1 Prepared for Hormone and Behavior 1 Does the syrinx, a peripheral structure, constrain effects of sex 2 steroids on behavioral sex reversal in adult canaries? 3 Ednei B. dos Santos1 (ORCID: 0000-0001-6772-7624), David M. Logue2 (ORCID: 0000-0003-3020-7101), 4 Gregory F. Ball3 (ORCID: 0000-0002-4784-0520), Charlotte A. Cornil1 (ORCID: 0000-0002-5536-7753), Jacques 5 Balthazart1* (ORCID:0000-0001-9492-2126) 6 7 1GIGA Neurosciences, Laboratory of Behavioral Neuroendocrinology, University of Liege, Belgium 8 2Department of Psychology, University of Lethbridge, Lethbridge AB, Canada. 9 3Department of Psychology, University of Maryland, College Park MD, USA. 10 11 Short title: syrinx structure constraints sex difference in trilling 12 13 14 15 16 17 18 19 20 * Corresponding author 21 Email: Jbalthazart@uliege.be (JB) 22 23 2 24 25 Abstract 26 We previously confirmed that effects of testosterone (T) on singing activity and on the volume 27 of brain song control nuclei are sexually differentiated in adult canaries: females are limited in 28 their ability to respond to T as males do. Here we expand on these results by focusing on sex 29 differences in the production and performance of trills, i.e., rapid repetitions of song elements. 30 We analyzed more than 42,000 trills recorded over a period of 6 weeks from 3 groups of 31 castrated males and 3 groups of photoregressed females that received SilasticTM implants 32 filled with T, T plus estradiol or left empty as control. Effects of T on the number of trills, trill 33 duration and percent of time spent trilling were all stronger in males than females. Irrespective 34 of endocrine treatment, trill performance assessed by vocal deviations from the trill rate 35 versus trill bandwidth trade-off was also higher in males than in females. Finally, inter-36 individual differences in syrinx mass were positively correlated with specific features of trills in 37 males but not in females. Given that T increases syrinx mass and syrinx fiber diameter in 38 males but not in females, these data indicate that sex differences in trilling behavior are 39 related to sex differences in syrinx mass and syrinx muscle fiber diameter that cannot be fully 40 suppressed by sex steroids in adulthood. Sexual differentiation of behavior thus reflects 41 organization not only of the brain but also of peripheral structures. 42 43 44 45 Keywords: sex differences – vocal behavior – sexual differentiation – behavioral constraints - 46 testosterone 47 48 49 50 3 51 52 Introduction 53 Sexual reproduction is often preceded by elaborate courtship displays of males that are used 54 by females to assess the quality of potential mates. Bird song is a textbook example of such 55 displays (Collins, 2004). Several song features that are thought to be reliable signals of male 56 quality in songbirds have been identified in various species. These features include song 57 repertoire size (Catchpole and Slater, 2018), stereotypy (Sakata and Vehrencamp, 2012), and 58 the occurrence of ‘special’ syllables (Rehsteiner et al., 1998). By mating with males that 59 produce songs displaying these characteristics, females can potentially obtain greater direct 60 benefits, such as better parental care and access to larger territories with more resources, or 61 indirect genetic benefits (better genes) for their offspring (Andersson, 1994; Kirkpatrick and 62 Ryan, 1991). 63 Vocal features that are physically challenging to produce can also indicate male quality 64 and these features have often been grouped under the term of vocal performance (Goller, 65 2022; Podos et al., 2009). Vocal performance has traditionally been measured as the rate of 66 vocal activity (number of vocalizations per unit time) but it can also be assessed by vocal 67 consistency (the degree of variation across renditions of the same song elements) and by the 68 fine-scale structure of trills (rapid repetitions of song elements in sequence). Podos proposed 69 that physiological constraints result in a trade-off between trill rate and trill frequency 70 bandwidth in the song of over 30 species of sparrows (Emberizidae) (Podos, 1997). Similar 71 trade-offs have since been reported in numerous other species (reviewed in (Goller, 2022; 72 Podos et al., 2009; Wilson et al., 1997). These tradeoffs emerge because there is a limit to 73 the speed with which a bird can modulate the fundamental frequency of its song. Both trill rate 74 and frequency bandwidth rely on frequency modulation over time. Therefore, as a bird 75 approaches the limit of frequency modulation speed, any increase in trill rate must be offset 76 by a decrease in frequency bandwidth, and vice versa. When plotted in an acoustic space 77 defining trill bandwidth as a function of trill rate, the distribution of trills forms a right triangle 78 with the hypotenuse defining a putative performance limit. Trills that are orthogonally closer to 79 this optimal limit have lower vocal deviation (i.e., higher performance) and are more 80 challenging to sing because they require faster movements or better coordination of the 81 respiratory system, syringeal muscles and the vocal tract. 82 4 In canaries (Serinus canaria), females exhibit copulation solicitation displays more 83 often and deposit more testosterone in eggs when exposed to playbacks of male songs 84 containing trills composed of broad frequency bandwidth two-note syllables repeated at a fast 85 rate of at least 15 elements s-1 that were labeled ‘sexy’ or ‘A’ syllables (Vallet et al., 1998; 86 Vallet and Kreutzer, 1995). Suthers and colleagues investigated the patterns of airflow in the 87 syrinx (the avian vocal organ) during song production and demonstrated that canaries employ 88 two different respiratory mechanisms to produce trills (Suthers et al., 2012). The rapid 89 production of syllables is normally constrained by the respiratory system that limits the rate at 90 which air can be replaced. Canaries circumvent this limitation by taking ‘mini-breaths’ 91 between syllables, allowing them to sing trills of protracted duration. However, to produce 92 faster trills canaries shift to ‘pulsatile respiration’, in which expiration is sustained for the entire 93 duration of trills, thus limiting trill duration to the available air supply. Thus, long sexy trills are 94 both preferred by females and challenging for males to produce. 95 The two independent sound sources in the syrinx associated with the left and right 96 bronchi are controlled by a minimum of 4 pairs of muscles that are among the fastest known 97 muscles in vertebrates (Elemans et al., 2008). The activity of these muscles is modulated by 98 steroid hormones. Both androgen receptors and estrogen receptor β are expressed in the 99 syringeal muscles of male and female zebra finches (Taeniopygia gutatta) (Veney and Wade, 100 2005). Furthermore, a study in male canaries in which syringeal androgen receptors were 101 blocked with bicalutamide (an androgen receptor antagonist that does not cross the blood-102 brain barrier) reported a significant decrease in syrinx mass associated with a reduction in trill 103 performance and complexity (Alward et al., 2016). Syrinx muscle fibers are additionally 104 affected by singing activity and muscles training (Adam and Elemans, 2019). Sex differences 105 in syrinx mass and muscle fiber composition have also been reported in several songbird 106 species (Christensen et al., 2017; Prince et al., 2011; Wade and Buhlman, 2000). However, 107 studies directly relating sex differences in syrinx mass or anatomy to sex differences in 108 singing behavior are scarce. 109 Canaries have frequently been used as a model to study effects of sex steroids on 110 singing behavior (see reviews in (Ball and Balthazart, 2007; Schlinger and Brenowitz, 2017)). 