggs-in-l-day four times after 9 May. Table IV contains data on the total number of clutches observed for those periods and the calculation of the Pe for each period in each year. Figure 2 shows the distribution of estimated parasitized clutches before and after 9 May for 1985, 1986 and 1987. There was a significant interaction between year and time (22=6.56, df=2, P - 0 . 0 3 7 6 ) indicating that our experimental manipulation had an effect on the likelihood of eonspecific nest parasitism. Furthermore, the likelihood of conspeciflc nest parasitism was greater after 9 May (Fig. 2), not before as would have been expected if early season opportunities for conspecific nest parasitism had inflated the percentage of non-directly descendant nestlings that we computed earlier in control areas before the time of nestbox removal. Contrasts analysis (Fig. 2) indicated that 1986 differed significantly from both 1985 and 1987 (1985-1986:22 =

4.94, P=0.0262; 1986-1987: z z = 10.4, P=0-0013; 1985-1987: Zz = 1.16, P=0-2819). To test the hypothesis that extra-pair fertilizations varied seasonally, we estimated non-directly descendant nestlings before and after nestbox removal in 1986 and before and after 9 May for 1985, so that the season was divided into spring broods and later broods. Synchrony of breeding females was strongest early in the breeding season (Gowaty 1980), so if successful extra-pair copulation was associated with the abilities of males to monopolize females (Emlen & Oring 1977), extrapair fertilization should be most common later in the breeding season. Table V shows the numbers of detections, the probability of detection from allozyme data, the estimated percentage of non-directly descendant nestlings and estimated percentage of parasites from two-eggs-in-1-day data and the total number of tested nestlings for each period of 1985 and 1986. Year had a significant effect on the frequencies of extra-pair fertilizations (22=13.29, df= 1, P=0.003); time, before and after 9 May during each year, had an almost significant effect (22=3.33, df=l, P=0.0681). The year and time interaction effect was also almost significant (Z 2 = 3.26, df= 1, P = 0-0709), reflecting the high relative rate of extra-pair fertilization before 9 May 1986, the year of the nestbox removals. Contrasts analysis indicated that the frequency of extra-pair fertilizations before 9 May in 1986 was significantly higher than each of the other tested times (1985 Before1985 After: g 2 = 0-00, df= 1, P = 0.989; 1985 Before~ 1986 Before: 21=20-32, df=l, P=0.0001; 1985 Before-1986 After: 22=1-17, df= 1, P=0.2794; 1985 After-1986 Before: 2z=25-69, df=l, P= 0-0001; 1985 After-1986 After: 22 = 1.33, df= 1, P = 0-2482; 1986 Before-1986 After: 2 z = 5.71, df= 1, P = 0.0168). Notably consistent with our previous analysis of variation in the frequencies of nondirectly descendant nestlings (Table II, Fig. 1), we observed no significant effect of nestbox removal on the frequency of extra-pair fertilization. These results are interesting in that extra-pair fertilization seems no more likely and perhaps less likely as female breeding synchrony decreases, something that should increase the ability of males to monopolize females, a result opposite to that predicted by the hypothesis that extra-pair fertilization is a function of the ability of males to monopolize females (Emlen & Oring 1977). Though subject to a variety of assumptions (see above), these comparisons also provide a method

Gowaty & Bridges: Bluebird breeding strategies of estimating the relative frequencies of extra-pair fertilization and conspecific nest parasitism. Over both years about 1% (computed from the data in Table V) of all nestlings could have resulted from conspecific nest parasitism. If our assumptions are valid, the data from Fig. 2 and Table V indicate that nestbox removal did have an effect on the likelihood of conspecific nest parasitism and a similar, though not significant, effect on extra-pair fertilization. Combining data from both years, 1% of nestlings could have been from parasitically laid eggs and 19.8% of nestlings from successful extrapair copulations (again computed from the data in Table V). Overall, extra-pair fertilization seems to have been significantly more frequent than conspecific nest parasitism. We speculate that conspecific nest parasitism may be an infrequent reproductive option of females in some years, but a regular option for all females when they lose their nests to predation.

