Differential access to social mates, to extrapair fertilizations,

Behavioral Ecology Vol. 10 No. 1: 80 90 Red coloration of male northern cardinals correlates with mate quality and territory quality L. LaReesa Wolfenbarger Section of Neurobiology and Behavior, Cornell University, Ithaca, NY 14853-2702, USA I investigated how mate quality and territory quality influence an extravagant ornament in a socially monogamous species that defends multipurpose territories. Northern cardinals (Cardinalis cardinalis) are a highly dichromatic, socially monogamous species, and males are a brilliant red. I conducted a 3-year field study of northern cardinals and found that redder males produced more offspring in a breeding season. Two selective factors mediated this fitness gain. Redder males were paired with earlier breeding females, an established measure of mate quality in birds. Second, redder males obtained territories of higher quality, as measured by vegetation density. Interactions among these factors were also important in explaining variance in male reproductive success. Multivariate analysis indicated that earlier breeding increased reproductive success independent of territory quality. In turn, territory quality contributed to male reproductive success through its effect on nest survival and possibly through its role in attracting an earlier breeding female. Key words: Cardinalis cardinalis, plumage coloration, laying date, mate quality, ornament, reproductive success, territory quality. [Behav Ecol 10:80 90 (1999)] Differential access to social mates, to extrapair fertilizations, and to mates of varying quality causes variance in reproductive success, and variability in these components of fitness has been proposed to explain the evolution and maintenance of extravagant, ornamental traits, often present only in males. In polygynous species, the number of mates per male ranges greatly among some species, and studies have shown that ornaments do influence the number of social mates (Andersson, 1994). In socially monogamous species, social bonds may preclude males from fertilizing multiple mates, except through extrapair copulations. Extrapair fertilizations are widespread in many species of birds (Birkhead and Møller, 1992). Studies have revealed associations between male ornaments and extrapair paternity (Hasselquist et al., 1996; Sundberg and Dixon, 1996; Yezerinac and Weatherhead, 1997), but lack of associations have also been reported (Hill et al., 1994; Rätti et al., 1995). Using independent contrasts of closely related species, Møller and Birkhead (1994) reported a positive correlation between male plumage coloration and extrapair paternity explaining 9% of the variation, indicating that other factors, such as mate quality, may account for variation in male plumage coloration. Darwin (1871) explained the possible importance of mate quality in promoting sexually selected traits and proposed that in monogamous species males with extravagant traits produce more offspring by pairing with early breeding females. Early breeding is often associated with higher reproductive success among birds, including both passerines and nonpasserines (Price et al., 1988). Several of the empirical studies that form the basis for our current understanding of male ornamentation in monogamous species have shown that females prefer more highly ornamented males and that early breeding females do pair with these males (Hill, 1990; Hill et al., 1994; Møller, 1988, 1990; Norris, 1990; O Donald, 1980). Although these studies have shown that female quality plays an impor- L. L. Wolfenbarger is now at the Department of Zoology, University of Maryland, College Park, MD 20742, USA. E-mail: LW137@ umail.umd.edu. Received 8 November 1997; accepted 9 July 1998. 1999 International Society for Behavioral Ecology tant role in generating variance in male reproductive success, there is no consensus about the importance of mate quality relative to other factors. Additionally, the interactions between male ornamentation, female quality, and territory acquisition and quality in determining reproductive success are not clearly understood, in spite of the fact that a large percentage of monogamous species are territorial. The classic studies of monogamous species have been conducted on species with either no territories or territories limited to the area immediately around the nest site (Davis and O Donald, 1976; Hill, 1990; Hill et al., 1994; Møller, 1988, 1990; O Donald, 1980). In short, male ornamentation is not understood in what may be the archetypal passerine social system: monogamy with males defending multipurpose territories (Emlen and Oring, 1977). In monogamous species with multipurpose territories, males may pair with high-quality females either by acquiring high-quality territories or by being chosen as mates by highquality females on the basis of some other male character, such as degree of ornamentation. Among monogamous male birds, one of the most widespread ornaments is conspicuous bright plumage color. The presence of conspicuous coloration is known to affect the acquisition and defense of resources (Flood, 1984; Røskaft and Rohwer, 1987; Rowland, 1984; but see Stutchbury, 1992), including high-quality territories (Hill, 1988; Studd and Robertson, 1985b). In addition to attracting a higher quality female, the quality of the breeding territory may affect the number of offspring a male produces because factors such as vegetation density and amount of food influence offspring survival (Best and Stauffer, 1980; Conner et al., 1986; but see Howlett and Stutchbury, 1996). In this study I examined how the red coloration of male northern cardinals (Cardinalis cardinalis) interacts with territory quality and mate quality to explain variance in male reproductive success. Male northern cardinals are socially monogamous and defend large, multipurpose territories throughout the breeding season (Gottfried, 1976; Kinser, 1973; Oberholser, 1974; Wolfenbarger, personal observation). The plumage of male cardinals is bright red ventrally and a reddish gray on the back, whereas females are primarily light brown. I found that redder male cardinals do indeed produce more offspring, as is the case in other ornamented species