111 In a previous study we compared the role of testosterone (T) on singing activity and brain 112 song control nuclei anatomy in both males and females (Dos Santos et al., 2022). That study 113 confirmed that the sexes respond differentially to T and there is a limit in the capacity of 114 5 females to respond to T in the same manner as males. Here we expand on these results by 115 focusing on sex differences in the production and performance of trills. First, we analyze in 116 detail the effects of T on trilling activity and trill features. Then we assess sex differences in T 117 effects on performance measured by vocal deviations from the trill rate versus trill bandwidth 118 trade-off. Finally, we explore how sex and inter-individual differences in syrinx mass and 119 syringeal muscle fiber diameter relate to trill acoustic features and performance. Although a 120 number of studies have demonstrated that expression of a given behavior can be limited by 121 peripheral structures. However since the discovery of sex differences in brain structure and 122 function in songbirds (Nottebohm and Arnold, 1976) and then in many other species (Tobet 123 and Fox, 1992), it is usually at the brain level that answers are searched to explain sex 124 differences in behavior. Based on what was already known about syrinx physiology, we 125 anticipated that relationships might be discovered between measures of syrinx volume or 126 syringeal muscles and some features of the trills, which represent in terms of complexity a 127 very challenging part of the song repertoire. 128 129 Material and methods 130 The birds used and the song recordings that were analyzed in the present study are the same 131 of those that were used in our previously published study on general song production (Dos 132 Santos et al., 2022). One-year old Fife Fancy canaries (24 females and 24 males) were 133 acquired from a local commercial breeder in Belgium and sexed by PCR at the Behavioral 134 Ecology and Ecophysiology lab of the University of Antwerp, Belgium (Griffiths et al., 1998). 135 All birds were housed in groups of six in visually (but not acoustically) isolated cages under an 136 8 L: 16 D photoperiod at the animal facility of University of Liege, Belgium, with food, water, 137 bath and grit ad libitum during the entire experiment. They were also fed egg food twice per 138 week. 139 All experimental procedures complied with Belgian laws concerning the Protection and 140 Welfare of Animals and the Protection of Experimental Animals, and experimental protocols 141 were approved by the Ethics Committee for the Use of Animals at the University of Liege 142 (Protocol number 2027). 143 144 General procedure 145 6 Six weeks after their arrival in the laboratory, all males were castrated and females were 146 laparotomized under general isoflurane anesthesia to confirm photoregression of their ovary 147 as described before (Dos Santos et al., 2022; Goldman and Nottebohm, 1983; Hartog et al., 148 2009; Madison et al., 2015; Shevchouk et al., 2017). Adult female canaries with regressed 149 gonads are commonly used to investigate the effects of T on song system neuroplasticity 150 (Hartog et al., 2009; Louissaint et al., 2002; Yamamura et al., 2011). Three weeks later, all 151 birds were weighed and the size of their cloacal protuberance area, an androgen-dependent 152 structure (Alward et al., 2013; Appeltants et al., 2003; Tramontin et al., 2003) was measured 153 with calipers before they received subcutaneous implants made of Silastic™ tube (Dow 154 corning, Midland, MI, USA; Degania Silicone; internal diameter 0.76 mm, external diameter 155 1.65 mm, length 12 mm) filled over a 10 mm length with either crystalline T or estradiol (E2) or 156 left empty as a control (C). These implants maintain concentrations of T and E2 that are in the 157 high physiological range of behaviorally active of doses (Appeltants et al., 2003; Cornez et al., 158 2020; Leboucher et al., 1994; Madison et al., 2015; Sartor et al., 2005). 159 Birds from each sex were randomly assigned to three experimental groups that 160 received either two empty implants (C group), one T and one empty implant (T group) or one 161 T and one E2 implant (T+E2 group). Birds were then moved into 8 cages (4 with males and 4 162 with females), each housing two C, two T and two T+E2 birds 163 We included a group treated with E2 in addition to T because previous studies have 164 reported that aromatase activity and that the induction of this enzymatic activity by T is lower 165 in the female than in the male brain and this consequently might explain the limited response 166 of females to exogenous T (for more detail, see (Dos Santos et al., 2022) ). 167 168 Song recording 169 Once every week for 6 weeks, birds were moved overnight individually inside custom-built 170 sound-attenuated boxes. Vocalizations produced by each bird were recorded for 3 h starting 171 immediately after lights on (0900 h) on the next morning using custom-made microphones 172 (Projects Unlimited/Audio Products Division) and an Allen & Heath ICE-16 multichannel 173 recorder.. 174 Sound files were acquired and saved as a .wav file by Raven v1.4 software 175 (Bioacoustics Research Program 2011; Raven Pro: Interactive Sound Analysis Software, 176 Version 1.4, Ithaca, NY: The Cornell Lab of Ornithology) at a sampling frequency of 44,100 Hz. 177 7 Sound files were then analyzed with a MATLAB script developed for canary song analysis by 178 Ed Smith and Robert Dooling (Department of Psychology, University of Maryland at College 179 Park, MD). Songs were defined as vocalizations at least 30 dB above background noise that 180 were at least 1 s long, and were preceded and followed by at least 0.4 s of silence. The 181 present report is based on these recordings (24 male and 24 female Fife fancy canaries). 182 Recordings were already used to analyze the sexually differentiated effects of sex steroids on 183 singing behavior and brain plasticity. These results have been published separately (Dos 184 Santos et al., 2022). 185 186 Tissue collection 187 Six weeks after the beginning of the steroid treatment, birds were weighed, their cloacal 188 protuberance area was measured again and they were deeply anaesthetized before being 189 euthanized by decapitation. Their brain was dissected out of the skull and fixed with 5% 190 acrolein, cryoprotected in sucrose, frozen on dry ice and stored at -80°C (See (Dos Santos et 191 al., 2022) for more detail). The syrinx of each subject was also extracted, fixed with acrolein, 192 frozen on dry ice and stored at -80°C until used. 193 194 Tissue processing 195 All syrinxes were defrosted and dissected by one cut of the trachea just dorsal to the 196 syringeal muscles and two cuts ventral to the third cartilaginous ring of the bronchi. Each 197 syrinx was then weighed to the nearest milligram. They were then cut with the cranial aspect 198 facing up on a cryostat (Thermo ScientificTM CryoStarTM NX70) in 15 µm coronal sections 199 that were mounted on superfrost microscopic slides. Sections were stained with hematoxylin-200 eosin and coverslipped with permount. The section containing the largest cross-sectional area 201 was photographed under a microscope at a 10X objective for each bird. A square (200x200 202 µm) was overlaid in each of the 4 quadrants of the syrinx photomicrograph (quadrants 1 and 203 2, left and right ventral side, including the muscles tracheobronchealis ventralis and 204 syringealis ventralis ; quadrants 3 and 4, left and right dorsal side, including the muscles 205 tracheobronchealis dorsalis and syringealis dorsalis) using the FIJI version of ImageJ 206 software (Schindelin et al., 2012). All fibers located within the square were identified and 207 measured (n= 3,401 fibers for all birds). The widest and narrowest diameter of each fiber was 208 measured. These 2 measures were averaged and considered as the mean diameter which 209 8 was then averaged across all quadrants and each bird before statistical analyses. The 210 distribution of muscle diameters within each experimental group was also investigated. 211 212 Trill analysis 213 Trills were identified within songs and defined as sequences of similar elements repeated at 214 least 4 times and separated by silence intervals of no more than 0.003 s. with the help of a 215 MATLAB routine especially developed by Ed. Smith and Robert Dooling (Department of 216 Psychology, University of Maryland at College Park) for the analysis of canary song. The 217 script computed the following metrics: total number of trills, percent time spent trilling (within 218 songs), trill duration, number of segments (per trill), segment duration, number of fast trills 219 (trills with a rate of at least 17 syllables s-1), interval duration (between segments), spectral 220 distance (between segments), trill entropy, trill bandwidth, trill center frequency and trill power. 