It has been argued that working memory contains a short-term phonologi
cal store that is vulnerable to decay within 1-2 sec unless it is refr
eshed by means of articulatory rehearsal. In Experiment 1, however, th
e phonemic similarity effect was found to persist through a 10-sec ret
ention interval during which the subjects engaged in articulatory supp
ression, although it was not significant when they had engaged in coun
ting backwards aloud. In Experiment 2, the phonemic similarity effect
was found to persist through a retention interval of either 10 or 20 s
ec during which the subjects engaged in either articulatory suppressio
n or counting backwards aloud. It is concluded that the phonological t
race in working memory may have a much longer duration than has previo
usly been assumed.