Wolfenbarger Red coloration, mate quality, and territory quality 81 (Andersson, 1994). I examined two hypotheses to explain this pattern: (1) that redder males were paired with earlier breeding females; and (2) that redder males defended higher quality territories. In an effort to understand how territory quality and female quality interact, I used multivariate techniques to examine the relative contribution of territory quality and mate quality to variance in male reproductive success. METHODS Study site and general methods I conducted fieldwork in the Fall Creek Nature Preserve of the Cornell University Plantations in Varna, New York, USA (42 27 N, 76 26 W). The preserve is a mixture of secondary deciduous forest habitat (populated by Quercus spp., Prunus spp., Planatus occidentalis, Acer spp.) and of riparian habitat [dominated primarily by honeysuckle (Lonicera spp.) and multiflora rose (Rosa multiflora)]. At this site and in other northern latitudes where cardinals are resident year round, nesting starts in mid- to late April, well after territory acquisition and pair formation, which occurs January through March (Bent, 1968; Kinser, 1973). Females build open-cup nests, and males are known to provide substantial amounts of food to females during nest building, the period of egg synthesis (Kinser, 1973). Females lay two to four eggs, and three eggs are most common (Kinser, 1973; Laskey, 1944). Both males and females feed nestlings (Bent, 1968; Filliater and Breitwisch, 1997). Nest replacement after a failure is rapid, on average 5 7 days (Scott et al., 1987). Cardinals attempt to produce multiple clutches of young throughout the breeding season, and nesting attempts may continue through August and early September (Bent, 1968; Kinser, 1973). At the end of the breeding season, cardinals undergo a single annual molt (Pyle et al., 1987). Between January and June of 1992 1995, I captured 57 male and 31 female cardinals using mist nets and funnel traps baited with sunflower seeds. Supplemental food could potentially influence the date a female begins breeding, a variable of interest; therefore, after early March, I ceased using baited traps and captured remaining individuals using mist nets. In every year a minimum of 6 weeks elapsed from the date that I stopped using baited traps to the date of the first recorded nest on the study site. Individuals were banded with a U.S. Fish and Wildlife Service metal band and two or three plastic colored leg bands for identification. Red color bands have been shown to influence reproductive success in some avian species, including red-winged blackbirds (Agelaius phoeniceus; Metz and Weatherhead, 1991), red-cockaded woodpeckers (Picoides borealis; Hagan and Reed, 1988), and zebra finches (Taeniopygia guttata; Burley et al., 1982). Because the study presented here addressed red coloration and its relationship to reproductive success, I did not use red or orange color bands. Measuring coloration To quantify male coloration, I used the color chip series of the Methuen Handbook of Color (Kornerup, 1967). The color chip series provides a measure of three components of color: hue, tone, and intensity. The hue component ranked color on a scale from yellow (5) to red (11). The tone component quantified the amount of black present [1 (all black) to 6 (none)]; the intensity component indicated the degree of saturation of pigment [from little (1) to complete (8)]. For each of these components, a higher score indicates a redder or brighter color. I analyzed hue, tone, and intensity scores separately. I measured coloration of males in the breast area. To measure breast coloration in 1993 1995, I used a grid that divided the breast into eight 1 4-cm rectangles and placed it immediately posterior to the black bib. In each region I recorded two color scores. From a distance male cardinals appear to have a uniformly bright red breast, and in each region, I quantified the brightest color (bright breast score). Close observation revealed that within each of the eight regions of the breast, patches of dull colors were often interspersed with a brighter red background, something that was not readily apparent in other body areas. Therefore, I also recorded the lowest color score (dull breast score) in each of the eight regions of the grid. I tested the repeatability of the color scoring method by using specimens in the Cornell Vertebrate Collections, and the method was highly repeatable (Wolfenbarger, 1996). In 1992 I used similar methods, except I quantified color in four larger regions (twice as large as in 1993 and later) rather than eight. Coloration scores obtained using both the 1992 and later methods were significantly correlated (n 7, hue: dull breast, r s.87, p.03; bright breast, r s.94, p.02; intensity: dull breast, r s.97, p.02; bright breast, r s.95, p.02; tone: all scores were maximum). For analyses, I summed the color scores for each color component (hue, tone, and intensity) of the eight (or four, in 1992) regions of the breast. In cardinals, hue and intensity were not consistently correlated in univariate relationships (bright breast hue intensity: r.134, ns, n 49; dull breast hue intensity: r.479, p.001, n 56), and the correlations explained a low amount of the variability in each (bright breast: r 2.02; dull breast: r 2.23 ). Therefore, a combination score could yield identical scores for two males that differed both in hue and in intensity, and for this reason I did not calculate combined scores. Other morphological measurements I also measured the following variables at the time of capture: tarsus length, crest length (from the proximate end of the exposed culmen to the tip of longest crest feather), length of black bib [from the base of the bill with the neck flattened to the maximum length of black extending toward the breast (after Møller, 1987)], and width of black bib at the widest region. All measurements were to the nearest 0.1 mm. Measuring territory quality Boundary locations and vegetation densities of all territories were recorded during 1992, 1993, and 1994. In approximately 700 field-h of censusing along with ad libitum observations in March and April of these years, locations of males and females were plotted on a scale map constructed from the location of markers that were placed on the study site approximately every 25 m. I identified territorial boundaries as the outer boundaries of singing perches of neighboring pairs; these boundaries did not overlap. For 39 territories, I measured horizontal foliage density at the ground, at 1 m, and at 2 m using methods based on Mac- Arthur and MacArthur (1961) and Conner et al. (1986). Using a gridded, 0.25-m 2 board, I measured the distance at which 50% of the board was obscured by vegetation. For each territory I chose three random sites, and at each site I measured distances in three compass directions (0, 120, 240 ), yielding nine measures at each height. Distance measurements were converted into density estimates using the formula density (ln 2/distance) from MacArthur and MacArthur (1961), and measurements were averaged to arrive at a single estimate for each height. Horizontal foliage density has been

