An experimental analysis of small clutch size in tropical House Wrens

An experimental analysis of small clutch size in tropical House Wrens

Bruce E. Young

INTRODUCTION

Variation in clutch size in birds has stimulated a great deal of research, both descriptive and experimental (reviewed in Martin 1987, Nur 1990, VanderWerf 1992).

Most attention has focused on understanding variation in clutch size within populations, leaving the question of geographic variation in clutch size mostly to a series of correlational studies that relate environmental variables to clutch size (Lack 1968, Ricklefs 1980, Moller 1984, Koenig 1986, Young 1994a). The observation that tropical species typically lay smaller clutches than temperate relatives (Hesse et al. 1937, Moreau 1944, Lack 1947, Cody 1966, Klomp 1970, Murphy 1989, Kulesza 1990, Young 1994a) still requires explanation. Experimental studies based on manipulation of brood size have rarely been performed on tropical birds (VanderWerf 1992), so the factors that limit clutch size in tropical birds and their relationship to geographical variation in clutch size are unknown.

Breeding birds in any geographic area are limited in offspring production because (1) their future fecundity may be reduced by large investment in current reproduction; (2) they are unable or unwilling to work hard enough to compensate for larger clutches, thus causing offspring survivorship to decline; or (3) a combination of these two factors (Lack 1968, Charnov and Krebs 1974). These limits are standard life history trade-offs and may be caused by food shortages or predation (Godfray et al. 1991, Stearns 1992). Future fecundity, as measured by adult survivorship or condition, is influenced by two factors that may increase with clutch size: food shortages and predation of adults. Parents that lay larger clutches and raise larger broods must forage more intensively to find food for themselves and their offspring (Zach and Smith 1981, Robinson 1986). Intensive foraging can lead to greater exposure to predators, and food shortages can cause deterioration in physiological condition, thus decreasing performance in, or lowering the likelihood of surviving to, future reproductive periods (Charnov and Krebs 1974, Nur 1984, Lima 1987, Tinbergen 1987). These effects of reproduction on future events are often called “costs of reproduction,” but I will use the term “future fecundity” to distinguish future-fecundity costs from immediate costs involving offspring survivorship. The future-fecundity hypothesis is supported if parents that raise extra young are less likely to survive to future breeding seasons, or if they produce smaller clutches in subsequent reproductive bouts, either within or between years.

Food shortages and predation can also cause offspring survivorship to decline and thus limit clutch size. Underfed nestlings can starve in the nest or fledge too small to survive after fledging (Lack 1947, 1968, Perrins 1977). Nestlings in larger broods might produce more odor, beg more loudly, or require more feeding visits, all of which increase the chances of revealing the nest location to predators (Skutch 1949, Ricklefs 1977, Slagsvold 1982, 1984, Lundberg 1985, Lima 1987, Redondo and Castro 1992). Observations that survivorship or growth of either nestlings or fledglings declines with increasing clutch size, or that the rate of nest predation increases with increasing clutch size, lend support to the offspring-survival hypothesis.

Many studies have examined these trade-offs in species that breed in temperate habitats (reviewed in Martin 1987, Nur 1990, VanderWerf 1992). Whether increased brood size decreases the future fecundity of parents or the production of offspring is variable, but a central finding is that the observed modal or mean clutch size is often somewhat smaller than that which produces the most offspring (Perrins 1965, Nur 1984, VanderWerf 1992). Tropical birds may exhibit the same pattern. Although mean clutch size is lower than in temperate birds, the observed mean clutch size may still be slightly smaller than the most productive clutch size. Demonstration of this pattern in tropical birds would indicate that the same ecological forces are at play in both temperate and tropical regions, but that the level of limitation on clutch size differs. The few brood manipulation studies that have been conducted on noncolonial, altricial tropical species differed in whether or not parents were successful in raising enlarged broods (Mader 1982, Beissinger 1990), so more study is needed to establish similarities or differences between temperate and tropical birds in their responses to brood manipulation.

To explain why actual clutch size is smaller than that shown by brood manipulation studies to be the most productive size in Great Tits (Parus major), Boyce and Perrins (1987) argued that individuals increase their geometric mean fitness by reducing annual variation in clutch size. Environmental conditions fluctuated at their study site in England; during bad years, individuals laying large clutches produced fewer young than individuals laying smaller clutches. This “bad-years effect” model applies when a small clutch strategy reduces variance in reproductive success such that it produces more offspring in the long term than do larger clutches (Gillespie 1977, Boyce and Perrins 1987). This same process may apply to tropical systems that also experience significant year-to-year variation in environmental conditions (e.g., Janzen 1974, Connell 1978, Foster 1982, Levings and Windsor 1982, Smith 1982, Karr and Freemark 1983, Grant and Grant 1989). If environmental variation is important, then observed modal clutch size should be smaller than the most productive clutch size in all years except those unfavorable for reproduction, and geometric mean fitness of individuals laying the observed clutch size should be higher than that of individuals laying any other clutch size. Hence, one factor, environmental variation, influencing clutch size in temperate populations may also be important in tropical populations.

In this study, I examine factors that limit clutch size in a population of House Wrens, Troglodytes aedon, in Costa Rica. Tropical House Wrens typically lay 3-4-egg clutches, whereas temperate House Wrens lay 6-egg clutches, on average (Young 1994a). Experimental studies of House Wrens at three north temperate sites (Finke et al. 1987 and Harper et al. 1992 in Illinois; Robinson and Rotenberry 1991 in Ohio; and Arnold 1993 in Saskatchewan) allow for comparisons among factors that are important in tropical and temperate populations of this species. Specifically, I test the predictions of the future-fecundity and offspring-survival hypotheses, and the relative influence of food shortage and nest predation in each of these trade-offs.

Simply showing the existence of trade-offs is one step, but assessing how each trade-off affects fitness is clearly more important. To examine fitness trade-offs, I built population models of House Wrens following strategies of laying larger than average, average, and smaller than average clutches. These models show how future-fecundity and offspring-survival trade-offs can affect fitness, and how these factors can work in concert to influence clutch size. By adding simulations of the effects of annual environmental variation on reproductive success and survivorship (based on variation observed during the 3-yr course of the study), I examine the importance of the bad-years effect on limiting clutch size in my study population. Finally, I discuss an alternative hypothesis, investment in offspring quality, to explain smaller clutches in tropical birds.

METHODS

Study organism and site

House Wrens are 10-13-g insectivorous passerines that nest in cavities throughout their breeding range from Canada to Tierra del Fuego. High latitude populations are migratory and lay clutches of 5-10 eggs, whereas tropical populations are sedentary and lay clutches of 3-4 or, occasionally, 5 eggs (Gross 1948, Skutch 1953, Young 1994a). In tropical regions, both members of a pair defend an all-purpose territory (Skutch 1953, Alvarez-Lopez et al. 1984, Freed 1987b, Winnett-Murray 1986). Incubation lasts 13-14 d and the nestling stage lasts 18 d. Fledglings remain in their natal territories in groups with their parents for 3-4 wk before dispersing. Reproduction begins for most wrens when they are 1-yr-old. At Monteverde, Costa Rica, pairs rear 0-3 broods/yr, although House Wrens can have up to 4 broods/yr in other parts of the country (Skutch 1953, Young 1994b). The breeding season in Monteverde lasts from late February or early March through early August (Winnett-Murray 1986), and clutch size shows little annual variation (1989-1991 range: 3.43-3.51 eggs).