221 Trill entropy reported here is the natural logarithm of the Wiener entropy based on the 222 power spectrum of the trill. A narrow power spectrum as present in a pure tone has a large 223 negative value while a broad band white noise has an entropy closer to zero. Entropy is thus 224 a measure of variability within the trill. The bandwidth is the frequency interval in Hertz 225 between the highest and lowest frequency of the vocalization that contains 90% of the total 226 trill power. The trill power reported here is a relative measure of power, expressed in decibels, 227 and can only be used to compare trills. Assessment of absolute power would require 228 measures of microphone and recorder sensitivity, preamplifier gain and information about 229 recording conditions such and distance and orientation with respect to the microphone. 230 One female bird died of natural causes during the experiment and was recorded only 231 for a 5-week period. 232 233 Performance analysis 234 Because the production of trills might be constrained by a number of physical factors, we also 235 quantified the trill rate (TR) vs. mean trill bandwidth (TBW) trade-offs, while controlling for 236 variation attributable to individuals. Trill rate (TR) was calculated as the number of segments 237 minus one divided by the time from the beginning of the first segment to the beginning of the 238 last segment. The final segment was excluded from this measure because it is impossible to 239 define the duration of the silent gap that follows this last segment and this would bias 240 estimates of trill rate for song with fewer segments. Mixed quantile regressions (99th) were 241 9 used to test for acoustic trade-offs (Logue et al., 2020; Wilson et al., 1997). A quantile 242 regression analysis generates a linear function to estimate a defined quantile variable Y over 243 a range of X (Cade and Noon, 2003). All models were run with both random intercepts and 244 random slopes, and to account for the non-independence of multiple data points from the 245 same individuals, we used individual id as a random variable. We then calculated deviation 246 scores (DS) as the orthogonal distance from the quantile regression line (Podos, 2001) 247 defined by: DS=(TWB intercept + TBW slope X TR –TBW)/sqrt (1+TBWslope2) X -1 248 249 Statistical analysis 250 Unless otherwise mentioned, data were analyzed by one- or two-way analyses of variance 251 (ANOVA) or by two-way mixed model analysis if a few data points were missing with the three 252 experimental groups, as appropriate. Post hoc comparisons were performed with the Tukey’s 253 multiple comparisons test. Nominal data were analysed by the 2 and Fisher exact probability 254 tests. Statistical analyses were performed using R Studio (Team, 2021) and GraphPad Prism 255 version 8.4 for Mac (GraphPad Software, San Diego, California USA). We used the lqmm 256 package (Geraci, 2014) to calculate mixed quantile regressions, and the package ggplot2 257 (Wickham and Chang, 2008) to graph the trill distributions with semi-transparent points and to 258 fit the mixed quantile regression line. Effect sizes (partial eta square p2) are represented by 259 the ratios of the relevant sums of squares in the two way ANOVA. We used an alpha level 260 of .05 for all statistical tests. All data are represented here by their mean ± SEM and when 261 feasible, individual data points are also presented. 262 263 Results 264 265 Analysis of trills 266 A total 864 hours of recordings (3h per week for 6 weeks for 48 experimental subjects) were 267 analyzed by the MATLAB script that detected and quantitatively characterized a total of 268 42,198 trills. Twelve dependent variables were analyzed by three types of two-way ANOVAs. 269 Data for each dependent variable were first averaged for each subject across the 6 weeks of 270 experiment and analyzed by a two-way ANOVA with endocrine treatment (3 conditions: C, T 271 and T+E2) and sex (male or female) as independent factors (Fig. 1, Fig. S1 and Table S1). 272 10 Over the entire experiment, steroid treatments significantly affected many aspects of 273 the trills including their total number, percent time trilling, trill duration (Fig. 1) and also the 274 number of segments per trill, the segment duration and the number of fast trills containing 275 more than 17 elements per second (Fig. S1; see Table S1 for detail of statistical analyses). In 276 most cases, posthoc analyses identified, as expected, significantly lower values in the C 277 compared to the T and/or the T+E2 groups (see detail in Fig. 1 and Fig. S1). Somewhat 278 surprisingly, these analyses also indicated a significantly lower total number of trills in the 279 T+E2 groups compared to the T groups. 280 The ANOVAs also detected significant sex differences for three variables: the total 281 number of trills, percent time trilling, and trill duration (Fig. 1, Table S1). A significant 282 interaction between treatment and sex was observed for trill duration only (sex difference was 283 larger in the T+E2 condition than in the two other conditions). 284 0 200 400 600 800 Number of trills 0 10 20 30 40 % time triling 0 500 1000 Trills duration Trt: *** Sex: ** Trt x Sex: ns Trt: *** Sex: *** Trt x Sex: * Trt: *** Sex: ** Trt x Sex: ns To ta l n u m b er P e rc e n ta g e M ill is e co n d s aaa aaa aaa aaa a,b C T+E2T Males Females C T+E2TC T+E2T 285 Fig. 1. Number of trills (A), trill duration (B) and percentage of time spent trilling (C) in male and 286 female canaries that were treated with Silastic™ implants left empty as control (C) or filled with 287 testosterone (T) or with testosterone plus estradiol (T+E2) . Bar graphs represent the mean ± SEM of 288 individual results that are the average of data collected during the 6 weeks of recording. Individual data 289 points are also indicated. Data were analyzed by two-way ANOVA with treatment (Trt) and Sex of the 290 subjects as independent factors and results are summarized in the insert for each panel. (***=p<0.001, 291 **=p<0.01, *=p<0.05, ns= not significant). Significant effects of treatment were further analyzed by Tukey’s 292 multiple comparison tests and their results are expressed as follows: aaa= p<0.001 versus C group, a (or b)= 293 p<0.05 versus C (or T) group. 294 295 Initially, females and castrated males did not produce trills; trills appeared 296 progressively during the treatment with steroids (Fig S2). Separate two-way ANOVAs within 297 each sex with time and treatment as factors confirmed the presence of treatment and time 298 11 effects for several trill features in males (Fig. S2 and Table S2). Furthermore, two-way 299 ANOVAs within each treatment with time (weeks) and sex as factors) confirmed the presence 300 of a sex difference (males > females) in trill numbers, percent time trilling and trill duration in 301 the T+E2 birds (Fig. S2 and Table S3). 302 303 Trill rate (TR) versus Trill Bandwidth (TBW) trade-off and performance analysis 304 The analysis of the 42,198 trills produced by all experimental birds across the 6 weeks of 305 experiment identified a trade-off between TR and mean TBW, presumably reflecting 306 physiological limits on frequency modulation and respiration in canary trilling activity (Fig. 2). 307 After transformation of data into z-scores to reduce the overall variance, a quantile regression 308 analysis on the pooled data from the 35 subjects that produced trills during the experiment 309 indicated the presence of a statistically significant negatively sloping upper boundary for TR 310 vs. TBW (intercept = 2.98, slope = −1.6, pslope < 0.001, Fig. 2A). 311 The deviation score (orthogonal distance from the regression line) for each individual 312 trill was then calculated, averaged within each subject and these individual scores were then 313 compared by a two-way mixed model ANOVA with the sex and treatment of the birds as 314 independent factors. This analysis identified a significant sex difference in the deviation 315 scores (Fig. 2B), with males on average having a smaller orthogonal distance from the 316 quantile regression line (F1,29= 4.487, p=0.043, p2=0.21). There was however no effect of the 317 endocrine treatments (F2,29= 0.460, p=0.636, p2=0.03) and no sex by treatment interaction 318 (F2,29= 0.174, p=0.841, p2=0.01) related to these scores. 