This weeks readings were verybroad (and long). So I've given up on providing a complete summary of everything that was said, and will instead endeavour to give a summary of what was said about the mechanisms of visual processing, the function of attention, and the models proposed. The specific models discussed (in some cases not so specific) we will hopefully draw a better picture of in class, for those who aren't clear. In a sense, I don't think a mere summary is good enough with some of these models, since there are some very clear flaws - and other models are vague enough that a summary would be empty.

So, a brief summary of the different authors' views on visual processing, and the function of attention.

I. Visual Processing - How it works.

LaBerge presents ample evidence, I think, that visual processing is modularized. There are two main pathways (although he makes it quite clear that this is a rough treatment - there is quite a bit of crosstalk, or the potential for it), one of which handles color, shape, identity, etc., and the other of which handles primarily location information; the "what" and "where" of visual perception. Each of the other authors we read for this week accepts this basic distinction, although at least Schneider wants to claim this is only a structural, not a functional modularity.

LaBerge takes this modularity merely as a note about where and how vision is actually processed, and goes in depth into the actual connections involved between the pathways to build his model. The rough picture he provides is one in which the level of interconnectivity, both direct and through other areas (e.g., the superior colliculus (SC), and the post-parietal cortex (PPC)) between these paths makes the distinction a little weaker.

Van der Heijden notes the modularity, and then declares that this is not a problem - that in fact, the modularity is just a division of processing. He spends significant time arguing for the absence of a binding problem - more precisely, claiming that the "what" and "where" don't need to be recombined. He then points out that his model does recombine them, by taking the output of the "where" path, and feeding it back into the visual system input. So he believes that his model provides a solution for the (nonexistant) binding problem.

Treisman argues that the binding problem does exist because of this modularity, and unsurprisingly thinks her model solves this problem.

Tsal, et. al, propose to divide up visual processing even more. While there is evidence presented in LaBerge for the further division of color, shape, and perhaps one or two other aspects of visual processing, Tsal and friends wish to discuss color, shape, curvature, orientation, and location "dimensions" of visual perception. Each of these they suggest evidence for, froma variety of experiments.

Schneider, lastly, views this modularity as a structural, not a functional, feature of the brain. He claims that information is in fact parsed "globally", and that this modularity show only how each "local" aspect of information is parsed in a different place. Why does he think he has something new to offer with this? Because he proposes that neuronal synchrony, that is, the synchronous firing of neurons, is what identifies the global structure that is being processed in these different places.

II. (Visual) Attentional Processing - What does it do?

LaBerge has the most common-sense view of what it is attention does. For him, attention is merely the enhancement of the attended area, and the degredation of the surround. So attention is essentially a contrastive process, which he thinks can occur by (1) enhancement of the attended area greater than the enhancement of the surround, or (2) degredation of the attended region less than the degredation of the surround, or (3) enhancement of the attended region and degredation of the surround. So for him, attention is what makes us "more" aware of one thing than the rest of our environment.

Van der Heijden thinks that attention is designed to solve two problems, namely the problem of selecting an action, and selecting an object for that act. He argues that these are not only nice goals for attention from a neurobiological point of view, but from an evolutionary standpoint. HIs model is designed with this in mind, and is a result of his quest to identify "attention" with "behavioral modification".

Treisman proposes that the single most important function of visual attention is solving the binding problem. Specifically, by telling us which object is where, and which one it is that we are interested in, amongst those out in the world. She calls this the "featural integration" model.

Tsal, et. al, have a more abstract conception of what it is that attention does, that may encompass the function of both Van der Heijden and Treisman's models. They propose that attention determines what location in the multi-dimensional visual-perception space that an object occupies. Namely, where it lies with respect to the axes of color, shape, spatial location, etc. The bulk of their paper is spent justifying the actual dimensions of this space; this function of attention is mostly a proposal, not a conclusion.

Schneider claims that attention is what solves the binding problem, among other things. In specific, he thinks attention accounts for "similarity effects", "local feature contrasts", and "conjunction search."

III. Attentional Processing - The models.

LaBerge proposes that attentional processing is carried out by thalamocortical circuits - by the transfer of information to and from the cortex from the thalamus. He thinks that the information coming into the visual system is selectively amplified when it comes through the thalamus, and that this amplified information is fed back to the system. It is somewhat unclear what the complete picture is, since he also discusses attentional processing occurring in the superior colliculus - where he thinks the location of attention is determined. Further, he mentions attentional processing going on in several other regions of the brain - although I may be comfusing this with expressions of attention.

Van der Heijden doesn't go into neurobiological models. His model is a functional one, proposing that the modularity of visual processing doesn't really affect the system. He thinks that the output of location processing is fed back into the input of the whole visual processing system, so that it is carried through along with the "what" information. This seems to be the real key of his model, and an easy point for dispute.

Treisman doesn't discuss her model in much detail at all, except to say that it is compatible with Van der Heijden's model to a great extent, and that it allows for attention to be directed by an external process.

Tsal, et. al, don't discuss neurobiological implications, either. They just discuss what they think the function of attention is, and propose that as their model of attentional processing.

Schneider, finally, proposes a model based on three functions: "stimulus driven and attentional segmentation, attentional selection via V1", and "what- and where- based attentional control." Segmentation is the parsing of visual input into shape, color, location, etc., for each specific "object" in the visual field. Attentional selection is accomplished by back-referencing to V1 to get location information lost in the "what" pathway. What- and where- based attentional control are proposed as solutions to identifying an object with and without distractors, respectively.