Abstract
Humans have the capacity to recognize the identity of thousands of faces (Jenkins, 2018), but do we remember where we saw a face? Recently, we have found that observers can remember the approximate location of >50 objects after viewing 300 objects (Wolfe & Lyu, VSS abstract). We call this “Spatial Massive Memory” (SMM). Here, we repeat the SMM experiment using unfamiliar human faces as stimuli. In the encoding phase, five upright full-color faces were placed randomly at different locations on a 7x7 grid, then each face was highlighted for 2 sec in random order (10 seconds total presentation time). This was followed by a test phase where, single faces were presented at a neutral location. Half were “old”. Observers (N=15) made 2AFC old/new responses about each face and clicked on the remembered location of any "old" face. Feedback was provided for each old/new response but not for location responses. This test sequence was repeated 40 times (200 faces), with new faces every time. At the end, observers were then retested for old/new discrimination and location using 400 faces: the 200 “old” faces and 200 completely “new” faces. Location was considered correct if it fell +/- 1 box from the true location. Guessing was corrected for by subtracting the number of "correct" localizations that could be obtained by random responses. The results show that observers were worse with faces than objects in old/new performance (d': 1.7 vs 2.7; localization: 1.9 vs 2.4 items/screen) in the test phase. On retest, both face and object old/new memory declined (d': 0.5 vs 1.5) while location memory was much worse for faces (10 total faces localized vs 57 objects). Thus, while observers show evidence that we are good at remembering the locations of many distinctive objects, memory for the locations of unfamiliar faces is poorer.