319 In addition, deviation scores were more variable in males than in females reflecting the 320 fact that male trills were more broadly distributed across the entire acoustic space. Analysis of 321 the deviation score coefficients of variation accordingly revealed an overall sex difference 322 (F1,29= 7.778, p=0.009, p2=0.21), but again no significant effect of treatment (F2,29= 1.750, 323 p=0.192, p2=0.11) and no interaction between sex and treatments (F2,29= 0.034, p= 0.966, 324 p2<0.01). 325 326 12 0 20 40 60 Deviation Score CV M F M FM F % 2.0 2.5 3.0 3.5 4.0 4.5 Deviation Score M F M FM F z s co re Trt: ns Sex: ** Trt x Sex: ns Trt: ns Sex: * Trt x Sex: ns T T+E2C T T+E2C B 327 328 Fig. 2. Trade-off between trill rate and trill bandwidth in songs of male (M) and female (F) canaries that 329 were treated with Silastic implants filled with testosterone (T), with testosterone plus estradiol (T+E2) or 330 left empty as control (C). Panel (A) shows the position of each of the individual trills with respect to the quantile 331 regression line with male data points indicated in turquoise and female points in red. Panel (B) shows the mean 332 ± SEM deviation scores and coefficient of variation of these scores for each bird in the six experimental groups. 333 Results of the two-way ANOVA of these data are summarized in the insert (*=p<0,05, **=p<0,01, ns= non-334 significant). 335 336 13 One interesting feature of data illustrated in Fig. 2A is that the points representing 337 individual trills were not randomly distributed in space but rather were grouped in relatively 338 discrete clusters. Additional plots of these trills in the TR vs. TBW space indicated that these 339 clusters are different for males and females (Fig. S3) as well as in individual subjects (Fig. 340 S4), suggesting they relate to different types of syllables used in the trills. This hypothesis 341 was confirmed by the qualitative analysis of spectrograms presented in Figure 3. 342 343 Fig. 3. Example of spectrograms from a typical song in a male (A) and a female (B) canary illustrating the 344 repetition of a same element within long trills in females and use of various elements in a male. In the 345 trade-off plots, female or384 displays a single cluster of data points while male bk020 shows multiple clusters 346 corresponding to the different elements of the trills. Some of these elements are easily identified by the combina-347 tion of their bandwidth and rate (circles and arrows). 348 14 349 Effects on syrinx mass and structure 350 In a previous study, we showed that T or T+E2 increase the volume of three song control 351 nuclei (HVC, used as a proper name,, RA, the robust nucleus of the arcopallium; and Area X 352 of the basal ganglia) in both sexes but values remained significantly smaller in females than in 353 males (Dos Santos et al., 2022). In parallel, T increased syrinx mass in males but not in 354 females (Dos Santos et al., 2022). 355 Correlatively, the syrinx muscles fiber diameter was differentially increased by T in 356 males and females (Fig. 4D) (Effect of treatment: F2,32= 3.87, p<0.001, p2=0.069; effect of 357 sex F1,32= 58.05, p<0.001, p2=0.64; interaction: F2,32= 16.40, p<0.001, p2=0.51). While 358 syrinx fibers average diameter was similar in control males and females (Post hoc Sidak’s 359 multiple comparisons: p=0.999), this diameter was significantly larger in males than in 360 females in the two T-treated groups (Sidak test: p<0.001 in both cases Fig. 4D). A similar 361 differential increase in males and females of the syrinx fiber diameter was actually observed 362 for each of the 4 quadrants of the syrinx (see Fig. S5A). Correlatively this increased fiber size 363 resulted in a decrease in the density of fibers (numbers per unit surface) in the two male 364 groups treated with T (Effect of treatment: F2,32= 7.77, p=0.002, p2=0.33; effect of sex F1,32= 365 6.48, p=0.02, p2=0.17; interaction: F2,32= 2.61, p=0.089, p2=0.14; Fig. S5B). 366 Plotting the distribution of all fibers measured in the 4 quadrants in each group 367 separately revealed a shift of the distribution to the right (towards larger diameters) for the two 368 groups of males that had been treated with T (T and T+E2; Fig. 4E) while this effect was not 369 present in females; their distribution still overlapped with the distribution in control birds, never 370 or very rarely reaching a size over 45 µm. This presence of larger fibers concerned all males 371 in the T groups: fibers with a mean diameter larger than 45 µm were observed in 6 males out 372 of 6 in the T and T+E2 groups. In contrast these larger fibers were completely absent in 373 control birds (0 out of 7 in MC and FC) and rare in females treated with T (1/7 in females T 374 and 2/5 in females T+E2). Overall, this distribution was significantly different from random (2 375 test= 29.39, df=5, p<0.001). Analyses confined to each treatment separately indicated the 376 presence of a significant difference between T males and T females (6/6 vs. 1/T ; Fisher exact 377 probability test : p=0.005) but not in T+E2 birds (6/6 vs 2/5, p=0.061) nor in controls (0/T vs. 378 0/7 ; p>0.999). 379 15 The number of large fibers (>45 µm) detected in the quantification square (200 x 200 380 µm) was similarly affected by the sex of the birds (F1,32=49.25, p<0.001, p2=0.61), the 381 endocrine treatments (F2,32=14.56, p<0.001, p2=0.48) and the interaction between the 2 382 factors (F2,32=13.16, p<0.001, p2=0.45). Posthoc Tukey tests within each sex demonstrated a 383 significant increase in the two males groups treated with T by comparison with the controls 384 (C :0±0, T :10.17±2,65, T+E2 : 9.83±1.19, p<0.001 in both cases) while such an effect was 385 not observed in females (C :0±0, T :0.14±0.14, T+E2 : 0.40±10.24, p<>0.968 in both cases). 386 387 0 10 20 30 40 50 F ib e r d ia m et e r (µ M ) Trt: *** Sex: *** Trt x Sex:*** aaa aaa Ventral Dorsal Left Right 50 µm A B C R e la ti v e f re q u e n cy i n % Q1 Q2 Q3 Q4 E D C T T+E2 0 20 40 60 80 0 10 20 30 40 Fiber diameter in µM (Bin Center) MC FC MT FT MT+E2 FT+E2 Males Females 388 Fig. 4. Effects of exogenous hormones on syrinx muscle fiber diameter and density of fibers per unit 389 surface. (A) Schematic view of the syrinx with the plane of section indicated by the dotted line, (B) Histological 390 section through the syrinx showing the 4 quadrants(Q1-4), (C) Example of fibers viewed at higher magnification, 391 (D) Mean diameter of the syrinx muscle fibers across the 4 quadrants, (E) Overall distribution of fiber sizes 392 measured in the 4 quadrants in the 6 experimental groups. Data in panel D were analyzed by two-way ANOVA 393 and results are summarized in the insert (***=p<0.001, **=p<0.01, *=p<0.05, ns= non-significant). 394 395 Correlations of vocal behavior with morphological measures 396 16 To further investigate whether these morphological sex differences could explain the 397 steroid resistant sex differences in trilling behavior, we next examined correlations between 398 syrinx muscle diameters and trilling behavior in males and females separately. This analysis 399 revealed in males significant correlations between numbers of trills, trill duration and 400 percentage of time trilling, on the one hand and syrinx muscles diameters, on another hand. 401 These correlations were not detected in females with the exception with a low correlation 402 between numbers of trills and fiber diameters (Fig. 5). 403 404 405 Fig. 5 Correlations between trilling behavior with syrinx muscle fiber diameters in males and females. 406 17 407 408 Then we developed multiple regression models for males and females in which the three 409 measures of trilling were the dependent variable and muscle fibers diameter (Table 1) was 410 the independent variables (predictors) together with the whole body mass. 411 The independent variables accounted for more than 50% of the variance in trills and 412 trill duration in males but less than 28% in females (Table 1). There was a significant positive 413 relationship between the number of trills and fiber diameters (t = 4.35 p < 0.001) and between 414 trill duration and fiber diameters (t = 3.93, p = 0.002). In females, these relationships were not 415 significant (Table 1). However, we did not find a significant relationship between the percent 416 time trilling and fiber diameters in either sex even if this effect was close to significance in 417 males. 