Purchase this article with an account.
Judith E. Fan, Nicholas B. Turk-Browne; Accessing visual memory distorts object representations. Journal of Vision 2012;12(9):1098. doi: 10.1167/12.9.1098.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
How does retrieving an object from visual memory affect its representation? Such operations could be inert, like reading a file from a hard drive. To the contrary, we propose that memory retrieval causally intervenes in shaping object representations. We presented observers with abstract symbols appearing in a random color, orientation, and angular location. Upon symbol offset, observers were cued to report one of these features from visual short-term memory (VSTM) by manipulating a black, canonically-oriented, centered version of the symbol (memory probe) until it matched the original symbol on the cued dimension. In a later test phase, only the memory probe was presented, requiring observers to report a cued feature from visual long-term memory (VLTM). In Experiment 1, we tested whether repeatedly probing the same feature of an object in VSTM would competitively bias VLTM representations, enhancing practiced information while suppressing unpracticed details. We found reduced error in reporting practiced features in the test phase relative to a passive viewing condition in which objects were equally familiar but probed only in the test phase, and increased error in reporting unpracticed features of objects for which another feature had been repeatedly practiced relative to passive viewing. In Experiment 2, we tested whether probing multiple features in VSTM would induce more inclusive VLTM representations, in which even information outside the scope of practice is enhanced. We found that practicing two different features yielded a more precise VLTM representation of an unpracticed third feature relative to accessing an unpracticed feature of objects for which one feature had been practiced twice. Taken together, these data suggest that accessing VSTM can induce two qualitatively different forms of learning, biasing object representations to repeatedly accessed features, and broadening them in the face of uncertainty about which features will be relevant in the future.
Meeting abstract presented at VSS 2012
This PDF is available to Subscribers Only