Early Condition, Song Learning, and the Volume of Song Brain Nuclei in the Zebra Finch (Taeniopygia guttata)

Early Condition, Song Learning, and the Volume of Song Brain Nuclei in the Zebra Finch (Taeniopygia guttata) Diego Gil, 1 Marc Naguib, 2 Katharina Riebel, 3 Alison Rutstein, 4 Manfred Gahr 5 1 Departamento de Ecología Evolutiva, Museo Nacional de Ciencias Naturales (CSIC), José Gutiérrez Abascal 2, 28006 Madrid, Spain 2 Department of Animal Behavior, University Bielefeld, PO Box 100 131, 33501 Bielefeld, Germany 3 Institute of Biology, Leiden University, PO Box 9516, 2300 RA Leiden, The Netherlands 4 School of Biology, University of St Andrews, Fife KY16 9TS, United Kingdom 5 Max-Planck-Institute of Ornithology, Department of Behavioral Neurobiology, 82319 Seewiesen, Germany Received 3 February 2006; revised 6 April 2006; accepted 1 May 2006 ABSTRACT: Songbirds are an important model system for the study of the neurological bases of song learning, but variation in song learning accuracy and adult song complexity remains poorly understood. Current models of sexual selection predict that signals such as song must be costly to develop or maintain to constitute honest indicators of male quality. It has been proposed that reductions of nestling condition during song development might limit the expression of song learning. Adult song could thus act as an indicator of early stress as only males that enjoy good condition during development could learn accurately and sing long songs or large repertoires. We tested this hypothesis in the zebra finch by modifying early condition through cross-fostering chicks to small, medium, and large broods. Song learning was very accurate and was found to reflect very closely tutor song characteristics and to depend on the number of males in the tutoring group. Although the brood size manipulation strongly affected several measures of nestling condition and adult biometry, we found no relationship between early condition and song learning scores or song characteristics. Similarly, brain mass and high vocal center (HVC), robust nucleus of the arcopallium (RA), and lateral magnocellular nucleus of the anterior nidopallium (LMAN) volumes did not covary with nestling condition and growth measurements. We found no significant relationship between song repertoire size and HVC and RA volumes, although there was a nonsignificant trend for HVC to increase with increasing proportion of learnt elements in a song. In conclusion, the results provide no evidence for song learning to be limited by nestling condition during the period of nutritional dependence from the parents in this species. ' 2006 Wiley Periodicals, Inc. J Neurobiol 66: 1602 1612, 2006 Keywords: birdsong; song brain nuclei; HVC; RA; LMAN; developmental stress Correspondence to: D. Gil (dgil@mncn.csic.es). Contract grant sponsor: Association for the Study of Animal Behavior (ASAB). ' 2006 Wiley Periodicals, Inc. Published online 20 October 2006 in Wiley InterScience (www. interscience.wiley.com). DOI 10.1002/neu.20312 INTRODUCTION Current knowledge suggests that sexual selection is at the base of the evolution of complex song repertoires (for reviews see Catchpole and Slater, 1995; Searcy 1602

Developmental Stress and Song Brain Nuclei 1603 and Yasukawa, 1996; Gil and Gahr, 2002; Riebel, 2003). Sexual selection theory predicts that for signals to constitute honest indicators of quality they must be costly to develop or maintain (Zahavi, 1975; Grafen, 1990). Otherwise, all individuals could acquire these signals irrespective of their quality and the signal would become uninformative. Individuals of the same species vary enormously in their song characteristics and song learning accuracy, and the origins of this variation remain poorly understood. One of the most popular hypotheses to explain this constraint is based on a presumed strong relationship between song repertoire size and the size of brain song nuclei (Nottebohm et al., 1981; Brenowitz et al., 1995; Gahr, 1997; Airey and DeVoogd, 2000). Neural song nuclei such as high vocal center (HVC), RA (robust nucleus of the arcopallium), and LMAN (lateral magnocellular nucleus of the anterior nidopallium) are involved in song production and/or learning (Nottebohm et al., 1976; Vu et al., 1994; Yu and Margoliash, 1996; Hahnloser et al., 2002). This brain space may be costly because of intrinsic costs of development or through competing demands for space or resources (Gil and Gahr, 2002). However, evidence for a link between song repertoire and brain space is not straightforward (Ward et al., 1998; Leitner et al., 2001; Garamszegi and Eens, 2004; Leitner and Catchpole, 2004; Sartor and Ball, 2005). Nowicki and coauthors have proposed that a relationship between the volume of song control nuclei and male quality (Nowicki et al., 1998, 2002) should be expected because song acquisition in songbirds coincides with a time of accelerated growth of the neuronal song system during which birds are most sensitive to nutritional stress. This hypothesis has been validated by several studies showing that when nestling or juvenile condition is reduced by means of some source of environmental stress, birds produce poor songs and have reduced song control brain nuclei (Nowicki et al., 2002; Buchanan et al., 2003, 2004; Spencer et al., 2003). However, the methods used to modify nestling condition so far have involved direct manipulation of food availability and corticosterone administration to nestlings. Although these are powerful techniques, it is difficult to know whether the levels of stress that are imposed are comparable with those experienced by nestlings in natural conditions. Here, we test the relationship between early condition and song learning in the zebra finch, addressing possible behavioral and neuronanatomical effects simultaneously. We take a different approach compared with previous studies by using a relevant experimental modification of condition on a large sample size of birds. We experimentally manipulated brood size, a method that has proven a useful tool for identifying reproductive trade-offs (Stearns, 1992), and that mimics natural variation in early condition. Even under laboratory conditions with ad libitum food, avian parents are constrained in providing food for large broods, and several studies in captivity show that increases in brood size result in delayed hatching, smaller growth, reduced immunocompetence, smaller