We examined several different methods for incorporating response bias
into a dual-process theory of recognition memory. Two high threshold c
orrection methods, which have previously been applied to the process d
issociation procedure, and a new, dual-process signal-detection method
, were assessed. An examination of receiver operating characteristics
(ROCs) showed that the threshold methods were inappropriate, but that
the signal-detection method provided a reasonable account of the obser
ved ROCs. Applying the corrections to a second data set showed that th
e different correction methods led to dramatically different conclusio
ns, demonstrating that selecting the correct correction method is crit
ical. Moreover, in agreement with the ROC analysis, the signal-detecti
on method was the only one to provide a reasonable account of the data
. (C) 1995 Academic Press, Inc.