Context-Dependent Effects of Castration and Testosterone Treatment on Song in Male European Starlings

Hormones and Behavior 42, 307–318 (2002) doi:10.1006/hbeh.2002.1824

Context-Dependent Effects of Castration and Testosterone Treatment on Song in Male European Starlings Rianne Pinxten,* ,1 Elke De Ridder,* Jacques Balthazart,† and Marcel Eens* *Department of Biology, University of Antwerp, U.I.A., B-2610 Wilrijk, Belgium; and †Research Group in Behavioral Neuroendocrinology, Center for Cellular and Molecular Neurobiology, University of Lie`ge, B-4020 Lie`ge, Belgium Received July 26, 2001, revised February 21, 2002, accepted March 29, 2002

Most seasonally breeding songbirds display dramatic seasonal fluctuations in plasma testosterone (T) levels and mate attraction behaviors, including song. However, males of some songbird species, such as the European starling (Sturnus vulgaris), continue to sing at high levels after the breeding season, when T levels are basal. In male starlings song during the breeding season functions mainly to attract mates, whereas song during the nonbreeding season appears unrelated to reproduction. This suggests that song produced in a context unrelated to female courtship, unlike song directed toward females, is not regulated by plasma T. In captive males housed in large outdoor aviaries we explored the relationship between plasma T and song produced during the breeding season within and outside a courtship context. This was achieved by determining the effects of castration and subsequent T treatment on song and mate attraction behaviors in both the presence and the absence of a female. Compared to intact males, castrated males did not show reduced song activity in the absence of a female for at least 6 months after the operation, strongly suggesting that the expression of noncourtship song is not regulated by plasma T. Likewise, we found that experimentally elevating T levels in castrated males did not affect noncourtship song rates. However, control castrated males receiving empty implants tended to show reduced noncourtship song rates after implantation. This may have been due to a suppressive effect caused by the presence of the T-implanted castrated males in the same aviary. In contrast, courtship singing was clearly controlled by plasma T: it was

1

To whom correspondence and reprint requests should be addressed at Department of Biology, University of Antwerp, U.I.A., Universiteitsplein 1, B-2610 Wilrijk, Belgium. Fax: 00 32 3 820 22 71. E-mail: [email protected].

0018-506X/02 $35.00 © 2002 Elsevier Science (USA) All rights reserved.

abolished by castration and restored by subsequent T replacement when males were housed both individually and in a group situation. High plasma levels of T also appeared necessary for the activation of three other behavioral traits critical for mate attraction, namely, nesthole occupancy, spending time (singing) in a nesthole, and carrying green nesting material into a nesthole. © 2002 Elsevier Science (USA)

Key Words: castration; testosterone; noncourtship song; courtship song; context dependence of song; Sturnus vulgaris.

The song of passerine birds is generally believed to be under the control of gonadal hormones. In numerous species males produce high levels of song during the breeding season, when gonads are active and plasma testosterone (T) levels are high (Nottebohm, Nottebohm, Crane, and Wingfield, 1987; Wingfield and Farner, 1993). Song produced during the breeding season has been shown to play an important role in mate attraction and/or territorial defense (Kroodsma and Beyers, 1991). Regression of gonadal activity in the winter nonbreeding season is usually accompanied by a strong decrease in or a cessation of singing and/or regression to subsong. In most species, singing is diminished or eliminated by castration and is reinstated by subsequent T treatment (Arnold, 1975; Balthazart, 1983; Heid, Guttinger, and Prove, 1985; Harding, Walters, Collado, and Sheridan, 1988; Dloniak and Deviche, 2001). However, in some songbird species, such as the European starling Sturnus vulgaris, song is also robustly expressed during the nonbreeding season, when T levels are basal (Eens, 1997). Male song during the nonbreeding season is thought to play

307

308

a role in flock cohesion (Eens, 1997) or in the maintenance of dominance hierarchies (Wiley, Archawaranon, and Thompson, 1993). Hence, these observations suggest that in starlings, and in other species in which males sing throughout the year, the androgenic stimulation of song might interact with the context in which song is produced (see also Walters, Collado, and Harding, 1991; Burmeister and Wilczynski, 2001). For example, song directed toward a female conspecific (courtship song) may be dependent on plasma T, whereas noncourtship song may not be plasma-Tdependent. The European starling is a facultatively polygynous hole-nesting songbird (Pinxten, Eens, and Verheyen, 1989; Pinxten and Eens, 1990) in which male song produced during the breeding season functions mainly to attract mates and to stimulate females to solicit copulations (Eens and Pinxten, 1990, 1995; Eens, Pinxten, and Verheyen, 1990, 1991, 1993; Pinxten and Eens, 1997, 1998). When a potential mate is present, unpaired males fly into the nesthole, where they begin singing as an invitation for the female to enter the nesthole and they also often carry green nest materials into the nest (Eens et al., 1993; Gwinner, 1997). As these courtship behaviors are nesthole-oriented, it is obvious that the occupancy of a nesthole is the most important initial step for mating. Although numerous laboratory experiments have established the role of androgens in the control of reproductive aggression and thus possibly in the maintenance/defense of a territory in birds, so far relatively few experimental manipulations seem to have been done to investigate whether plasma T is indeed a causal factor in the establishment of a territory (Balthazart, 1983; Beletsky, Gori, Freeman, and Wingfield, 1995). In male starlings, T levels have been shown to peak during the spring breeding season when males are defending nestholes (Ball and Wingfield, 1987; Riters, Eens, Pinxten, and Ball, 2002), suggesting that T influences the type of aggression involved in territorial contests. However, as in most other songbird species, it is not clear which role T plays in all phases of territory establishment and maintenance. As male starlings only defend a small territory around the nesthole and do not maintain a feeding territory, and easily breed under seminatural conditions in captivity (see Eens, 1997), they are ideal subjects for experimental hormonal manipulations. In this study we examined whether within the spring breeding season the androgenic stimulation of song interacts with contextual variables by exploring in captive males the relationship between plasma T and song

Pinxten et al.

produced respectively within and outside a courtship context. We also investigated whether T is a causal factor in the establishment of a territory (nestbox) and whether T affects nestbox-oriented courtship behaviors (see above). This was all achieved by determining the effects of castration and subsequent T treatment on the occurrence of song produced in the absence of a female conspecific (noncourtship song) and on the occurrence of courtship behaviors (including courtship song) in the presence of a female. In previous experiments, we have shown that in a captive situation, unpaired reproductively active males respond to the introduction of a female as they would to a potential mate in the wild by significantly increasing their song rate (Eens et al., 1990, 1993).