418 419 420 Table 1. Multiple linear regression models of trilling behavior (rate, duration, percent time trilling) on 421 predictor variables (fiber diameter with whole body mass), including for each predictor the estimate and 422 size (η2)*. 423 424 Males Females Number of trills (F2,16 = 13.52, Adj. R2 = 0.58, p < 0.001)*** (F2,16 = 3.13, Adj. R2 = 0.19, p = 0.071) Predictors Estimate (SE) t p η2 Estimate (SE) t p η2 Fiber diameters 112.4 (25.8) 4.35 <0.001 *** 0.63 300.5 (120.3) 2.50 0.024 0.28 Body mass 11.3 (80.5) 0.14 0.89 0.001 124.5 (132.8) 0.94 0.363 0.05 Intercept -2586.2 (1359.7) -1.9 0.075 ---- -9115.8 (4502) -2.02 0.060 --- Trill duration (F2,13 = 7.73, Adj. R2 = 0.47, p = 0.006)** (F2,9 = 0.33, Adj. R2 = 0.06, p = 0.725) Predictors Estimate (SE) t p η2 Estimate (SE) t p η2 Fiber diameters 25.9 (6.6) 3.93 0.002 ** 0.54 1.5 (21.9) 0.07 0.946 0.0005 Body mass -32.3 (17.7) -1.83 0.091 0.2 -20.5 (25.4) -0.81 0.439 0.068 Intercept 540.5 (315.5) 1.71 0.11 ---- 965.2 (750.4) 1.29 0.23 --- % time trilling (F2,16 = 2.25, Adj. R2 = 0.12, p = 0.137) (F2,16 = 1.47, Adj. R2 = 0.05, p = 0.25) Predictors Estimate (SE) t p η2 Estimate (SE) t p η2 Fiber diameters 0.9 (0.5) 2 0.063 0.22 3.7 (2.2) 1.68 0.112 0.15 Body mass -0.6 (1.4) -0.44 0.665 0.01 0.6 (2.4) 0.24 0.811 0.004 Intercept 7.2 (24.1) 0.3 0.768 ---- -82.4 (81.1) -1 0.33 --- 18 * Effect size interpretation for η2: small=0.01; medium=0.06; large=0.14. 425 426 In our previous study based on the same experimental subjects (Dos Santos et al., 427 2022), we had also measured the volume of the 4 telencephalic vocal control nuclei. There-428 fore we wondered whether the volume of these nuclei was also related to trill features quanti-429 fied here. As done with fiber muscle diameters, we computed separate multiple regression 430 models separately for males and females in which the three measures of trilling were the de-431 pendent variables and volumes of vocal control nuclei and body mass (Table S4) were the 432 independent variables (predictors). No relationship was detected in females and in males two 433 of the three trill features (duration and % time trilling) also displayed no relationship. The 434 number of trills produced was however predicted by RA volume and even more by Area X 435 volume. 436 The availability of measurements of fiber diameter for the syrinx muscles also provided 437 an opportunity to assess the relationship between these measures and the different features 438 of song that had been quantified in our previous paper based on the same subjects (dos 439 Santo et al., 2022). As shown in Table S5, these measures were not related in females but in 440 males fiber diameters were significantly correlated to all features of song except bandwidth 441 and measures of energy distribution across the bandwidth (1st, 2nd, and 3rd quartile). 442 443 Discussion 444 The present study shows that treatments with sex steroids (T or T+E2) markedly increase the 445 number and affect several features of trills produced by both male and female canaries. T 446 also increases syrinx mass and syrinx muscles fiber diameter in males but not females and 447 inter-individual differences in syrinx mass correlate positively with multiple aspects of trill 448 activity in males but not females. These results are in line with our previously reported data on 449 the effects of steroid treatments on singing activity (Dos Santos et al., 2022) and additionally 450 strongly suggest that sex differences in trill features are caused, at least in part, by the sex 451 difference in the peripheral vocal organ, the syrinx. The prominent sex differences in these 452 effects deserve further discussion. 453 454 Sex differences in trilling activity 455 19 Analysis of the entire data set over 6 weeks (Fig. 1, S1 and Table S1) revealed significant sex 456 differences for three variables: the total number of trills, percent time trilling, and trill duration. 457 A significant interaction between treatment and sex was also observed for trill duration (sex 458 difference was larger in the T+E2 condition than in the two other conditions). Qualitative 459 inspection of the data indicated that sex differences were mostly the result of larger effects of 460 T and T+E2 in males than in females and accordingly, analysis of the effects of treatments 461 week by week (Fig. S2 and Table S3) showed that sex differences are exclusively present in 462 the T or T+E2 condition. Note also that trilling increased soon after the initiation of steroid 463 treatments and plateaued after 3 or 4 weeks of treatment in both sexes. These results 464 indicate that responses of females to T treatments are not slower than those of males, and 465 also that the sex differences in responses are not due to the lower aromatization of T into 466 estrogens in females since addition of exogenous E2 does not suppress the sex difference 467 (Dos Santos et al., 2022). 468 Sex differences were particularly prominent for “fast trills” which contain more than 17 469 segments per second). Fast trills were almost completely absent in the T+E2 female 470 vocalizations (Fig. S1). Female canaries exhibit copulation solicitation displays with higher 471 frequency when exposed to playbacks of male songs containing ‘sexy’ syllables (Vallet et al., 472 1998; Vallet and Kreutzer, 1995). Additionally, the expression of immediate early genes ZENK 473 and c-Fos in two auditory forebrain regions, the caudal mesopallium and nidocaudal 474 mesopallium (areas analogous to secondary auditory cortices in mammals), is higher in 475 female canaries exposed to male songs including ‘sexy’ trills than to songs without these sexy 476 trills (Leitner et al., 2005; Monbureau et al., 2015), although these responses may depend on 477 the acoustic context (Haakenson et al., 2019). 478 479 Sex difference in trill performance 480 The TR versus mean TBW trade-off in canaries agrees with previous studies reporting similar 481 findings for different species (see reviews in (Podos and Sung, 2020; Wilson et al., 1997). 482 This trade-off probably results from constraints in frequency modulation and in rate of syllable 483 repetition (Logue et al., 2020). The deviation scores from the TR vs. TBW trade-off were 484 sexually differentiated (Fig. 2B), with males on average having a smaller but more variable 485 orthogonal distance from the quantile regression line than females. However, there was no 486 significant effect of the endocrine treatments and no sex by treatment interaction related to 487 20 these scores. The higher dispersion in the trade-off space of trills produced by males 488 compared to females (Fig. 2A) is especially prominent for the Y axis (trill bandwidth). The trill 489 data points from females are mostly positioned in one or two clusters at a relatively low 490 bandwidth. 491 This more variable trill bandwidth in males is functionally important since females have 492 a preference for syllables with broad bandwidth (Draganoiu et al., 2002). They increase their 493 copulation solicitation displays when exposed to syllables with a broad bandwidth 494 broadcasted at an artificially increased rate. This preference for supranormal vocalizations in 495 terms of bandwidth and repetition rate is consistent with the notion that these features are 496 honest-signals of male quality (Draganoiu et al., 2002) that vary across male subjects while 497 females lack the capacity to modulate trills to the same extent as males. This raises the 498 obvious question of the mechanism(s) mediating this behavioral sex difference. 499 500 Correlations with syrinx mass and structure 501 We previously reported that syrinx mass increases in response to steroid treatments in males 502 but not in females (Dos Santos et al., 2022). These observations are indeed consistent with 503 previous work that has shown that although testosterone does increase syrinx mass in adult 504 songbirds, variation in T cannot explain the general sex differences affecting this structure 505 and its function (reviewed in (Adam and Elemans, 2019). Here we show that syrinx mass and 506 syringeal muscle fiber diameter are positively correlated with the number of trills and trill 507 duration in males but not in females. Similarly, the diameter of syringeal muscle fibers 508 increases in response to T in males but not in females. Larger diameters of syringeal muscles 509 are known to be linked to the ability to produce faster rates of song (Christensen et al., 2017). 