82 Behavioral Ecology Vol. 10 No. 1 Table 1 Variation in coloration, vegetation density, first-nest initiation date, and reproductive success among years 1992 Mean SD 1993 1994 1995 n Mean SD n Mean SD n Mean SD n Dull breast hue 36.7 3.0 7 69.5 5.0 8 73.1 6.6 14 76.1 4.3 8 Bright breast hue 40.0 0.0 7 80.4 3.7 8 81.0 4.9 14 81.5 2.7 8 Dull breast intensity 29.0 1.0 7 61.5 2.2 8 61.7 1.5 14 62.9 0.8 8 Bright breast intensity 31.4 0.8 7 63.6 0.7 8 63.4 1.1 14 62.9 0.8 8 Dull breast tone 23.9 0.4 7 48.0 0.0 8 47.9 0.3 14 48.0 0.0 8 Bright breast tone 24.0 0.0 7 48.0 0.0 8 48.0 0.0 14 48.0 0.0 8 Vegetation density at 0 m (m 2 /m 3 ) 0.37 0.12 9 0.52 0.41 13 0.54 0.27 17 Vegetation density at 1 m (m 2 /m 3 ) 0.29 0.09 9 0.33 0.13 13 0.36 0.22 17 Vegetation density at 2 m (m 2 /m 3 ) 0.24 0.09 9 0.27 0.19 13 0.21 0.15 17 First-clutch initiation date (1 January 1) 137.4 6.8 8 124.6 6.4 10 119.6 6.5 8 Reproductive success 0.4 0.9 7 1.4 1.0 15 1.5 1.2 17 Significant differences in first-clutch initiation date occurred among years (Kruskall Wallis test, H 15.3, p.001). All others p.05. shown to be an appropriate index of territory quality for cardinals by Conner et al. (1986), who demonstrated that denser horizontal foliage (at ground, 1 m, 2 m, as well as total density from 0 3 m) on a territory, as well as insect abundance, resulted in higher fledging success from cardinal nests. Other characteristics of a territory, such as size and vegetation around a nest site, have not been found to be related to nesting success (Conner et al., 1986; Filliater et al., 1994). Vegetation density measurements were taken 28 July 5 August 1992, 12 July 21 July 1993, and 6 August 15 August 1994. The dominant vegetation at the heights measured consisted of deciduous, perennial shrubs (primarily Lonicera and Rosa) on all territories in the study site. Assigning first-clutch initiation date First nests were found during the 1993, 1994, and 1995 breeding seasons. First clutch initiation date was defined as the date of the first egg that a female laid in a breeding season. To determine this date, I searched for nests by locating females and by watching their behavior. Nests were found by following females that were carrying nesting material or by identifying areas in which pairs spent large amounts of time. Due to the difficulty of reliably assigning a nesting status to some females, I used two criteria to determine that a nest was not present: (1) a female was readily locatable and was not in a repeatable location in the 2- to 4-h period I spent searching for a nest; and (2) her mate did not follow her when she flew to different locations around her territory. In no case were these criteria later found to be incorrect. Twenty-six nests were considered first nests, and 20 (73%) of these were found before completion of egg laying. In all cases where nests were visited during laying, only one egg per day was added to the nest. If a nest with a completed clutch was found, first clutch initiation date was calculated by subtracting the average incubation period ( 12 days, SD 0, n 9, where incubation period is defined as the period between the laying of the penultimate egg and the hatching of the first egg) and the appropriate number of days for egg laying from the day of hatching. If I was unable to locate a female for 4 consecutive days, I did not consider any subsequent nest to be a first nest. Clutches on my study site have been completed and preyed upon in 4 days. The methods used for assigning first-clutch initiation date are conservative and ensure that only first nests are included in the analyses involving first-clutch initiation date. Determining reproductive success in a season Once all possible first nests had been located, I searched territories of marked males for nests once every 7 days and relied on the conspicuous activity associated with feeding nestlings to locate successful nests. Therefore, I obtained minimum estimates of nest failure and of number of nesting attempts. The conspicuous activity of older nestlings and recently fledged offspring minimized the possibility of missing successful nests. The number of eggs and the number of young in the nest could be counted by looking directly into the open-cup nest or by using a ladder and a pole with a mirror on the end. The average nest height in the population was 1.7 m (SD 0.5, n 53). I visited nests once every 3 days after the eggs hatched until the nestlings fledged or disappeared. I attributed nest failure to predation if the contents disappeared completely between successive nest checks, and often bits of egg shell or remains of nestlings were present. I categorized the nest as abandoned if the contents remained intact between two successive nest checks but the female had discontinued incubating the nest as evidenced by cold eggs. I used the category of abandonment only in 1994 for 4 nests, and the eggs remained in the nest for an average of 5 days (SD 1.5). I determined each pair s reproductive success in the 1992, 1993, and 1994 seasons by totaling the number of young that left nests tended by a pair. In the analyses using reproductive success per season, I included all successful nests that were found on each pair s territory. Data analyses For color measures with limited variability (bright breast intensity) or only two categories of scores (crest hue and crest tone), I compared seasonal reproductive success of males with the maximum score to those that had lower scores using the Mann-Whitney test. Spearman rank correlation tests were used for other univariate analyses unless otherwise noted (Conover, 1980). Data for multiple years were pooled, and relationships were consistent among years unless otherwise noted. The data in each year were standardized to a mean of 0 and a standard deviation of 1, and the original means and standard deviation are shown in Table 1. I standardized coloration data to pool data collected in 1992 with that collected in 1993 1995. Also, first-clutch initiation dates varied substantially among years, as did reproductive success per season. Therefore, by standardizing the data I eliminated seasonal differences to test relationships between coloration and variables of interest.