METHODS Origin and Maintenance of Starlings Thirty-five free-living male starlings were captured at several sites around Antwerp, Belgium, during the first week of October. Thirteen of these 35 males were adults (in at least their second calendar year, but their precise age was unknown), while the other 22 were, judging from their juvenile plumage characteristics, young birds (yearlings for the rest of the paper) hatched during the preceding breeding season (and thus between 4 and 5 months old). Birds were randomly assigned to treatment groups. Males that were going to be castrated (15 yearlings and 7 adults) and control males (7 yearlings and 6 adults) that were going to be sham-operated, were then housed in two separate outdoor aviaries that were visually and acoustically isolated from one another. On October 9, males were bilaterally castrated or sham-operated while under Hypnodil anesthesia (Janssen Pharmaceutica, Beerse, Belgium, 15 mg/kg; see also Pinxten, De Ridder, Balthazart, Berghman, and Eens, 2000). All males had fully regressed testes at that time. From October 9 onward the males were housed in three separate outdoor aviaries. Castrated males were housed in two different aviaries, with 11 castrated males (3 adults and 8 yearlings) in aviary 1 (L ⫻ W ⫻ H, 5 ⫻ 2 ⫻ 2.5 m) and the remaining 11 castrated males (4 adults and 7 yearlings) in aviary 2 (6 ⫻ 3 ⫻ 2.5 m). The aviaries were visually isolated from each other. All 13 sham-operated males were housed in aviary 3 (6 ⫻ 4 ⫻ 2.5 m), which was visually and acoustically isolated from aviaries 1 and 2. All three aviaries contained eight nestboxes. A 20-cm perch be-

309

Context-Dependent Effects of Testosterone on Song

low the nesthole allowed the birds to sit in front of the nestbox. Nest materials were available in excess. Food (Orlux pate´ ) and water were provided ad libitum. Behavioral observations started March 11 and ended April 30. During the winter period four intact males (1 adult and 3 yearlings) and five castrated yearlings (four from aviary 1 and one from aviary 2) died, and one adult castrated male was added to aviary 2. Table 1 gives an overview of the number of males of each age/treatment category that were still available in the three aviaries when behavioral observations started. Note that in aviary 2, unlike in aviary 1, there were more males than available nestboxes. However, because only T-treated males occupied nestboxes (see Results) and there were always more nestboxes in the aviary than the number of T-treated males, there was no indication that this resulted in consistent differences in the behaviors of the birds housed in these two aviaries. On April 2, males were given implants filled with crystalline testosterone or empty implants (see below). During the breeding season, the bill color of each male was checked regularly. In starlings, the bill remains black in the absence of T and turns yellow soon after subjects are exposed to levels of T only slightly above baseline levels (Ball and Wingfield, 1987). Testosterone Implantations On April 2 nine castrated males received subcutaneous testosterone implants in the neck region under local anesthesia (Xylocaine), while the other eight remaining castrated males and the nine control males received empty tubes (Table 1). We packed 12-mmlong capsules of Silastic tubing (Down Corning, i.d., 1.47 mm; o.d., 1.96 mm) with crystalline testosterone. This dose has previously been demonstrated to result in T concentrations below the maximal levels determined in both free-living and captive male starlings (about 2.5 to 3 ng/ml: Dawson, 1983; Ball and Wingfield, 1987; Gwinner, Gwinner, and Dittami, 1987). Because preliminary statistical analyses revealed that in the period before T implantation the song activity of castrated males in aviary 1 was significantly higher than the song activity of castrated males in aviary 2, we decided to implant a comparable proportion of castrated males from both aviaries with T instead of implanting all castrated males from one aviary and treating the castrated males housed in the other aviary as controls (see Table 1). Throughout the paper castrated males are referred to as Cx males, castrated males that later received testosterone as Cx ⫹ T males, and control males as Ctrl males.

The Ethical Advisory Committee of the University of Antwerp approved the experimental procedure, following Belgian and Flemish laws concerning the protection of animal welfare. Behavioral Observations Group observations. The behavior of the males in each of the three aviaries was observed both before and after T implantation. Initially, the order in which the three aviaries were observed each day was randomized. For subsequent observations, we used a rotating schedule. Prior to T implantation, the birds were observed for 30 min in the morning on 9 different days between March 11 and 24. After T implantation (April 2), the birds were observed for 45 min in the morning on April 15, 16, and 17. Birds were always observed between 0800 and 1200 h. On March 13 and 20 (before T implantation) and on April 15–17 (after T implantation) the behavior of the males was recorded for 30 – 45 min after a female was introduced into each of the three aviaries. On April 15–17 when, due to time constraints, males were observed on the same day both without and with a female present in the aviary (see above), they were first observed in the absence of a female and then after introduction of a female. We used seven stimulus females, which were housed together in one aviary that was visually separated from the aviaries housing males. Each observation day we captured a different female which was then subsequently introduced in all three male aviaries. Hence, all male groups were confronted with the same female on a particular observation day. After the response of the males was recorded, the female was quickly (within 1 min) captured and removed from the aviary. During each observation session, an account of each male’s behavior was recorded on cassette tape and later transcribed. We recorded which male occupied one or more nestboxes. A male was considered to have occupied a particular nestbox when he frequently (at least five times during 2 successive days) visited this nestbox, sang in this nestbox or on the perch in front of it, and defended it against other males (see Pinxten et al., 1989). Nestbox occupancy is very obvious in male starlings and could always be clearly observed. The activity and location of each male were recorded every minute at the signal of a timer, whereas the carrying of green nesting material into the nestbox was described continuously. Before each observation period, we always introduced green nesting material in the aviaries. For statistical analyses, we later calculated the

310

Pinxten et al.

