Title: The Memorability of Voices is Consistent and Predictable
Cambria Revsine, doctoral student in the Bainbridge Lab, Department of Psychology, University of Chicago
Abstract: Memorability, or the likelihood that an item is remembered, is an intrinsic stimulus property that is highly consistent across viewers. In other words, people tend to remember and forget the same faces, scenes, objects, and more. However, stimulus memorability research until now has been limited to the visual domain. In this talk, I will present the first exploration of auditory memorability, in which we investigated whether this consistency in what individuals remember extends to speakers’ voices, and if so, what makes a voice memorable. Across three experiments, over 3000 online participants heard a sequence of different speakers from a largescale voice database saying the same sentences. Participants indicated whenever they heard a repeated voice clip (Exp. 1 and 2) or a repeated speaker speaking a different sentence (Exp. 3). We found that participants were significantly consistent in their memory performance for voice clips, and for speakers across different utterances. Next, we tested regression models of voice memorability incorporating both low-level properties (e.g., pitch) and high-level properties measured in a separate experiment (e.g., perceived confidence). The final models, which contained primarily low-level predictors, were significantly predictive and cross-validated out-of-sample. These results provide the first evidence that people are similar in their memory for speakers’ voices, regardless of what the speaker is saying, and that this memory performance can be reliably predicted by a mix of voice features.