Model-Driven Study of Visual Memory

Abstract

Short-term episodic visual recognition memory is crucial to success in many everyday activities. We synthesized concepts, insights, and methods from memory research, and from vision research, working within a coherent, quantitative framework for understanding episodic visual recognition memory. Seven experiments were carried out in two related sub-projects. One sub-project confirmed that high-dimensional stimuli (synthetic human faces) afford important insights into episodic recognition memory. The results were well accommodated by a summed similarity theory of recognition memory (Kahana & Sekuler, 2002). The second sub-project supported coordinated experiments on recognition memory and item identification (source memory). Source identification errors were deterministic rather than stochastic, and their causes were identified. Receiver operating characteristics (ROCs) compared recognition across experiments. Combining signal detection theory and a summed similarity model explained the unusual properties of the z-transformed ROCS.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 01, 2004
Accession Number
ADA435772

Entities

People

  • Robert Sekuler

Organizations

  • Brandeis University

Tags

Communities of Interest

  • Human Systems

DTIC Thesaurus Topics

  • Accuracy
  • Algorithms
  • Brain
  • Cognitive Neuroscience
  • Cognitive Science
  • Computers
  • Databases
  • Detection
  • Identification
  • Information Science
  • Monte Carlo Method
  • Neurosciences
  • Psychology
  • Recognition
  • Signal Detection
  • Three Dimensional
  • Two Dimensional

Readers

  • Artificial Intelligence
  • Clinical Trial Research.
  • Computer Vision.