It is often found that faces which are rated as being distinctive
are more memorable than faces rated as being typical. We used unfamiliar
exemplars from a familiar class as stimuli -- ie cars -- to test
if distinctiveness effects generalise to different object classes. We
created a 'shape space' with thirty original car images by producing morphs
between each original car and an average car of the set. Our stimulus
set consisted of three different morph images between each original car
and the average, each at a distance of 30%, 60%, and 90% from the average.
We predicted that car morphs which were further away from the average
(ie 90% morphs) would be rated as more distinctive than car morphs which
were closer to the average (ie 30% morphs). We found a monotonic increase
in distinctiveness ratings to images which were a greater distance from
average. Our second prediction was that car morphs which were further
from the average would be more memorable than those nearer to the average.
We conducted an old/new recognition test and found that car images
were more memorable the further away the morph was from the average.
A further study revealed that memorability was dependent on the distance
from the average of the set, and not any other image within the set. Our
results support a 'shape space' model of object representation where object
locations are determined by the degree of inter-item similarity.
© 2001 Pion Ltd