Interdependence of feature dimensions in the representation of 3D objects

S. Edelmar, H. H. Bülthoff, I. Bülthoff

Research output: Contribution to journalReview articlepeer-review


Purpose. The dimensions of the representation space of 3D objects may be independent, if nonaccidental - generic or qualitative shape contrasts serve as the distinguishing features. Alternatively, the dimensions can be interdependent, as predicted by some theories that postulate metric feature-space representations. To explore this issue, we studied human performance in forced-choice classification of objects composed of 4 geon-like parts, emanating from a common center. Methods. The two class prototypes were distinguished by qualitative contrasts (cross-section shape; bulge/waist), and by metric parameters (degree of bulge/waist, taper ratio). Subjects were trained to discriminate between the two prototypes (shown briefly, from a number of viewpoints, in stereo) in a 1-interval forced-choice task, until they reached a 90% correct-response performance level. Subsequent trials involved both original and modified versions of the prototypes; the latter were obtained by varying the metric parameters both orthogonally (ORTHO) and in parallel (PARA) to the line connecting the prototypes in the parameter space. Results. 8 out of 11 subjects succeeded to learn the task within the allotted time. For these subjects, the error rates increased progressively with the parameter-space displacement between the stimulus and the corresponding prototype. The effect of ORTHO displacement was significant: F(1, 68) = 3.6, p < 0.06. There was also a hint of a marginal PARA displacement effect: F(1, 68) = 1.9, p = 0.17 Conclusions. Theories that postulate exclusive reliance on qualitative contrasts (such as Biederman's Recognition By Components) predict near-perfect discrimination performance for stimuli derived from the prototypes both by PARA and by ORTHO parameter-space displacement. Our results contradict this prediction, and support the notion of a metric representation space, in which any displacement away from the familiar region incurs performance costs.

Original languageEnglish
Pages (from-to)S1125
JournalInvestigative Ophthalmology and Visual Science
Issue number3
Publication statusPublished - 1996 Feb 15

ASJC Scopus subject areas

  • Ophthalmology
  • Sensory Systems
  • Cellular and Molecular Neuroscience


Dive into the research topics of 'Interdependence of feature dimensions in the representation of 3D objects'. Together they form a unique fingerprint.

Cite this