An architecture for rapid, hierarchical structural description

John E. Hummel, Brian J. Stankiewicz

Research output: Contribution to journalArticlepeer-review


Dynamic binding is a necessary prerequisite to structural description, but it is not straightforward to represent structural descriptions suitable for object recognition, given the limitations of dynamic binding (specifically, limitations of speed and reliability). This chapter presents a model of object recognition addressed to this problem. The model generates hierarchical representations that function as structural descriptions in the event of correct dynamic binding but that still support object recognition in the event of dynamic binding errors. The model can recognize objects rapidly (i.e., before dynamic binding is established) but has the properties of a structural description when dynamic binding is established correctly (e.g., it can recognize objects from novel viewpoints). The model provides an account of several phenomena in human object recognition, including the time course of recognition and viewpoint invariance and the role of attention in object recognition.

Original languageEnglish (US)
Pages (from-to)92-121
Number of pages30
JournalAttention and Performance
StatePublished - 1996
Externally publishedYes

ASJC Scopus subject areas

  • Neuropsychology and Physiological Psychology
  • Experimental and Cognitive Psychology


Dive into the research topics of 'An architecture for rapid, hierarchical structural description'. Together they form a unique fingerprint.

Cite this