Sequence-Level Vocal Convergence in Common Marmosets
Public preprint · manuscript in revision
This manuscript is available on bioRxiv and is currently in revision. It has not completed peer review.
Study objective
The study tests whether pair formation is followed by convergence in sequence organization, individual phee-call acoustics, or both, and whether any change depends on the audience to which calls are directed.
Study design
Three male and three female common marmosets were recorded in all nine possible male–female combinations. After three permanent pairs were formed and co-housed, the recording design was repeated four months later. Calls were therefore compared before and after pair formation and separately for partner and non-partner audiences.
Acoustic and sequence representations
Sequence similarity was quantified using transition probabilities, bigram distributions, repeat-length distributions, and local sequence alignment. Individual phee-call similarity was quantified using spectro-temporal measurements, MFCC representations, constrained dynamic time warping, and a variational autoencoder.
The use of multiple representations at both levels tests whether the estimated pattern is robust to a particular feature set or distance metric.
Results reported in the preprint
Following pair formation, partners showed increased sequence similarity when calling to non-partners. The combined model and three of four sequence measures supported this pattern; repeat-length distributions showed no clear change in the non-partner condition. No comparable convergence was detected in partner-directed sequences. Individual phee-call acoustics showed no clear convergence in either audience context across the four representations.
The estimated sequence-level change was therefore audience dependent rather than a general increase in similarity between partners. This pattern supports socially contingent plasticity in sequence production while leaving the underlying mechanism unresolved.
One interpretation is that sequence organization forms a pair-associated signature expressed toward non-partners. The production data support sequence-level flexibility, but do not establish receiver recognition, shared meaning, or a learned convention.
Interpretation
Studies of vocal accommodation often emphasize the acoustic form of individual calls. These results show that sequence order and repetition can vary independently of individual-call acoustics and that audience context is necessary for interpreting convergence estimates.
Limitations and planned tests
The study includes three pairs, and one pair lacked partner-after observations. The four-month interval between recording periods was not accompanied by an unpaired control group or repeated pre-pairing baseline; temporal drift, arousal, and other pair-specific changes therefore cannot be excluded. Playback or interactive experiments are required to test receiver perception and whether the convergent sequence features communicate pair affiliation.
Research themes
Preprint
Wewhare, N., Burkart, J. M., and Wierucka, K. (2026). Sequence-level vocal convergence in common marmosets. bioRxiv. Read the preprint.