Four experiments investigated item and order memory for sequences of seen unfamiliar faces and heard nonwords. Experiments 1 and 3 found bowed serial position curves using the serial reconstruction test of order with faces and nonwords, respectively. Experiments 2 and 4 found limited recency, no primacy, and above chance performance on all items using a two-alternative forced choice (2AFC) test of item recognition, again with faces and nonwords. These results suggest that the different serial position curves typically found using traditional paradigms for exploring visual and verbal short-term memory are due to differences in the methods used rather than modality-specific mechanisms.

Five experiments examined item and order memory for short lists of novel visual patterns. Memory was tested either by an item recognition test, choosing between a target and a similar foil (Experiments 1, 3a, and 4), or by a relative recency decision between two patterns that occupied adjacent list positions (Experiments 2, 3b, and 5). For both item recognition and relative recency tasks, accuracy was in most cases constant across serial positions, except for a recency advantage that was usually restricted to the most recent item or recency decision. Only a small and marginally significant effect of list length was observed for item recognition. Relative recency was more sensitive to list length and fell to near-chance levels with lists of eight items. We conclude that for these materials, prerecency item recognition depends on stable, context-free descriptions of items. Relative recency judgements are sensitive to list properties, but fail to show evidence of primacy or extended recency that are observed when other techniques are used to study serial order memory. We discuss the results in relation to four current models of serial order memory that embody different assumptions in the way that serial order is represented.