From encoding to retrieval: Change in level of unitization resolves debate about Unitization's effect on associative recognition.

Although it is widely accepted that familiarity could support associative recognition when the to-be-learn items are 'unitized' into a new representation, the effects of unitization on associative recognition and recollection remain much debated. The current study aimed to explain these debates by exploring when and how unitization benefits associative recognition using event-related potentials (ERPs). During the encoding phase, participants learned compound words and unrelated word pairs (i.e., High vs. Low level of unitization). At retrieval, the compound words were rearranged into new compound words (i.e., no-change) and unrelated word pairs (i.e., change). Similarly, the unrelated word pairs were rearranged into new unrelated word pairs (i.e., no-change) and compound words (i.e., change). Results showed that under the no-change condition, unitization did not affect associative recognition, nor its underlying processes. In contrast, under the change condition, unitization improved associative recognition by increasing both familiarity-related FN400 effect and recollection-related LPC effect. In addition, a planned comparison between the compound-change and unrelated-no change conditions-a common index for unitization effect in past studies-revealed that unitization could not only elicit significant FN400 effect, but also improve associative recognition by increasing LPC effect. Collectively, these results not only allowed to explain the current discrepancies in the literature concerning the effect of unitization on associative recognition, but also emphasized the importance of matching the level of unitization between the studied and rearranged word pairs.