TABLE 1 Treatment and Age of the Males Housed in Each of the Three Large Outdoor Aviaries during Behavioral Observations Aviary 1

Cx males* Cx ⫹ T males Ctrl males

Aviary 2

Aviary 3

Adult

Yearling

Total

Adult

Yearling

Total

Adult

Yearling

Total

1 2 0

2 2 0

3 4 0

3 2 0

3 3 0

6 5 0

0 0 5

0 0 4

0 0 9

Note. Birds were implanted on April 2. Cx males are castrated males, Cx ⫹ T males are castrated males that later received T-filled implants, and Ctrl males are control intact males. The behavior of the males was observed March 11–24 before implantation and April 15–17 after implantation. *One yearling Cx male from aviary 1 escaped 2 days after implantation and 1 adult Cx male from aviary 2 escaped 1 day before implantation, implying that their behavior could not be observed in the period after implantation. All of the other 7 Cx males and 9 Cx ⫹ T males and 7 of the 9 Ctrl males (excluding 1 adult and 1 yearling Ctrl male) were also tested individually in a small aviary in the period before (March 25–April 2) and after April 18 –April 30) T implantation.

number of times per hour each male carried green nesting material into a nestbox. The proportion of time spent singing and the number of times per hour each male carried green nesting material in a nestbox were recorded both without and with a female present in the aviary. In a captive situation, unmated reproductively active male starlings possessing a nestbox respond to the introduction of a female by flying to their nestbox and spending a high proportion of the time in the nestbox (while they are usually singing) in an attempt to attract the female to the nestbox (Eens et al., 1990, 1993). The latter behavior occurs at a much lower rate when no females are present. Therefore, the proportion of time a male spent in the nestbox was analyzed only during the observations sessions when a female was present (free flying) in the aviary. Isolated birds. When males are housed in a group, it is possible that the song activity of one male stimulates the song activity of other males or that the song activity of subdominant males is suppressed by the presence of dominant males. Therefore, 23 of the males that were tested in a group situation (7 Cx males, 9 Cx ⫹ T males, and 7 Ctrl males, see Table 1) were also tested individually for 30 min in a small outdoor aviary both before and after T implantation. These males were always tested individually after behavioral observations in the group situation were terminated. Each of the three identical small aviaries (1.75 ⫻ 1.35 ⫻ 2 m, L ⫻ W ⫻ H) that were used for this individual testing contained a single nestbox with a perch and a few branches to perch on. As males usually show only low frequencies of song or no song at all and very rarely mate attraction behaviors in such a small aviary when no female is present (see Eens et al., 1993), the behavior of the males was recorded only

after introduction of a female in the aviary. Males were tested individually a first time between March 25 and April 2 and for the second time between April 18 and April 30 (after T implantation). During most observation days, we tested three males: one Cx male, one Cx ⫹ T male, and one Ctrl male of the same age class (yearling or adult). Males were always tested for 30 min between 0800 and 1030 h. Before each observation period we introduced green nesting materials into the aviaries. The order in which each of the three males was tested was randomized. Every 30 s, we recorded the activity and location of the male. A different stimulus female was used each observation day, both before and after T implantation. Before males were tested individually, they were removed from the large outdoor aviary and housed in the test aviary for 1 or 2 days to get them familiarized with the small aviary. Testosterone Assays To compare T levels among the three experimental groups both before and after T implantation, we collected blood samples from the males just before they were tested individually. Blood samples were collected when males were removed from the large outdoor aviary to be housed in the small test aviaries (see above). Blood samples (250 – 400 ␮l) were taken from the brachial vein into 75-␮l heparinized hematocrit capillary tubes within 2 min of removal from the large outdoor aviary, transferred into Eppendorf tubes, and centrifuged at 7000 rpm for 15 min within 3 h, and the plasma was stored at ⫺70°C. T levels were determined from the plasma using an 125 iodine double-antibody kit purchased from ICN Biomedicals, Inc. (Costa

311

Context-Dependent Effects of Testosterone on Song

Mesa, CA). The primary antibody used in this assay does not cross-react significantly with other androgens beside T (5␣-DHT, 3.4%; 5␣-androstane-3␤, 17␤diol, 2.2%; 11-oxotestosterone, 2%; all other steroids, ⬍ 1%). Briefly, in order to avoid interference from plasma proteins, 100 ␮l of plasma samples was extracted with 400 ␮l of a 50:50 mixture of cyclohexane: ethylacetate, of which 300 ␮l was then evaporated by vacuum centrifugation. The dried samples were then diluted with PBS containing 0.1% BSA and further treated as recommended by the manual provided by the manufacturer. In some cases only between 40 and 90 ␮l of plasma was extracted. Testosterone standards ranged from 9.7 pg/ml to 10 ng/ml, yielding an effective working interval between 1.73 pg/ml and 1.73 ng/ml, owing to the concentration effect of the extraction procedure (plasma samples were extracted while standards were not; see also Eens, Van Duyse, Berghman, and Pinxten, 2000). All samples were measured in one assay; intra-assay variation was 6.6%. This assay has been validated for use in starlings (see Duffy, Bentley, Drazen, and Ball, 2000). The samples of three Cx ⫹ T birds and one Ctrl bird after T implantation contained T levels above the maximum detection limit (1.73 ng/ml). For statistical purposes these samples were assigned a value equivalent to this detection limit. Statistical Analyses Data were analyzed using the statistical software programs StatXact-Turbo (Mehta and Patel, 1995), SPSS/PC (SPSS, 1986), and SAS (SAS, 1988), following procedures outlined in Siegel (1956) and Sokal and Rohlf (1981). SAS procedures were used to carry out three-way ANOVAs (mixed ANOVA model) with the Time (before and after T implantation) as a repeated factor and the Groups (Cx males, Cx ⫹ T males, Ctrl males) and Age (yearling, older) as independent factors. Because repeated observations were made on single individuals, residual values may be correlated. We therefore explicitly modeled these potential correlations and compared five types of correlation structure (i.e., no correlation, compound symmetry, serial autocorrelation, Toeplitz, and unstructured) by likelihood ratio test. After selecting the “most likely” correlation structure the number of degrees of freedom of the F tests of the fixed effects were adjusted using sattertwaite formulas (see Littell, Milliken, Stroup, and Wolfinger, 1996, for more details). We calculated average values for each period for each individual bird to avoid problems with pseudoreplication. The model

