Memory for faces: testing method, encoding strategy, and confidence.

Signal detection theory predicts a square root 2 recognition memory performance advantage for the two-alternative forced choice (2AFC) procedure over the yes-no (YN) procedure. In auditory psychophysics this advantage has been related to greater demands on memory in the YN task. The present experiment tested this prediction by assessing face-recognition accuracy and confidence in 72 college student subjects. Testing method (2AFC or YN) and encoding instructions (standard, overall gestalt, or distinctive feature scan) were varied, the latter in an effort to vary trace strength to see whether stronger traces would yield a lesser 2AFC advantage. A 1-week retention test revealed an overall 2AFC advantage of 1.61 and superiority of gestalt and feature-scan instructions over standard ones. While confidence and accuracy were related both within and across subjects, 2AFC subjects were significantly more confident than YN ones. Intriguingly, more efficient encodings resulted in a greater, rather than a lesser 2AFC advantage.