Existing talker knowledge may make convergence more difficult.

Abstract

Over time, interlocutors tend to sound more and more like each other in conversation. This “convergence” can also be tracked on the acoustic-phonetic level in the lab. Here, we present two experiments that shed light on the role of existing talker information (i.e., familiarity) on the tendency to converge to voice fundamental frequency (F0). In Experiment 1, participants (N = 32) first read 40 sentences without auditory input to provide baseline F0 values. Then, in a synchronous speech task, participants were asked to speak at the same time with a pre-recorded model talker. The model talker’s voice was pit ch-shifted ±2 semitones to create high and low F0 conditions. Onset synchrony was achieved by visually presenting the sentence to the participants during a 3-second countdown before playback. The data revealed that 75% of participants across both groups significantly changed their F0 from the reading task to the synchronous speech task, in the direction of the model talker. Thus, Experiment 1 provided a valid paradigm for measuring F0 convergence in the lab. In Experiment 2, another group of participants (N = 32) was split evenly into high F0 and low F0 exposure groups. After a reading task identical to that of Experiment 1, participants listened to a pitch-shifted (i.e., ±2 semitones; per talker F0 group) model talker speak for 20 minutes. Finally, they performed a synchronous speech task where the model talker’s F0 was contradictory to their knowledge. That is, participants in the high F0 talker group had to synchronize with the low pitch-shifted version of the model talker. I n Experiment 2, only 50% of the participants converged to the model talker. Taken at face value, our findings suggest that conflicting talker information may interfere with convergence, highlighting the potential susceptibility of the production system to talker-bound changes in the perceptual system.