Recognition memory for Spanish-Catalan cognate and noncognate words was testedatretentionintervalsof20minutes,1hour,and24hours(Experiment1) using a remember/know response procedure, and requiring a confidence judgement on the yes/no response. Noncognate words were accompanied by more ``remem- ber'' responses than cognates, and overall A9 was significantly different from remember A9, except in the cognate condition at the longest retention interval. A strong mirror effect for the cognate±noncognate stimulus class was found for overall responding, and for high but not low confidence, indicating a differential use of recollection and familiarity in recognition. In general, the pattern of results was inconsistent with Donaldson's (1996) signal detection model, indicating that, when available, subjects use two different sources of information for discrimina- tion. The examination of individual hits and false alarms as a function of con- fidence indicated that ``remember'' is uniformly associated with high confidence, but ``know'' shows a bipolar pattern. In Experiment 2, new and old words were repeated at test 2 and 3. Repetition greatly affected the difference between the discrimination indices, indicating that an increase in the familiarity of new words prevented the use of a dual source of information in recognition. Results are discussed in terms of Rajaram's distinctiveness (1996, 1998) and Reder, Nhouy- vanisvong, Schunn, Ayers, Angstadt, and Hiraki's (2000) SAC theories.