Vocal convergence in the multi-level society of Guinea baboons 1 2

The extent to which nonhuman primate vocalizations are amenable to modification through experience is relevant for understanding the substrate from which human speech evolved. One road to investigate the flexibility in vocal production has been to study differences in vocal behaviour between different social groups. We examined the vocal behaviour of Guinea baboons, Papio papio, ranging in the Niokolo Koba National Park in Senegal. Guinea baboons live in a multi-level society, with parties nested within gangs. We investigated whether the acoustic structure of grunts of 30 male baboons of five gangs differed in relation to social level and genetic relatedness. Males in this species are philopatric, resulting in increased male relatedness within gangs and parties. Grunts from members of the same gang were more similar to each other than across gangs (N = 435 dyads), but for parties within gangs we found no evidence for higher similarity (N = 169 dyads). Acoustic similarity did not correlate with genetic relatedness. Our study provides evidence for acoustic convergence in male Guinea baboon grunts; the observed nonlinear relationship between social level and acoustic similarity may reflect the limits of the extent to which vocal accommodation is possible, or even advantageous.


ABSTRACT 23
The extent to which nonhuman primate vocalizations are amenable to modification 24 through experience is relevant for understanding the substrate from which human 25 speech evolved. One road to investigate the flexibility in vocal production has been to 26 study differences in vocal behaviour between different social groups. We examined the 27 vocal behaviour of Guinea baboons, Papio papio, ranging in the Niokolo Koba National 28 Park in Senegal. Guinea baboons live in a multi-level society, with parties nested within 29 gangs. We investigated whether the acoustic structure of grunts of 30 male baboons of 30 five gangs differed in relation to social level and genetic relatedness. Males in this 31 species are philopatric, resulting in increased male relatedness within gangs and 32 parties. Grunts from members of the same gang were more similar to each other than 33 across gangs (N = 435 dyads), but for parties within gangs we found no evidence for 34 higher similarity (N = 169 dyads). Acoustic similarity did not correlate with genetic 35 relatedness. Our study provides evidence for acoustic convergence in male Guinea 36 baboon grunts; the observed nonlinear relationship between social level and acoustic 37 similarity may reflect the limits of the extent to which vocal accommodation is possible, 38 or even advantageous. The structure of nonhuman primate vocalisations appears to be highly conserved [1]. 45 Nevertheless, modifications in vocal output in relation to experience appear to be possible 46 within species-specific constraints [2][3][4], which spurred debates about whether or not 47 nonhuman primates show evidence for vocal learning [5,6]. To overcome the futile 48 arguments about whether or not modification in vocal output should be taken as evidence for 49 imitation, we recently proposed a framework that aims to distinguish between different 50 mechanisms that may contribute to acoustic variation as a result of auditory experience [7]; 51 see also [8] for similar considerations about vocal learning more generally. 52

53
Of particular relevance in the present context are two of the mechanisms, namely 54 'specific auditory facilitation' and 'learning from success'. Specific auditory facilitation is 55 based on the idea of a "common coding" framework, according to which perception and 56 action are represented by the same system [9,10]. So, listening to specific calls may increase 57 the likelihood of producing corresponding call variants by the listener. Variation in the 58 exposure to specific variants should translate into corresponding variation in acoustic 59 similarity. In other words, subjects that interact more frequently should be more similar to 60 each other than those that interact less frequently. Acoustic similarity may also result from 61 genetic relatedness, however. For instance, subjects that are highly related may also have a 62 similar morphology of the vocal production apparatus [11]. Before conclusions about the role 63 of experience can be drawn, it is therefore necessary to check whether potential acoustic 64 variation between individuals can (also) be explained by genetic distance. 65

66
We here set out to test the idea of auditory facilitation by comparing the acoustic 67 variation in male Guinea baboon, Papio papio, grunts. Guinea baboons are an intriguing 68 model to examine the influence of auditory experience and social group membership, as they 69 live in a nested multi-level society with male philopatry [12]. Males associate with specific 70 females and young in 'units' [13]. Around four to five such units form 'parties', and 2-3 parties 71 team up to form a 'gang' [14]. Males are spatially highly tolerant and show low levels of 72 aggression [15]. They also form differentiated and stable relationships with other males [16]. 73 During affiliative interactions with other group members, males produce low frequency tonal 74 grunts (Fig. 1). Such grunts have been shown to increase the likelihood of a subsequent 75 affiliative interaction or infant handling [17]. 76 77 If specific auditory facilitation affects the structure of calls, subjects that interact 78 frequently with one another should produce calls that are more similar to each other. In other 79 words, members of the same party should have the greatest similarity, and members of the 80 same gang should produce calls that are more similar to each other than to calls produced 81 by members of another gang. If genetic relatedness affects vocal structure, dyads that are 82 more highly related should reveal greater acoustic similarity. (frequency resolution 5 Hz, temporal resolution 6.4ms) and a second FFT resulting in a 103 frequency range of 500 Hz (frequency resolution of 1 Hz, and a temporal resolution of 104 16 ms). The resulting frequency time spectra were analyzed with a custom software program 105 5 LMA 2019, which allows a visual control of the accuracy of parameter estimation [18,19]. We 106 focused on the following acoustic features: fundamental frequency (F0), peak frequency, 107 distribution of frequency amplitudes and noisiness expressed as Wiener entropy (see 108 Supplementary Material Table S1). In total, we included 780 grunts in the acoustic analysis. 109 On average, we used 26 calls per subject in the analysis (range: 5-127). The Mare and 110 Simenti gang males were represented by 390 and 366 grunts, respectively; the other three 111 gangs by 7-10 grunts. 112

