Divergent acoustic properties of gelada and baboon vocalizations and their implications for the evolution of human speech.

Human speech has many complex spectral and temporal features traditionally thought to be absent in the vocalizations of other primates. Recent explorations of the vocal capabilities of non-human primates are challenging this view. Here, we continue this trend by exploring the spectro-temporal properties of gelada (Theropithecus gelada) vocalizations. First, we made cross-species comparisons of geladas, chacma baboons, and human vowel space area. We found that adult male and female gelada exhaled grunts-a call type shared with baboons-have formant profiles that overlap more with human vowel space than do baboon grunts. These gelada grunts also contained more modulation of fundamental and formant frequencies than did baboon grunts. Second, we compared formant profiles and modulation of exhaled grunts to the derived call types (those not shared with baboons) produced by gelada males. These derived calls contained divergent formant profiles, and a subset of them, notably wobbles and vocalized yawns, were more modulated than grunts. Third, we investigated the rhythmic patterns of wobbles, a call type shown previously to contain cycles that match the 3-8 Hz tempo of speech. We use a larger dataset to show that the wobble rhythm overlaps more with speech rhythm than previously thought. We also found that variation in cycle duration depends on the production modality; specifically, exhaled wobbles were produced at a slower tempo than inhaled wobbles. Moreover, the variability in cycle duration within wobbles aligns with a linguistic property known as 'Menzerath's law' in that there was a negative association between cycle duration and wobble size (i.e. the number of cycles). Taken together, our results add to growing evidence that non-human primates are anatomically capable of producing modulated sounds. Our results also support and expand on current hypotheses of speech evolution, including the 'neural hypothesis' and the 'bimodal speech rhythm hypothesis'.