Wolfenbarger Red coloration, mate quality, and territory quality 83 In examining the relationships between male coloration, horizontal foliage density, and seasonal reproductive success in 1992 1994, 5 of the 37 males had data for more than one year. To maintain independence among years, each male was represented only once in both summary statistics and tests. For individuals with multiple years of data, I randomly selected one year of data to include. In testing hypotheses about the relationships between color, first-nest initiation date, vegetation density, and reproductive success, I used one-tailed tests for comparisons because the hypotheses tested made a priori predictions. For parametric multiple regression analysis, assumptions of normality were tested using the Anderson-Darling statistic (Snedecor and Cochran, 1989). Variables used in parametric tests satisfied assumptions of normality. I used logistic regression analysis to test whether the occurrence of nest predation was related to nest initiation date (Hosmer and Lemeshow, 1989). When I tested a single hypothesis with multiple statistical tests, I adjusted the level using a sequential Bonferroni adjustment for multiple correlations (Rice, 1989). In particular, I used a Bonferroni adjustment when I tested a relationship between one color component (i.e., hue) and another variable using both the bright and dull measurement of color so that the level for the lower p value was 0.025 and was 0.05 for the higher p value. Second, I used a Bonferroni adjustment when I tested relationships between vegetation density and bright and dull breast hue because I had three measures of vegetation density. Therefore, the levels for these six tests were 0.0083, 0.01, 0.013, 0.017, 0.025, and 0.05 for the relationship with the lowest p value to the highest p value. RESULTS Coloration and reproductive success Dull breast hue strongly correlated with reproductive success in a breeding season (r s.534, n 29, p.0024; Figure 1a), and bright breast hue tended to correlate with reproductive success (r s.405, n 22, p.030; Figure 1b). Therefore, redder males produced more offspring in a breeding season. No other measures of color were significantly related to reproductive success, although males with the maximum brightbreast intensity scores tended to produce more offspring than those with lower scores (standardized reproductive success: x maximum 0.21 0.96, n 22; x other 0.66 0.66, n 7); these differences were not significant when using a sequential Bonferroni adjustment for the level (Z 1.81, n 29, p.035). There was no relationship between intensity using the dull breast measurement and reproductive success (r s.063, n 29, p.37). The relationship between breast tone and seasonal reproductive success was not tested because breast tone did not vary for the bright measurement, and 28 of 29 males had identical tone scores for the dull measurement. Because dull breast hue was significantly correlated with seasonal reproductive success and bright breast hue marginally so, I used both dull and bright breast hue as the coloration measures in all subsequent examinations of the relationships between coloration and territory density and breeding date. Red coloration, early breeding, and vegetation density Male breast hue negatively correlated with first-clutch initiation date (dull breast: r s.64, n 23, p.0014; bright breast: r s.42, n 23, p.024; Figure 2): males with redder breasts were paired with earlier breeding females. Figure 1 Relationship between standardized breast hue and standardized seasonal reproductive success. Symbols denote data by year: circles 1992, squares 1993, triangles 1994. (A) Dull breast measurement is positively correlated with seasonal reproductive success in 1992 1994 (Spearman rank correlation with sequential Bonferroni adjustment, 0.025, r s.53, n 29, one-tailed, p.0024). (B) Bright breast measurement is positively correlated with seasonal reproductive success in 1993 and 1994 ( 0.05, r s.41, one-tailed, p.030, n 22). Data from 1992 were excluded because there was no variation in bright breast hue in that year.