sexually selected traits, or reduced maternal investment (Coleman and Whittall, 1988; De Kogel and Prijs, 1996; De Kogel, 1997; Gil et al., 2004; Naguib et al., 2004; Naguib and Gil, 2005). These patterns likely arise because parents do not fully compensate their levels of food provisioning when broods are experimentally enlarged (Stearns, 1992), and additionally because increased nestling begging and competition among nestlings incur costs of its own (Neuenschwander et al., 2003). Our experimental design included cross-fostering to control for common environment and genetic relatedness. We measured song learning and song characteristics in the males that survived to adulthood, and we control for environmental learning effects such as tutor song and number of males in the tutoring group (Slater et al., 1988; Tchernichovski and Nottebohm, 1998). MATERIALS AND METHODS Subjects, Experimental Treatment, and Measures of Nutritional Stress The zebra finches used in this experiment were a F4 generation from wild Australian origin, housed in captivity at the University Bielefeld, Germany. Details of the experimental manipulation are given in Naguib et al. (2004). Briefly, nestlings of experimental broods were cross-fostered in mixed parentage broods at the age of 2 6 1 days. The experimental treatment consisted in varying the size of these broods to small (2 3 nestlings), medium (4 nestlings), and large (5 6 nestlings). We arrived at these figures after checking that the average number of hatchlings in our captive population is 4.45 (SD ¼ 1.07) and the average number of fledglings is 3.6 (SD ¼ 1.19). This means that our manipulations correspond to a range of 1 to 2 SD below and above the natural mean, depending on whether we consider hatchlings or fledglings, and is similar to previous studies on the same species (De Kogel, 1997). To control for parental differences in rearing, we made sure that there was no correlation between initial and experimental brood sizes. Foster parents were provided with ad libitum standard finch food mixture, as well as daily supplements of protein and vitamins. In previous papers, we showed that this experimental manipulation affected nestling growth, testosterone, and immunocompetence levels, as well as adult body size

1604 Gil et al. (Naguib et al., 2004) and female reproductive investment once they had reached sexual maturity (Gil et al., 2004; Naguib and Gil, 2005; Naguib et al., 2006). Both male and female nestlings suffered reduced growth with increasing brood size. Nestling body condition at 10 days of age, which was defined as nestling body mass when size was corrected for, was 25% lower for birds from large broods than for birds from small broods (Naguib et al., 2004). These effects were still present in adult body size: weight, tarsus, and wing length decreased with increasing experimental brood size. Since parents had access to ad libitum food, this suite of effects was most likely due to parental limitations at feeding large brood sizes within the natural range and to increased costs derived from high levels of begging and nestling competition (Stearns, 1992; Neuenschwander et al., 2003; Naguib et al., 2004). In this paper, we test whether this experimental manipulation of brood size covaried with song learning, adult song characteristics, and the volume of song brain nuclei in adulthood, as predicted by the nutritional stress hypothesis. Body size was recorded by measuring tarsus length with calipers to the nearest 0.01 mm, body weight was recorded in a Sartorius PT120 balance to the nearest 0.01 g, and body condition was defined as the effect of body weight when tarsus length was included in the model to control for size (García-Berthou, 2001). Song Tutoring Regime When nestlings were 35 days old, they were taken from their foster parents and housed in 19 mixed-sex groups of three to nine tutees each. We standardized tutor groups by introducing all males within a given group at a maximum of 2 days apart to limit within-group heterogeneity in social experience. However, this procedure lead to variation in tutor group size as a consequence of asynchronic breeding. As tutors we used adult males that had not been in contact with the tutees, and we allocated one tutor per group. Tutoring groups were housed in aviaries (0.92 1.8 1.85 or 1 3 3.3 m) that were set up so that the different tutor groups were visually isolated from each other. We made sure that each group contained a mixture of offspring coming from different experimental groups (average size ¼ 3.5 tutees, SD ¼ 1.0). The offspring were kept in these groups until they were 90 days of age after which they were transferred to single-sex groups. Measures of Song Song was recorded using a Sennheiser ME66/K6 microphone and a SONY TCD5M tape recorder when birds were 3 to 6 months of age. We removed males from common cages to a sound recording room and placed them in individual cages in visual contact with a female located in an adjacent cage. Taped songs were digitized with Cool Edit (Syntrillium, Phoenix, USA) at 32kHz (16 bits) and PC SoundBlaster Live! Soundcard and further edited with SASLab Pro software (R. Specht, Berlin, Germany). Zebra finches sing one individually distinctive song motif. For a typical \song," males sing a few introductory elements followed by an individually distinctive sequence of different elements, the so-called song motif which is normally delivered in a few repetitions in quick succession (a song) separated by a short pause from the next song. We selected the most common variant of the individually distinctive song motif from a sample of five songs for each male (since elements are occasionally added or omitted in some phrases). The mean motif duration in ms was determined, when possible, by averaging the duration of five stereotyped motifs from different song bouts. We also counted the number of different elements (element repertoire) and the number of elements that were good copies from elements in the tutor song. Initially, elements were given scores of 1, 2, or 3 based on their resemblance to the tutor s element in terms of duration, fundamental frequency, and frequency modulation, where 3 ¼ virtually identical, 2 ¼ good match but slight differences in fundamental frequency/duration or frequency modulation, and 1 ¼ recognizable as that element but poor similarity. In the analysis, only those elements receiving scores of 2 or 3 were