Monteverde is in Puntarenas Province of Costa Rica (10 [degrees] 18[minutes] N, 84 [degrees] 45[minutes] W), at 1500-m elevation near the continental divide in the Cordillera de Tilaran. House Wrens occur in dairy pastures and gardens that are surrounded by either 15-30-yr-old second growth or old growth cloud forest, classified as Lower Montane Wet by Holdridge (1967). A pronounced wet season lasts from mid-May to mid-December, but the so-called dry season is characterized by northeast trade winds that blow in Caribbean moisture in the form of mist (Young 1994b). Thus, pasture grasses remain green year-round and few trees are deciduous.

In September 1988, I set up wooden nest boxes (inside dimensions 10 x 10 x 12.5 cm, with 2.8 cm diameter holes) in 50 House Wren territories. To avoid artificially increasing the density of breeding wrens and the problems in interpretation this brings about (Hagvar et al. 1989, Alatalo et al. 1990, Bock et al. 1992), I first mapped the territories and then put up one box per territory. During the three years of the study, I increased the sample to 60 territories by expanding the study area slightly. From time to time, I changed the location of boxes within territories or put up two boxes in a territory in attempts to entice birds to nest in boxes. These modifications never led to changes in nesting density. All boxes were initially attached to remnant trees in pastures, but 15 were attached to 1.9 cm diameter steel posts in 1990 and 10 additional boxes were moved to post mounts in 1991. At least some predators were able to climb the posts: (1) predation occurred in several post-mounted nests and (2) on two occasions, I found mouse opossums (Marmosa mexicana), common nest predators, asleep in the boxes. I treated nests not built in boxes (hereafter “natural” nests) as equivalent to nests built in boxes in all aspects of the study.

Throughout, I refer to House Wrens living within the study area as the Monteverde population. I use this term for convenience and do not wish to imply that birds in the study area lived in a closed population. The boundary of the study area cut through a continuous population of House Wrens that extended for many kilometres to the north and west.

Nest checks

I checked all boxes weekly for nesting activity during the 1989-1991 breeding seasons. As nests neared completion, I checked them twice weekly to estimate clutch initiation date, and daily when hatching was due. All eggs from a nest hatched within a 24-h period, so I considered nest mates to be the same age. Hatching began on day 0, and I subsequently visited nests on days 3, 6, 9, and 12 in all years, and also on day 15 in 1990 and 1991, to monitor progress and measure growth. After fledging, or 3-4 wk without progress from nests in which no eggs were laid, I removed all contents of the boxes.

Brood manipulations

Both trade-off hypotheses make predictions about consequences, to offspring or parents, of broods that are larger than normal, normal, or smaller than normal. I performed brood manipulations by moving two 1-2-d-old nestlings between like-aged nests. I left other broods undisturbed to serve as controls. The goal of the moves was to create (1) broods with two nestlings fewer than the number of eggs laid (reduced), (2) broods the same size as the number of eggs laid (control), and (3) broods with two more than the number of eggs laid (enlarged). In the cases in which not all eggs hatched, I added or removed additional nestlings to make the final brood size exactly fit one of the brood manipulation categories. I assigned nests to treatment groups as randomly as possible. However, I could move nestlings only when at least two nests hatched on the same day. In these cases, I used the flip of a coin to assign treatments. When hatching occurred in only a single nest on a given day, I assigned the nest to either the control or reduced treatment. Opportunities for brood additions dwindled as nesting activity declined at the end of the breeding season, so I excluded data from nests that hatched later than the last experimental nest. After this adjustment, the mean hatching date of the three treatment groups did not differ (Kruskal-Wallis tests: 1989, H = 2.717, P = 0.26; 1990, H = 0.666, P = 0.26; 1991, H = 3.117, P = 0.21). For pairs that produced more than one brood in a season, I assigned treatments to each brood independently.

Due to a shortage of synchronously hatched broods, I could not perform reciprocal transfers of nestlings to control for the effects of parental recognition of young and the possible differential provisioning of foster vs. nonfoster young. However, body mass of foster and nonfoster nest mates did not differ at 12 d of age (Wilcoxon signed-ranks tests, P [greater than or equal to] 0.14 for each year), indicating that parents of enlarged broods did not appear to distinguish their biological offspring from foster chicks.

For analyses of nestling growth and survivorship, I treated all nesting attempts as independent events and assigned treatments randomly. This assumption was necessary to include nesting attempts throughout the breeding season. In fact, the brood treatment in the first nesting attempt had no effect on mass of 12-d-old nestlings in the second attempt by the same parents (Kruskal-Wallis test on changes in mass of chicks in reduced, control, and enlarged treatment groups, N = 46 pairs, H = 0.141, df = 2, P = 0.93). Also, repeatability in growth measures of successive broods raised by the same parents was very low (range -0.096-0.189; calculations based on Lessells and Boag 1987). Furthermore, neither clutch size nor reproductive success varied seasonally (Young 1994a), so I assumed that nesting attempts throughout the breeding season are comparable.

Future fecundity

The future-fecundity hypothesis assumes that the physiological condition of parent birds is negatively correlated, and that exposure to predation is positively correlated, with the number of young they raise. Parents trade off the number of young produced in one period of reproduction with survivorship to, and production of young in, future reproductive periods. To monitor the fates of individual wrens, I marked each with a unique combination of color bands.

I first tested the assumption that parents work in proportion to the number of young they raise. To measure work, I counted nest visits by parents during a single 60-min observation period at each nest. To control for the effect of nestling age on visit rate, I only observed nests with nestlings 6-14-d-old; at these ages, visit rate is unrelated to nestling age (multiple regression with chick age and brood size as factors: chick age [F.sub.1,101] = 2.517; P = 0.12).

To measure adult survivorship, I censused adults on the study area during March and April in 1990 and 1991, and during April in 1992. For each adult that participated in a nesting attempt one year, I scored whether or not it survived to the following year. I used playbacks of House Wren songs recorded in Monteverde to draw all male and most female wrens close enough to read band combinations. I also trapped incubating females to read their band combinations.

During the course of the study, I never observed adult territory owners move farther than to an adjacent territory, so I assumed all disappearances were due to mortality. I used simple enumeration to measure survivorship (see Pollock et al. 1990) because (1) the adult population was operationally closed, in that adults never dispersed far from their breeding territories, and (2) I never missed a marked individual during one census and then saw it on a subsequent census. For each parent, I tallied the total number of nestlings it raised to at least 9-d of age. I arbitrarily chose 9 d as a point at which parents have made a significant investment in their young. I assessed the effect on adult survivorship of the total number of brood raised in two ways. First, I fit cubic splines to visualize the shape of the survivorship curve for parent wrens as a function of the total number of young raised. The cubic spline is a non-parametric curve-fitting routine that approximates a weighted moving regression (Schluter 1988, Linden et al. 1992). Second, for survivorship curves that were either sigmoid in shape or formed part of a sigmoid curve, I used logistic regression, with total number of young raised as the independent variable and survivorship as the dependent variable (Nur 1990, Trexler and Travis 1993), and employed a maximum likelihood routine for parameter estimation (Wilkinson 1989). To assess the importance of brood additions, I added a covariate to the logistic regression model coding for whether or not the individual raised an enlarged brood.

To examine the effects of brood treatment on clutch size in future attempts within years, I classified females based on whether they raised reduced, control, or enlarged broods in their first attempt of the year, and measured the change in clutch size from the first to the second clutch. The future-fecundity hypothesis makes an ordered prediction about the outcome: clutch size should be increasingly negatively affected as the number of nestlings raised in the first brood increases. I therefore analyzed the data with isotonic regression, which is an appropriate method of testing experiments with ordered predictions (Gaines and Rice 1990). Because sample sizes were too small to examine within-year differences in fecundity patterns, I pooled the three years for analysis. For between-year effects on clutch size, I included only those females that were present at the beginning of two consecutive nesting seasons. I compared the change in size of the first clutch in year t to the first clutch in year t + 1 in individuals that raised an enlarged brood at any time during year t vs. individuals that raised only control and/or reduced broods in year t.