84 Behavioral Ecology Vol. 10 No. 1 Figure 2 The relationship between standardized dull breast hue of males and standardized date of first-clutch initiation date by mates. Symbols denote data by year: circles 1993, squares 1994, triangles 1995. (A) Dull breast measurement was negatively correlated with first-clutch initiation date (r s.64, n 23, one-tailed, p.0014). (B) Bright breast measurement was negatively correlated with firstclutch initiation date (r s.42, n 23, one-tailed, p.024). Male breast hue was positively correlated with the density of horizontal foliage at 2 m on territories (dull breast hue: r s.60, n 29, p.0010; bright breast hue: r s.34, n 22, p.014; Figure 3), indicating that redder males had territories with denser vegetation at this height. Vegetation density at 2 m is near the average nest height in the population of 1.7 m (SD 0.5 m). Male breast hue and vegetation density at the ground and at 1 m were not significantly correlated (0 Figure 3 Relationship between standardized breast hue of males and the vegetation density at 2 m on breeding territories. Symbols denote data by year: circles 1992, squares 1993, triangles 1994. (A) Dull breast measurement (r s.60, n 29, one-tailed, p.0010). (B) Bright breast measurement (r s.34, n 22, one-tailed, p.014). m: dull breast, r s.15, n 29, p.21; bright breast, r s.30, one-tailed, n 22, p.083; 1 m: dull breast, r s.22, n 29, p.12; bright breast, r s.063, n 22, p.30). Vegetation density at 2 m was negatively but not significantly correlated with first-clutch initiation date (r s.41, n 18, two tailed, p.088); pairs on denser territories tended to breed earlier. In cardinals, territory acquisition and settlement occurs 4 10 weeks before nest initiation; therefore territory characteristics may influence nest initiation date. In a partial correlation analysis, dull breast hue and first-

Wolfenbarger Red coloration, mate quality, and territory quality 85 clutch initiation date were significantly correlated while controlling for vegetation density (Spearman rank partial correlation, hue, date.veg 0.65, n 15, p.030). However, bright breast hue and first-clutch initiation date were not significantly correlated when effects of vegetation density were controlled (Spearman rank partial correlation, hue, date.veg 0.16, n 15, ns). Thus, the correlation between dull breast hue and early breeding was independent of any association between vegetation density and breeding date, but bright breast hue and early breeding were not significantly independent of territory effects. First-clutch initiation date, vegetation density, and reproductive success Early breeding pairs had higher reproductive success over an entire breeding season (r s.64, n 18, p.0043; Figure 4a). Across years, first-clutch initiation date was negatively correlated with reproductive success of first-nests (r s.42, n 26, two tailed, p.038; Figure 4b). However, the relationship was not present in the 1994 breeding season (r s.17, n 8, two tailed, p.65). Nest abandonment and hatching failure occurred only in 1994 and coincided with a cold, rainy period just after the onset of breeding. Early breeding pairs had more successful clutches of young than later breeding pairs (r s.58, n 18, two-tailed, p.017), as later breeding pairs were most likely to have no successful clutch. Early breeding pairs did not produce significantly more offspring per successful clutch than later breeding pairs (r s.39, n 11, two tailed, p.22). In 1992 and 1993, pairs with denser foliage density at 2 m on their territories fledged significantly more offspring (r s.39, n 22, p.037; Figure 5a) but not in 1994 (r s.097, n 17, p.35; Figure 5b). And, if data are pooled across years, denser horizontal foliage at 2 m tended to correlate with higher reproductive success but not significantly (r s.26, n 39, p.053). The strong correlations between vegetation density at 2 m and reproductive success were associated with relative high predation rates and low fledging success. In 1992 and 1993 nest failures attributed to predation were 88% and 69%, and fledging success was 12% and 21%, respectively. In contrast, in 1994, the year in which there is no relationship between vegetation density and reproductive success, only 24% of nest failures were attributed to predation, and 88% of nests fledged offspring. There was no evidence that nest predation throughout the breeding season was influenced by nest initiation date (logistic regression: y occurrence of nest predation, x 1 year, x 2 nest initiation date; G 2.90, df 2, p.23). The multiple regression of reproductive success on initiation date and vegetation density was highly significant (r 2.51, n 18, F 2,15 7.85, p.0046). The standardized partial regression coefficients ( ) of this analysis show that earlier breeding pairs fledged more young, independent of the effects of vegetation density ( date 0.46, p.046). Similarly, for vegetation, while not significant, indicates that vegetation density is positively correlated with number of young ( veg 0.35, p.12). These results confirm that an earlier breeding date is associated with producing more offspring in a season regardless of any possible effect vegetation density may have on nest initiation date. Other variables correlated with breast hue For all males captured at the study site, neither dull breast hue nor bright breast hue correlated significantly with tarsus length, width or length of the black bib, or crest length (all Figure 4 (A)The relationship between standardized first-nest initiation date and standardized reproductive success in a breeding season. Pairs with the earliest first-nest initiation dates had the highest seasonal reproductive success (r s.64, n 18, one-tailed, p.0043). (B) The relationship between standardized first-nest initiation date and standardized number of offspring produced by first nests. Across years, nests initiated earliest in the breeding season produced the highest number of offspring (r s.42, n 26, two-tailed, p.038). However, there is no relationship between these two variables in 1994 (r s.17, n 8, two-tailed, p.65). Symbols denote data by year: circles 1993, squares 1994, triangles 1995. p.05, between 47 and 54 males included in each correlation). For the 8 individuals captured in 2 years, dull breast hue and bright breast hue of individuals differed significantly between years (paired t test, dull: t 3.70, df 7, p.008; bright: t 3.51, df 7, p.01). On average, dull breast hue