Pathology or adaptation? Because the population of bluebirds we studied uses nestboxes rather than natural cavities, it seemed possible that the conspecific nest parasitism we reported originally was somehow an 'artefact' of the habit of using nestboxes. We believe our nestsite manipulations have put this idea to rest in as much as (1) non-directly descendant nestlings occur in standard and low density treatments as well as in the high one and (2) conspecific nest parasitism increases when females are deprived of their nests, consistent with adaptive explanations. Our original observations were on study sites with nestboxes a 'standard' distance apart. Popularizers of bluebird trails (Zelany 1976) recommend placing boxes no closer than 100m apart so that typical bluebird trails, such as our earlier study sites, have nestboxes from 10(~150 m apart. There are other reasons that it is unlikely that the previous records of non-directly descendant nestlings in bluebirds (Gowaty & Karlin 1984; P. A. Gowaty, unpublished data) result from'pathology', such as that seen in egg-dumping wood ducks (Semel & Sherman 1986, 1988). Nestboxes designed for use by wood ducks that are close together, easily spotted, and clumped apparently exacerbate parasitic tendencies of wood duck females while decreasing the effectiveness of surreptitious behaviour of host females. Conspecific nest parasitism in wood ducks has, in some situations, reached what some observers (Semel & Sherman 1986) have called

673

'pathologically high levels' that should be distinguished from 'normal' frequencies of parasitism before evaluating hypotheses concerning the adaptive significance of conspeeific nest parasitism in wood ducks. Data on bluebirds is inconsistent with a similar interpretation of pathology. First, clutch sizes are never double the average clutch size except in the rare case of cooperative polygyny (Gowaty 1980), which is readily distinguished from parasitic egg laying. Second, eggs are almost always laid within 0-7 days of one another, and seldom after the second day of incubation. Third, parasitism rates are never extreme; explosive increases in clutch size, frequent nest abandonment and parasitic egg laying at inappropriate times or places, which have been observed in wood ducks (Semel & Sherman 1986, 1988), are never observed in our bluebird populations. Lastly, the rates of parasitism reported in the current study are hardly consistent with pathology. Our evidence suggests that females behave adaptively in seeking conspecifics' nests in which to lay. While it is not clear that bluebird females engage in parasitism only when they do not have access to their own nests, our nestbox removal experiment does indicate one situation when females will lay eggs parasitically. Previous estimates of the frequency of parasitism within the study population (Gowaty & Davies 1986; P. A. Gowaty, unpublished data) indicate that variation between years in the occurrence of parasitism may be large. Ecological features that would seem to favour parasitism in some species may include cavity nesting, because nesting cavities are often easy to find (Yom-Tov 1980; Andersson 1984), and highly synchronous egg laying (at least during early nesting attempts). Evaluations of the idea that cavity-nesting species are more vulnerable to parasitism will require comparative studies of species that nest in cavities with those that do not. Conspecific nest parasitism in American robins, Turdus migratorius, which like bluebirds are thrushes (Muscicapidae) but lay their eggs in open cup nests, was at least twice as high as in eastern bluebirds (Gowaty & Davies 1986), weakening the hypothesis that nest parasitism is a simple function of the visibility of nests, in these two thrush species at least. We have never observed copulations that appear to be forced; thus, we assume that females copulate willingly and may even solicit extra-pair copulations. The pattern of frequencies in extra-pair fertilizations relative to variation in female breeding synchrony is consistent with the idea that

Animal Behaviour, 41, 4

674

females rather than males control extra-pair fertilizations in this species. This idea that females may control extra-pair fertilizations and the rate of successful extra-pair fertilization that we report here suggests that there may be important advantages of extra-pair fertilization to females. These advantages may include fertility assurance, higher quality offspring, bet-hedging and decreases in the likelihood of infanticide should a female's territorial partner be replaced.

