Does source memory exist for unrecognized items?


There is a longstanding debate in episodic memory research about whether it is possible to remember contextual details of an event if the event itself is not recognized. Many notable models of recognition and source memory have predicted that such a phenomenon is not possible, while others have suggested that it is entirely possible. To complicate the matter, several empirical studies utilizing different methodologies have reported source memory for unrecognized items that is both at chance-level and above chance.

Research Questions / Hypotheses

The aim of our research is to determine whether source memory exists for unrecognized items. It is hypothesized that we will observe above-chance source accuracy for unrecognized items across the majority of conditions in the experiment.


A total of 89 REP participants completed the study.


Participants completed 6 cycles of a memory task, with each involving a study phase and test phase. In the study phase, participants were presented with several words on a computer screen, one at a time. Half the words were presented in the bottom-left corner and half the words were presented in the top-right corner. During the test phase, participants were presented with each of the words from the study phase and an equal amount of words that had not been seen before. For each word, participants were first asked to determine whether it was a “NEW” word or an “OLD” word. Following this, they were asked to determine whether the word was initially seen in the bottom-left corner or the top-right corner of the screen.


When the source memory trials are conditioned on items that were not recognized (i.e., they were labelled as “NEW”), it appears that accuracy is quite low, but still slightly above chance.


The findings of the experiment suggest that it is possible to retrieve contextual (i.e., source) details of an event, even if the event itself is not recognized. These findings support models of recognition and source memory that involve multivariate signal detection theory, as they allow for such a phenomenon to occur. The results from this study will be written up as a journal article for publication, and will be presented at various conferences on psychological research.