86 Behavioral Ecology Vol. 10 No. 1 Figure 5 Relationships between standardized vegetation density at 2 m and seasonal reproductive success on territories. Symbols denote data by year: circles 1992, squares 1993, triangles 1994. (A) In 1992 1993 more offspring were produced in a breeding season on territories with higher vegetation density (r s.39, n 22, onetailed, p.037). (B) In 1994 there was no significant relationship between these two variables (r s.097, n 17, one-tailed, p.35). scores increased 9.3 7.1 (13%), and bright breast hue scores increased 3.4 2.7, or 4%. Dull breast intensity of males breast color tended to increase between years (x 1.75, SD 2.3, t 2.28, df 7, p.064), but bright breast intensity showed no significant increase (x 0.25, SD 1.8, t 0.40, df 7, p.70). Tone scores did not change significantly in relation to age (dull: t 1.43, df 7, p.20; bright: no variation in males.). DISCUSSION Coloration and reproductive success In northern cardinals, males with redder breast coloration, and specifically with a homogenous distribution of the reddest hue, had higher reproductive success within a breeding season, whereas bright breast hue and bright breast intensity showed positive but nonsignificant correlations with reproductive success (Figure 1). Most studies of plumage coloration have measured the brightest coloration present, its extent, or age-based plumage characteristics (Hill 1988, 1990, 1991; Järvi et al., 1987; Møller, 1987; Norris, 1990; Rohwer, 1975; Sundberg, 1995). Cardinals, however, exhibit very low amounts of phenotypic variation in bright plumage coloration and its extent and have no age-specific plumage. In cardinals, the most variable plumage character was dull breast hue, which measured the coloration of small patches of dull color interspersed within the larger area of bright red coloration on the breast, and this trait correlated more strongly with all components of reproductive success. The difference in the strength of correlations when using dull breast hue and bright breast hue likely reflect the differences in variability of these two measure of breast coloration. Red coloration in cardinals and most other birds is attributed primarily to carotenoid-based pigments (Fox and Vevers, 1960; Hudon, 1991) obtained from diet (Hill 1992, 1993b; Witmer, 1996). Carotenoid-based coloration is associated with male reproductive success in a variety of taxa (Bakker and Mundwiler, 1994; Endler and Houde, 1995; Hill, 1988, 1990; Johnson et al., 1993; Kodric-Brown, 1985; Sundberg, 1995; Zuk et al., 1990; reviewed by Andersson, 1994) and may function as an indicator of condition, foraging ability, parental care, genetic quality, age, or some other male characteristic (reviewed by Andersson, 1994; Gray, 1996; but for alternative functions, see Andersson, 1994). In male cardinals, red coloration is not associated with absolute effort in feeding nestlings (Linville et al., 1998). Other possible functions of male color have not been studied in this species. Red coloration did not correlate with measures of size, crest length, or black bib size; therefore, these morphological traits do not explain the correlations between coloration, breeding date, and territory quality discussed below. Age-related effects on reproduction are widespread on birds (Sæther, 1990). Although cardinals are not considered to have age-specific plumage (Pyle et al., 1987), variation in red coloration may be associated with male age. Males recaptured in subsequent years became redder, but this could be a result of a changing environment rather than age. For example, in southwestern Ohio, a year with low fruit availability (a possible source of dietary carotenoids) was associated with a subsequent population-wide decrease in mean coloration scores (Linville and Breitwisch, 1997). Over the course of the study presented here, only 7 of the 57 males banded were observed breeding in more than 1 year, and data were collected on 5 of these in both years. This small sample makes it impossible to incorporate the effects of age into the present analysis. Disappearance rates in the population were associated with severe spring snowstorms, and vacated territories were resettled by new males (Wolfenbarger, unpublished data). If disappearance rates indicated mortality, then survivorship between years was low, suggesting that breeding experience may be of limited importance in explaining the variance in reproductive success measured here. Other data indicate that breeding experience is not related to parental care performance of male cardinals, one possible way breeding experience could influence reproductive success. Male provisioning of nestlings but not active nest defense represented a significant investment of paternal care (Filliater and Breitwisch, 1997; Nealen and Breitwisch, 1997), and male provisioning of nestlings was not related to breeding experience (Linville et. al., 1998). However, the effects of breeding experience and age on reproductive success in cardinals remain untested.

Wolfenbarger Red coloration, mate quality, and territory quality 87 Contribution of mate quality and territory quality to reproductive success The association between degree of ornamentation and reproductive success described here conforms to the general pattern seen in other species (Andersson, 1994). Less attention has been paid to the relative importance of the factors that contribute to variance in male reproductive success. I examined two hypotheses to explain how males with redder plumage accrued this benefit: (1) redder males were paired with earlier breeding females; and (2) redder males defended higher quality territories. The results of this study provide support for both hypotheses. Redder males were paired with earlier breeding females. On my study site, early breeding pairs consistently produced more offspring per season than later breeding pairs (Figure 4), supporting the hypothesis that mate quality is an important determinant of reproductive success. Second, redder males obtained territories with higher foliage density at 2 m, a height just above the average nest height in the population (Figure 3). In 2 of the 3 years, pairs on territories with denser vegetation at this height produced significantly more offspring than those on territories with lower foliage density at 2 m (Figure 5), corroborating previous evidence that vegetation density is a measure of territory quality (Conner et al., 1986). Among birds, there is a well-established relationship between earlier breeding and reproductive success (Price et al., 1988). Variation among females in date of first egg-laying has been linked to aspects of quality, such as better nutritional condition (Arcese and Smith, 1988; Meijer et al., 1990), age (Perrins, 1970), and breeding experience (Davis, 1976; Forslund and Larsson, 1992; Pärt, 1995; Smith, 1993). In addition to female quality, environmental influences may also contribute to earlier breeders having higher reproductive success (see Winkler and Allen, 1996). However, in cardinals, this relationship most likely arises from differences in mate quality rather than from differences