To test our second hypothesis regarding the association between left midfusiform activity and episodic memory for novel characters, we first examined whether the midfusiform activity could predict remembered vs. forgotten characters within subjects, using the subsequent memory approach [30], [31].
More importantly, it has been shown that the leftward lateralization of midfusiform activation during initial learning strongly predicted the outcome and long-term (six-month) retention of a two-week training regimen [24], [25], [26].
From both within- and cross-subject analyses, we found that weaker midfusiform activation during learning was associated with worse recognition memory.
Second, activation in the left midfusiform cortex predicted memory for novel characters both across and within subjects (i.e., across items), with stronger midfusiform activation associated with better recognition.
Although many studies have revealed strong midfusiform activation when a novel script is being processed, the functional significance of the activity is not clear.
To explain the observed increase of midfusiform activation for familiar words in comparison to nonwords, we have proposed that such increased activation might have resulted from the associations between visual form and other linguistic factors, such as phonology and semantics [16], [17], [18].
Strong midfusiform activation has been observed during processing of words as compared to nonwords in both alphabetic and logographic writing systems [1], [2], [3].
In particular, orthographic learning leads to decreased rather than increased midfusiform activation [16], [17], [18].
Based on results from previous subsequent memory studies on one-shot learning [30], [31], [32], we hypothesized that stronger midfusiform activation during learning (across repetitions) would be associated with better long-term memory both within and across individuals.