113
To quantify the acoustic similarity, we used a discriminant function analysis with 114 which we obtained a pairwise F-value for the acoustic distance for each dyad. This approach 115 has been applied in different studies examining relationships between acoustic structure and 116 genetic or geographic distance [20][21][22] [24] and the PopGenReport R package -version 3.0.0 [25], respectively). We used the 134 R package "related version 1.0" [26,27] to estimate relatedness using R version 3. 4

.4 and 135
RStudio version 1.1.456. The Queller-Goodnight estimator [28] showed the best performance 136 and was thus chosen in the present analysis. This estimator ranges from -1 to 1. Negative 137 values indicate that dyads are less related than on average, while positive values indicate 138 that they are more highly related than on average (see [16] for a detailed description of the 139 analysis). Confirming previous analyses, males were more highly related within gangs than between 157 gangs (P = 0.009, N = 435), though the average (mean ± se) relatedness did not differ much 158 within gangs (QG estimator: -0.003 ± 0.011, N = 169 dyads) and between gangs (-0.054 ± 159 0.014, N = 266 dyads). Within gangs, males in the same party were more highly related on 7 average (QG estimator: 0.024 ±0.024) than males that were not members of the same party 161 (-0.029 ± 0.015, P = 0.035). 162 163 Grunts could be assigned to the correct individual significantly more frequently than 164 by chance, with an average correct assignment of 38.1 % (chance level 3.3%, pDFA, P < 165 0.001). Across the five gangs, the acoustic similarity did not correlate with genetic similarity (r 166 = -0.002, P = 0.497, Fig. S1). Yet, grunts of males within gangs were more similar to each 167 other than between gangs (categorical Mantel test, P = 0.001). The variation in acoustic 168 dissimilarity for within and between gangs is shown in Fig. 2A. Grunts of dyads in the same 169 gang had a (mean ± se) dissimilarity of F = 1.08 ± 0.02, while grunts of dyads in different 170 gangs had a mean dissimilarity of F = 1.25 ± 0.02. We found no evidence that grunts from 171 males within parties were more similar to each other than between parties in the same gang 172 The structure of male grunts varied clearly between members of different gangs, but not 176 between members of parties within a gang. Males in the same gang were also more highly 177 related to one another, but this did not seem to account for the acoustic variation between 178 gangs, as evidenced by the lack of correlation between genetic and acoustic similarity. 179 Interaction frequency did not appear to account for the emergence of acoustic similarity 180 either, as grunts of males in the same party were not acoustically more similar to each other 181 than grunts of males in different parties in the same gang. This was somewhat surprising, 182 given that in a previous study, dyads within parties were observed within 100 m proximity 183 The absence of a high degree of similarity within parties discounts the idea of a linear 188 relationship between auditory experience and vocal production. One possible explanation for 189 the observed nonlinear relationship between social level and acoustic similarity may be that 190 there are limits to the extent to which vocal accommodation is possible, or even 191 advantageous. Indeed, there may also be selective pressure to maintain individual identity. 192 Since the acoustic structure of grunts is relatively simple, this may additionally place a barrier 193 on vocal convergence. 194 195 Alternatively, gang membership may have a special social value that promotes the 196 adjustment of calls to that particular level. With the present data and in the absence of 197 experimental manipulation, we are not able to distinguish between these two hypotheses, 198 although the latter seems rather unlikely. Either way, it would be illuminating to track the 199 vocal output of males that transfer between different levels, but since males are philopatric 200 and dispersal is female-biased, such transfers are rare and rather unpredictable. We 201 therefore have no data to test to which degree and over which time course males would 202 adjust the structure of their calls. 203

204
The degree of variation in grunts between different social levels is relatively small 205 though and it is not clear whether subjects pay attention to this variation. In a previous study 206 [31], we tested male responses to the playbacks of grunts of males that share the same 207 home range as the study males ("neighbours") vs. to grunts of males living 50 km away 208 ("strangers"). As control, we played back the grunts of males from the own gang. 209 Surprisingly, males responded strongly only to the grunts from males of the own gang, but 210 largely ignored neighbour or stranger males' calls (see also [32]). In principle, these 211 responses could be explained by recognition of the males' individual characteristics. Yet, it 212 might also be the case that males recognize the 'sound' of their own gang. Playbacks 213 presenting artificially created grunts bearing the own gang's characteristics vs. another 214 gang's characteristics would be needed to test this conjecture.