considered to be good copies of tutor song. Song scoring was done blindly with respect to the experimental treatment of the males and independently by two different observers (AR and KR) experienced with zebra finch song. Both observers showed high repeatability in scoring elements (element repertoire: r ¼ 0.89, n ¼ 45, p < 0.001; number of matching elements r ¼ 0.765, n ¼ 45, p < 0.001), and final scoring of a small number of nonmatching cases was determined by consensus after common consideration of each song by the two scorers. Song similarity between tutor and tutee was measured in two different ways, each of them providing different biological meanings. First, we calculated the percentage of tutor elements that had been copied by a tutee. This is a measure of learning performance and will be referred to as \learning score." Second, we calculated the percentage of tutee elements that had been copied from the tutor. This is a measure of whether tutees copy tutor elements or add new elements to the song, and will be refer to in the rest of the paper as \song similarity." Note that these two measures are independent of each other (r p ¼ 0.04, n ¼ 46, p ¼ 0.75) and that a tutee may have a low similarity score with tutor song despite having a very high learning score if he has additional elements in his repertoire, i.e., a high level of improvisation. Automated procedures for song comparison are increasingly used in the literature (Tchernichovski et al., 2000). We conducted a preliminary test of this option, and found that assessment by eye was a more reliable option for our material. Indeed, measurements of song similarity by human eye are typically used for validation of automated procedures, and evidence show highly reliability between the two methods (Tchernichovski et al., 2000). We measured song activity at the age of 14 months by exposing birds to a previously unseen female and recording the number of songs during 45 min. We used several different females as stimulus for different males, but each female was used with more than one male so as to be able to con-

Developmental Stress and Song Brain Nuclei 1605 Figure 1 Tutees produced very high accurate copies of their tutor song. Sonogram A shows a tutor song, sonogram B a good copy of the tutor song, and sonogram C a bad copy of the tutor song. trol for possible effects of female identity on male singing activity. Indeed, male singing activity was significantly influenced by female identity (F 15,19.9 ¼ 9.71, p < 0.001), and therefore this factor was included as a random factor in models of song activity. Measurements of Song Brain Nuclei Birds were sacrificed with an overdose of isofluran and decapitated. We removed the brains and froze them on liquid nitrogen and stored at 808C until analysis. Body and

1606 Gil et al. Table 1 The Effect of Tutor Song and Number of Males in the Tutoring Group in Song Learning and Song Characteristics Tutee Song Measurement Effect of Tutor Effect of Number of Males in Group F Test r est. (SE) F Test r est. (SE) Learning score F 16,22.2 ¼ 0.99, p ¼ 0.53 n/a F 1,42.6 ¼ 0.64, p ¼ 0.42 0.03 (0.04) Song similarity F 16,28 ¼ 2.16, p ¼ 0.03 n/a F 1,40.9 ¼ 2.49, p ¼ 0.12 0.04 (0.03) Song duration F 1,38.4 ¼ 7.92, p ¼ 0.007 0.54 (0.19) F 1,43.8 ¼ 6.88, p ¼ 0.01 94.4 (36.1) Element repertoire F 1,43.9 ¼ 34.5, p ¼ 0.001 0.76 (0.12) F 1,36.3 ¼ 23.1, p ¼ 0.001 1.3 (0.27) Song rate F 16,15.3 ¼ 2.56, p ¼ 0.03 n/a F 1,29 ¼ 0.0, p ¼ 0.9 0.01 (0.15) In the case of song duration and element repertoire, data show the covariance with the respective tutor characteristic, whereas in the rest of measurements, data show the effect of tutor identity. n ¼ 45 (song rate), and n ¼ 46 (rest of measurements). n/a ¼ non applicable. brain masses were measured to the nearest 0.1 g. Brains were cut on a cryostat into 20-m parasagittal sections. Sections were mounted onto Superfrost Plus slides (Fisher) in five different series. First, we Nissl-stained (0.1% Thionin) one series of sections of both the left and right hemisphere of six randomly chosen males for morphometric analysis. Since there were no left right differences in the volume of HVC, RA, and LMAN (data not shown), we subsequently stained and analyzed only the right hemisphere. Measurements were done using an image analysis system on a video screen (Spot, Visitron, Germany). For volume measurement, the perimeter of the region of interest in each section was drawn on digitized images (2.58 m/ pixel) and the area was calculated by a built-in function of the software. The volume of each song nuclei was calculated as the sum of these measurements multiplied by section thickness. Neuroanatomical measures were performed by one of us (MG), blind to the treatment that the birds had been subjected to. Although the morphometric results are dependent on the delineation method (Gahr, 1997), we used the Nisslmethod throughout for economic reasons. Statistics All measurements were checked for normality before the analysis, and transformed when necessary, thus all brain and song nuclei measurements were log transformed. We used mixed linear models for most analyses, using Proc Mixed (SAS), and declared original and foster nest as random factors to control for covariance due to common genetic origin or early maternal effects. When testing the effect of brood size, we corrected nondirectional p values from F tests with two-tailed ordered heterogeneity tests, since we predicted medium brood sizes to be intermediate in song characteristics and song brain nuclei volumes between large and small brood sizes (Rice and Gaines, 1994). Since there is disagreement as to whether the absolute or the relative size of brain nuclei is the most relevant measure, we tested the relationship between song nuclei and song characteristics in two ways: with and without including brain or body mass as covariates in the analysis. The results were the same irrespective of whether overall size was included or not. Descriptive statistics (mean and SE) per experimental group are provided for all variables (untransformed) in Appendix (Table A1). RESULTS Song Characteristics As expected from previous work in this species, song learning was strongly affected by two main social Figure 2 Relationship between tutor and tutee element repertoire size. See Table 1 for stats. Figure 3 Relationship between number of males in the tutoring group and tutee element repertoire size. See Table 1 for stats.