510 These differences in the response to T in the male vs. female syrinx suggest that aspects of 511 the structure that vary in males and females in response to T relate to sex differences in 512 trilling rate and performance. Differences in syrinx morphology have already been proposed 513 to explain sex differences in vocal behavior. In European Starlings (Sturnus vulgaris), for 514 example, males have greater muscle mass in the syrinx, sing at higher rates and have larger 515 vocal repertoires than females (Prince et al., 2011). 516 Sex differences in syrinx morphology might be explained by differences in testosterone 517 sensitivity, mediated by androgen receptor (AR) expression in the syringeal muscles. In zebra 518 finches, AR expression, as measured by in situ hybridization of the corresponding mRNA, is 519 21 indeed denser in male than in female syrinxes (Veney and Wade, 2004, 2005). In songbirds, 520 androgen sensitivity in specific parts of the vocal control system and periphery mediates the 521 display of physically elaborate sexual displays (Fuxjager et al., 2015) and AR expression in 522 the syrinx muscles is thus likely a sexually selected trait. Sexual selection might have 523 increased the sensitivity of syringeal muscles to androgenic hormones specifically in males 524 facilitating the production of vocal traits that are favored by females. This hypothesis is 525 reinforced by a study in male canaries assessing effects of blocking androgen action in the 526 syrinx by a treatment with the anti-androgen bicalutamide that does not cross the blood-brain 527 barrier and thus cannot affect the brain but can act on peripheral structures such as the syrinx. 528 Bicalutamide decreased the trill repetition rate without affecting other song features such as 529 song rate, therefore linking androgen action in the syrinx directly to features of song that are 530 known to be sexually selected (Alward et al., 2016). Additional studies can asses whether 531 those birds that are able to produce trills at higher rates express greater levels of androgen 532 receptors in the syrinx. Overall the significant correlation between trilling activity and syrinx 533 mass or syrinx muscle fibers diameter in males supports the view that the syrinx limits trill 534 production and explains the trilling sex difference. 535 An alternative, non-mutually exclusive, explanation for the sex difference in syrinx 536 mass relates to the fact that T-treated males sang many more songs than T-treated females 537 so that the larger syrinx mass in males could be activity-dependent, i.e., the result of more 538 exercise (Alvarez-Borda and Nottebohm, 2002; Alward et al., 2013; Alward et al., 2016; Li et 539 al., 2000; Maxwell et al., 2021). Indeed, experiments in Starlings have identified a 20–40% 540 decrease in syringeal muscle mass after vocal denervation (Prince et al., 2011), suggesting a 541 connection between vocal activity and syrinx mass. Taking advantage of data from a previous 542 study on the same birds, we also demonstrate here that many aspects of vocal behavior are 543 highly correlated with the fiber diameters of syrinx muscles in males but not in females (Table 544 S5). This was in particular the case for measures that reflect the amount of singing (song rate, 545 song duration and time vocalizing). These data are therefore consistent with the notion that 546 the sex difference in syrinx structure (overall mass and muscle fiber diameter) is induced, at 547 least in part, by the higher singing activity of males. Based on available data, it is however 548 impossible to determine whether this indirect control via increased vocal activity is more or 549 less important than the direct control by steroids. 550 551 22 Correlations with song control nuclei 552 The sex differences in trilling activity could obviously also result from differences in vocal 553 control nuclei anatomy. Our previously published study based on the same subjects had 554 indeed identified stable sex differences in three vocal control nuclei that were not suppressed 555 by the treatments with T or T+E2 (Dos Santos et al., 2022). Therefore we analyzed here, 556 separately in the two sexes, the relationships between the volume of vocal control nuclei and 557 the three features of trills that are different in males and females (Table S4). No relationship 558 could be identified in females but in males, trill rate was predicted by the Area X volume and 559 to a lower extent by RA volume. The association with RA could make sense given that song 560 production is well known to be controlled by the caudal motor pathway that includes HVC 561 projections to RA that in turn projects to the motor nucleus controlling the syrinx. 562 However, the relationship with Area X is at first glance difficult to explain. The anterior 563 forebrain pathway of the song system that includes a prominent projection from HVC to Area 564 X of the basal ganglia is more specifically associated with the auditory feedback needed for 565 song learning and song stability (Brainard and Doupe, 2000; Kao et al., 2005). Recent studies 566 have suggested, however, that the activity of dopamine neurons in Area X, in addition to 567 encoding short term performance error (Gadagkar et al., 2016), might also control the 568 initiation and performance of focused repetitive behavior (Umemoto et al., 2022), similar to 569 the involvement in the performance of sequential behavior reported in mammals (Howe et al., 570 2013). Additional work in songbirds is needed to elucidate the role of neural activity in the 571 basal ganglia during the production of repetitive behaviors such as trills. 572 573 Conclusions 574 The present study reveals that multiple sex differences in trilling behavior and performance as 575 well as individual differences among males are correlated with and might be caused, at least 576 in part, by differences in syrinx mass and structure. The proposed relationship between a 577 sophisticated learned vocal behavior and a peripheral vocal organ is reminiscent of the sex 578 dimorphism in the vocal-fold and vocal tract lengths that emerges during adolescence and is 579 mediated by testosterone in humans. In males, during adolescence, mean voice F0 and 580 formant frequencies decrease around 50-60% and 80-90% respectively when compared to 581 female values (Owren et al., 2007). 582 23 The modern field of the neuroendocrine basis of sexual differentiation was established 583 by the seminal paper of Phoenix et al. (Phoenix et al., 1959) who argued that perinatal steroid 584 hormones organize the brain in a male-typical or female-typical manner that in many cases 585 can not be reversed in adulthood by steroid hormone manipulations. Based on this paper the 586 notion was established that sex-typical reproductive behaviors are differentiated by brain 587 changes induced by perinatal androgens or their metabolites. At that time Beach (reviewed in 588 (Beach, 1971; Beach, 1981) challenged the notion that this sexual differentiation was 589 dependent on the organization of the brain by steroid hormones but rather it was the perinatal 590 effect of steroids on the development of effector organs such as the penis and other 591 peripheral organs that was critical for the sexual differentiation of these critical reproductive 592 behaviors. Our data show that like so many controversies in science there is evidence for 593 both positions. It seems clear that the neural circuit regulating song is sexually differentiated 594 at least in part by perinatal steroid hormone action but we establish here that a peripheral 595 effector organ, such as the syrinx, also plays a critical role in explaining sexually differentiated 596 courtship song. Future research should be designed to investigate the relative importance of 597 these two modes of control of sex differences. 598 599 See online Supplementary results for Figures S1 to S5 and tables S1 to S5. 600 601 Online supporting information 602 Fig. S1. Additional song features that were recorded and analyzed in male and female canaries that were 603 treated with Silastic™ implants filled with testosterone (T) or with testosterone plus estradiol (T+E2) or 604 left empty as control (C). Bar graphs represent the mean ± SEM of individual results that are the average of 605 data collected during the 6 weeks of recording. Individual data points are also indicated. Data were analyzed by 606 two-way ANOVA with treatment (Trt) and Sex of the subjects as independent factors and results are summarized 607 in the insert for each panel that is shaded in gray when significant effects were detected. (***=p<0.001, 608 **=p<0.01, *=p<0.05, ns= not significant). Significant effects of treatment were further analyzed by Tukey’s 609 multiple comparison tests and their results are expressed as follows: a= p<0.05 versus C group. 