Offspring survivorship

To test the offspring-survivorship trade-off, I measured the effects of food limitation and predation on the growth and survival of juvenile House Wrens. During the nestling stage, I was able to separate food limitation and predation as factors affecting survivorship. For the fledgling stage, I could only measure their combined effects on survivorship.

Nestlings. – To examine survivorship of nestlings, I calculated the fraction of nestlings in a nest on day 3 (after all brood manipulations were complete) that survived to day 15, the last nest check before fledging. I assumed that all nestlings alive at day 15 fledged successfully, and compared nestling survivorship in the three brood manipulation groups. To focus this analysis on mortality caused by starvation, I eliminated all nests lost to predation (defined as entire broods disappearing between successive nest visits) and broods lost to abandonment (entire broods that were healthy in one visit, then all dead in the nest at the next visit, with the concurrent disappearance of one of the parents).

I measured growth by weighing nestlings on days 9, 12, and 15 with a pesola scale, accurate to [+ or -] 0.1 g. On day 15, I also measured tarsus length using dial calipers accurate to [+ or -] 0.1 mm and wing chord with a ruler accurate to [+ or -] 1 mm. To control for time of day, I made all measurements between 0500 and 0600 local time. For parents and offspring to be affected by a brood size change, the manipulation must be maintained for a good portion of the nestling stage. I therefore excluded from growth analyses all nests in which the full complement of chicks did not survive at least through day 9. Because growth measures varied from year to year, I analyzed data for each year separately. The data were not consistently normally distributed (based on Lilliefors’ test, Wilkinson 1989) and no single transformation significantly improved normality, so I analyzed the data with Kruskal-Wallis tests. In 1989, I did not take any of the day 15 measurements. The sample of reduced nests in 1989 was too small for inclusion in the analysis; thus, I performed Mann-Whitney tests on day 9 and day 12 mean mass per nest data for the control and enlarged broods only.

I monitored predation of nestlings by visiting nests every 3 d. In Monteverde, predation was an all-or-nothing event: I had no evidence that a predator ever took less than an entire brood, as sometimes happens in larger species (Schaub et al. 1992). Thus, to compare predation rates in the three brood manipulation groups, I compared Mayfield (1975) daily survivorship estimates for broods in each treatment, using the chi-square statistic suggested by Sauer and Williams (1989). In the comparison, I eliminated all nests that failed because of abandonment or infanticide (see Freed 1986). Because the nest predation hypothesis makes no prediction about predation during the egg stage, I examined survivorship of nests in the nestling stage only, between nest visits on days 3 (after which brood manipulations were complete) and 12 (1989) or 15 (1990 and 1991). Natural, post-mounted, and tree-mounted nests did not suffer different rates of predation ([[Chi].sup.2] = 0.35, df = 2, P = 0.84), so I pooled nests of all types in the analysis.

Fledglings. – To measure survivorship of fledglings, I revisited a subset of the territories (in 1990 and 1991 only) 10-14 d after fledging to score which fledglings were still alive. If I could not account for all of the fledglings in the first visit, I revisited the territory up to two more times in the next 2-4 d to search for missing fledglings. I assumed that fledglings not seen on any of the visits were dead. To examine the effect of brood size on survivorship during this period, I compared the fraction of fledglings surviving among brood manipulation groups.

Some factors that influence postfledging survival may not be independent of prefledging events. For example, size at fledging may influence fledgling survivorship (Hochachka and Smith 1991, Magrath 1991, Linden et al. 1992). To understand the importance of fledgling size, I examined the effects of the various growth measures on postfledging survivorship, using both cubic spline and logistic regression techniques. In the logistic regression models, I added a covariate for fledgling group size to control for the independent effects of this factor on survivorship.

Population and simulation models

To examine the combined effects of trade-offs for future fecundity and offspring survivorship on fitness and to compare life history strategies, I constructed a simple model of House Wren population dynamics (Nur 1984, Caswell 1989). I considered three hypothetical populations of House Wrens, in which females had fixed strategies for laying 1-2 eggs, 3-4 eggs, and 56 eggs. I used demographic data, gathered from females forced to raise broods of these sizes during this study, in a general female-only model of House Wren population dynamics. To compare fitnesses of the three strategies, I calculated the population growth rate ([Lambda], such that [N.sub.t+1] = [Lambda][N.sub.t], where [N.sub.t] is the population size in year t) of each strategy (Schaffer 1974, Nur 1984, Caswell 1989).

The model is among the simplest possible for an iteroparous species, taking the form

[Lambda] = S + [G.sub.1][G.sub.2][G.sub.3]F, (1)

where S is the adult survival rate, [G.sub.1] is the probability of a nestling surviving to fledge (taking into account starvation, natural abandonment, infanticide, and predation), [G.sub.2] is the probability of a fledged juvenile surviving to independence (postfledging survivorship), [G.sub.3] is the probability of surviving from independence to age 1 yr, at which time breeding begins, and F is the fecundity parameter. The model ignores the cost of laying eggs, which is not well understood but is probably not great in passerinc birds (Walsberg 1983).

A single-stage model is appropriate in the case of tropical House Wrens because yearling females are as capable as older females in raising young, in terms of clutch size, number of nesting attempts, egg hatching success, and nestling growth (Freed 1987b, Young 1993). This is in contrast to many temperate species (Jarvinen 1991). To use the nestling survivorship data, the period over which [G.sub.1] was measured began on day 3 of the nestling stage, just after the brood manipulations took place. Consequently, I calculated F for each brood size group as the product of the average number of breeding attempts in which eggs survive the incubation period per season per female (accounting for nesting attempts that fail during the incubation stage); average clutch size for the group (1.5, 3.5, 5.5 eggs); 0.5 (because only half of the eggs will produce females); and the hatching success rate of eggs not killed by abandonment, infanticide, or predation. [G.sub.3] was based on sightings of yearlings (banded as nestlings) on or within a 0.75-km radius of the study site during intensive annual searches. Unfortunately, females are more likely to disperse out of the study area than are males at Monteverde (B. E. Young, unpublished data), and thus are less likely to be detected by my censusing. To counteract the dispersal bias, I assumed that females survived at the same rate as males.

As one might expect, values of the model parameters varied from year to year. Some of the values changed dramatically. Nestling, juvenile, and adult survivorship measures showed much annual variation, although other parameters such as the number of nesting attempts per female and hatching success did not vary much over the course of the study. Simply averaging all of the parameters over all of the years that they were calculated would mask the true dynamics of the system. Instead, I calculated parameter sets for typically “good” and typically “bad” years and ran separate models.

I calculated [Lambda] for each clutch size group based on demographic data for both good and bad years. I calculated an approximate variance for [Lambda] as

[Mathematical Expression Omitted],

for all demographic parameters x, and assumed no co-variance between parameters (Lande 1988). The partial derivative term is also called the sensitivity, and [Mathematical Expression Omitted] is the sampling variance of parameter x. The standard error of [Lambda] is simply the square root of [Mathematical Expression Omitted]. A crude confidence limit can be calculated as two standard errors about the mean (Lande 1988). I performed a sensitivity analysis to look at the effect of moderate changes (arbitrarily set at 20%) in each demographic parameter on [Lambda]. One at a time, I increased each parameter by 20% and recalculated [Lambda]. The percent change in [Lambda] is a measure of the relative importance of each parameter to the fitness of the clutch size strategy in question (Caswell 1989). Thus, one can compare the fitness consequences of trade-offs for adult survivorship (S), clutch size (F), or offspring survivorship ([G.sub.1], [G.sub.2], [G.sub.3]).

