Measurement and comparison of acoustic space use in vocalizations of humans and close primate relatives

Technical Summary: Measurement and Comparison of Acoustic Space Use in Vocalizations of Humans and Close Primate Relatives

Problem Statement
The evolutionary trajectory of human vocal morphology is characterized by specific anatomical shifts, including a deeper larynx position and the absence of vocal membranes and air sacs found in closely related primates. Historically, these morphological changes have been analyzed primarily through the lens of vowel-like vocalizations, under the prevailing hypothesis that the evolution of speech required an expansion of vocal range driven by vocal tract modifications. However, this focus is limited because vowels represent only one class of phoneme, and speech constitutes merely one category of human utterance. Consequently, a broader bioacoustic comparison is necessary to fully understand the evolutionary underpinnings of these anatomical shifts. The central problem addressed is whether human speech and song actually occupy a significantly larger acoustic-feature space compared to non-linguistic human vocalizations and the repertoires of extant close primate relatives.

Methodology
To address this, the authors conducted a comprehensive measurement of the acoustic-feature space occupied by diverse vocalizations. The study analyzed:

• Human Vocalizations: Speech, non-linguistic vocalizations, and musical vocalizations (song).
• Non-Human Primate Vocalizations: Calls from chimpanzees, bonobos, and chacma baboons.

The dataset comprised over 750,000 brief vocal segments sourced from published databases and other verified sources. To quantify the acoustic space, the researchers utilized Mel-frequency cepstral coefficients (MFCCs) to create a representation of the spectro-temporal features of these vocalizations. This approach allowed for the calculation of the volume of acoustic space occupied by each category of vocalization.

Key Results
The analysis yielded two primary findings regarding the volume of acoustic space:

1. Human Internal Comparison: Speech and song occupied significantly less volume in the acoustic-feature space than human non-linguistic vocalizations.
2. Cross-Species Comparison: The acoustic-feature volumes of human speech and song were not statistically distinct from those of the non-human primates studied (chimpanzees, bonobos, and baboons).

Key Contributions
The paper provides a systematic, large-scale quantitative comparison of vocal acoustic space across humans and close primate relatives, moving beyond the traditional focus on vowels. By utilizing MFCCs on a massive dataset of over 750,000 segments, the study challenges the assumption that human vocal evolution was defined by an expansion of the available acoustic space for learned vocalizations. Instead, it demonstrates that the unique anatomical shifts in humans may have facilitated an elaboration of non-linguistic utterances, while learned vocalizations (speech and song) utilize a surprisingly small fraction of the total available acoustic space.

Significance and Claims
The authors claim that these results suggest speech was not enabled by an expansion of human vocal acoustic space, contrary to long-held theories linking vocal tract modifications directly to an expanded range for speech. The paper posits that understanding human vocal evolution requires a shift in focus toward the function and homology of non-speech vocalizations. Furthermore, the authors note that future understanding will be further informed by the collection and incorporation of more complete non-human primate vocal datasets, specifically highlighting the need for data from Gorilla and Orangutan, alongside continued systematic comparisons of non-speech vocalizations.