610 611 Fig. S2. Trill rate and trill features quantified by a MATLAB script in songs recorded from male and fe-612 male canaries that were treated with Silastic™ implants filled with testosterone (T) or with testosterone 613 plus estradiol (T+E2) or left empty as control (C). The different graphs represent the mean ± SEM of individ-614 ual results collected during the 6 successive weeks of recording. Data were analyzed by two two-way ANOVAs 615 separately for males and females with the treatment (TRT) and Time as factors and results, when significant, are 616 summarized at the top left for each panel. (***=p<0.001, **=p<0.01, *=p<0.05). Significant effects of treatments 617 were further analyzed by Tukey’s multiple comparison tests and their results are expressed as follows: aaa (or 618 bbb)= p<0.001 versus C (or T) group, aa (or bb)= p<0.01 versus C (or T) group, a (or b)= p<0.05 versus C (or T) 619 group. Significant interactions are described in the text. 620 24 621 Fig. S3. Trade off between trill rate and trill bandwidth as observed separately in songs of all males (M; 622 left panel) and all females (F; right panel) canaries that were treated with Silastic implants filled with 623 testosterone (T) or with testosterone plus estradiol (T+E2). 624 625 Fig. S4. Trade off between trill rate and trill bandwidth as observed separately in individual subjects that 626 had produced at least 10 trills during the entire experiment. 627 628 Fig. S5. Differential effects in males and females of exogenous testosterone (T) associated or not with 629 estradiol (E2) on the fiber diameter in the 4 quadrants of the syrinx (Q1 to Q4) (A) and on the overall fiber 630 density (number/mm2) over the 4 quadrants (B). Data were analyzed by two-way ANOVAs with treatment 631 (Trt) and Sex of the subjects as independent factors and results are summarized in the insert for each panel. 632 (***=p<0.001, **=p<0.01, *=p<0.05). 633 634 Table S1. Trill analysis: Means of 6 weeks. The table presents the results (F with degrees of freedom and 635 associated probabilities) of the two-way ANOVAs of the means of data collected over the six weeks of experi-636 ment. 637 638 639 Table S2. Trill analysis: group differences. The table presents the results (F with degrees of freedom and 640 associated probabilities) of the two-way ANOVAs of data collected each week to assess treatment effects 641 separately in males and females. 642 643 Table S3. Trill analysis: Sex differences. The table presents the results (F with degrees of freedom and asso-644 ciated probabilities) of the two-way ANOVAs of data collected each week separately to assess sex differences 645 separately in each experimental group. 646 647 Table S4. Multiple linear regression models of trilling behavior (rate, duration, percent time trilling) on 648 predictor variables (volume of 4 vocal control nuclei, whole body mass), including for each predictor the 649 estimate and size (η2)*. 650 651 Table S5. Correlations between the diameter of synringeal muscles fibers and multiple features of sing-652 ing behavior. Correlations were calculated separately in males and females based on the fiber diameters 653 measured in the present study and features of songs as quantified in our previous study (Dos Santos et al., 654 2022). 655 656 657 Acknowledgments 658 This work was supported by a Grant from the National Institute of Neurological Disorders and 659 Stroke Grant RO1NS104008 (to G.F.B., J.B., and C.A.C.). We thank Pr. Robert Dooling and 660 25 Ed. Smith, department of Psychology, University of Maryland in College Park for providing the 661 MATLAB script used to analyze canary song. C.A.C. is F.R.S.-FNRS Research Director. 662 663 Authors contributions 664 Conceptualization: Ednei B. dos Santos, Gregory F. Ball, Charlotte A. Cornil, Jacques 665 Balthazart 666 Investigation: Ednei B. dos Santos 667 Formal analysis: Ednei B. dos Santos, David M. Logue, Jacques Balthazart 668 Writing-original draft: Ednei B. dos Santos, Jacques Balthazart 669 Writing-review & edition: Ednei B. dos Santos, David M. Logue,GFB, Charlotte A. Cornil, 670 Jacques Balthazart 671 672 Declaration of interests 673 The authors declare no competing interests. 674 675 676 References 677 Adam, I., Elemans, C.P.H., 2019. Vocal Motor Performance in Birdsong Requires Brain-Body Interac-678 tion. eNeuro 6. 679 Alvarez-Borda, B., Nottebohm, F., 2002. Gonads and singing play separate, additive roles in new neu-680 ron recruitment in adult canary brain. J Neurosci 22, 8684-8690. 681 Alward, B.A., Balthazart, J., Ball, G.F., 2013. Differential effects of global versus local testosterone on 682 singing behavior and its underlying neural substrate. Proceedings of the National Academy of 683 Sciences of the United States of America 110, 19573-19578. 684 Alward, B.A., Madison, F.N., Gravley, W.T., Ball, G.F., 2016. Antagonism of syringeal androgen re-685 ceptors reduces the quality of female-preferred male song in canaries. Animal Behaviour 119, 686 201-212. 687 Andersson, M., 1994. Sexual selection. Princeton University Press, Princeton NJ. 688 Appeltants, D., Ball, G.F., Balthazart, J., 2003. Song activation by testosterone is associated with an 689 increased catecholaminergic innervation of the song control system in female canaries. Neurosci-690 ence 121, 801-814. 691 Ball, G.F., Balthazart, J., 2007. The neuroendocrinology and neurochemistry of birdsong, in: Lajtha, A. 692 (Ed.), Handbook of Neurochemistry and Molecualr Neurobiology, 3rd Edition, J.D. Blaustein 693 (Volume Ed.). Springer, New York, pp. 419-457. 694 26 Beach, F.A., 1971. Hormonal factors controlling the differentiation, development and display of copu-695 latory behavior in the ramstergig and related species, The Biopsychology of Development. Aca-696 demic Press, New York, pp. 249-296. 697 Beach, F.A., 1981. Historical origins of modern research on hormones and behavior. Horm Behav 15, 698 325-376. 699 Brainard, M.S., Doupe, A.J., 2000. Auditory feedback in learning and maintenance of vocal behaviour. 700 Nat Rev Neurosci 1, 31-40. 701 Cade, B.S., Noon, B.R., 2003. A gentle introduction to quantile regression for ecologists. Front. Ecol. 702 Environ. 1, 412-420. 703 Catchpole, C.K., Slater, P.J.B., 2018. Bird song. Biological themes and Variations. Cambridge Univer-704 sity Press, Cambridge UK. 705 Christensen, L.A., Allred, L.M., Goller, F., Meyers, R.A., 2017. Is sexual dimorphism in singing be-706 havior related to syringeal muscle composition? . Auk 134, 710-720. 707 Collins, S., 2004. Vocal fighting and flirting: the functions of birdsong, in: Marler, P., Slabbekoorn, H. 708 (Eds.), Nature's Lusic, The Science of Birdsong. Elsevier, Amsterdam, pp. 39-79. 709 Cornez, G., Collignon, C., Muller, W., Cornil, C.A., Ball, G.F., Balthazart, J., 2020. Development of 710 Perineuronal Nets during Ontogeny Correlates with Sensorimotor Vocal Learning in Canaries. 711 eNeuro 7. 712 Dos Santos, E.B., Ball, G.F., Cornil, C.A., Balthazart, J., 2022. Treatment with androgens plus estro-713 gens cannot reverse sex differences in song and the song control nuclei in adult canaries. Horm 714 Behav 143, 105197. 715 Draganoiu, T.I., Nagle, L., Kreutzer, M., 2002. Directional female preference for an exaggerated male 716 trait in canary (Serinus canaria) song. Proc Biol Sci 269, 2525-2531. 717 Elemans, C.P., Mead, A.F., Rome, L.C., Goller, F., 2008. Superfast vocal muscles control song produc-718 tion in songbirds. PLoS One 3, e2581. 719 Fuxjager, M.J., Eaton, J., Lindsay, W.R., Salwiczek, L.H., Rensel, M.A., Barske, J., Sorenson, L., Day, 720 L.B., Schlinger, B.A., 2015. Evolutionary patterns of adaptive acrobatics and physical perfor-721 mance predict expression profiles of androgen receptor - but not oestrogen receptor - in the fore-722 limb musculature. Funct Ecol 29, 1197-1208. 723 Gadagkar, V., Puzerey, P.A., Chen, R., Baird-Daniel, E., Farhang, A.R., Goldberg, J.H., 2016. Dopa-724 mine neurons encode performance error in singing birds. Science 354, 1278-1282. 