Table 2 Tutee Song Measurement Summary of Models Analyzing the Effect of Experimental Brood Size in Song Characteristics Experimental Brood Size Developmental Stress and Song Brain Nuclei 1607 Factor or Covariate in Each Model Effect of Tutor Number of Males in Group Learning score F 2,24.9 ¼ 0.03, p ¼ 0.96 n/a n/a Song similarity F 2,27 ¼ 1.09, p ¼ 0.35 F ¼ 1.26, p ¼ 0.10 n/a Song duration F 2,23.8 ¼ 0.12, p ¼ 0.87 F 1,33.7 ¼ 3.88, p ¼ 0.07 F 1,41 ¼ 2.85, p ¼ 0.09 Element repertoire F 2,23 ¼ 0.85, p ¼ 0.43 F 1,17.2 ¼ 10.3, p ¼ 0.005 F 1,15.4 ¼ 6.19, p ¼ 0.02 Song rate F 2,14.7 ¼ 0.73, p ¼ 0.50 F 16,13.4 ¼ 2.11, p ¼ 0.08 n/a Models include tutor song and number of males in the tutoring group when these variables were found to significantly affect song in previous analysis (Table 1). In the case of song duration and element repertoire, the corresponding tutor variable is included as covariate, in the case of proportion of learnt elements and song rate, tutor identity is included as a random factor. n ¼ 45 (song rate), and n ¼ 46 (rest of measurements). n/a ¼ non applicable. factors. First, birds copied their song very accurately from their respective tutors (e.g., Fig. 1). The mean learning score was 0.86 (SD ¼ 0.16), and song similarity averaged 0.94 (SD ¼ 0.1), showing respectively that most tutor elements were copied by tutees and that most learned song consisted of elements copied from tutees. Tutor identity significantly influenced song similarity, song duration, element repertoire, and song rate (Table 1). In the case of song duration and element repertoire, we found a strong positive covariance with the respective tutor song measurement (see for instance, element repertoire: Fig. 2). We tested whether tutor song characteristics influenced learning, to determine, for instance, whether longer songs were more difficult to learn than shorter ones or whether tutees selected certain song characteristics. However, neither learning score nor song similarity were found to covary with any tutor song characteristic (mixed ANCOVA, all p > 0.1, data not shown). A second factor that influenced song characteristics was the number of males present in the tutor group. This variable negatively affected both song duration and element repertoire (see Fig. 3), implying that songs became shorter and repertoires smaller as the number of males in the tutoring group increased (Table 1). The effect of number of males on learning score, song similarity, and song rate was not significant. Similarly, we tested the effect of total number of birds in the group instead of number of males, but this variable was not significant in exploratory models, and was dropped in subsequent analyses. In the following analyses, we controlled for these social effects when necessary, by including song tutor characteristics or number of males in the tutoring group as covariates, or tutor identity as random factor. The main objective of this experiment was to test whether the differences in early environmental condition imposed by experimental brood size influenced song learning. The results of the mixed model analyses (Table 2) show that manipulated brood size had no effect on song or learning or any measured song characteristic (see for example, learning score: Fig. 4), whereas tutor song and number of males in group significantly explained part of this variation. Although manipulated brood size was found to be a significant predictor of nestling condition in the study from which the tutees originated (Naguib et al., 2004), we explicitly entered nestling condition as a predicting covariate of song development to provide higher statistical power for detecting a relationship. The results confirm the previous analyses in showing that nestling condition at 10 days of age did not explain variance in song characteristics of tutee song (mixed ANCOVA, all p > 0.4, data not shown). Volume of Brain and Song Control Nuclei Although body mass decreased with increasing experimental brood size (F 2,28.1 ¼ 2.1, p < 0.05), brain mass did not differ between brood size categories, irrespective of whether its effect was tested alone, or Figure 4 The effect of manipulated brood size in song learning score. See Table 2 for stats.

