A Caltech Library Service

What you can see outside the focus of attention


Wen, Jiajun (2000) What you can see outside the focus of attention. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/9z53-3157.


Visual attention is widely considered to be first and foremost a means of controlling the flow of information between different levels of processing. Stimuli selected by attention are thought to gain access to higher levels of processing, including awareness, recognition, memory, and the generation of voluntary responses, while stimuli rejected by attention are deemed to be denied such access. However, visual attention remains poorly understood in many aspects. In this study, we try to investigate some part of the problem. Several behavioral methods have been devised for the study of visual attention, one of them - the concurrent task paradigm - has gained more and more recognition and is widely used in different studies. In the first part of the thesis, we use this paradigm to investigate an important question: whether attentional capacity is differentiated or not. In non-human primates, neuronal sensitivities to stimulus dimensions such as form, color, or motion are concentrated in different visual cortical areas (Felleman and Van Essen, 1991; Van Essen and Gallant 1994), and functional imaging studies show similar functional specializations in the visual cortex of humans (Corbetta et al., 1990, 1995; Van Essen and Drury, 1997). It has been proposed that attentional capacity is "differentiated" (Sperling and Dosher, 1986; Pashler, 1997) in that discriminations concerning different dimensions draw at least in part on different attentional capacities and are thus less liable to interfere than discriminations concerning the same stimulus dimension. However, Duncan and colleagues have argued at length for a contrary point of view (Desimone and Duncan, 1995; Duncan et al., 1997). In their view, due to the strong interdependence of the subsystems that process different stimulus dimensions attentional capacity is "undifferentiated". By conducting experiments on the concurrent discrimination of form, color and motion attributes, we test which hypothesis is true. We quantify interference between concurrent discriminations by establishing the attention-operating characteristic. Interference is indistinguishable for similar and dissimilar task combinations (form-form, color-color, motion-motion, and color-form, color-motion, motion-color, motionform, respectively) for both demanding and less-demanding tasks. These results suggest strongly that different visual discriminations draw on the same attentional capacity, in other words, that the capacity of visual attention is undifferentiated. After establishing that a task is equally attentional demanding no matter what the stimulus dimension the second task is based on, we use the concurrent task paradigm to investigate how attention affects spatial vision. We measure thresholds for discriminating the contrast, orientation, and spatial frequency of simple patterns that are either fully or poorly attended. We observe differences of about 20% in contrast detection thresholds, 40-50% in contrast discrimination thresholds (and appearance of the “dipper”), 60-70% in orientation and spatial frequency discrimination thresholds, and up to 50% in contrast masking thresholds. These observations tightly constrain any effect attention may have on the visual filters and/or the interactions amongst filters that are thought to underlie basic spatial vision. Comparison with a computational model due to Laurent Itti shows that the observed effects of attention are consistent with stronger interactions amongst filters, but not with a change in noise parameters, as is sometimes thought (Bonnel and Miller, 1994; Lu and Dosher, 1998. Essentially, the effects of attention on different thresholds are too disparate to be accommodated by a single change in noise parameters. In the framework of our model, the strength of the interaction amongst filters is controlled by the exponents of a power law, γ and δ. The larger exponents activate what is best described as a winner-take-all competition amongst visual filters. Attention accentuates existing differences between filter responses, boosting filters that respond relatively well to a given stimulus, while suppressing filters that respond relatively poorly. This explains the perceptual advantage conferred by attention: attention enhances the sensory representation by restricting responses to the filters tuned best to the stimulus at hand. Our experimental and modeling results show that the activation of a winner-take-all competition amongst overlapping visual filters explains many basic perceptual consequences of attention.

Item Type:Thesis (Dissertation (Ph.D.))
Subject Keywords:Computation and Neural Systems
Degree Grantor:California Institute of Technology
Division:Engineering and Applied Science
Major Option:Computation and Neural Systems
Thesis Availability:Public (worldwide access)
Research Advisor(s):
  • Koch, Christof
Thesis Committee:
  • Unknown, Unknown
Defense Date:21 May 1999
Record Number:CaltechTHESIS:10012010-144224980
Persistent URL:
Default Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:6089
Deposited On:06 Oct 2010 18:15
Last Modified:16 Apr 2021 23:04

Thesis Files

PDF - Final Version
See Usage Policy.


Repository Staff Only: item control page