Because no single clutch size strategy emerged as optimal in both good and bad years, overall fitness depended on the frequency of good (or bad) reproductive years. To determine the range of frequencies over which a particular strategy is optimal, and the importance of bad years, I used a computer simulation of an initial population of arbitrary size, all members employing the same clutch size strategy for 100 yr. Each year was randomly chosen as good or bad depending on a preset frequency of good (as opposed to bad) years ranging from 0 to 1. I then calculated a geometric mean [Lambda] for the specific strategy and frequency of good years as

[Lambda] = 50 [square root of ([N.sub.t=100] / [N.sub.t=50])].

The outcome of the simulation is independent of the initial population size because the population will have reached a stable age distribution in [less than] 50 yr. At the stable age distribution, the population changes each year by its multiplication rate, [Lambda]. I ran the simulation 100 times for each clutch size strategy and frequency of good years, and calculated an overall mean [Lambda] for each combination of parameters.

RESULTS

Future fecundity

Both males and females increased their frequency of nest visits when they had larger broods [ILLUSTRATION FOR FIG. 1 OMITTED], indicating they had to work harder to raise extra young. During the course of the study, adult survivorship varied greatly between years (for females, 1989: 0.57, N = 44; 1990: 0.45, N = 58; 1991: 0.42, N = 64; for males, 1989: 0.66, N = 29; 1990: 0.40, N = 52; 1991: 0.41, N = 58). The cubic splines showed no clear pattern, however, of lower adult survivorship being associated with raising large numbers of young [ILLUSTRATION FOR FIG. 2 OMITTED]. Only one (1990 females) of the six curves shows a decline in survivorship for adults raising the largest broods. In the logistic regression analyses, none of the coefficients for total brood raised was significant (P [greater than or equal to] 0.15), and two of the four were positive, opposite the prediction of the future fecundity hypothesis. Whether or not individuals raised enlarged broods also had no influence on adult survivorship (no coefficients were significant).

The brood treatment in the first nesting attempt negatively affected clutch size in the second nesting attempt within a year, as predicted (isotonic regression, [Mathematical Expression Omitted], P = 0.050). Females that raised reduced or control first broods increased clutch size slightly in the second attempt ([Mathematical Expression Omitted], reduced: 0.10 [+ or -] 0.12 egg increase, N = 21; control: 0.10 [+ or -] 0.13, N = 30). Females that raised enlarged first broods had slightly smaller second clutches (-0.24 [+ or -] 0.14 egg decrease, N = 21), causing a net difference of about one-third of an egg.

Between-year fecundity also was influenced by brood manipulation ([Mathematical Expression Omitted], P = 0.020). Females that raised an enlarged brood one year had a first clutch the following year that was, on average, 0.15 [+ or -] 0.22 (N = 13) egg smaller than the first clutch the previous year. In contrast, females that raised only control or reduced broods increased their clutches by 0.32 [+ or -] 0.11 (N = 25) egg. In both the within-year and between-year comparisons, the change in clutch size is in the direction predicted by the future-fecundity hypothesis.

Offspring growth and survivorship

In 1989 and 1990, survivorship of nestlings was unrelated to brood size (Table 1) and low overall. In 1991, nestlings raised in the largest broods had the greatest likelihood of starving and the trend was marginally significant (P = 0.05, Table 1). In all years, nests with the most nestlings on day 3 were also the nests that produced the most fledglings [ILLUSTRATION FOR FIG. 3 OMITTED].

Growth measurements also varied between years and across brood groups (Table 2). In 1989 and 1990, nestlings in the three manipulation treatments did not differ in any measure. In 1991, however, nestlings in enlarged broods grew more slowly and fledged lighter than nestlings in other groups; tarsus and wing length were unrelated to treatment group (Table 2). Although 1991 was a difficult year for nestlings raised in enlarged broods, nestlings in control broods grew normally. Most growth measures for control nestlings were remarkably similar across the three years of the study.

Predation of nests in the nestling stage was generally uncommon, and no brood size was more susceptible to predation than any other ([[Chi].sup.2] = 0.957, df = 2, P = 0.62). The daily survivorship estimate was 0.9911 [+ or -] 0.0036 ([Mathematical Expression Omitted], N = 675 exposure days) for reduced broods, 0.9952 [+ or -] 0.0021 (N = 1042.5) for control broods, and 0.9944 [+ or -] 0.0032 (N = 532.5) for enlarged broods.

In 1990, brood size did not affect survivorship during the first 2 wk of the postfledging period [ILLUSTRATION FOR FIG. 4 OMITTED]. In 1991, however, survivorship was marginally lower for young fledged from enlarged broods than for young fledged from control and reduced broods (P = 0.08, [ILLUSTRATION FOR FIG. 4 OMITTED]). Although group size per se could have been responsible for the decline in fledgling survivorship, the lower survivorship more likely resulted from the low fledging mass of chicks in enlarged broods that year. Cubic splines [ILLUSTRATION FOR FIGS. 5 and 6 OMITTED] and logistic regressions (Table 3) show that the sizes attained by nestlings can affect their future survivorship such that larger nestlings are more likely to survive the postfledging period than smaller nestlings. Among the size characters, mass, especially at fledging (day 15), most affected survivorship through the postfledging period in both years studied. The growth terms were the only significant terms in the logistic regression models (Table 3). The term for fledgling group size never significantly improved the fit of any of the models (Table 3), so fledgling group size seemed to have no independent influence on postfledging survivorship.

Population model

Details on calculation of the model parameters, together with their estimates are presented in the Appendix. The fitnesses (i.e., [Lambda] values) of the different clutch size strategies varied markedly in both good years and bad (Table 4). During good years, females laying 5-6-egg clutches far outperformed females with smaller clutches. In bad years, however, females laying 3-4-egg clutches appeared to be the most productive, although confidence limits based on standard errors broadly overlap for the 3-4- and 5-6-egg strategies. Females laying 3-4-egg clutches, which account for 96% of Monteverde clutches (N = 322), had about the same fitness, independent of whether the year was good or bad. Females laying 1-2-egg clutches had the lowest fitness of the three groups in both good and bad years.

The sensitivity analysis showed that changes in adult survivorship, S, have more influence on fitness than identical changes in any other parameter (Table 4). Increases in the juvenile survivorship parameters, [G.sub.1], [G.sub.2], [G.sub.3], or the fecundity, F, had the largest relative influence on [Lambda] when the product of these four parameters was greatest. This was the case in the 5-6-egg strategy in both good and bad years and the 3-4-egg strategy in bad years. Here, a 20% increase in any parameter had about the same effect on [Lambda].

TABLE 1. Survivorship of tropical House Wren nestlings. H is the

Kruskal-Wallis test statistic, corrected for ties, for comparisons

among treatment groups of the fraction of those nestlings alive on

day 3 that survived to day 15.

Second, survivorship costs of reproduction are also difficult to detect in an expanding population, which can be caused by providing nest boxes to a population of secondary cavity nesters initially limited by nest site availability (Linden and Moller 1989). Although my study does rely on data from a population nesting in boxes, the outcome was probably unaffected by box use because demographic data show that the study population was far from expanding. For control (3-4-egg) clutches, [Lambda] [less than] 1 and the 95% confidence interval around [Lambda] barely included values greater than unity (Table 4), whereas an expanding population would have [Lambda] [greater than] 1. Setting up the boxes at a density matching the preexisting population density of territorial wrens may also have succeeded in preventing the population from artificially expanding. I therefore conclude that raising two extra young in a breeding attempt does not significantly increase the risk of mortality to adult wrens.

