hide
Free keywords:
-
Abstract:
1.How do monkeys classify parameterized visual stimuli in comparison to humans?
2.Which models best describe the classification probabilities? 3.How are the parameterized
stimuli represented in the subjects’ psychological space? 4.Are the psychological
representations necessary to explain the classification performance?
The first set of stimuli consisted of 34 line drawings of Brunswik faces varying along
four dimensions: the height of the eyes (EH), the separation of the eyes (ES), the length
of the nose (NL) and the height of the mouth (MH). The second set of stimuli consisted
of 20 line drawings of fish also varying along four dimensions: the shape of the dorsal fin
(DF), the shape of the tail (T), the shape of the ventral fin (VF) and the shape of the
mouth (M). The subjects (6 humans and 3 monkeys) first learned two classes of exemplars
for each stimulus set and then assigned new exemplars to these two classes in an
unsupervised manner. During a second task they rated the similarity of all the stimuli presented
in triads. This data was used to derive distances between the stimuli and to determine
the psychological representation of the stimuli using Multidimensional Scaling
(MDS). The results of the classification task were combined with psychological and parametric
stimulus representations to assess the performance of four broad classes of models:
exemplar based models, prototype models, cue validity models and linear boundary
models.
1.Monkeys classify parameterized visual stimuli similarly to humans. 2.The exemplar
based models and the linear boundary models seem to describe best the classification
probabilities both for monkeys and humans. 3.The four varying dimensions of the stimuli
are not represented uniformly neither in the monkeys’ nor in the humans’ psychological
space. Specifically, two dimensions of the schematic faces (EH, ES), and three dimensions
of the schematic fish (M, DF, T) were more strongly correlated with the reported
similarities than the other dimensions of the stimulus sets. 4.The psychological distances
derived from the subjects’ similarity ratings do not improve substantially the fit of the
classification models.
We established that monkeys classify and represent the visual parameterized stimuli studied
here like humans do. Exemplar based models and linear boundary models best
describe the classification data both for monkeys and humans. However it is not evident
that the psychological distances are necessary in addition to the parametric distances to
explain the classification data. The similarity of classification performance in the two
species promise interesting insights in the neural mechanism of recognition in primates.
