In a free-emission procedure participants were asked to generate insta
nces of a given category and to report, retrospectively, the strategie
s that they were aware of using in retrieving instances. In two studie
s reported here, participants generated instances for common categorie
s (e.g. fruit) and for ad hoc categories (e.g. things people keep in t
heir pockets) for 90 seconds and for each category described how they
had proceeded in doing so. Analysis of the protocols identified three
broad classes of strategy: (1) experiential, where memories of specifi
c or generic personal experiences involving interactions with the cate
gory instances acted as cues; (2) semantic, where a consideration of a
bstract conceptual characteristics of a category were employed to retr
ieve category exemplars; (3) unmediated, where instances were effortle
ssly retrieved without mediating cognitions of which subjects were awa
re. Experiential strategies outnumbered semantic strategies (on averag
e 4 to 1) not only for ad hoc categories but also for common categorie
s. This pattern was noticeably reversed for ad hoc categories that sub
jects were unlikely to have experienced personally (e.g. things sold o
n the black market in Russia). Whereas more traditional accounts of se
mantic memory have favoured decontextualised abstract representations
of category knowledge, to the extent that mode of access informs us of
knowledge structures, our data suggest that category knowledge is sig
nificantly grounded in terms of everyday contexts where category insta
nces are encountered.