Although the effects of brood manipulations on adult survivorship have not been examined in temperate House Wrens, some studies on other temperate species have shown that adult survivorship is inversely related to the number of young raised, whereas other studies have not (Nur 1990). In what situations should survivorship costs be most easily detected? Reproduction is only one of several major energy-demanding activities that birds undergo in their annual cycle (Walsberg 1983). Birds must also molt, and many migrate as well. When molting and migrating occur soon after breeding, it is reasonable to expect that individuals stressed from reproduction may not have the necessary reserves to survive the subsequent energetically demanding activities. Birds that do not undertake long-distance migrations and that do not need to molt immediately after breeding, such as those living in the tropics, may have more time to replenish reserves. Thus, the negative effects of increased reproductive effort on survivorship may be more moderate in resident tropical birds than in migratory temperate birds.

[TABULAR DATA FOR TABLE 4 OMITTED]

How should a tropical bird be most affected by caring for enlarged broods? Many tropical birds attempt to raise more than one brood in a season, as do some temperate birds. The energetic demands of raising a second brood come on the heels of the demands of the first clutch. An individual stressed by rearing the first brood may not be able to invest heavily in the second nesting attempt. Clutch size may be lower or parents may be less able to raise extra young than in the first attempt. In this study, clutch sizes were indeed slightly smaller in second clutches of females that cared for extra young in the first attempt.

Surprisingly, raising enlarged broods one year caused females to be more likely to lay smaller clutches the next year. Given the length of time between breeding seasons, it is hard to understand why females would not recover completely after each breeding season, especially considering how minor the within-season effects were. Perhaps the trauma of making so many feeding visits and caring for so many fledglings caused females to adjust by laying fewer eggs in all future clutches, whether within the same breeding season or later. Studies of some temperate species have shown this fecundity trade-off (Roskaft 1985, Gustafsson and Sutherland 1988, Orell 1990) but, prior to this study, no study has documented this cost in a tropical species.

In Illinois House Wrens, manipulating the first brood of the season did not affect the likelihood of a second clutch attempt, time between clutches, size of the second clutch, or mass of females (Finke et al. 1987). Results were similar in Ohio, except for the observation that chicks from second broods of parents that raised enlarged first broods fledged lighter than chicks from second broods of parents that raised control or reduced first broods (Robinson and Rotenberry 1991). Thus, brood enlargements in temperate House Wrens, just as in tropical House Wrens, had only slight consequences on future fecundity.

Of the two mechanisms that can influence future fecundity, food limitation and predation, only food limitation may have been operating in the Monteverde study population. The decline in clutch size after raising enlarged broods suggests that females were not completely replenishing nutrient reserves after rearing a brood. Because adult survivorship was unrelated to brood manipulation treatment, predation on adults evidently did not increase with the number of young raised. Aerial predators, such as Barred Forest-Falcons (Micrastur ruficollis), occurred in the study area and were observed to attack House Wrens, but apparently did not pose an increased threat to parent wrens that made extra flights to their nests to raise enlarged broods.

Offspring survivorship

Food was not a consistent factor causing an upper limit to clutch size for tropical House Wrens. In only one year of three (1991) did parents fail to raise broods enlarged by 50-67% at a rate that allowed normal growth (Tables 1 and 2). Yet, even in the one year of apparent food shortage, chicks raised in control broods grew at the same rate as in previous years (Table 2). I could detect the existence of a food shortage only by the failure of chicks in enlarged broods to grow as quickly as in other years. Nevertheless, more chicks fledged from enlarged broods than control broods, even during the year when food was less plentiful [ILLUSTRATION FOR FIG. 3 OMITTED]. This is not surprising, considering that starvation was as rare in this study (Table 1) as it has been in other studies of tropical birds (Ricklefs 1969). The moderate level of the food shortage in 1991 caused most nestlings in enlarged broods to grow a little more slowly and to survive less well, but it did not prevent the largest broods from being the most productive in terms of numbers of chicks fledged.

In studies of temperate House Wrens, enlarged broods also produced the greatest number of fledglings (Finke et al. 1987, Robinson and Rotenberry 1991, Harper et al. 1992, Arnold 1993). Chicks from enlarged broods, however, fledged up to 7% lighter than chicks from control or reduced broods. Food may be a more important limit to raising unusually large broods in temperate populations, in which growth was retarded in seven of eight years studied: four of five years in Illinois (Finke et al. 1987, Harper et al. 1992); two of two in Ohio (Robinson and Rotenberry 1991); one of one in Saskatchewan (Arnold 1993), than it is in tropical populations (growth retarded in one of three years; this study).

Nest predation was infrequent and unrelated to brood size in the Monteverde population. Skutch (1949, 1985) predicted that large broods would produce more odor and be noisier than small broods and, therefore, would be more likely to attract predators. Except for the effects of smell, the nest predation hypothesis would only apply to birds persecuted primarily by diurnal predators, since there is no activity at nests during the night that might attract a nocturnal predator. In Monteverde, however, most nest predators that I could detect (mouse opossums) hunted at night and rested by day, sometimes even in nest boxes. I once witnessed a diurnal predator, a coati (Nassua narica), take a brood, but it did so at dawn before parents had begun to forage. The only report of a diurnal attack, that I know of, was by an unidentified snake entering a nest box at a site 4 km from Monteverde. These predators, with the possible exception of the coati, do not hunt visually and thus would not be attracted by activity at a nest.

Survivorship during the postfledging period was relatively low and influenced by fledging mass. Typically, only 65-80% of individuals alive on day 15 survived through their first 2 wk out of the nest [ILLUSTRATION FOR FIG. 4 OMITTED]. By comparison, survivorship during the nestling stage between days 3 and 15 averaged 75-90% (cf. [G.sub.1] values in Appendix). Nestling mass, especially at day 15, was the most important factor influencing postfledging survivorship. Fledging mass has repeatedly been shown to correlate with survivorship in temperate species (Hochachka and Smith 1991, Magrath 1991, Linden et al. 1992), and it is not surprising that a tropical species shows the same pattern.

Fledgling group size was not an important factor influencing survivorship during the postfledging period. Several authors have suggested that clutch size may be limited by the ability of birds to protect their young after leaving the nest (Safriel 1975, Lessells 1986, Karr et al. 1990), but fledging group size was not a factor in postfledging survivorship in this study. Perhaps the secretive behavior of House Wrens sufficiently disguises the number of fledglings in an area, such that predators are equally likely to be attracted to areas with few or many fledglings.

In this study, adults with enlarged brood sizes traded off both future fecundity (in terms of clutch size) and offspring survival. The offspring-survival trade-off seemed to have a greater influence on fitness than did future fecundity. During a lean year, young in larger than normal broods survived less well than young in control or reduced broods. Although enlarged broods still fledged the most young in 1991, a bad year [ILLUSTRATION FOR FIG. 3 OMITTED], these young fledged lighter and had lower subsequent survival, such that 3-4-nestling broods actually contributed more to the next generation (Table 4). In contrast, the one-third of an egg future-fecundity cost of raising an enlarged brood (which applied in both good and bad years) was far outweighed by the advantages of rearing more young in good years (Table 4). In bad years, the two factors worked in concert to depress fitness of parents raising enlarged broods.

Brood manipulation studies, in general, tend to show that enlarged broods produce more fledglings than control broods, especially in altricial species (VanderWerf 1992). Yet, in half of the studies on altricial species that were reviewed by VanderWerf (1992), fledglings from enlarged broods were smaller than fledglings from control broods. As studies increased in length, they were more likely to show that food can limit clutch size (VanderWerf 1992). These patterns, drawn largely from a sample of temperate studies, are remarkably similar to the results of this study on a tropical species. Enlarged broods produced more fledgling House Wrens than control broods, and food limitation manifested itself only in the third year of the study. These simulations suggest that tropical House Wrens are affected by the same processes that control clutch size in many temperate species. These processes (trade-offs for future fecundity and offspring survivorship) may exert a different level of control on tropical birds and, thus, select for smaller clutch sizes, but are not fundamentally different than those acting in the temperate zone. The observation that, in many years in both temperate and tropical systems, larger than modal clutch sizes produce the most surviving offspring still needs clarification.