The Genetically Effective Mating System Genetic mismatches between care-givers and nestlings have been observed in almost every avian species examined for them. Because of mixed reproductive strategies of both females and males, sociographic monogamy (Wickler & Seibt 1982) is characterized by genetic conflicts of interest between individuals that cooperate in raising offspring (Davies 1989). Trivers' (1972) seminal prediction seems a truism! Importantly, the growing list of sociographically monogamous birds with evidence of non-trivial frequencies of genetically successful extra-pair fertilization or conspecific nest parasitism may further challenge (Davies 1989) traditional hypotheses (Emlen & Oring 1977) for the evolution and maintenance of particular mating systems. For example, it seems possible that 'monogamous' and 'polygynous' species may be characterized by similar gametic contribution ratios (Gowaty 1981; Wickler & Seibt 1982). It seems likely that new models may be required to explain the diversity of sociographic mating systems in birds.

ACKNOWLEDGMENTS We thank J. H. Plissner, T. Williams, K. Gauthreaux, and L. Hansen for tireless work as assistants during the field parts of this study; Bernie May at the Cornell Laboratory of Ecological and Evolutionary Genetics (CLEEG), Department of Natural Resources, College of Agriculture and Life Sciences, Cornell University, Ithaca, NY 14853 for submitting our samples to electrophoresis and J. Beltoff, D. Heckel and D. Westneat for comments on a previous draft. P.A.G. also thanks C. Helms and H. Wheeler for kind allocation of space from their laboratories and offices. The National Science Foundation grant, 8417500, to P.A.G., supported

this research; an NIMH RCDA, l-K02 MH0070601A2, supported P.A.G. during the final preparation of the manuscript for publication.

REFERENCES Andersson, M. 1984. Brood parasitism within species. In: Producers and Scroungers: Strategies of Exploitation and Parasitism (Ed. by C. J. Barnard), pp. 195-227. London: Croom Helm. Brown, C. R. & Brown, M. B. 1989. Genetic evidence of multiple parentage in broods of cliff swallows. Behav. Ecol. Sociobiol., 23, 37%387. Davies, N. B. 1988. Dumping eggs on conspecifics. Nature, Lond., 331, 9. Davies, N. B. 1989. Sexual conflict and the polygamy threshold. Anim. Behav., 38, 226-234. Emlen, S. T. & Oring, L. W. 1977. Ecology, sexual selection, and the evolution of mating systems. Science, 197, 215-223. Emlen, S. T. & Wrege, P. H. 1986.Forced copulations and intraspecificparasitism: two costs of social living in the white-fronted bee-eater. Ethology, 71, 2-29. Fredericks, P. & Shields, M. A. 1986. Correction for the underestimation of brood parasitism frequencyderived from daily nest inspection. J. Field Ornithol., 57, 224-226. Gavin, T. A. & Bollinger, E. K. 1985. Multiple paternity in a territorial passerine: the bobolink. Auk, 102, 550-555. Gowaty, P. A. 1980. Origin of mating system variability and behavioral and demographic correlations of the mating system of eastern bluebirds (Sialia sialis). Ph.D. thesis, Clemson University, Clernson, South Carolina. Gowaty, P, A. 1981. An extension of the Orians VernerWillson model to account for mating systems besides polygyny. Am. Nat., 118, 581-589. Gowaty, P. A. 1983. Male parental care and apparent monogamy in eastern bluebirds (Sialia sialis). Am. Nat., 121, 149-157. Gowaty, P. A. 1985. Multiple parentage and apparent monogamy in birds. In: Avian Monogamy (Ed. by P. A. Gowaty & D. W. Mock), pp. 11-21.Washington, D.C.: American Ornithologists' Union. Gowaty, P. A. & Davies, J. C. 1986. Uncertain maternity in American robins. In: Behavioural Ecology and Population Biology (Ed. by L. C. Drickamer), pp. 65-70. Toulouse: Privat. Gowaty, P. A. & Karlin, A. A. 1984. Multiple maternity and paternity in single broods of apparently monogamous eastern bluebirds (Sialia sialis). Behav. Ecol. Sociobiol., 15, 91-95. Gowaty, P. A., Plissner, J. H. & Williams, T. G. 1989. Behavioural correlates of uncertain parentage: mate guarding and nest guarding by eastern bluebirds, Sialia sialis. Anim. Behav. , 38, 272-284. Grizzle, J. E., Starmer, C. F. & Koch, G. G. 1969.Analysis of categorical data by linear models. Biometrics, 25, 45%504. Holland, B. S. & Copenhaver, M. D. 1987. An improved sequentially rejective Bonferoni test procedure. Biometrics, 43, 417M23.

