| |
Abstract:
Recent word-recognition experiments using signal detection
analyses indicate that bias to respond "old" is greatest for
("critical") words that are strongly associated with previously
encoded word lists, and lesser for ("related") words with weaker
associations to previously encoded lists and ("unrelated") words
not related to previously encoded lists. Those experiments also
suggest that discrimination of old and new words (sensitivity) is
not significantly different between those three conditions, fueling
debate over how to measure and interpret false memory effects. We
suggest that the conclusion of equivalent sensitivity across
conditions was premature. In a new experiment, participants heard
lists of related words and lists of unrelated words during encoding
and then produced old/new recognition judgments (with confidence
ratings) for critical, related, and unrelated words during test.
Using signal detection measures estimated from receiver operator
characteristics (ROCs), we found bias results (c2) that replicate
the previous findings, but we found reduced sensitivity (d[a]) for
critical words compared with other words. The reason we found
sensitivity differences that were not observed in the previous
experiments may be that we used the slope of each condition's ROC
to calculate sensitivity for that condition (as recommended in
Macmillan & Creelman, 1991); when we used a single slope from
data collapsed across conditions, we replicated the (potentially
spurious) previous pattern of no sensitivity differences across
conditions. Implications for how to measure and interpret false
memory effects will be offered.
|