Environmental variability and the bad-years effect

One explanation may revolve around year-to-year variability in environmental conditions. The simulation model showed that if the frequency of good reproductive years were low in the Monteverde population of House Wrens, a strategy of laying 3-4 eggs would yield the highest fitness [ILLUSTRATION FOR FIG. 7 OMITTED]. In fact, the reproductive success of individuals rearing control (3-4-nestling) broods was remarkably constant over the three years of the study, in terms of nestling survivorship (Table 1), nestling growth (Table 2), or fledgling survivorship [ILLUSTRATION FOR FIG. 4 OMITTED]. Combining offspring and adult survivorship, the population model showed very similar fitnesses of laying 3-4-egg clutches in both good and bad years (Table 4). Variance in reproductive success in good and bad years was much higher, in all of these measures, for parents raising enlarged broods. When bad years were frequent enough, this strategy became suboptimal.

The simulation model exercise demonstrates that the optimal clutch size strategy depends on the frequency and intensity of environmental variation. Parameters used in the model reflect the actual variation in these parameters observed over the 3-yr course of the study. A longer term study might reveal greater variation, which can be caused, for example, by El Nino weather patterns (Freed 1987a, Grant and Grant 1989). Thus, the parameters used surely do not represent the complete range of effects that environmental variation can have on clutch size strategies in Monteverde House Wrens. However, the simulation model does show that the observed clutch size may be an adaptation to a variable environment. This is the same pattern that occurs in Great Tits in Britain (Boyce and Perrins 1987), and may generally explain clutch sizes in tropical and temperate birds.

Offspring quality

Without data on longer term variation in the environment, this study cannot demonstrate unequivocally that clutch size in Monteverde House Wrens is, in fact, an adaptation to environmental variation. Indeed, the simulation model showed that a strategy of laying 5-6-egg clutches would be more profitable than the other strategies when the frequency of good reproductive years is high [ILLUSTRATION FOR FIG. 7 OMITTED]. This strategy is within the realm of physiological possibility. A small percentage of all clutches laid in Monteverde had 5 eggs (Appendix), clutches of up to 10 eggs have been reported for House Wrens elsewhere (Gross 1948), and egg removal experiments show that temperate House Wrens can produce many more eggs than they typically lay in a clutch (Kennedy and Power 1990). As an alternative to explain small clutches in tropical birds, offspring quality, as an extension to the offspring-survivorship trade-off, may be important.

Many tropical birds live in resident populations in habitats with less annual variation and more predictability of food resources than temperate habitats. Annual survivorship, which perhaps is no different from that of temperate birds (Karr et al. 1990), may be less influenced by random events such as bad weather during migration or during the winter. Instead, survivorship may depend on holding on to a territory in the face of competition from conspecifics. Thus, strong competitors would be most likely to maintain their territories and have the opportunity to reproduce. Tropical birds may, therefore, raise fewer young in order to produce higher quality offspring that would be more likely to win contests for territories (Fretwell 1969, Garnett 1981, Smith et al. 1989, Smith 1994). This offspring-quality hypothesis predicts that juveniles from smaller broods survive better and are more likely to gain territory ownership by winning social contests once they disperse from their natal territories. Offspring quality would also be important to temperate House Wrens, but not to the degree that it is in tropical wrens. Temperate wrens, whose offspring are more subject to uncontrollable sources of mortality (due to unpredictable weather during migration or on the wintering grounds), may be better off producing more offspring to increase the chance that any survive to breed. The observation that food is a more important limit on reproduction in temperate than tropical House Wrens further supports this notion that temperate wrens produce as many offspring as resources allow.

The offspring-quality trade-off is the same as the offspring-survivorship trade-off, with the addition of an extra factor for the ability of offspring to obtain breeding territories. To accommodate the offspring-quality hypothesis in the population model, an additional variable T, the probability of a young bird getting a breeding territory, is needed in the fecundity term of Eq. 1. The offspring-quality hypothesis predicts that T will be higher for young from smaller broods, even if [G.sub.2] and [G.sub.3] do not differ between offspring of small and large broods.

The data gathered in this study are insufficient to test the offspring-quality hypothesis, but a comparison between temperate and tropical wrens of time spent in each stage of the nesting cycle and fledgling mass suggests that parental investment in offspring is higher in tropical wrens (Table 5). Tropical wrens devote more time to incubation, nestling, and postfledging stages of parental care than do temperate species. Thus, the amount of time spent being a parent is much greater in tropical House Wrens. Whether or not this investment translates into higher quality offspring and can influence clutch size is unknown and awaits further experimentation.

[TABULAR DATA FOR TABLE 5 OMITTED]

ACKNOWLEDGMENTS

For help with field work, I am indebted to C. Darmstadt, III, B. Guindon, and P. Rose. I thank the many landowners in Monteverde who kindly allowed me to work on their property. N. Nur helped to analyze the costs of reproduction data, and J. Karr, G. Orians, S. Rohwer, D. Schemsky, E. VanderWerf, T. Wootton, and two anonymous reviewers provided many helpful comments on earlier drafts of this manuscript. For financial assistance, I am grateful to have received fellowships from the National Science Foundation and the Fulbright Program for Study Abroad, and grants from the Jesse Smith Noyes Foundation and Pew Charitable Trusts (administered by the Organization for Tropical Studies), the Chapman Fund of the American Museum of Natural History, and the Burke Museum Endowment for Ornithology.

LITERATURE CITED

Alatalo, R., A. Carlson, and A. Lundberg. 1990. Pages 323-330 in J. Blondel, A. Gosler, J. B. Lebreton, and R. McCleery, editors. Population biology of passerine birds, an integrated approach. Springer-Verlag, Berlin, Germany.

Alvarez-Lopez, H., M. D. Heredia-Flores, and M. C. Hernandez-Pizarro. 1984. Reproduccion del cucarachero comun (Troglodytes aedon, Aves, Troglodytidae) en el Valle del Cauca. Caldasia 14:85-123.

Arnold, T. 1993. Fledging success in experimentally manipulated broods of House Wrens. Wilson Bulletin 105:448-454.

Baltz, M. E., and C. F. Thompson. 1988. Successful incubation of experimentally enlarged clutches by House Wrens. Wilson Bulletin 100:70-79.

Beissinger, S. R. 1990. Experimental brood manipulations and the monoparental threshold in Snail Kites. American Naturalist 136:20-38.

Beissinger, S. R., and J. R. Waltman. 1991. Extraordinary clutch size and hatching asynchrony of a neotropical parrot. Auk 108:863-871.

Bock, C. E., A. Cruz, Jr., M. C. Grant, C. S. Aid, and T. R. Strong. 1992. Field experimental evidence for diffuse competition among southwestern riparian birds. American Naturalist 140:815-828.

Boyce, M. S., and C. M. Perrins. 1987. Optimizing Great Tit clutch size in a fluctuating environment. Ecology 68: 142-153.

Caswell, H. 1989. Matrix population models. Sinauer Associates, Sunderland, Massachusetts, USA.

Charnov, E. L., and J. H. Krebs. 1974. On clutch-size and fitness. Ibis 116:217-219.

Cody, M. L. 1966. A general theory of clutch size. Evolution 20:174-184.

Connell, J. H. 1978. Diversity in tropical rainforests and coral reefs. Science 199:1302-1310.

Drilling, N. E., and C. F. Thompson. 1988. Natal and breeding dispersal in House Wrens (Troglodytes aedon). Auk 105:480-491.