first tested whether there were significant three-way or two-way interaction effects or significant main effects. In this experiment, no significant effects of Age were found, indicating that age differences were of minor importance. We were particularly interested in the Time ⫻ Group interaction effect. If this interaction effect was significant, the model analyzed the effect of experimental group for each period separately using the t statistic adapted for posthoc testing. As this analysis is done within the same SAS model, degrees of freedom differ from degrees of freedom used in standard t tests (see SAS, 1988, procedure). P values were adjusted (P a) for multiple testing using sequential Bonferroni corrections (Rice, 1989). The model also compared the effect of Time for each group separately using the t statistic. For conciseness (1) main effects are usually not shown when the Time ⫻ Group interaction was significant and (2) results of comparisons that did not reach statistical significance are usually not shown. To satisfy the assumptions required for parametric statistics (normally distributed data, homogeneity of variance) proportions were subjected to arcsine square root transformation. When transformation did not result in normally distributed data (as determined by the Shapiro–Wilks test, W ⬍ 0.9), we used nonparametric tests. Kruskal–Wallis tests, followed by Dunn’s multiple comparisons tests (posthoc tests) when appropriate, were used to compare behaviors among experimental groups and Wilcoxon matchedpairs signed-ranks tests were used to compare behaviors between the two time periods for each group. Two-tailed statistics were used with ␣ ⫽ 0.05. Values presented in the text and figures are means ⫾ SE.

RESULTS Testosterone Levels In all Cx males the bill remained completely black throughout the breeding season, while the bills of Cx ⫹ T males, which were completely black before T implantation, became completely yellow afterward. These results indicate that the castrations and T treatments had been fully successful. Before T implantation, Ctrl males had significantly higher T levels than castrated males (combining Cx males and Cx ⫹ T males which had comparable levels; t test, P ⫽ 0.013; Fig. 1). After T implantation. T levels differed significantly among the three experimental groups (one-way ANOVA, F (2,20) ⫽ 18.60, P ⬍ 0.0001): Cx ⫹ T males had significantly higher T levels than both Cx and Ctrl

312

Pinxten et al.

of 9; Fisher’s exact test, P a ⫽ 0.26). After T implantation, the proportion of males singing did not differ significantly among experimental groups (5 of the 9 Ctrl males, 4 of the 7 Cx males, 9 of the 9 Cx ⫹ T males). However, analysis of average song rates identified a highly significant Time ⫻ Group interaction effect (F (2,24) ⫽ 23.81, P ⫽ 0.0001). Before T implantation Ctrl males sang about 25% of the time and had a

FIG. 1. Plasma testosterone (T) levels (mean ⫾ SEM) in control intact males (Ctrl males), castrated males (Cx males), and castrated males receiving T-filled implants (Cx ⫹ T males) in the period before and after T implantation. Numbers under the bars denote sample sizes. Differences between male groups within each period are indicated by different lowercase letters above the bars, P ⬍ 0.05.

males, while Ctrl males had significantly higher levels than Cx males (Fig. 1). Song Activity Without a female present. The proportion of males singing during at least one of the observation sessions without a female present in the aviary did not differ significantly among the three experimental groups, either prior to (8 of 9 Cx males, 9 of 9 Cx ⫹ T males, 8 of 9 Ctrl males) or after (5 of 7 Cx males, 9 of 9 Cx ⫹ T males, 8 of 9 Ctrl males; Fisher’s exact tests, NS) T implantation. Before implantation, noncourtship song rates clearly did not differ among experimental groups (Fig. 2a). However, after implantation, song rates differed among groups, being lowest in control castrated males, intermediate in control intact males, and highest in T-implanted castrated males (see Fig. 2a). A three-way ANOVA revealed a significant Time effect (F (1,23) ⫽ 6.12, P ⫽ 0.02), but no significant Group effect (F (2,22) ⫽ 2,23, P ⫽ 0.13) and no Time ⫻ Group interaction (F (2,23) ⫽ 2.49, P ⫽ 0.08), even if the interaction tended to be significant. With a female present. Before T implantation the proportion of Ctrl males (8 of 9) singing during at least one of the observation sessions when a female was present was significantly higher than the proportion of Cx males (2 of 9: Fisher’s exact test, P a ⫽ 0.016), but did not differ from the proportion of Cx ⫹ T males (4

FIG. 2. Proportion of time spent singing (mean ⫾ SEM) by control intact males (Ctrl males), castrated males (Cx males), and castrated males receiving T-filled implants (Cx ⫹ T males) in the period before and after T implantation: (a) housed in a group without a female present, (b) housed in a group with a female present, and (c) individually housed with a female present. Numbers under the bars denote sample sizes. Differences between male groups within each period are indicated by different lowercase letters above the bars, P ⬍ 0.05.