1608 Gil et al. Table 3 Effects of Experimental Group on Brain Mass and Song Nuclei Volumes, Either as Direct Measurements or Controlling for Size Brain Measurement Effect of Experimental Treatment on Direct Measurement Brain mass F 2,22.4 ¼ 1.91, p ¼ 0.17 HVC volume F 2,43 ¼ 0.09, p ¼ 0.91 RA volume F 2,24.3 ¼ 0.19, p ¼ 0.82 LMAN volume F 2,43 ¼ 1.15, p ¼ 0.32 n ¼ 46. Effect of Experimental Treatment F 2,24.1 ¼ 1.18, p ¼ 0.32 F 2,42 ¼ 0.76, p ¼ 0.47 F 2,20.2 ¼ 0.75, p ¼ 0.46 F 2,42 ¼ 1.24, p ¼ 0.30 Effect of Experimental Treatment on Measurements Controlling for Size Effect of Size Covariate Body mass: F 1,41.5 ¼ 3.98, p ¼ 0.05, r est. (SE) ¼ 0.21 (0.10) Brain volume: F 1,42 ¼ 6.29, p ¼ 0.01, r est. (SE) ¼ 1.02 (0.4) Brain volume: F 2,32.8 ¼ 6.77, p ¼ 0.01, r est. (SE) ¼ 0.94 (0.36) Brain volume: F 1,42 ¼ 0.59, p ¼ 0.44, r est. (SE) ¼ 0.33 (0.43) corrected for body size (Table 3). Similarly, none of the HVC, RA, or LMAN volume differed between experimental groups (see for instance, HVC: Fig. 5), also irrespective of whether we tested raw volumes or included brain size as a covariate to control for overall size (Table 3). Mirroring the results of song measurements, nestling condition at 10 days of age did not explain variance in brain mass or volume of song nuclei (mixed ANCOVA, all p > 0.3, data not shown). The mean (SD) volumes of the song nuclei were respectively: 0.3105 (0.060) mm 3 for HVC, 0.2130 (0.034) mm 3 for RA, and 0.0939 (0.0181) mm 3 for LMAN. Surprising was the large range of volumes (HVC: 0.2065 to 0.4568 mm 3 ; RA: 0.1430 to 0.2803 mm 3 ; LMAN: 0.0607 to 0.1340 mm 3 ), comprising on average a coefficient of variation of 18%. There is a nonsignificant trend for HVC to be positively related to RA (r s ¼ 0.28, n ¼ 46, p ¼ 0.055), but the rest of correlations among the three song nuclei are not significant (HVC and LMAN: r s ¼ 0.15, n ¼ 46, p ¼ 0.31; RA and LMAN: r s ¼ 0.16, n ¼ 46, p ¼ 0.28). Figure 5 The effect of manipulated brood size in HVC volume. See Table 3 for stats. Relationship among Song Learning, Song Characteristics, and Song Nuclei Volume We analyzed the relationship of each song learning estimate and song characteristic with the volume of brain song nuclei by running mixed ANCOVA models where song was the dependent variable and song nuclei volumes the predicting covariates, together with the social factors explained above. Final models (Table 4) dropped the effect of tutor, and show that brain mass had a positive relationship with song duration and repertoire size (Fig. 6), but that neither RA nor HVC explained variation in these song variables. Song similarity was marginally predicted by variation in HVC size (p ¼ 0.07), but learning score was not related to any song nuclei volume that we studied. DISCUSSION The present study showed that the substantial differences in nestling and adult body condition and morphology induced in zebra finches by means of brood size manipulation (Naguib et al., 2004) had no effects on song copying accuracy or the volume of song nuclei. We predicted that if there is a cost to song learning (Nowicki et al., 1998) experimental males should sing poorer copies of their tutor song and have smaller song brain nuclei. However, after controlling for social effects on learning, we found that song learning, song characteristics, or song brain nuclei were not affected by experimental variation in early condition. Given that our manipulation had been successful in inducing long-term strong differences in condition and fitness-related traits (Gil et al., 2004;

Table 4 Relationship Between Each Song Measurement and Song Brain Nuclei, including Social Predictors in the Model Song Measurement Number of Males in Group Developmental Stress and Song Brain Nuclei 1609 Brain Nuclei Predictors HVC RA LMAN Brain Mass Learning score NS NS NS NS NS Song similarity NS F 1,33.4 ¼ 3.5, p ¼ 0.07, r est (SE) ¼ 0.4 (0.2) NS NS F 1,37.9 ¼ 5.1, p ¼ 0.03, r est (SE) ¼ 1.4 (0.6) Song duration F 1,42.9 ¼ 8.2, p ¼ 0.006, r est (SE) ¼ 98.9 (34.6) NS NS NS F 1,43 ¼ 5.1, p ¼ 0.03, r est (SE) ¼ 1181.1 (521) Element repertoire F 1,13.9 ¼ 14.9, p ¼ 0.002, r est (SE) ¼ 1.3 (0.3) NS NS NS F 1,39.3 ¼ 5.7, p ¼ 0.02, r est (SE) ¼ 9.7 (3.9) Song rate NS NS NS NS NS Naguib et al., 2004; Naguib and Gil, 2005; Naguib et al., 2006), we interpret our results as evidence that song learning, song characteristics, and brain song nuclei in the zebra finch are not affected by this kind of experimentally induced variation in condition during the period of nutritional dependence from the parents. The concept of allostasis refers to the adaptive process for actively maintaining stability through change (for a recent review see Korte et al., 2005). In an allostasis context, our results would imply that the development of the brain and the song system in particular are highly buffered against early allostatic load in this species, stressing their high survival value (Schew and Ricklefs, 1998). Song learning and song characteristics were found to be influenced by two social factors. First, tutees produced extremely good copies of their tutor song, independently of the characteristics of these songs, suggesting strong selection for young birds to sing exact copies of their tutor song. Since birds were randomly assigned to tutors, this raises the question of whether tutor choice itself would have been the same Figure 6 Relationship between log brain mass and element repertoire corrected for the number of males in the social group. See Table 4 for stats. for all birds if these would have had the opportunity to choose tutors. A possibility is that tutor choice might be influenced by tutee condition, and that birds in low condition might have chosen more simple or shorter