PROJECT SUMMARY
Humans encode and store vast numbers of memories. An inevitable consequence of our impressive memory
capacity is that many of the memories we store are similar to other memories, which can lead to interference
between memories. The hippocampus plays an essential role not only in forming new memories, but in
preventing interference between similar memories. In rodents, patterns of activity in the hippocampus have been
shown to dramatically change—or remap—with subtle changes in the environment. While it has been speculated
that similar remapping may occur in human episodic memory and play a key role in reducing interference, direct
evidence of remapping in human episodic memory remains limited. Our preliminary data demonstrate that as
memories for highly similar experiences are acquired, human hippocampal activity patterns exhibit changes that
exaggerate the representational distance between similar memories and that are temporally-coupled with
behavioral expressions of successful memory disambiguation. Therefore, my central hypothesis is that –
similar to remapping of spatial environments in the rodent hippocampus – the human hippocampus
remaps episodic memories in order to reduce interference between similar memories. To test this
hypothesis, I will gain new training in applying advanced computational approaches (Hidden Markov Models)
that are ideally suited for studying remapping in the hippocampus. These methods have not yet been widely
applied in human neuroimaging research, but they represent a promising approach for understanding how
hippocampal representations change with experience and the relationship of these changes to the resolution of
memory interference. Aim 1 will apply Hidden Markov Models to human fMRI data to detect hippocampal state
changes (remapping) and will test whether the timing of these state changes predicts (a) the timing of behavioral
expressions of learning and (b) changes in the representational structure of competing episodic memories. Aim
2 will test for hippocampal remapping in the context of a spatial learning task in which internal representations
of the environment are experimentally manipulated while holding the objective environment constant. Again,
Hidden Markov Models will be used to detect state changes, but in this case to determine whether remapping is
coupled to experimentally manipulated changes in internal representations. Collectively, the proposed research
will leverage novel, sophisticated computational methods to provide new insight into the neural mechanisms that
resolve episodic memory interference. These findings will also be relevant for understanding memory-related
impairments associated with aging and neurodegenerative diseases.