Context-Dependent Effects of Testosterone on Song

significantly higher song rate than Cx males and Cx ⫹ T males, which both sang only about 3% of the time (Fig. 2b). After T implantation, Cx ⫹ T males sang significantly more than both Ctrl males and Cx males, while Ctrl males sang significantly more than Cx males (Fig. 2b). Comparing both periods, the song activity of Cx males did not differ significantly between periods (P ⫽ 0.9), while after implantation Cx ⫹ T males sang significantly more (t ⫽ ⫺4.37, df ⫽ 23, P ⫽ 0.0002) and Ctrl males sang significantly less (t ⫽ 5.37, df ⫽ 23, P ⫽ 0.0001). Before T implantation, the proportion of Ctrl males singing (7 of 7) during tests performed when birds were individually housed was significantly higher than both the proportion of Cx males (1 of 7) and Cx ⫹ T males (0 of 9; Fisher’s exact test, P a ⬍ 0.03 in both cases). After T implantation, the proportions of both Ctrl males (7 of the 7) and Cx ⫹ T males (7 of the 9) singing were significantly higher than the proportion of Cx males singing (none of the 7; Fisher’s exact test, P a ⬍ 0.01 in both cases). A Kruskal–Wallis test revealed that in the periods both before (␹ 2 ⫽ 8.051, P ⫽ 0.017) and after (␹ 2 ⫽ 12.88, P a ⫽ 0.0016) implantation, song rates differed significantly among experimental groups (Fig. 2c). Before implantation Ctrl males, which sang on average only about 5% of the time, sang significantly more than Cx ⫹ T males (Fig. 2c) and tended to sing more than Cx males (P ⫽0.07). After implantation both Cx ⫹ T and Ctrl males sang significantly more than Cx males, which did not show any song activity (Fig. 2c). Cx ⫹ T males sang significantly more after implantation than before (Wilcoxon test; z ⫽ ⫺2.52, P ⫽ 0.01), while this was not the case for Cx males (z ⫽ ⫺1, P ⫽ 0.32) and Ctrl males (z ⫽ ⫺1.35, P ⫽ 0.17). Nestbox Occupancy Before implantation, none of the 18 castrated males (combining Cx and Cx ⫹ T males) occupied a nestbox, compared to 5 of the 9 Ctrl males (Fisher’s exact test, P ⫽ 0.0016). After implantation, the proportion of Cx males occupying a nestbox (0 of 7) was significantly lower than the proportion of Cx ⫹ T males (7 of 9; Fisher’s exact test, P a ⫽ 0.009) and tended to be lower than the proportion of Ctrl males (still 5 of 9; Fisher’s exact test, P a ⫽ 0.06). Time Spent Inside a Nestbox The proportion of time males spent (singing) inside a nestbox when a female was present differed signif-

313

FIG. 3. Proportion of time spent in a nestbox (mean ⫾ SEM) by Ctrl males, Cx males, and Cx ⫹ T males in the period before and after T implantation: (a) housed in a group with a female present and (b) individually housed with a female present. Numbers under the bars denote sample sizes. Differences between male groups within each period are indicated by different lowercase letters above the bars, P ⬍ 0.05.

icantly among groups, both before (Kruskal–Wallis test, ␹ 2 ⫽ 11.66, P a ⫽ 0.006) and after (Kruskal–Wallis test, ␹ 2 ⫽ 10.48, P ⫽ 0.005) T implantation. Before T implantation Ctrl males spent significantly more time in a nestbox than Cx and Cx ⫹ T males, neither of which ever entered a nestbox (Fig. 3a). After T implantation, both Cx ⫹ T and Ctrl males spent significantly more time in a nestbox than Cx males. Cx ⫹ T males spent significantly more time inside a nestbox after T implantation than before (Wilcoxon test, z ⫽ ⫺2.66, P ⫽ 0.007), while this was not the case for either Cx or Ctrl males (Wilcoxon tests, P ⬎ 0.25; see Fig. 3a). The analysis by a three-way ANOVA of data collected on individually housed males revealed a significant Group effect (F (2,19) ⫽ 6.91, P ⫽ 0.005), but no

314

Pinxten et al.

significant Time ⫻ Group interaction (P ⬎ 0.15; see Fig. 3b). Overall, Ctrl males spent significantly more time in a nestbox than both Cx and Cx ⫹ T males (t ⫽ ⫺3.65, df ⫽ 19, P ⫽ 0.0017, and t ⫽ ⫺2.55, df ⫽ 19, P ⫽ 0.019). Although Cx ⫹ T males spent more time in a nestbox than Cx males, differences between the two groups were not significant (t ⫽ ⫺1.33, df ⫽ 19, P ⫽ 0.2). Gathering of Green Nest Materials Without a female present. Both before and after implantation all males occupying a nestbox (see above) were observed carrying green nest materials into this nestbox, but the frequencies were very low in the absence of a female (maximum observed frequency, three times/hour; see Fig. 4a). With a female present. The fraction of males in each group carrying green nest materials in a nestbox after introduction of a female was comparable to that when no female was present (see above), but frequencies were higher (Fig. 4b). The number of times per hour that males carried green nesting material in a nestbox differed significantly among groups, both before and after T implantation (Kruskal–Wallis tests, ␹ 2 ⫽ 11.66, P a ⫽ 0.006, and ␹ 2 ⫽ 10.37, P ⫽ 0.006, respectively). Before T implantation Ctrl males carried significantly more green nest material into the nestbox than Cx and Cx ⫹ T males, both of which never showed this behavior (Fig. 4b). After T implantation, both Cx ⫹ T and Ctrl males carried significantly more green nesting material into the nestbox than Cx males, which still did not show this behavior. Cx ⫹ T males carried significantly more green nest material in the nestbox after T implantation than before (Wilcoxon test, z ⫽ ⫺2.66, P ⫽ 0.007), while frequencies did not differ significantly between the two periods in Cx males (z ⫽ 0, P ⫽ 1) and Ctrl males (z ⫽ ⫺0.13, P ⫽ 0.89). Before implantation none of the 16 castrated males that were tested individually carried green nesting material into the nestbox compared to 4 of the 7 Ctrl males (Fisher’s exact test, P ⫽ 0.004). After implantation, 6 of the 9 Cx ⫹ T males and all 7 Ctrl males carried green nesting material into the nestbox, compared to none of the 7 Cx males (Fisher’s exact tests, P a ⫽ 0.02 and P a ⫽ 0.002, respectively). Comparing frequencies per hour provided results comparable to those in the group situation: frequencies differed significantly among experimental groups both before and after T implantation (Kruskal–Wallis test, ␹ 2 ⫽ 7.49, P ⫽ 0.02, and ␹ 2 ⫽ 10.55, P a ⫽ 0.01, respectively; see

FIG. 4. The number of times per hour (mean ⫾ SEM) Ctrl males, Cx males, and Cx ⫹ T males carried green nesting material in a nestbox in the period before and after T implantation: (a) housed in a group without a female present, (b) housed in a group with a female present, and (c) individually housed with a female present. Numbers under the bars denote sample sizes. Differences between male groups within each period are indicated by different lowercase letters above the bars, P ⬍ 0.05.

