Tuesday

Devlin, J.T., Gonnerman, L.M., Anderson, E.S., & Seidenberg, M.S. (1998). Category-specific semantic deficits in focal and widespread brain damage: A computational account.  Journal of Cognitive Neuroscience, 10, 77-94.

Thursday

Ashby, F.G., Noble, S., Filoteo, J.V., Waldron, E.M., & Ell, S.W. (2003) Category Learning Deficits in Parkinson’s Disease. Neuropsychology, 17, 1, 115-124.

Common Threads, Reactions, and Future Research

A central question this week was whether people have separate systems for learning and memory structures.

The Devlin article provided compelling evidence for  a connectionist computational account of how similar patterns of category-specific semantic impairment could arise both from localized/focal and widespread damage.  A summary of the findings to take with us was that the model reflected the real impairment patients show. In the model, an initial small amount of damage was applied. The model mirrored the impairment of people in that there was a loss of the ability to remember artifacts. Then as the amount of damage increases, natural kinds were impaired more than artifacts. This occurred because initial  damage causes random impairment in connections between semantic units and categories that contain many informative features and few intercorrelated ones and patients lose the ability to remember individual items. This, of course, applies to artifacts. As more damage is applied, whole clusters of items are damaged which means that shared features are lost. This applies to natural kinds. A short brain-storming session confirmed natural kinds of categories have  many more common features than artifacts.

It also helps to explain the variability and dissociations observed from neurologically impaired subjects.  The graphs were wonderful. The present findings also indicated that the following four properties of the semantic system were relevant to category-specific deficits:

1. Perceptual features are more important to natural kinds than artifacts and functional features are more important to artifacts than natural kinds.

2. There is a topographic organization of perceptual and functional features.

3. There is a variation in the informativeness of features.

4. There are more intercorrelations amount features in natural kinds than artifacts.

The intercorrelation may explain the pattern and time course of deficits in AD.

The discussion involved how we classify categories. One of the most difficult things to understand is how people view art. There is so much debate over what is art and what is junk or pornography. For example, some view the work of Robert Mapplethorpe (1946 - 1980) as art, but history has shown that others clearly viewed his work as pornography. It all depends on what features individuals focus on. Or perhaps it is how individuals perceive those features. In many cases experts can distinguish art from junk, but even then, expertise is not always the authority and lay people often believe that experts have misjudged junk as art.  Individuals have strong opinions. Taking an object such as a urinal out of its usual context and putting it in an art gallery, makes the statement that it is art. Is this in fact true? Does the function change if the context changes?

In the Ashby paper the question was whether there are separate systems for rule-based and information integration structures. We related this to the previous discussions we had had in common and ad hoc categories, simple and complex features, and natural categories and artifacts. We spent most of the time discussing how the two systems represent a related way of thinking about natural categories and artifacts.

A point to consider in the learning of new categories is why older people have more difficulty. Is true that the learning of new categories is a skill that requires practice and that as people age, they actually no longer get as much practice because they have formed most categories and seldom have the need to categorize into novel categories. They lose the skill that is so essential in children. Most of the discussion for this paper centered on the flaws we saw with the methodology.

We were surprised to note that PD patients were severely impaired relative to the ONC subjects when learning rule-based category structures, but not in information-integration category learning. We felt our understanding in go why this occurred would have been facilitated by asking the PD  subjects and ONC what rules they were using to make decisions.  We also queried why participants in the experiments were presented with different tones to indicate incorrect and correct responses. This seemed an unnecessary complication in the experiment because older people lose the ability to discriminate tones easily as they get older. Adding tones to the experiment also required more memorization on the part of the subjects, as well as more working memory and this may have affected the outcome. We could not understand why a simple "yes" or "no" was not used.  The response stimulus interval of 1 s did not seem at all appropriate for patients with Parkinson's disease, given that one of the most debilitating symptoms of the disease is perseverance.

The problem of studying patients with PD as a single group was also discussed. We have all had some experience with this tragic disease and it is clear that there are many factors to consider. The disease is progressive, and the severity as well as the effectiveness of the medications have to be considered.

Although both papers dealt with impairment in memory and category formation, they both took different directions.

This page was last updated:
07/18/2006 00:36