People remember face pareidolia more than human face images during naturalistic encoding

Poster Presentation 63.332: Wednesday, May 24, 2023, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Visual Memory: Capacity, encoding, retrieval

There is a Poster PDF for this presentation, but you must be a current member or registered to attend VSS 2023 to view it.
Please go to your Account Home page to register.

Olga Kreichman1,2 (), Limor Brook1,2, Susan Wardle3, Sharon Gilaie-Dotan1,2,4; 1School of Optometry and Vision Science, Bar Ilan University, Israel, 2The Gonda Multidisciplinary Brain Research Center, Bar Ilan University, Israel, 3National Institute of Mental Health, 4UCL Institute of Cognitive Neuroscience, London, UK

It was recently found that bigger images and higher-contrast images are better remembered during naturalistic encoding. In these studies, it was also found that human faces were best remembered and outdoor images the least. Face pareidolia is the phenomenon of spontaneously perceiving illusory faces in inanimate objects or scenes. Here we assumed that face pareidolia images activate both face- and non-face processing networks and therefore will be remembered better than human face images during naturalistic encoding. In a set of 4 online experiments (n=188) we consistently found that during naturalistic encoding face pareidolia images were remembered more than their control images and importantly more than human face images (even more than face pareidolia images with low faceness scores). Image memorability scores (based on participants’ performance) reflected these between-category memory results. However, testing memorability with ResMem (dedicated artificial neural network (ANN) trained to estimate image memorability) revealed that while this ANN was able to capture that pareidolia images are more memorable than human face images, it did not capture the difference in memorability between face pareidolia and their control images. These results highlight the complexity of human visual representations and further suggest that human memory during naturalistic encoding may indirectly indicate about the extent of activity elicited during initial image encoding.

Acknowledgements: ISF 1485/18 to SGD, Lev-Zion fellowship to OK