songs. Studies in the laboratory show that tutor choice after 35 days is guided by similarity of the tutor s song to the song heard before 35 days, regardless of whether this song belongs to the father or a foster father (Clayton, 1987; Slater and Mann, 1990; Mann and Slater, 1994). In a study made under seminatural conditions, 40% of males sang songs from other males despite the father s song being available as a model (Zann, 1990). It is thus possible that in natural conditions social hierarchies between offspring may result in some birds not learning the father s song. Second, birds produced shorter and simpler copies of the tutor song with increasing number of males present in the tutoring group. This phenomenon has been described before (Volman and Khanna, 1995; Jones et al., 1996; Tchernichovski and Nottebohm, 1998), and has been explained as a consequence of birds copying incomplete songs from other pupils before song crystallization. Similarly to previous studies (Tchernichovski et al., 1999), total number of tutees (males or females) did not explain variation in song learning, suggesting that it is the behavior of males that affects learning and not the size of the group. In contrast to our results, a previous test of the developmental stress hypothesis found that zebra finch nestlings treated with corticosterone or whose parents had restricted access to food produced shorter songs and had smaller HVC volume than control birds (Spencer et al., 2003; Buchanan et al., 2004). We can identify at least three main differences between the studies that could explain this discrepancy. First, in our study, birds from different treatments were moved into groups

1610 Gil et al. with new tutors, while Spencer et al. (2003) continued housing the broods with the foster fathers. Thus, tutors and tutees were simultaneously exposed to the experimental treatment (Spencer et al., 2003; Buchanan et al., 2004). If males rearing experimental broods were affected by the stress imposed on the nestlings, their singing behavior may have been modified, possibly becoming poorer tutors. Second, the studies by Buchanan, Spencer, and colleagues used a manipulation of nutritional stress based on corticosterone administration and food availability instead of a brood size manipulation. It is difficult to know whether direct food and corticosterone manipulations result in the same kind of variation in condition as does manipulation of brood size. Especially the administration of corticosterone during the nestling phase is problematic. Although adults can cope with sudden increases in corticosteroid titers through a series of metabolic mechanisms (Breuner and Wingfield, 2000; Dufty et al., 2002), nestlings may not have developed this capacity in full and thus may have been exposed to unnaturally high levels of corticosterone. Indeed, the treatment produced a twofold increase in basal corticosterone levels (Spencer et al., 2003). Similarly, mean HVC size was much smaller than the normal range found in males of the species (e.g., Airey et al., 2000). Third, previous studies did not measure song learning but general song characteristics without relating them to the tutor s song (Spencer et al., 2003; Buchanan et al., 2004). However, since tutor song is so precisely copied in this species, song development should always be measured in relation to the particular tutor song that the young bird was exposed to. The study of Buchanan and colleagues (2004) could be taken to indicate that early severe stress affects the development of song areas in the zebra finch, although the required level of stress might be higher than the normal variation that birds are subject to in the wild. Nowicki and colleagues (2002) found in their study on swamp sparrows Melospiza georgiana effects of nutritional treatment on song copying accuracy and on the size of song control nuclei RA and HVC. However, the latter was due to the differences in the forebrain size of stressed males. Since the brain is an organ with a very high degree of developmental buffering against nutritional shortage (Schew and Ricklefs, 1998), it would also seem that the manipulation of the latter study was particularly strong. Thus, both studies show that early stress present before the onset of the sensory-motor period of song learning can affect song sensory learning. Alternatively, it should be considered that severe stress might not affect song sensory learning per se but modify physiological parameters such as sex hormone production that play a role in the differentiation of song control areas, therefore indirectly affecting song sensory-motor learning. Three previous studies on the relationship between song and brain anatomy in the zebra finch had found alternatively (a) weakly significant or no correlation between size of HVC and RA and song repertoire (MacDougall-Shackleton et al., 1998; Ward et al., 1998), (b) a strong correlation between HVC and song repertoire (Airey and DeVoogd, 2000), and (c) a strong correlation between these areas and the size of the learned part of the repertoire (Ward et al., 1998). Our data, based on a larger sample size than any previous study on brain song nuclei and song characteristics, provide no evidence for a positive correlation between HVC, RA, or LMAN size and either song repertoire size or song duration. In particular, we expected a correlation between RA volume and element repertoire in light of the correlated electrical activity of HVC- and RA-circuits with syllable and element identity, respectively (Yu and Margoliash, 1996; Hahnloser et al., 2002). At best, there was a nonsignificant trend for HVC volume to be positively related to the proportion of tutee song learnt from tutor, as found by Ward and