related to environmental factors, such as predation rates or food abundance. There was no evidence that nest predation occurred more frequently later in the season, and it is unlikely that food was most abundant early in the nesting season. Cardinals on my study site began nesting in late April, when abundances of insects, fruits, and seeds were relatively low due to continuing cold weather (Wolfenbarger, personal observation). The relationship between earlier breeding and cumulative reproductive success over a breeding season is stronger than that between breeding date and success of first nests (Figure 4) because earlier breeding pairs had more successful clutches of young than later breeding pairs. The longer breeding season available to early breeding females may lead to higher reproductive success by increasing the number of nesting attempts possible, generally an under appreciated component of reproductive success (but see Filliater et al., 1994; Verhulst et al., 1997). The ability of earlier breeding females to produce more successful broods per season emphasizes the benefits to redder males of being paired with females capable of initiating breeding early in the season. Predation was the single largest cause of nest failure in this population, a result that corroborates other research on cardinals (Filliater et al., 1994). The importance of nest predation may explain why vegetation density is a good measure of territory quality in cardinals. Average nest height in the population is 1.7 m, and only vegetation density at 2 m was correlated with reproductive success; therefore, vegetation density near the average nest height may reduce the probability of nest predation. Variation in predation rates among years also supports this hypothesis. In 1992 and 1993, the 2 years with higher predation rates, denser territories did produce more offspring (Figure 5a). In contrast, in 1994, a year with a relatively lower predation rate, nearly all territories produced offspring, and there was no significant correlation between territory density and reproductive success (Figure 5b). The importance of vegetation density for predation rate most likely depends on the type and abundance of nest predators. While dense vegetation in the cardinal s habitat was associated with higher nest success, dense vegetation in the habitat of the closely related black-headed grosbeak (Pheucticus melanocephalus) was associated with more nest predators and lower nest success (Hill, 1988). As a result, younger, less brightly colored black-headed grosbeaks established territories in dense habitat, the opposite of what one would predict for northern cardinals. Although mate quality and territory quality individually explain higher reproductive success among redder males, these two factors may interact to explain variance in male reproductive success. Vegetation density could potentially influence breeding date if territories with denser foliage also contain more food resources before breeding that increase the nutritional condition of females. However, the partial correlation coefficient revealed that redder males were paired with earlier breeding females independent of the effects of vegetation density at 2 m. Likewise, regressing first initiation date and vegetation density on reproductive success showed that earlier breeding pairs had higher seasonal reproductive success regardless of the relationship between vegetation density and breeding date. The partial regression coefficient for vegetation density was positive but not significant. Although the small sample size limits the power of this analysis, the result does indicate that vegetation density is relatively less important than is first nest initiation date in explaining variance in reproductive success, indicating the importance of mate quality to male reproductive success. I have argued here that coloration affects territory acquisition, but the statistical relationships, by themselves, do not eliminate the possibility that territories influence male coloration. However, this is unlikely. Cardinals undergo a single molt after the breeding season, when they no longer actively defend territories at my study site (Wolfenbarger, personal observation). The amount of carotenoid-based food obtained during the time of molting is known to affect coloration in other species with carotenoid-based coloration (Brush and Power, 1976; Hill, 1992, 1993b; Witmer, 1996). Only if carotenoids obtained during the breeding season were stored until molt would the previous year s territory influence color. Finally, any effect of territories on color in the subsequent year will be of limited importance in populations, such as the one I studied; few males bred in more than 1 year, and those that did moved territories between years. The data from this study, combined with data from other studies, suggest that the higher reproductive success of redder male cardinals is most influenced by mate quality. In other species where highly ornamented males pair with earlier breeding females, mate acquisition of both social and extrapair mates as well as female quality have been shown to create variance in male reproductive success, but the relative importance of female quality is unclear (Hill, 1990, 1993a; Hill et al., 1994; Møller, 1988, 1991; O Donald, 1980). In this study, finding mates did not appear to be a constraint on male cardinals. All males included in the present analysis had a single social mate. Despite extensive observations during the prebreeding and breeding period, no unpaired males were observed in this population. Redder males may also have paired more quickly, as I did not systematically census for both males and females on territories but rather concentrated on males. Pairing occurs 2 4 months before nesting (Kinser, 1973; Wolfenbarger, personal observation), and such early pairing in