Fig. 4c). However, in pairwise comparisons, frequencies did not differ significantly between groups before implantation. After implantation, Cx ⫹ T males carried significantly more nest material in the nestbox than Cx males, which never showed this behavior, while the frequencies of Ctrl males were significantly higher than those of both Cx ⫹ T and Cx males (Fig.

Context-Dependent Effects of Testosterone on Song

4c). Both Cx ⫹ T and Ctrl males carried significantly more green nest materials into a nestbox after implantation than before (Wilcoxon test, z ⫽ ⫺2.20, P ⫽ 0.02, and z ⫽ ⫺1.99, P ⫽ 0.04) while Cx males did not show this behavior in either period.

DISCUSSION Context-Dependent Control of Song The present data clearly highlight, in male European starlings, the complexity of the relationships between plasma T levels and song produced during the breeding season within and outside a courtship context. Song produced in a context with no immediate relationship to female courtship did not appear to be dependent upon plasma T. Before implantation, castrated and intact males clearly had similar song rates in the absence of a female, although intact males had significantly higher T levels. These results not only confirm the results of Davis (1957), who reported that male starlings castrated during the fall continued singing for at least 6 weeks, but also show that castrated males continue singing for at least 6 months after the operation. That castration does not abolish song has been demonstrated in a few other bird species (see Balthazart, 1983). After T implantation, noncourtship song rates differed among experimental groups, being lowest in Cx males, intermediate in intact males, and highest in Cx ⫹ T males (see Fig. 2a), but a three-way ANOVA revealed a significant Time effect but no significant Time ⫻ Group interaction effect. Although we cannot entirely exclude that a significant interaction might have been found if a larger number of subjects had been studied, it should be noted that the higher song rate in Cx ⫹ T males compared to Cx males was not due to experimentally elevated T levels increasing song rates in Cx ⫹ T males (see Fig. 2a), which would clearly have supported the idea that noncourtship song is stimulated by T, but resulted from song rates of Cx males having decreased in the period after T implantation. The latter may have been due to a suppressive effect on song caused by the presence of the, probably dominant, T-implanted castrated males in the same aviary (see Wiley et al., 1993). Alternatively, the decrease in singing occurrences seen here may reflect normal seasonal changes. These possibilities certainly need further investigation before the relationship between plasma T and noncourtship song in male starlings is completely resolved. However, the results obtained here before implantation,

315

paired with the observation that male starlings produce large amounts of song during the nonbreeding season when gonads are regressed, strongly suggest that plasma T and other gonadal hormones are not essential for the production of noncourtship song, irrespective of whether it is produced within or outside the breeding season. In contrast, in male starlings, as in several other bird species (Harding, Sheridan, and Walters, 1983; see Schlinger, 1997, for review), the production of courtship song depends on the presence of high plasma T levels: castration abolished and T replacement reinstated courtship song under both group and individual housing conditions. In a recent field study investigating the role of T in the trade-off between parental and mate attraction behavior, T-implanted male starlings also spent significantly more time singing to attract additional females than control males during the incubation period (De Ridder, Pinxten, and Eens, 2000). The present results thus suggest that the activation of song by T interacts with the context in which song is produced in male starlings. Likewise, in green treefrogs, Hyla cinerea, it was recently shown that social context influences androgenic effects on calling behavior (Burmeister and Wilczynski, 2001). Earlier, Walters et al. (1991) showed that in zebra finches, Taeniopygia guttata, in which male song can be divided into three types depending on the context in which it is sung, hormone treatments modulate the three types of song differently. While all three types of song depend on gonadal hormones, the conversion of androgens to estrogens is important in activating both female-directed and male-directed, but not undirected, songs. Walters et al. (1991) emphasized that when studying the physiological bases of singing behavior one should always take into account the circumstances under which songs are recorded and should not consider song as a unitary phenomenon. Although we did not explicitly differentiate between male-directed and undirected song when measuring noncourtship song, most observed noncourtship song should be considered undirected song as it was not clearly directed toward another male(s). However, as male-directed song also occurs in starlings, albeit infrequently (Eens et al. 1993), additional research examining whether castration and/or T implantation affects both types of song differently would be valuable. It would also be interesting to examine whether during the breeding season noncourtship song produced by intact male starlings (or by castrated males) is qualitatively different from courtship song. Riters,

316

Eens, Pinxten, Duffy, Balthazart, and Ball (2000) recently demonstrated that male starlings sang shorter song bouts during the winter period, when song is unrelated to female courtship, than during the breeding season. Male song sparrows (Melospiza melodia morphna) have been found to sing a more stereotyped song during the breeding season in spring when T levels are high than in fall when T levels are low (Smith, Brenowitz, Beecher, and Wingfield, 1997). Interestingly, it has been suggested that in song sparrows in which territorial aggression is high throughout the year, androgens of unknown nongonadal origin activate aggression after being aromatized during periods with low plasma T levels (Soma, Tramontin, and Wingfield, 2000; Soma and Wingfield, 2001). In male starlings, Riters, Baillien, Eens, Pinxten, Foidart, Ball, and Balthazart (2001) analyzed the seasonal changes in the activity of four T-metabolizing enzymes in the brain and found that high aromatase activity is maintained in the telencephalon, but not the diencephalon, throughout the fall and winter. Together, these results are compatible with the idea that singing in male starlings when plasma T levels are basal (outside the breeding season or by castrated males) may continue to be regulated by steroids from a nongonadal source that activate the behavior after being aromatized in the brain. More studies are needed to experimentally test (e.g., with aromatase inhibitors) the role of androgen aromatization in the activation of noncourtship singing in male starlings.

