What makes something memorable?: Analyzing the Memorability of Objects
Max Kramer
A growing body of research has demonstrated that certain stimuli are consistently remembered more often than others, even across large heterogeneous populations (Isola et al, 2011), leading many to ask, “what makes something memorable?” This consistency in what is remembered and what is forgotten has been hypothesized to reflect an intrinsic and measurable property of stimuli known as memorability (Bainbridge, 2019). In attempting to determine why we remember certain things and forget others, some researchers have hypothesized that the most memorable stimuli are the most atypical or distinctive (Valentine, 1991) while others suggest that the most typical items are most often remembered (Bainbridge, Dilks, & Oliva, 2017; Bainbridge & Rissman, 2018). Here, we examine THINGS, a hierarchical naturalistic object image database that systematically samples all concrete object concepts to determine whether the most typical or atypical items are most often remembered. We collect behavioral ratings of memorability from 13,946 AMT participants to compare to three different types of typicality ratings. We collect behavioral ratings of typicality to capture human intuition, similarity scores across the dimensions of an object space associated with THINGS, and similarities across features in a deep neural network. We find a spread of memorability that persists across all levels of THINGS and determine that there is a bias towards the most typical items being most often remembered, though there are counterexamples across the dataset. These results run counter to decades of research in memory, suggesting potential targets for future analyses.
*****************************************************
Perception and evaluation of courtship song in female songbirds during mate choice
Andrew Savoy
Mate choice is a complex psychological and behavioral process. It is also a central agent of the evolutionary theory of sexual selection. Insights from cognitive neuroscience are essential for understanding mate choice but are largely absent from studies of it. I study courtship display preferences and the corresponding neural signals of perception and evaluation. For this workshop I will present my rationale and methodology for testing two hypotheses pertaining female zebra finch responses to courtship song—one regarding temporal regularity in song and the other regarding song familiarity. This research has the potential to valuably extend the scope of our knowledge about sexual selection mechanisms while also deepening our neurobiological understanding of fundamental cognitive processes.