88 Behavioral Ecology Vol. 10 No. 1 cardinals may be due to competition for mates and territories as occurs in pied flycatchers (Ficedula hypoleuca; Slagsvold and Lifjeld, 1994). Monogamous males can potentially add significantly to their reproductive success through extrapair paternity (Webster et al., 1995; Westneat et al., 1990). However, results from socially monogamous species to date do not reveal any consistent relationship between male color and extrapair fertilizations (Hill et al., 1994; Rätti et al., 1995; Sundberg and Dixon, 1996; Yezerinac and Weatherhead, 1997). The relatively low rate of extrapair fertilizations in cardinals provides a limited role for extrapair fertilizations as a source of variance in reproductive success. For example, in northern cardinals only 13.5% of nestlings were from extrapair fertilizations (Ritchison et al., 1994), compared to a 35% rate in the indigo bunting, Cyanea passerina, a close relative that is also socially monogamous (Westneat, 1990). Furthermore, duller male cardinals do not have higher rates of cuckoldry than brighter males (Linville et al., 1998). These data alone do not eliminate the possibility that brighter males may have higher extrapair paternity rates than duller males (i.e., brighter males may gain more extrapair fertilization than they lose in their nest). Rather than investing in extrapair copulations, male cardinals may benefit more from caring for within-pair offspring and maximizing the number of broods in a season (Linville et al., 1998; Ritchison et al., 1994). The results of this study present a complicated picture for future work on mate choice in northern cardinals. The importance of red coloration in this species presents the possibility that females are choosing males based on color, as demonstrated in other species (Bakker and Mundwiler, 1994; Endler and Houde, 1995; Hill, 1990; Sætre et al., 1994; Sundberg, 1995; reviewed by Andersson, 1994). However, laboratory experiments using both unmanipulated and manipulated plumage of males have revealed no evidence for female mating preferences for redder males (Wolfenbarger, 1996). The prolonged period of interactions before breeding ( January through May) may limit the reliance of females on simple indicators of male quality, relative to species such as barn swallows where pairing occurs within days of arrival on the breeding grounds (Møller, 1990). At the same time, territory quality is a direct correlate of nesting success in cardinals (this study; Conner et al., 1986). Because pair formation may occur after males have established breeding territories (Kinser 1973), females may choose mates based on differences in territory quality, as has been suggested for polygynous species (Searcy and Yasukawa, 1995; Turner and McCarty, 1998). This may be the case in some monogamous species, such as arctic skuas (Stercorarius parasiticus), where the relationship between male color and mating success is complicated by the effects of territory quality (Davis and O Donald, 1976). Alternatively, females may use a combination of male and territory characteristics. In pied flycatchers both male coloration and nestsite quality influence female mate choice, but the relative importance of these is not well understood (Alatalo et al., 1986; Dale and Slagsvold, 1996; Sætre et al., 1994; but see also Norris, 1990). Finally, the relationship between coloration and territory quality suggests a role for male male competition in the cardinal s mating system. Red plumage may function as an honest signal of fighting ability in the acquisition and defense of territories, as has been found for coloration in other species (Røskaft and Rohwer, 1987; Studd and Robertson, 1985a; but see Belthoff et al., 1994; Lozano and Lemon, 1996). Consistent with these studies, laboratory experiments have revealed that males with redder (unmanipulated) coloration are more dominant (Wolfenbarger, 1996), but whether red coloration per se causes males to acquire higher quality territories remains an untested question. Based on this study, mate quality emerges as the primary factor generating variance in male reproductive success and maintaining the red plumage coloration of male cardinals. Territory quality may directly influence male reproductive success or may also contribute to a male s ability to pair with a high-quality female. The operation of sexual selection in monogamous species that defend large, multipurpose territories is likely to be more complicated due to the interaction between multiple factors, such as mate quality and territory quality. Stephen T. Emlen, John P. McCarty, and Paul W. Sherman provided valuable criticism and encouragement from the inception through the publication of this research. I benefited from discussions with D. Able, S. Birks, G. Borgia, R. Cocroft, C. Freeman-Gallant, S. Kain, and C. Murphy. I received expert field assistance from C. Foster, N. Gonzalez, and A. Washburn. I thank R. Breitwisch, S. Emlen, G. Hill, J. McCarty, P. Sherman, T. Slagsvold, and D. Winkler for comments on a previous version of the manuscript. This work was supported by the Walter E. Benning Fund of the Cornell Laboratory of Ornithology, by Grants-in-Aid-of-Research from Sigma Xi, by the Frank M. Chapman Memorial Fund of the American Museum of Natural History, and by a Dissertation Improvement Grant (IBN-9321801) from the National Science Foundation. REFERENCES Alatalo RV, Glynn C, Lundberg A, 1986. Female pied flycatchers choose territory quality and not male characteristics. Nature 323: 152 153. Andersson MB, 1994. Sexual selection. Princeton, New Jersey: Princeton University Press. Arcese P, Smith JNM, 1988. Effects of population density and supplemental food on reproduction in song sparrows. J Anim Ecol 57: 119 136. Bakker TCM, Mundwiler B, 1994. Female mate choice and male red coloration in a natural three-spined stickleback (Gasterosteus aculeatus) population. Behav Ecol 5:74 80. Belthoff JR, Duffy AM Jr, Gauthreaux SA Jr, 1994. Plumage variation, plasma steroids and social dominance in male house finches. Condor 96:614 625. Bent AC, 1968. Life histories of North American cardinals, grosbeaks, buntings, towhees, finches, sparrows and allies. Part I. New York: Dover Publications. Best LB, Stauffer DF, 1980. Factors affecting nesting success in riparian bird communities. Condor 82:149 158. Birkhead TR, Møller AP, 1992. Sperm competition in birds: evolutionary causes and consequences. London: Academic Press. Brush, AH, Power, DM, 1976. House finch pigmentation: carotenoid metabolism and the effect of diet. Auk 93:725 739. Burley N, Krantzberg G, Radman P, 1982. Influence of colour-banding on the conspecific preferences of zebra finches. Anim Behav 30: 444 455. Conner RN, Anderson ME, Dickson JG, 1986. Relationships among territory size, habitat, song, and nesting success of northern cardinals. Auk 103:23 31. Conover WJ, 1980. Practical non-parametric statistics, 2nd ed. New York: John Wiley and Sons, Inc. Dale S, Slagsvold, T, 1996. Mate choice on multiple cues, decision rules and sampling strategies in female pied flycatchers. Behaviour 133:903 944. Darwin C, [1871] 1981. The descent of man, and selection in relation to sex. Princeton, New Jersey: Princeton University Press. Davis JWF, 1976. Breeding success and experience in the Arctic skua, Stercorarius parasiticus (L.). J Anim Ecol 45:531 535. Davis JWF, O Donald P. 1976. Territory size, breeding time and mating preference in the Arctic skua. Nature 260:774 775. Emlen ST, Oring L, 1977. Ecology, sexual selection, and the evolution of mating systems. Science 197:215 223. Endler JA, Houde AE, 1995. Geographic variation in female preferences for male traits in Poecilia reticulata. Evolution 49:456 468.