| |
Abstract:
Dual process models of recognition memory assume that
recognition is the product of two distinct memory processes:
recollection and familiarity. One implication of these models is
that single-parameter measurements of recognition accuracy, such as
d' or proportion correct, should be insufficient to characterize
performance. If these models are correct, then in order to measure
recognition accurately, it is necessary to rely on test paradigms
that include more than a single measure of performance, such as the
process dissociation, remember/know, or ROC procedures. In the
current study, dual process theory is tested by examining ROCs
under conditions in which performance is expected to rely on (a)
both processes (standard recognition), (b) primarily on
recollection (tests of associative recognition), and (c) primarily
on familiarity (recognition by amnesics). As predicted by the dual
process model, the standard recognition ROCs were curved and
asymmetrical, recollection-based ROCs were relatively linear, and
familiarity-based ROCs were curved and symmetrical. The results of
these studies show that single-factor models of recognition are not
capable of describing recognition memory performance, and that
models that are at least as complex as current dual process models
are necessary. Furthermore, the results highlight the erroneous
conclusions that can arise when a single factor measure of
recognition is used.
|
