Serial position effects in visual short term memory

Andrea Reinecke and Jeremy M Wolfe

After being briefly presented with an array of simple items, observers seem to be able to
retain a small number of these (~4) in a visual short-term memory (e.g. Luck & Vogel,
Nature, 199, 390, 279). We looked for serial position effects in this type of memory by
endogenously cueing a subset of items and then probing for memory of one of those
items. In Experiment One, subjects viewed 20 colored spots. On each trial, 3, 6, or 8 of
these spots were cued by a luminance increment. Items were cued, one after another,
every 50, 150 or 300 ms. After the cue sequence, a single item was masked and subjects
were asked to make a 5AFC decision about the color of the now hidden item. On 90% of
trials, the probed item was chosen from the cued set. Ss responded at chance levels to the
10% of uncued probed items but at above chance level at all cued locations. Results show
a strong "recency effect" with the last two cued items recalled more accurately than the
other cued items. However, performance was above chance even for items that were early
in strings of 8 Cues. There was no significant "primacy" effect favoring the item cued
first. A visual memory capacity can be estimated by multiplying overall accuracy by the
length of the string of cues. Estimates approach the "magic number" of 4, especially for
longer strings at slower SOAs. We obtained comparable results in a second experiment
using real objects in composed scenes. This supports the assumption that integrated
objects, not merely single features are stored in visual short term memory. In an
additional control experiment, verbal labeling was prevented by a distraction task.
Results were similar, suggesting that verbal recoding is not the basis for the pattern of
results.