Finke, M. A., D. J. Milinkovich, and C. F. Thompson. 1987. Evolution of clutch size: an experimental test in the House Wren (Troglodytes aedon). Journal of Animal Ecology 56: 99-114.

Foster, R. B. 1982. Famine on Barro Colorado Island. Pages 201-212 in E.G. Leigh, Jr., A. S. Rand, and D. M. Windsor, editors. The ecology of a tropical forest: seasonal rhythms and long-term changes. Smithsonian Institution Press, Washington, D.C., USA.

Freed, L. A. 1986. Territory takeover and sexually selected infanticide in tropical House Wrens. Behavioral Ecology and Sociobiology 19:197-206.

—–. 1987a. Rufous-and-white Wrens kill House Wren nestlings during a food shortage. Condor 89:195-197.

—–. 1987b. The long-term pair bond of tropical House Wrens: advantage or constraint? American Naturalist 130: 507-525.

Fretwell, S. D. 1969. The adjustment of birth rate to mortality in birds. Ibis 111:624-627.

Gaines, S. D., and W. R. Rice. 1990. Analysis of biological data when there are ordered expectations. American Naturalist 135:310-317.

Garnett, M. C. 1981. Body size, heritability and influence on juvenile survival among Great Tits, Parus major. Ibis 123:31-41.

Gillespie, J. H. 1977. Natural selection for variances in offspring numbers: a new evolutionary principle. American Naturalist 111:1010-1014.

Godfray, H. C. J., L. Partridge, and P. H. Harvey. 1991. Clutch size. Annual Review of Ecology and Systematics 22:409-429.

Grant, B. R., and P. R. Grant. 1989. Evolutionary dynamics of a natural population. University of Chicago Press, Chicago, Illinois, USA.

Graves, J. 1991. Comments on the sample sizes used to test the effect of experimental brood enlargement on adult survival. Auk 108:967-969.

Gross, A. O. 1948. Eastern House Wren. Pages 113-141 in A. C. Bent, editor. Life histories of North American nuthatches, wrens, thrashers and their allies. Bulletin of the United States National Museum 195.

Gustafsson, L., and Sutherland, W. J. 1988. The costs of reproduction in the Collared Flycatcher Ficedula albicollis. Nature 335:813-815.

Hagvar, S., O. J. Sorensen, and E. Lund. 1989. Nest box studies on hole-nesting passetines in spruce forest. I. Effect of nest hole availability on population density. Fauna Norvegica Series C, Cinclus 13:33-41.

Harper, R. G., S. A. Juliano, and C. E Thompson. 1992. Hatching asynchrony in the House Wren, Troglodytes aedon: a test of the brood reduction hypothesis. Behavioral Ecology 3:76-83.

Hesse, R., W. C. Allee, and K. P. Schmidt. 1937. Ecological animal geography. John Wiley, New York, New York, USA.

Hochachka, W., and J. N.M. Smith. 1991. Determinants and consequences of nestling condition in Song Sparrows. Journal of Animal Ecology 60:995-1008.

Holdridge, L. R. 1967. Life zone ecology. Revised edition. Tropical Science Center, San Jose, Costa Rica.

Janzen, D. H. 1974. Tropical blackwater rivers, animals, and mast fruiting by the Dipterocarpaceae. Biotropica 6:69-103.

Jarvinen, A. 1991. A meta-analytic study of the effects of female age on laying-date and clutch-size in the Great Tit Parus major and the Pied Flycatcher Ficedula hypoleuca. Ibis 133:62-66.

Karr, J. R., and K. E. Freemark. 1983. Habitat selection and environmental gradients: dynamics in the “stable” tropics. Ecology 64:1481-1494.

Karr, J. R., J. D. Nichols, M. K. Klimkiewicz, and J. D. Brawn. 1990. Survival rates of birds of tropical and temperate forest: will the dogma survive? American Naturalist 136:277-291.

Kennedy, E. D., and H. W. Power. 1990. Experiments on indeterminate laying in House Wrens and European Starlings. Condor 92:861-865.

Klomp, H. 1970. The determination of clutch size in birds. Ardea 58:1-124.

Koenig, W. D. 1986. Geographical ecology of clutch size variation in North American woodpeckers. Condor 88:499-504.

Kulesza, G. 1990. An analysis of clutch-size in New World passerine birds. Ibis 132:407-422.

Lack, D. 1947. The significance of clutch size, parts I and II. Ibis 89:302-352.

—–. 1968. Ecological adaptations for breeding in birds. Methuen, London, UK.

Lande, R. 1988. Demographic models of the northern spotted owl (Strix occidentalis caurina). Oecologia 75:601-607.

Lessells, C. M. 1986. Brood size in Canada Geese: a manipulation experiment. Journal of Animal Ecology 55:669-689.

Lessells, C. M, and P. T. Boag. 1987. Unrepeatable repeatabilities: a common mistake. Auk 104:116-121.

Levings, S. C., and D. M. Windsor. 1982. Seasonal and annual variation in litter arthropods. Pages 355-387 in E. G. Leigh, Jr., A. S. Rand, and D. M. Windsor, editors. The ecology of a tropical forest: seasonal rhythms and long-term changes. Smithsonian Institution Press, Washington, D.C., USA.

Lima, S. L. 1987. Clutch size in birds: a predation perspective. Ecology 68:1062-1070.

Linden, M., L. Gustafsson, and T Part. 1992. Selection on fledging mass in the Collared Flycatcher and the Great Tit. Ecology 73:336-343.

Linden, M., and A. P. Moller. 1989. Cost of reproduction and covariation of life history traits in birds. Trends in Ecology and Evolution 4:367-371.

Lundberg, S. 1985. The importance of egg hatchability and nest predation in clutch size evolution in altricial birds. Oikos 45:110-117.

Mader, W. J. 1982. Ecology and breeding habits of the Savanna Hawk in the llanos of Venezuela. Condor 84:261-271.

Magrath, R. D. 1991. Nestling weight and juvenile survival in the Blackbird, Turdus merula. Journal of Animal Ecology 60:335-351.

Martin, T. E. 1987. Food as a limit on breeding birds: a life history perspective. Annual Review of Ecology and Systematics 18:453-487.

Mayfield, H. F. 1975. Suggestions for calculating nest success. Wilson Bulletin 87:456-466.

Moller, A. P. 1984. Geographical trends in breeding parameters of Swallows Hirundo rustica and House Martins Delichon urbica. Ornis Scandinavica 15:43-54.

Moreau, R. E. 1944. Clutch size: a comparative study, with special reference to African birds. Ibis 86:286-313.

Murphy, M. T. 1989. Life history variability in North American breeding tyrant flycatchers: phylogeny, size, or ecology? Oikos 54:3-14.

Nur, N. 1984. The consequences of brood size for breeding Blue Tits II. Nestling weight, offspring survival and optimal brood size. Journal of Animal Ecology 53:497-517.

—–. 1990. The cost of reproduction in birds, evaluating the evidence from manipulative and nonmanipulative studies. Pages 281-296 in J. Blondel, A. Gosler, J. B. Lebreton, and R. McCleery, editors. Population biology of passerine birds, an integrated approach. Springer-Verlag, Berlin, Germany.

Orell, M. 1990. Brood size manipulations and reproductive success in the Willow Tit. Pages 297-306 in J. Blondel, A. Gosler, J. B. Lebreton, and R. McCleery, editors. Population biology of passerine birds, an integrated approach. Springer-Verlag, Berlin, Germany.

Perrins, C. M. 1965. Population fluctuations and clutch-size in the Great Tit, Parus major L. Journal of Animal Ecology 34:601-647.

—–. 1977. The role of predation in the evolution of clutch size. Pages 181-191 in B. Stonehouse and C. Perrins, editors. Evolutionary ecology. University Park Press, Baltimore, Maryland, USA.