coauthors. It is difficult to understand this discrepancy of results as far as a relationship between brain areas and song is concerned. A recent meta-analysis of withinspecies studies found that there is a positive relationship between HVC and song repertoire size (Garamszegi and Eens, 2004). This relationship is assumed to be based in a role of HVC in song control, which is suggested by studies that have found consistent differences in HVC volume between sexes, seasons, or species song repertoires (e.g., Nottebohm et al., 1981; DeVoogd et al., 1993; Airey and DeVoogd, 2000). However, several studies have shown inconsistencies in these patterns (for a review see Gil and Gahr, 2002), which could be due to song output not being a perfect correlate of learned song or to the fact that the unit of stored motor memories is unknown. To conclude, our study differs from previous tests of the developmental stress hypothesis in the zebra finch (Spencer et al., 2003; Buchanan et al., 2004) in carefully controlling additional factors important for learning, and by using an experimental paradigm that imposed variation in nestling condition within a natural range. Also, unlike most studies, we specifically tested that learning, and not just song characteristics, was condition dependent. Our interpretation is that, given our large sample size, if there is an effect of variation in early condition on song development in this species, it is weak at best. We feel that one of the most interesting conclusions that can be derived from our study is the overriding effect of social effects

Developmental Stress and Song Brain Nuclei 1611 over variation in early condition. The strong resemblance between tutor and tutee song suggests that for a young bird it is important to produce accurate tutor learning, rather than obtaining longer songs or larger repertoires. In addition, the large degree of variation in the size of brain song nuclei that we found was not related to variation in song repertoire, suggesting that further tests of this relationship should take into account additional hypotheses (Gahr et al., 1998; Gil and Gahr, 2002; Sartor and Ball, 2005). The authors thank Isabelle Bauthian, Christine Brenninkmeyer, and Sandra Wilhelm for help in bird rearing, Edda Geissler for recording the birds, and Roger Cue for statistical advice. APPENDIX Table A1 Table with Mean (SE) of all Measured Variables in Experimental Birds, per Experimental Group Small Broods Medium Broods Large Broods Nestling tarsus (mm) 12.43 (0.26) 11.98 (0.24) 11.97 (0.25) Nestling mass (g) 8.55 (0.64) 8.03 (0.38) 6.29 (0.31) Mass when sacrificed (g) 11.07 (0.21) 11.02 (0.22) 10.11 (0.3) Brain mass (mg) 475.1 (12.3) 487.2 (6.06) 459.2 (6.49) HVC volume (mm 3 ) 0.32 (0.02) 0.31 (0.01) 0.32 (0.02) RA volume (mm 3 ) 0.22 (0.01) 0.21 (0.01) 0.21 (0.01) lman volume (mm 3 ) 0.1 (0.001) 0.09 (0.001) 0.09 (0.001) Song duration (ms) 771.25 (79.36) 800.2 (56.92) 742 (72.78) Number of elements 9.5 (0.93) 9.65 (0.49) 8.64 (0.95) Element repertoire 8.58 (0.83) 9.3 (0.46) 7.86 (0.69) Matching elements 8.33 (0.8) 8.4 (0.46) 7.5 (0.63) Learning score 0.84 (0.06) 0.89 (0.03) 0.84 (0.05) Similarity score 0.97 (0.01) 0.91 (0.03) 0.96 (0.02) Song rate (songs/min) 1.98 (1.07) 1.91 (0.55) 1.03 (0.22) n ¼ 12 n ¼ 20 n ¼ 14 REFERENCES Airey DC, Castillo-Juarez H, Casella G, Pollak EJ, DeVoogd TJ. 2000. Variation in the volume of zebra finch song control nuclei is heritable: Developmental and evolutionary implications. Proc R Soc London Ser B 267:2099 2104. Airey DC, DeVoogd TJ. 2000. Greater song complexity is associated with augmented song system anatomy in zebra finches. Neuroreport 11:2339 2344. Brenowitz EA, Lent K, Kroodsma DE. 1995. Brain space for learned song in birds develops independently of song learning. J Neurosci 15:6281 6286. Breuner CW, Wingfield JC. 2000. Rapid behavioral response to corticosterone varies with photoperiod and dose. Horm Behav 37:23 30. Buchanan KL, Leitner S, Spencer KA, Goldsmith AR, Catchpole CK. 2004. Developmental stress selectively affects the song control nucleus HVC in the zebra finch. Proc R Soc London Ser B 271:2381 2386. Buchanan KL, Spencer KA, Goldsmith AR, Catchpole CK. 2003. Song as an honest signal of past developmental stress in the European starling (Sturnus vulgaris). Proc R Soc London Ser B 270:1149 1156. Catchpole CK, Slater PJB. 1995. Bird Song: Biological Themes and Variations. Cambridge, UK: Cambridge University Press, p 248. Clayton NS. 1987. Song tutor choice in zebra finches. Anim Behav 35:714 721. Coleman RM, Whittall RD. 1988. Clutch size and the cost of incubation in the Bengalese finch (Lonchura striata var. domestica). Behav Ecol Sociobiol 23:367 372. De Kogel CH. 1997. Long-term effects of brood size manipulation on morphological development and sexspecific mortality of offspring. J Anim Ecol 66:167 178. De Kogel CH, Prijs HJ. 1996. Effects of brood size manipulations on sexual attractiveness of offspring in the zebra finch. Anim Behav 51:699 708. DeVoogd TJ, Krebs JR, Healy SD, Purvis A. 1993. Relations between song repertoire size and the volume of brain nuclei related to song: Comparative evolutionary analyses amongst oscine birds. Proc R Soc London Ser B 254:75 82. Dufty AM, Clobert J, Møller AP. 2002. Hormones, developmental plasticity and adaptation. Trends Ecol Evol 17:190 196. Gahr M. 1997. How should brain nuclei be delineated? Consequences for developmental mechanisms and for correlations of area size, neuron numbers and functions of brain nuclei. Trends Neurosci 20:58 62. Gahr M, Sonnenschein E, Wickler W. 1998. Sex difference in the size of the neural song control regions in a dueting songbird with similar song repertoire size of males and females. J Neurosci 18:1124 1131.

