Abstract
When a consolidated memory is reactivated, it becomes unstable and can be modified before being reconsolidated (e.g. Schiller & Phelps 2011; Elsey, Van Ast et al., 2018). In the new learning paradigm—the most commonly used in human studies of memory reconsolidation— a new learning interferes with an old one when learned as long as a 6 hours after the reminder of the old learning (Nader et al., 2000). This interference increases with the similarity between old and new learning sets on a number of high-level attributes (for a review, see Crowder, 2014). Here, we tested for the first time if this extends to the similarity between low-level visual properties of objects from an old and a new learning set. Two subject groups learned to discriminate two sets of target from non-target faces over 3 sessions 24 hours apart: on day 1, subjects learned face set A; on day 2, they recalled face set A and learned face set B; on day 3, they recalled face set A again. Stimuli were 185 filtered neutral faces from the Chicago Face Database. In subject group 1, the spatial frequencies of face set A and face set B were filtered with the same log-polar checkerboard in the Fourier domain (like the one described in Yue et al., 2006) whereas, in subject group 2, the spatial frequencies of face set A and face set B were filtered with complementary log-polar checkerboards in the Fourier domain. Results indicate that the similarity of orientations and frequencies between face sets increases significantly the number of interferences.