Pinxten et al.

ior and larger territories, suggesting that T also exerts a dose-dependent effect on territorial establishment (Searcy, 1981; Watson and Parr, 1981; Wingfield, 1984; Gwinner and Gwinner, 1994). Nestbox-Oriented Courtship Behaviors The present results also indicate that two important courtship behaviors, singing in a nestbox and carrying green nesting material into a nestbox, are expressed only when high plasma T levels are present. It is not surprising that castrated males did not show these two nestbox-oriented courtship behaviors, since they never occupied nestboxes. We did find that individually housed castrated males spent about 3% of the time in the nestbox when a female was present, but this certainly did not serve a courtship function as it appeared to be due to disturbance caused by the introduction of the female into the small aviary. Moreover, the males clearly were not singing in the nestbox. Castrated male zebra finches also display a significant decrease in courtship behaviors, compared to intact males (Harding et al., 1983). There is increasing evidence suggesting that the effects of T on courtship behaviors are actually produced via its conversion in the brain into estrogen by aromatase (Walters and Harding, 1988; Riters et al., 2000). Further experimental work blocking aromatase activity is necessary to investigate the hormonal regulation of courtship behaviors, including song, in male starlings (see also Riters et al., 2000, 2001).

Nestbox Occupancy In most avian species studied, plasma levels of T are high during the period when males are establishing and defending territories (Wingfield, Hegner, Dufty, and Ball, 1990; Beletsky et al., 1995). Although it is now well established that T influences the type of aggressive behavior involved in territorial contests, relatively few experimental manipulations seem to have been done to establish whether T is a causal factor in the establishment of a territory. The present results clearly demonstrate a causal link between plasma T levels and the establishment of a territory, since nestbox occupancy was abolished by castration and reinstated by subsequent T replacement. They also agree with field observations showing that male starlings may show nestbox occupation behavior in September (”autumn sexuality”), when they experience a small peak in T levels (Dawson, 1983). In several studies, including one in starlings, increased T levels have been shown to lead to increased levels of territorial behav-

ACKNOWLEDGMENTS We thank K. Verbeeck for his help in taking care of the captive starlings. We are also grateful to Lieve Geenen for measuring T levels and to Lauren Riters for her valuable comments on an earlier draft. R.P. is supported by Research Project G.0075.98 of the Fund for Scientific Research (FWO) Flanders, Belgium. This study was also made possible through financial support from the Research Council of the University of Antwerp (NOI-BOF-UA 97). In addition, it was supported by grants from the Belgian FRFC (No. 2.4555.01) and the NINDS (NS-35467) to J.B. The collaboration between M.E. and J.B. was supported by a Scientific Research Network (WO.007.96N) of the FWO, Flanders.

REFERENCES Arnold, A. P. (1975). The effects of castration and androgen replacement on song, courtship, and aggression in zebra finches (Poephila guttata). J. Exp. Zool. 191, 309 –326. Ball, G. F. (1999). The neuroendocrine basis of seasonal changes in

Context-Dependent Effects of Testosterone on Song

vocal behavior among songbirds. In M. D. Hauser and M. Konishi (Eds.), The Design of Animal Communication, pp. 213–253. MIT Press, Cambridge. Ball, G. F., and Wingfield, J. C. (1987). Changes in plasma levels of luteinizing hormone and sex steroid hormones in relation to multiple-broodedness and nest-site density in male starlings. Physiol. Zool. 60, 191–199. Balthazart, J. (1983). Hormonal correlates of behavior. In D. S. Farner, J. R. King, and K. C. Parkes (Eds.), Avian Biology, Vol. 7, pp. 221–365. Academic Press, New York. Beletsky, L. D., Gori, D. F., Freeman, S., and Wingfield, J. C. (1995). Testosterone and polygyny in birds. In D. M. Power (Ed.), Current Ornithology, Vol. 12, pp. 1– 41. Plenum, New York. Burmeister, S. S., and Wilczynski, W. (2001). Social context influences androgenic effects on calling in the Green Treefrog (Hyla cinerea). Horm. Behav. 40, 550 –558. Davis, D. E. (1957). Aggressive behavior in castrated starlings. Science 126, 253. Dawson, A. (1983). Plasma gonadal steroid levels in wild starlings (Sturnus vulgaris) during the annual cycle and in relation to the stages of breeding. Gen. Comp. Endocrinol. 49, 286 –294. De Ridder, E., Pinxten, R., and Eens, M. (2000). Experimental evidence of a testosterone-induced shift from paternal to mating behaviour in a facultatively polygynous songbird. Behav. Ecol. Sociobiol. 49, 24 –30. Dloniak, S. M., and Deviche P. (2001). Effects of testosterone and photoperiodic condition on song production and vocal control region volumes in adult male Dark-Eyed juncos (Junco hyemalis). Horm. Behav. 39, 95–105. Duffy, D. L., Bentley, G. E., Drazen, D. L., and Ball, G. F. (2000). Effects of testosterone on cell-mediated and humoral immunity in non-breeding adult European starlings. Behav. Ecol. 11, 654 – 662. Eens, M. (1997). Understanding the complex song of the European starling: An integrated ethological approach. Adv. Study Behav. 26, 355– 434. Eens, M., and Pinxten, R. (1990). Extra-pair courtship in the Starling Sturnus vulgaris. Ibis 132, 618 – 619. Eens, M., and Pinxten, R. (1995). Inter-sexual conflicts over copulations in the European starling: Evidence for the female mateguarding hypothesis. Behav. Ecol. Sociobiol. 36, 71– 81. Eens, M., and Pinxten, R. (1996). Female European starlings increase their copulation solicitation rate when faced with the risk of polygyny. Anim. Behav. 51, 1141–1147. Eens, M., Pinxten, R., and Verheyen, R. F. (1990). On the function of singing and wing-waving in the European starling Sturnus vulgaris. Bird Study 37, 48 –52. Eens, M., Pinxten, R., and Verheyen, R. F. (1991). Male song as a cue for mate choice in the European Starling. Behaviour 116, 210 –238. Eens, M., Pinxten, R., and Verheyen, R. F. (1993). Function of the song and song repertoire in the European starling (Sturnus vulgaris): An aviary experiment. Behaviour 125, 51– 66. Eens, M., Van Duyse, E., Berghman, L., and Pinxten, R. (2000). Shield characteristics are testosterone-dependent in both male and female moorhens. Horm. Behav. 37, 126 –134. Gwinner, H. (1997). The function of green plants in nests of European starlings (Sturnus vulgaris). Behaviour 134, 337–351. Gwinner, H., and Gwinner, E. (1994). Effects of testosterone on nest-box occupation and associated behaviours by male European starlings (Sturnus vulgaris). Behaviour 129, 141–148. Gwinner, H., Gwinner, E., and Dittami, J. P. (1987). Effects of nestboxes on LH, testosterone, testicular size, and the reproductive