725 Geraci, M., 2014. Linear Quantile Mixed Models: The lqmm Package for Laplace Quantile Regression. 726 J. Stat. Softw. 57, 1-29. 727 Goldman, S.A., Nottebohm, F., 1983. Neuronal production, migration, and differentiation in a vocal 728 control nucleus of the adult female canary brain. Proceedings of the National Academy of Sci-729 ences of the United States of America 80, 2390-2394. 730 Goller, F., 2022. Vocal atheletics - from birdsong production mechanisms to sexy songs. Anim Behav 731 184, 173-184. 732 Griffiths, R., Double, M.C., Orr, K., Dawson, R.J., 1998. A DNA test to sex most birds. Mol Ecol 7, 733 1071-1075. 734 Haakenson, C.M., Madison, F.N., Ball, G.F., 2019. Effects of Song Experience and Song Quality on 735 Immediate Early Gene Expression in Female Canaries (Serinus canaria). Dev Neurobiol 79, 521-736 535. 737 Hartog, T.E., Dittrich, F., Pieneman, A.W., Jansen, R.F., Frankl-Vilches, C., Lessmann, V., Lilliehook, 738 C., Goldman, S.A., Gahr, M., 2009. Brain-derived neurotrophic factor signaling in the HVC is 739 required for testosterone-induced song of female canaries. J Neurosci 29, 15511-15519. 740 Howe, M.W., Tierney, P.L., Sandberg, S.G., Phillips, P.E., Graybiel, A.M., 2013. Prolonged dopamine 741 signalling in striatum signals proximity and value of distant rewards. Nature 500, 575-579. 742 27 Kao, M.H., Doupe, A.J., Brainard, M.S., 2005. Contributions of an avian basal ganglia-forebrain circuit 743 to real-time modulation of song. Nature 433, 638-643. 744 Kirkpatrick, M., Ryan, M.J., 1991. The evolution of mating preferences and the paradox of the lek. 745 NAture 350, 33-38. 746 Leboucher, G., Kreutzer, M., Dittami, J., 1994. Copulation-solicitation displays in female canaries 747 (Serinus canaria): are oestradiol implants necessary? Ethology 97, 190-197. 748 Leitner, S., Voigt, C., Metzdorf, R., Catchpole, C.K., 2005. Immediate early gene (ZENK, Arc) expres-749 sion in the auditory forebrain of female canaries varies in response to male song quality. Journal 750 of neurobiology 64, 275-284. 751 Li, X.C., Jarvis, E.D., Alvarez-Borda, B., Lim, D.A., Nottebohm, F., 2000. A relationship between be-752 havior, neurotrophin expression, and new neuron survival. PNAS 97, 8584-8589. 753 Logue, D.M., Sheppard, J.A., Walton, B., Brinkman, B.E., Medina, O.J., 2020. An analysis of avian 754 vocal performance at the note and song levels. Bioacoustics 29, 709-730. 755 Louissaint, A.J., Rao, S., Leventhal, C., Goldman, S.A., 2002. Coordinated interaction of neurogenesis 756 and angiogenesis in the adult songbird brain. Neuron 34, 945-960. 757 Madison, F.N., Rouse, M.L., Jr., Balthazart, J., Ball, G.F., 2015. Reversing song behavior phenotype: 758 Testosterone driven induction of singing and measures of song quality in adult male and female 759 canaries (Serinus canaria). Gen Comp Endocrinol 215, 61-75. 760 Maxwell, A., Adam, I., Larsen, P.S., Sorensen, P.G., Elemans, C.P.H., 2021. Syringeal vocal folds do 761 not have a voice in zebra finch vocal development. Sci Rep 11, 6469. 762 Monbureau, M., Barker, J.M., Leboucher, G., Balthazart, J., 2015. Male song quality modulates c-Fos 763 expression in the auditory forebrain of the female canary. Physiology & behavior 147, 7-15. 764 Nottebohm, F., Arnold, A.P., 1976. Sexual dimorphism in vocal control areas of the songbird brain. 765 Science 194, 211-213. 766 Owren, M.J., Berkowitz, M., Bachorowski, J.A., 2007. Listeners judge talker sex more efficiently from 767 male than from female vowels. Percept Psychophys 69, 930-941. 768 Phoenix, C.H., Goy, R.W., Gerall, A.A., Young, W.C., 1959. Organizational action of prenatally ad-769 ministered testosterone propionate on the tissues mediating behavior in the female guinea pig. 770 Endocrinology 65, 369-382. 771 Podos, J., 1997. A Performance Constraint on the Evolution of Trilled Vocalizations in a Songbird 772 Family (Passeriformes: Emberizidae). Evolution 51, 537-551. 773 Podos, J., 2001. Correlated evolution of morphology and vocal signal structure in Darwin's finches. 774 Nature 409, 185-188. 775 Podos, J., Lahti, D.C., Moseley, D.L., 2009. Vocal Performance and Sensorimotor Learning in Song-776 birds. Advances in the Study of Animal Behavior (Academic Press), 159-195. 777 Podos, J., Sung, H.-C., 2020. Vocal Performance in Songbirds: From Mechanisms to Evolution, in: 778 Sakata, J.T., Woolley, S.C., Fay, R.R., Popper, A.N. (Eds.), The Neuroethology of Birdsong, 779 Springer Handbook of Auditory Research., pp. 245-268. 780 Prince, B., Riede, T., Goller, F., 2011. Sexual dimorphism and bilateral asymmetry of syrinx and vocal 781 tract in the European starling (Sturnus vulgaris). J Morphol 272, 1527-1536. 782 Rehsteiner, U., Geisser, H., Reyer, H., 1998. Singing and mating success in water pipits: one specific 783 song element makes all the difference. Anim Behav 55, 1471-1481. 784 Sakata, J.T., Vehrencamp, S.L., 2012. Integrating perspectives on vocal performance and consistency. J 785 Exp Biol 215, 201-209. 786 Sartor, J.J., Balthazart, J., Ball, G.F., 2005. Coordinated and dissociated effects of testosterone on sing-787 ing behavior and song control nuclei in canaries (Serinus canaria). Hormones and Behavior 47, 788 467-476. 789 28 Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Preibisch, S., 790 Rueden, C., Saalfeld, S., Schmid, B., Tinevez, J.Y., White, D.J., Hartenstein, V., Eliceiri, K., 791 Tomancak, P., Cardona, A., 2012. Fiji: an open-source platform for biological-image analysis. 792 Nat Methods 9, 676-682. 793 Schlinger, B.A., Brenowitz, E.A., 2017. Neural and hormonal control of birdsong, in: Pfaff, D.W., Jo-794 els, M. (Eds.), Hormones, brain and behavior, 3rd ed. Academic Press, Oxford, pp. 255-290. 795 Shevchouk, O.T., Ball, G.F., Cornil, C.A., Balthazart, J., 2017. Studies of HVC Plasticity in Adult Ca-796 naries Reveal Social Effects and Sex Differences as Well as Limitations of Multiple Markers 797 Available to Assess Adult Neurogenesis. PLoS One 12, e0170938. 798 Suthers, R.A., Vallet, E., Kreutzer, M., 2012. Bilateral coordination and the motor basis of female pref-799 erence for sexual signals in canary song. J Exp Biol 215, 2950-2959. 800 Team, R., 2021. RStudio: integrated development for R. 801 Tobet, S.A., Fox, T.O., 1992. Sex differences in neuronal morphology influenced hormonally through-802 out life. Handbook of behavioral neurobiology. Vol 11. Sexual differentiation, in: Gerall, A.A., 803 Moltz, H., Ward, I.L. (Eds.), 0 ed. Plenum Press, New York, pp. 41-83. 804 Tramontin, A.D., Wingfield, J.C., Brenowitz, E.A., 2003. Androgens and estrogens induce seasonal-805 like growth of song nuclei in the adult songbird brain. Journal of neurobiology 57, 130-140. 806 Umemoto, S., Yanagihara, S., Okanoya, K., 2022. Durations of preparatory motor activity in the avian 807 basal ganglia for songs and calls in a species of songbirds. Neurosci Res 181, 66-73. 808 Vallet, E., Beme, I.I., Kreutzer, M., 1998. Two-note syllables in canary songs elicit high levels of sexu-809 al display. Anim Behav 55, 291-297. 810 Vallet, E., Kreutzer, M., 1995. Female canaries are sexually responsive to special song phrases. 811 Anim.Behav. 49, 1603-1610. 812 Veney, S.L., Wade, J., 2004. Steroid receptors in the adult zebra finch syrinx: a sex difference in an-813 drogen receptor mRNA, minimal expression of estrogen receptor alpha and aromatase. Gen 814 Comp Endocrinol 136, 192-199. 815 Veney, S.L., Wade, J., 2005. Post-hatching syrinx development in the zebra finch: an analysis of an-816 drogen receptor, aromatase, estrogen receptor alpha and estrogen receptor beta mRNAs. J Comp 817 Physiol A Neuroethol Sens Neural Behav Physiol 191, 97-104. 818 Wade, J., Buhlman, L., 2000. Lateralization and effects of adult androgen in a sexually dimorphic neu-819 romuscular system controlling song in zebra finches. J.Comp.Neurol. 426, 154-164. 820 Wickham, H., Chang, W., 2008. ggplot2: an implementation of the grammar of graphics 821 (https://ggplot2-book.org/introduction.html). 822 Wilson, D.R., Bitton, P.-P., Podos, J., 1997. Uneven Sampling and the Analysis of Vocal Performance 823 Constraints. Am. Nat. 183, 214-228. 824 Yamamura, T., Barker, J.M., Balthazart, J., Ball, G.F., 2011. Androgens and estrogens synergistically 825 regulate the expression of doublecortin and enhance neuronal recruitment in the song system of 826 adult female canaries. Journal of Neuroscience 31, 9649-9657. 827 828