Gowaty & Bridges: Bluebird breeding strategies May, T. P., Wright, J. E. & Stoneking, M. 1979. Joint segregation of biochemical loci in Salmonidae: results from experiments with Salvelinus and review of the literature on other species. J. Fish Res. Bd Can., 36, 1114-1126. Pinkowski, B. 1977. Breeding adaptations in the eastern bluebird. Condor, 79, 289-302. Plissner, J. H. & Gowaty, P. A. 1988. Evidence of reproductive error in the adoption of nestling eastern bluebirds. Auk, 105, 575 578. Price, D. K., Collier, G. E. & Thompson, C. F, 1989. Multiple parentage in broods of house wrens: genetic evidence. J. Heredity, 80, 1-5. Selander, R. K., Smith, M. H., Yang, S. Y., Johnson, W. E. & Gentry, J. B. 1971. Biochemical polymorphism and systematics in the genus Peromyscus, I. Variation in the old-field mouse (Peromyscus polionotus). Stud. Genet., Univ. Texas Publ., 7103, 49-90. Semel, B. & Sherman, P. W. 1986. Dynamics of nest parasitism in wood ducks. Auk, 103, 813-816. Semel, B. & Sherman, P. W. 1988. Effects of brood parasitism and nest box placement on wood duck breeding ecology. Condor, 90, 920-930. Sherman, P. W. & Morton, M. L. 1988. Extra-pair fertilizations in mountain white-crowned sparrows. Behav. EcoL Sociobiol, 22, 413-420. Trivers, R. L. 1972. Parental investment and sexual

675

selection. In: Sexual Selection and the Descent of Man (Ed. by B. G. Campbell), pp. 136-179. Chicago: Aldine. Verner, J. & Willson, M. F. 1969. Mating systems, sexual dimorphism and the role of male North American passerine birds in the nesting cycle. Ornithol. Monogr., 9, 1-76. Westneat, D. F. 1987a. Extra-pair copulations in a predominantly monogamous bird: observations of behavionr. Anim. Behav., 35, 865-876. Westneat, D. F. 1987b. Extra-pair fertilizations in a predominantly monogamous bird: genetic evidence. Anim. Behav., 35, 877-886. Westneat, D. F., Frederick, P. C. & Wiley, R. H. 1987, The use of genetic markers to estimate the frequency of successful alternative reproductive tactics. Behav. Ecol. Sociobiol., 21, 35-45. Wickler, W. & Seibt, U. 1982. Monogamy: an ambiguous concept. In: Mate Choice (Ed. by P. Bateson), pp. 3352. Cambridge: Cambridge University Press. Wrege, P. H. & Emlen, S. T. 1987. Biochemical determination of parental uncertainty in white-fronted beeeaters. Behav. Ecol. Soeiobiol., 20, 153-160. Yom-Tov, Y. 1980. Intraspecific nest parasitism in birds. Biol. Rev., 55, 93-108. Zelany, L. 1976. The Bluebird. Bloomington: Indiana University Press.