Tuesday, February 5, 2008

Semantics and Brain - Is semantic dementia an impairment to "conceptual knowledge"?

The claim is that semantic dementia, true to its name, is a selective deficit of semantic knowledge. The papers we read over last week constitute at least some of the evidence supporting this claim. So is it true? Well, it depends on what you mean by "semantic knowledge." As long as we restrict our definition to knowledge of relatively concrete concepts, then I think the evidence is decent (not rock solid, though).

So what is the evidence? Here's my outsider's assessment based on what we've read so far. I would, of course, welcome comment/correction from the experts!

1. Naming impairment is largely independent of input modality (picture, word, verbal description, environmental sound), and output modality (spoken, written response).

2. Relatedly, not only is naming affected, but so too is comprehension. Here there is a bit of a dissociation in that naming problems tend to be more severe than comprehension problems, but this is not surprising given that production in general seems to be more difficult than recognition. The paper we read by Matt Lambon Ralph et al. (2001, JoCN, 13:341-56) had a nice analysis of the relation between naming and comprehension in SD, particularly in relation to primarily left vs. right dominant atrophy.

3. Performance on "semantic judgment" tasks (such as the pyramid and palm trees test) is (sorta-kinda) modality independent as well. I say sorta-kinda because there are modality differences, but these, like the naming-comprehension asymmetry, are assumed to be easily explainable. For example, both Bozeat et al. (2000, Neuropsychologia, 38:1207-15) and Benedet et al. (2006, Neurocase, 12:15-26) report better performance on the picture than word version of the pyramid & palm tree/camel & cactus test. On the face of it, this is a bit puzzling because if the deficit involves central semantic representations and not the mechanisms that access them, then one should not expect any difference. This difference is explained away, however, by assuming that pictures have a more direct access route (a more systematic mapping) to the semantic system than do words (see Bozeat et al. p. 1213 for a paragraph on the topic). Kind of sounds like a mapping problem rather than a representation problem, doesn't it? But I suppose the idea is that a picture will produce a bigger jolt of activation than a word, thus making it more likely that the correct semantic representation in a degraded network is activated. Probably open to discussion, but let's grant it to them for now.

3. Breakdown of semantic abilities is hierarchical. SD patients tend to make superordinate errors (calling a FROG an animal) and tend to have more difficulty with subordinate features. These deficits transcend modality, affecting both input and output. I find myself having a hard time specifically stating how this observation argues for a semantic knowledge deficit above and beyond the modality effect, but it feels like it does. Maybe Warrington (1975; Quart. J. of Exp. Psy., 27:635-57) had the same sense when she wrote, without further argument, "The findings using probe recognition tasks [that assess judgments of super- vs. sub-ordinate semantic knowledge] suggest that the deficit is not merely in the retrieval of semantic information but that the storage systems are damaged." p.652. I guess the point is that since these patients can retrieve some semantic information successfully (i.e., the superordinate stuff), then retrieval per se cannot be affected. But why couldn't retrieval of details be differentially affected? Or maybe it's just the general sense that this pattern of deficit really feels like a degradation of semantically specific features, which in turn feels like a general knowledge problem (not exactly a crisp argument). If someone can clarify, please do!

Taken together, evidence of this sort feels like a pretty strong case for a general semantic deficit in the sense that it is clearly not just aphasia, or visual object agnosia.

Here's what I'm wondering though: is the semantic information that is lost, specific to the visual and auditory modalities? That is, how truly general is it?

The tests of modality generality of these deficits have involved predominantly, visual, auditory, and language functions. All of these abilities have been associated with temporal lobe function. Is it possible that the (no so focal) temporal lobe dysfunction in SD is causing disruptions to somewhat separable visual object systems, auditory object systems, and language systems? Or are these deficits a result of damage to a single semantic hub? I know that many SD researchers believe and argue for the later, but I'm not convinced yet that at least a portion of the deficit profile isn't due to modality specific dysfunction. Murray Grossman appears to have similar thoughts (based on looking ahead to future readings).

I'll elaborate on these thoughts in subsequent posts, in the context of summarizing some of the papers we read last week. For example, the paper by Hodges et al. (2000, Brain, 123:1913-25) showing that SD patients have deficits in object use (a less-visual/non-auditory/non-language ability that presumably involves parietal-frontal circuits) would seem to make a strong case for a more general deficit. But I think that study, while very interesting, is flawed. I'll explain in the next post.


Jonathan Peelle said...

This is precisely one of the thoughts we have. Let's assume some sort of sensory-functional distributed storage account of semantics in which modality-specific semantic information is stored in areas related to that processing. Visual information would tend to be dependent on inferior temporal regions related to visual object knowledge etc.

Presumably, this information is stored in the same place, regardless of how a concept is activated. So a concept like APPLE has a lot of visual features associated with it, but these can be recalled whether you read the word "apple", see a line drawing, or (importantly) hear the word apple.

Although I don't have any actual data (!), I would bet that most items used to test semantic memory are concrete nouns that are high in visual feature knowledge (Boston Naming Test, Snodgrass Pictures, Pyramids & Palm Trees, etc). Within these items, deficits in visual feature knowledge would present the same as an amodal semantic impairment.

In theory this should be easy to test. If the above is true, then nouns that rely more on knowledge from other modalities should be relatively spared, since SD preferentially affects visual areas.

Testing SD patients through multiple modalities is very important in terms of testing modality-independence of access, but it may not test modality of storage.

(I would be glad to know of any reasons that what I just said is utter hogwash, too.)

Greg Hickok said...

Sounds right on to me! Thanks for your thoughts!