Two face-recognition studies were run, in which different groups of su
bjects were presented with photographs of fully exposed faces, faces w
ith their lower halves masked, and faces with their upper halves maske
d. In the first experiment, subjects were shown a series of 50 picture
s of 50 different people and had to keep a mental tally of how many di
fferent people were shown. The presentation of such a series produced
a confusion among faces, resulting in errors' being made in the final
tallies given. The greatest number of errors (lowest tallies) was asso
ciated with masking the upper halves of the faces, and the fewest (hig
hest tallies) occurred with no masking. Similar results were obtained
in the second experiment, when subjects had to sort a set of 25 differ
ent facial stimuli into piles, with each pile supposedly containing pi
ctures of the same person. The smallest number of piles was created wh
en the upper halves of the faces were masked, and the largest number,
with no masking. The finding that intrusion errors were associated pri
marily with the lower region of the face supports the proposition that
faces are most distinctive in the upper half and most similar in the
lower portion.