Use this URL to cite or link to this record in EThOS:
Title: Computational modelling of visual search experiments with the selective attention for identification model (SAIM)
Author: Backhaus, Andreas
ISNI:       0000 0001 3434 8994
Awarding Body: University of Birmingham
Current Institution: University of Birmingham
Date of Award: 2007
Availability of Full Text:
Access from EThOS:
Visual search is a commonly used experimental procedure to explore human processing of visual multiple object scene (Wolfe, 1998b). This thesis introduces a new computational model of visual search, termed Selective Attention for Identification model (SAIM). SAIM aims to solve translation invariant object identification in a connectionist modelling framework. The thesis demonstrates that SAIM can successfully simulate the following experimental evidence: symmetric searches (L amongst Ts and/or upside down Ts (Duncan & Humphreys, 1989; Egeth & Dagenbach, 1991)), asymmetric searches of oriented lines (Treisman & Gormican, 1988), line size (Treisman & Gormican, 1988), item complexity (Treisman & Souther, 1985; Rauschenberger & Yantis, 2(06)) and familiarity (Wang, Cavanagh, & Green, 1994; Shen & Reingold, 2(01)), the influence of distractor (non-targets) orientation (Foster & Ward, 1991; Foster & Westland, 1995), effects of priming (Hodsoll & Humphreys, 2001; Mueller, Reimann, & Krummenacher, 2003; Wolfe, Butcher, Lee, & Hyle, 2003; Anderson, Heinke, & Humphreys, 2(06)) and 'contextual cueing' (Chun & Jiang, 1998). Crucially, SAIM's success emerges from the competition of objects for object identification. This competition is chiefly influenced by three factors: the similarity between search targets and non-targets (distractors), the visual features of the distractors, e.g., line orientation, and the influence of the object identification stage on the selection process (top-down modulation). On the other hand, a detailed reView in this thesis highlights that none of the existing computational models and theories can satisfactorily account for these experimental results. Instead, each theoretical account contains only a subset of the factors suggested by SAIM and, therefore, can explain only a subset of the experimental data. In SAIM these factors are pulled together into a unifying approach of parallel competitive interaction towards visual search.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available