Learning to overcome the binding problem

Rappaport, Sarah J. (2013). Learning to overcome the binding problem. University of Birmingham. Ph.D.

Preview

Rappaport13PhD.pdf
PDF - Accepted Version
Download (3MB)

Abstract

The work presented in this thesis considered the binding problem; how segregated visual information becomes unified and our perception coherent. This thesis is divided into two parts, each focused on proposed binding solutions.

Part 1 examined the binding processes that parse information at early stages of visual processing, perhaps prior to the engagement of attention. These grouping processes were examined in visual extinction patients, as an assay of pre-attentive processing. I demonstrated that item similarity can exacerbate as well as alleviate extinction, depending on the display conditions. Exacerbation of spatial deficits was shown when the array parameters encourage stimuli to be processed as distinct elements (Chapter 2). Where displays promote grouping of items both spatial (Chapter 2) and temporal (Chapter 3) impairments in extinction may be overcome.

Part 2 considered the effects of learned associations on feature binding. I focused specifically on the processing of objects with an associated, so called ‘diagnostic’, colour. I demonstrated that for such stimuli, binding of these features is automatic (Chapter 4) and modulated performance even to ones determinant. Subsequently I examined how this binding interacts with attention in a series of visual search experiments. I found that, under conditions requiring colour-shape feature conjunction, appropriately coloured targets ‘popped-out’ whereas incorrectly coloured depictions did not (Chapter 5). Converging neuropsychological evidence from visual extinction suggested that binding of learned features might not engage serial attentive processes (Chapter 6). Finally, an fMRI study was carried out to examine brain areas sensitive to the congruency of colour-shape associations; binding consistency affects the earliest stage of processing (Chapter 7). I provide evidence for efficient conjunction coding of associated features highlighting learning as one way the brain may overcome the binding problem.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):