EAGER: Characterizing the intrinsic memorability of voices

Voices are one of the primary ways in which we communicate, and they convey important information about the speaker, such as their thoughts and emotions. However, not all voices are created equal—some intuitively stick in our memories better than others. In the visual domain, previous research has found that despite all of our distinctive personal experiences, people tend to generally remember and forget the same faces, and so people’s visual memories are predictable to an extent. It is not yet known if the same is true for voices—whether voices have an intrinsic “memorability” that makes us universally remember some better than others. This project examines these open questions. Understanding what makes a voice easy to remember has resounding real-world applications. From designing learning materials, to creating voice assistants, understanding what makes a voice more memorable may be able to improve learning, memory and attention. By understanding what influences memory for voices, we also advance scientific knowledge about memory more broadly. For example, this research sheds light on whether or not there are similar principles across both vision and audition that determine what types of memories are preserved and discarded by our brains. Towards this goal, this project consists of three experiments to take place over a 15-month period. First, we test thousands of people’s memories for 630 diverse voices saying the same sentence, determining which voices emerge as the most memorable and forgettable, and if there are similarities across listeners in terms of what they remember. Second, we examine whether there are voices that are memorable or forgettable regardless of what they are saying. Third, we determine the features that make a voice memorable. Computational tools, that take in a voice recording and predict the chance someone will remember that voice, are being generated and made publicly available, and findings are being incorporated into an undergraduate textbook on Big Data in the Psychological Sciences. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.