Event-related potentials and the recollection of low and high frequency words

A recognition memory test was conducted in which low and high frequency words were initially presented in one of two different study tasks. A word was defined as recollected if, at test, it was both confidently judged 'old', and confidently assigned to its correct study context. Low frequency words were more accurately recognised than high frequency items, and were also more likely to be assigned to their correct study context. The results are consistent with the view that low frequency words are better recognised because they are more likely to be recollected, rather than because they engender higher levels of relative familiarity. Event-related potentials (ERPs) evoked at test by correctly classified new words were contrasted with those evoked by old, recollected words. The ERPs to low frequency words exhibited large and reliable 'old/new' effects, in that from approx. 300 msec post-stimulus, waveforms were more positive-going for old than for new items. These effects were markedly smaller, and indeed non-significant, in the ERPs evoked by high frequency items. The results show that the interaction between word frequency and old/new differences in ERPs does not arise because of a confound between frequency and the probability of recollection. Together with other findings, they suggest that recollection is better conceived of as a graded, rather than as an all-or-none phenomenon.