Previous studies of memory for novel visual patterns have reported how
item recognition varies across serial positions, but have neglected s
erial order memory. In Expts 1 and 2 participants reported the order i
n which short series of novel patterns had been presented. The results
showed markedly bowed serial position curves similar to those reporte
d for sequences of verbal items or spatial locations. Concurrent artic
ulation reduced performance in the serial report task relative to the
memory task alone (Expt 2) or concurrent capping (Expt 3) suggesting t
hat a verbal component was involved. When two-alternative forced choic
e tests were used to test memory for the configuration of each pattern
in the series, no primacy or recency was found (Expt 4). In Expt 5 th
e presentation of a series of five items was followed at random by eit
her a serial report or the two-choice item recognition task. The seria
l position curves for the two tasks remained different, suggesting tha
t encoding strategies were not responsible. The results show that bowe
d serial position curves are found when judgments of serial order are
required, even when phonological coding is discouraged, whereas memory
for item descriptions is independent of serial position. The implicat
ions for current conceptualizations of short-term memory are discussed
.