317

behavior of male European starlings in spring. Behaviour 103, 68 – 82. Harding, C. F., Sheridan, K., and Walters, M. (1983). Hormonal specificity and activation of social behavior in male zebra finches. Horm. Behav. 17, 111–133. Harding, C. F., Walters, M. J., Collado, D., and Sheridan, K. (1988). Hormonal specificity and activation of social behavior in male red-winged blackbirds. Horm. Behav. 22, 402– 418. Heid, P., Guttinger, H. R., and Prove, E. (1985). The influence of castration and testosterone replacement on the song architecture of canaries (Serinus canarius). Z. Tierpsychol. 69, 224 –236. Kroodsma, D. E., and Byers, B. E. (1991). The function(s) of bird song. Am. Zool. 31, 318 –328. Littell, R. C., Milliken, G. A., Stroup, W. W., and Wolfinger, R. D. (1996). SAS Systems for Mixed Models. SAS Institute. Mehta, C., and Patel, N. (1995). StatXact-3 for Windows: User’s Manual. Cytel Software Corporation, Cambridge, MA. Nottebohm, F., Nottebohm, M. E., Crane, L. A., and Wingfield, J. C. (1987). Seasonal changes in gonadal hormone levels of adult male canaries and their relation to song. Behav. Neur. Biol. 47, 197–211. Nottebohm, F., Stokes, T. M., and Leonard, C. M. (1976). Central control of song in the canary. J. Comp. Neurol. 165, 457– 486. Pinxten, R., De Ridder, E., Balthazart, J., Berghman, L., and Eens, M. (2000). The effect of castration on aggression in the nonbreeding season is age-dependent in male European starlings. Behaviour 137, 647– 661. Pinxten, R., and Eens, M. (1990). Polygyny in the European Starling: Effect on female reproductive success. Anim. Behav. 40, 1035–1047. Pinxten, R., and Eens, M. (1997). Copulation and mate-guarding patterns in polygynous European starlings. Anim. Behav. 54, 45– 58. Pinxten, R., and Eens, M. (1998). Male starlings sing most in the late morning, following egg-laying: A strategy to protect their paternity? Behaviour 135, 1197–1211. Pinxten, R., Eens, M., and Verheyen, R. F. (1989). Polygyny in the European starling. Behaviour 111, 234 –256. Rice, W. R. (1989). Analyzing tables of statistical tests. Evolution 43, 223–225. Riters, L. V., and Ball, G. F. (1999). Lesions to the medial preoptic area affect singing in the male European starling. Horm. Behav. 36, 276 –286. Riters, L. V., Baillien, M., Eens, M., Pinxten, R., Foidart, A., Ball, G. F., and Balthazart, J. (2001). Seasonal variation in androgenmetabolizing enzymes in the diencephalon and telencephalon of the male European starling (Sturnus vulgaris). J. Neuroendocrinol. 13, 985–997. Riters, L. V., Eens, M., Pinxten, R., Duffy, D. L., Balthazart, J., and Ball, G. F. (2000). Seasonal changes in courtship song and the medial preoptic area in male European starlings (Sturnus vulgaris). Horm. Behav. 38, 250 –261. Riters, L. V., Eens, M., Pinxten, R. and Ball, G. F. (2002). Seasonal changes in the densities of ␣2-noradrenergic receptors are inversely related to changes in testosterone and the volumes of song control nuclei in male European starlings. J. Comp. Neurol. 444, 63–74. SAS (1988). SAS/STAT User’s Guide, release 6.03 edition. SAS Institute Inc., North Carolina. Schlinger, B. A. (1997). Sex steroids and their actions on the birdsong system. J. Neurobiol. 33, 619 – 631. Siegel, S. (1956). Nonparametrical Statistics. McGraw–Hill, New York. Smith, G. T., Brenowitz, E. A., Beecher, M. D., and Wingfield, J. C. (1997). Seasonal changes in testosterone, neural attributes of song

318

control nuclei, and song structure in wild songbirds. J. Neurosci. 17, 6001–10. Sokal, R. R., and Rohlf, F. J. (1981). Biometry, 2nd ed. Freeman, New York. Soma, K. K., Tramontin, A. D., and Wingfield, J. C. (2000). Oestrogen regulates male aggression in the nonbreeding season. Proc. R. Soc. London Ser. B 267, 1089 –1096. Soma, K. K., and Wingfield, J. C. (2001). Dehydroepiandrosterone in songbird plasma: Seasonal regulation and relationship to territorial aggression. Gen. Comp. Endocrinol. 123, 144 –155. SPSS (1986). SPSS Reference Guide, Chicago. Walters, M. J., Collada, D., and Harding, C. F. (1991). Oestrogenic modulation of singing in male zebra finches: Differential effects on directed and undirected songs. Anim. Behav. 42, 445– 452. Watson, A., and Parr, R. (1981). Hormone implants affecting terri-

Pinxten et al.

tory size and aggressive and sexual behavior in red grouse. Ornis Scand. 12, 55– 61. Wiley, R. H., Archawaranon, M., and Thompson, E. W. (1993). Singing in relation to social dominance and testosterone in whitethroated sparrows. Behaviour 129, 175–190. Wingfield, J. C. (1984). Environmental and endocrine control of reproduction in the song sparrow, Melospiza melodia: II. Agonistic interactions as environmental information stimulating secretion of testosterone. Gen. Comp. Endocrinol. 56, 417– 424. Wingfield, J. C., and Farner, D. S. (1993). Endocrinology of reproduction in wild species. Avian Biol. 9, 163–327. Wingfield, J. C., Hegner, R. E., Dufty, A. M., Jr., and Ball, G. F. (1990). The ‘challenge hypothesis’: Theoretical implications for patterns of testosterone secretion, mating systems and breeding strategies. Am. Nat. 136, 829 – 846.