ABSTRACT
Models of sequence memory typically rely on either item–to–item associations, or position–to–item associations with a rule for how positions are retrieved and updated. Both classes of models can account for the changes in serial position curves across trials. The serial position curve measures the average recall probability at each list position for successive trials of learning. This function typically shows a large primacy e ect—an advantage for early list items (Drewnowski & Murdock, 1980). The associative chaining model, which relies on item–to–item associations, produces the primacy e ect because the previous recalled item serves as the current recall cue, thus the probability of an error in recall accumulates over output positions. The positional coding model, which uses position–to–item associations, produces the primacy e ect due to edge effects. Items in terminal serial positions can be perturbed to positions in only a single direction, while middle list items can be perturbed to positions in both the forward and backward directions. This results in early list items having a higher probability of recall in the correct serial position.
