Boundary memory and image memorability are systematically related.

Boundary extension refers to the tendency to remember scenes as having wider boundaries than were physically present. A separate line of research has shown that image memorability, defined as the likelihood that an image will be remembered across observers, varies substantially across images. Previous research suggests that both boundary extension and image memorability heavily depend on image characteristics, yet the possibility that the two phenomena are related has not been directly tested. In the current study, we asked whether images that produce stronger boundary extension also tend to be more memorable. First, we compared boundary transformation scores from an existing dataset with memorability estimates generated by a deep-learning model. Results revealed that images showing greater boundary extension tended to have higher predicted memorability. Next, we conducted behavioral experiments measuring boundary memory for scene images and directly assessed recognition memory for the same images. Consistent with the first analysis, images associated with greater boundary extension showed higher memorability. Finally, analyses of experimentally manipulated stimuli showed that image manipulations known to increase boundary extension also increased predicted memorability. Taken together, these findings reveal a systematic relationship between boundary memory and memorability, suggesting that shared image properties may influence both the tendency to extend scene boundaries and the strength of memory encoding.