Converging evidence from behavioral, ERP, and single unit studies, as well as computational modeling, supports a “standard model” of the visual system in which ventral cortical areas produce a final visual representation consisting of cells having significant translation invariance and tuned to features of intermediate complexity. Such a “feature list” representation is beneficial for the recognition of individual objects but seems at odds with our ability to understand visual structure – that is, our ability to parse the world into objects and parts and to understand the spatial relations among them. In a series of fMRI-adaptation experiments, I test predictions of this standard model. Subjects viewed a sequence of two minimal scenes, each composed of a pair of separated objects. Swapping relational roles of the objects (e.g. elephant above bus → bus above elephant) resulted in greater release from fMRI-adaptation in the lateral occipital complex (LOC) than when objects were translated an equivalent extent. This pattern of results is not consistent with standard model predictions, and could not be explained by variations in global or inter-object features, foveal bias, small receptive fields, or attentional shifts.
In another fMRI experiment I ask if this sensitivity to spatial relations is due to scene-centered receptive fields, or if it is better characterized by a structural description model where each object's feature list is assigned a separate “slot” in LOC. In such a slot model, a swap of relational roles produces a wholesale swapping of feature lists between LOC slots. Results supported the structural description slot model.
To explain these results in terms of neural circuitry I propose the Multiple Slots Multiple Spotlights (MS)2 model which is based on the standard model with the added assumption that the ventral visual stream contains multiple semi-independent feature hierarchies and can thus create multiple simultaneously-active feature lists. The (MS)2 model's predictions coincide with the standard model's for existing physiological data, yet can also explain a host of cognitive psychological results which the standard model cannot. I provide an artificial neural network implementation of this (MS)2 model and show that it can correctly simulate my experimental results.
|Commitee:||Lu, Zhong-Lin, Mel, Bartlett, Nayak, Krishna, Tjan, Bosco|
|School:||University of Southern California|
|School Location:||United States -- California|
|Source:||DAI-B 71/01, Dissertation Abstracts International|
|Subjects:||Neurosciences, Cognitive psychology|
|Keywords:||Neuroimaging, Spatial relations, Visual system|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be