Pollock, K. H., J. D. Nichols, C. Brownie, and J. E. Hines. 1990. Statistical inference for capture-recapture experiments. Wildlife Monographs 107:1-97.

Redondo, T, and F. Castro. 1992. The increase in risk of predation with begging activity in broods of magpies, Pica pica. Ibis 134:180-187.

Ricklefs, R. E. 1969. An analysis of nesting mortality in birds. Smithsonian Contributions to Zoology 9.

—–. 1977. A note on the evolution of clutch size in altricial birds. Pages 193-214 in B. Stonehouse and C. Perrins, editors. Evolutionary ecology. University Park Press, Baltimore, Maryland, USA.

—–. 1980. Geographic variation in clutch size among passerine birds: Ashmole’s Hypothesis. Auk 97:38-49.

Robinson, K. D., and J. T Rotenberry. 1991. Clutch size and reproductive success of House Wrens rearing natural and manipulated broods. Auk 108:277-284.

Robinson, S. K. 1986. Three-speed foraging during the breeding cycle of Yellow-rumped Caciques (Icterinae: Cacicus cela). Ecology 67:394-405.

Roskaft. E. 1985. The effect of enlarged brood size on the future reproductive potential of the Rook. Journal of Animal Ecology 54:255-260.

Safriel, U. N. 1975. On the significance of clutch size in nidifugous birds. Ecology 56:703-708.

Sauer, J. R., and B. K. Williams. 1989. Generalized procedures for testing hypotheses about survival or recovery rates. Journal of Wildlife Management 53:137-142.

Schaffer, W. M. 1974. Optimal reproductive effort in fluctuating environments. American Naturalist 108:783-790.

Schaub, R., R. L. Mumme, and G. E. Woolfenden. 1992. Predation on the eggs and nestlings of Florida Scrub Jays. Auk 109:585-593.

Schluter, D. 1988. Estimating the form of natural selection on a quantitative trait. Evolution 42:849-861.

Skutch, A. F. 1949. Do tropical birds rear as many birds as they can nourish? Ibis 91:430-455.

—–. 1953. Life history of the Southern House Wren. Condor 55:121-149.

—–. 1985. Clutch size, nesting success, and predation on nests of Neotropical birds, reviewed. Pages 575-594 in P. A. Buckley, M. S. Foster, E. S. Morton, R. S. Ridgely, and F. G. Buckley, editors. Neotropical ornithology. Ornithological Monographs 36.

Slagsvoid, T. 1982. Clutch size variation in passerine birds: the nest predation hypothesis. Oecologia (Berlin) 54:159-169.

—–. 1984. Clutch size variation of birds in relation to nest predation: on the cost of reproduction. Journal of Animal Ecology 53:945-953.

Smith, H. G., H. Kallander, and J.-A. Nilsson. 1989. The trade-off between offspring number and quality in the Great Tit Parus major. Journal of Animal Ecology 58:383-401.

Smith, N. G. 1982. Population irruptions and periodic migrations in the day-flying moth Urania fulgens. Pages 331-344 in E. G. Leigh, Jr., A. S. Rand, and D. M. Windsor, editors. The ecology of a tropical forest: seasonal rhythms and long-term changes. Smithsonian Institution Press, Washington, D.C., USA.

Smith, S. M. 1994. Social influences on the dynamics of a northeastern Black-capped Chickadee population. Ecology 75:2043-2051.

Stearns, S. C. 1992. The evolution of life histories. Oxford University Press, New York, New York, USA.

Tinbergen, J. M. 1987. Costs of reproduction in the Great Tit: intraseasonal costs associated with brood size. Ardea 75:111-122.

Trexler, J. C., and J. Travis. 1993. Nontraditional regression analyses. Ecology 74:1629-1637.

VanderWerf, E. 1992. Lack’s clutch size hypothesis: an explanation of the evidence using meta-analysis. Ecology 73: 1699-1705.

Walsberg, G. E. 1983. Avian ecological energetics. Pages 161-220 in D. S. Farner, J. R. King, and K. C. Parkes, editors. Avian biology. Volume 7. Academic Press, New York, New York, USA.

Wilkinson, L. 1989. SYSTAT: the system for statistics. SYSTAT, Inc., Evanston, Illinois, USA.

Winnett-Murray, K. 1986. Variation in the behavior and food supply of four Neotropical wrens. Dissertation. University of Florida, Gainesville, Florida, USA.

Young, B. E. 1993. Geographical variation in avian clutch size: the case of the tropical House Wren, Troglodytes aedon. Dissertation. University of Washington, Seattle, Washington, USA.

—–. 1994a. Geographical and seasonal patterns of clutch size variation in House Wrens. Auk 111:545-555.

—–. 1994b. The effects of food, nest predation, and weather on the timing of breeding in tropical House Wrens, Troglodytes aedon. Condor 96:341-353.

Zach, R. 1982. Nestling House Wrens: weight and feather growth. Canadian Journal of Zoology 60:1417-1425.

Zach, R., and J. N. M. Smith. 1981. Optimal foraging in wild birds? Pages 95-113 in A. C. Kamil and T. D. Sargent, editors. Foraging behavior: ecological, ethological, and psychological approaches. Garland, New York, New York, USA.

APPENDIX

PARAMETER CALCULATION AND ESTIMATES FOR POPULATION MODEL

I calculated the parameters for the model from survivorship and reproductive data collected during the Monteverde House Wren study. Good year parameters represent the 1989 and 1990 breeding seasons ([G.sub.1], [G.sub.2]) and 1989-1990 survivorship ([G.sub.3], S). Bad year parameters represent the 1991 breeding season ([G.sub.1], [G.sub.2]) and 1990-1991 and 1991-1992 survivorship ([G.sub.3], S).

Details of parameter calculations are as follows. I calculated [G.sub.1] separately for each clutch size strategy from the survivorship of chicks in each brood treatment group. In 1990, a good year for postfledging survivorship, the three treatment groups did not differ, so [G.sub.2] represents the pooled data from all groups. 1991 was a bad year for postfledging survivorship, with significant differences among treatment groups, so [G.sub.2] reflects actual survivorship of each group. Survivorship from independence to the following breeding season, [G.sub.3], varied among brood treatment groups, so I calculated it separately for each. To estimate [G.sub.3] (good years: 1989-1990, 1990-1991; bad year: 1991-1992), I divided the survivorship of all male offspring (assuming a 1:1 sex ratio at fledging and equal survivorship of females and males) from fledging to the following breeding season by [G.sub.2].

Because brood treatment had no effect on adult survivorship, S was the survivorship of adult females regardless of the number of nestlings they raised. One year (1989-1990) was good for adult survivorship and two were bad (1990-1991 and 1991-1992). For F, the number of breeding attempts that produce eggs surviving the incubation period averaged 1.3333 attempts per female; it did not vary among brood treatment groups or years. The 3-yr mean hatching success rate was 0.8884. Second clutches of females raising enlarged first broods averaged 6.8% smaller than first clutches, so average clutch size in the five to six-egg group was adjusted to account for this factor. For the purposes of comparing clutch size strategies, clutch size did not change in good and bad years. Actual clutch size varied little in Monteverde. During the three years of the study, annual mean clutch size ranged from 3.43 (N = 113, 1990) to 3.51 (N = 119, 1991).

To calculate the variance in F, I calculated separate sensitivities and variances for the factors that made up F. Hence, [F.sub.1] is clutch size, [F.sub.2] is the number of breeding attempts per female, and [F.sub.3] is hatching success. The variances of these three parameters reflect their measure during the 1989-1991 breeding seasons.

[TABULAR DATA FOR TABLE A1 OMITTED]

COPYRIGHT 1996 Ecological Society of America

COPYRIGHT 2004 Gale Group