1612 Gil et al. Garamszegi LZ, Eens M. 2004. Brain space for a learned task: Strong intraspecific evidence for neural correlates of singing behavior in songbirds. Brain Res Rev 44:187 193. García-Berthou E. 2001. On the misuse of residuals in ecology: Testing regression residuals vs. the analysis of covariance. J Anim Ecol 70:708 711. Gil D, Gahr M. 2002. The honesty of bird song: Multiple constraints for multiple traits. Trends Ecol Evol 17:133 141. Gil D, Heim C, Bulmer E, Rocha M, Puerta M, Naguib M. 2004. Negative effects of early developmental stress on adult yolk testosterone levels in a passerine bird. J Exp Biol 207:2215 2220. Grafen A. 1990. Biological signals as handicaps. J Theor Biol 144:517 546. Hahnloser RHR, Kozhevnikov AA, Fee MS. 2002. An ultra-sparse code underlies the generation of neural sequences in a songbird. Nature 419:65 70. Jones AE, ten Cate C, Slater PJB. 1996. Early experience and plasticity of song in adult male zebra finches (Taeniopygia guttata). J Comp Psychol 110:354 369. Korte SM, Koolhaas JM, Wingfield JC, McEwen BS. 2005. The Darwinian concept of stress: Benefits of allostasis and costs of allostatic load and the trade-offs in health and disease. Neurosci Biobehav Rev 29:3 38. Leitner S, Catchpole CK. 2004. Syllable repertoire and the size of the song control system in captive canaries (Serinus canaria). J Neurobiol 60:21 27. Leitner S, Voigt C, Garcia-Segura LM, Van t-hof T, Gahr M. 2001. Seasonal activation and inactivation of song motor memories in wild canaries is not reflected in neuroanatomical changes of forebrain song areas. Horm Behav 40:160 168. MacDougall-Shackleton SA, Hulse SH, Ball GF. 1998. Neural correlates of singing behaviour in male zebra finches (Taeniopygia guttata). J Neurobiol 36:421 430. Mann NI, Slater PJB. 1994. What causes young male zebra finches, Taeniopygia guttata, to choose their father as song tutor? Anim Behav 47:671 677. Naguib M, Gil D. 2005. Transgenerational body size effects caused by early developmental stress in zebra finches. Biol Lett 1:95 97. Naguib M, Riebel K, Marzal A, Gil D. 2004. Nestling immunocompetence and testosterone covary with brood size in a songbird. Proc R Soc London Ser B 271:833 838. Naguib M, Nemitz A, Gil D. 2006. Maternal developmental stress reduces reproductive success of female offspring in zebra finches. Proc R Soc London Ser B 273:1901 1905. Neuenschwander S, Brinkhof MWG, Kölliker M, Richner H. 2003. Brood size, sibling competition, and the cost of begging in great tits (Parus major). Behav Ecol 14:457 462. Nottebohm F, Kasparian S, Pandazis C. 1981. Brain space for a learned task. Brain Res 213:99 109. Nottebohm F, Stokes TM, Leonard CM. 1976. Central control of song in canary, Serinus canarius. J Comp Neurol 165:457 486. Nowicki S, Peters S, Podos J. 1998. Song learning, early nutrition and sexual selection in songbirds. Am Zool 38:179 190. Nowicki S, Searcy WA, Peters S. 2002. Brain development, song learning and mate choice in birds: A review and experimental test of the \nutritional stress hypothesis". J Comp Physiol A 188:1003 1014. Rice WR, Gaines SD. 1994. Extending nondirectionaly heterogeneity tests to evaluate simply ordered alternative hypothesis. Proc Natl Acad Sci USA 91:225 226. Riebel K. 2003. The \mute" sex revisited: Vocal production and perception learning in female songbirds. Adv Study Behav 33:49 86. Sartor JJ, Ball GF. 2005. Social suppression of song is associated with a reduction in volume of a song-control nucleus in European starlings (Sturnus vulgaris). Behav Neurosci 119:233 244. Schew WA, Ricklefs RE. 1998. Developmental plasticity. In: Starck JM, Ricklefs RE, editors. Avian Growth and Development. New York: Oxford University Press, pp 288 304. Searcy WA, Yasukawa K. 1996. Song and female choice. In: Kroodsma DE, Miller EH, editors. Ecology and Evolution of Acoustic Communication in Birds. Ithaca, NY: Cornell University Press, pp 454 473. Slater PJB, Eales LA, Clayton NS. 1988. Song learning in zebra finches (Taeniopygia guttata): Progress and prospects. Adv Study Behav 18:1 34. Slater PJB, Mann NI. 1990. Do male zebra finches learn their fathers songs? Trends Ecol Evol 5:415 417. Spencer KA, Buchanan KL, Goldsmith AR, Catchpole CK. 2003. Song as an honest signal of developmental stress in the zebra finch (Taeniopygia guttata). Horm Behav 44:132 139. Stearns SC. 1992. The Evolution of Life Histories. Oxford, UK: Oxford University Press, p 249. Tchernichovski O, Lints T, Mitra PP, Nottebohm F. 1999. Vocal imitation in zebra finches is inversely related to model abundance. Proc Natl Acad Sci USA 22:12901 12904. Tchernichovski O, Nottebohm F. 1998. Social inhibition of song imitation among sibling male zebra finches. Proc Natl Acad Sci USA 95:8951 8956. Tchernichovski O, Nottebohm F, Ho CE, Pesaran B, Mitra PP. 2000. A procedure for an automated measurement of song similarity. Anim Behav 59:1167 1176. Volman SF, Khanna H. 1995. Convergence of untutored song in group reared zebra finches (Taeniopygia guttata). J Comp Psychol 109:211 221. Vu ET, Mazurek ME, Kuo YC. 1994. Identification of a forebrain motor programming network for the learned song of zebra finches. J Neurosci 14:6924 6934. Ward BC, Nordeen EJ, Nordeen KW. 1998. Individual variation in neuron number predicts differences in the propensity foravianvocalimitation. ProcNatlAcadSciUSA95:1277 1282. Yu AC, Margoliash D. 1996. Temporal hierarchical control of singing in birds. Science 273:1871 1875. Zahavi A. 1975. Mate selection: A selection for a handicap. J Theor Biol 53:205 214. Zann R. 1990. Song and call learning in wild zebra finches in South East Australia. Anim Behav 40:811 828.