Tuesday, October 14, 2008

Does Parkinson's disease impair action verb processing?

I've been slogging through the evidence typically cited as support for an embodied cognition view of language processing. Much of this research focuses on processing actions verbs, which according to the "EC" view, critically involve motor representations as part of their semantics. In previous posts I've discussed studies that use TMS, ALS, and stroke data to make the case for an embodied view of action word processing. None of it, I argued, was particularly compelling.

Here we have a close look at a recent paper involving Parkinson's disease (PD) patients (Boulenger et al., 2008). These authors used a lexical-decision, masked, identity-priming paradigm: primes were identical to targets (= identity-priming) and were presented rapidly, followed by a mask which precludes conscious awareness of the prime (= masked); priming effects were assessed relative to a control condition where the "prime" was a string of consonants. Priming was compared for visually presented nouns and verbs in PD patients both on and off medication. This is an interesting design because it allowed the team to assess processing when the basal ganglia circuit was relatively functional compared to when it was not. Control subjects were also tested.

So what did they find? On medication, PD patients showed priming for both nouns and verbs (middle panel in figure below), whereas off medication, PD patients only showed priming for nouns. Since nouns primed even off medication, this argues against generalized attentional, perceptual, etc. explanations of the failure of verbs to prime off medication.

(White circles are nouns, black circles are verbs.)

This is a pretty cool result and is interpreted as "compelling evidence that processing lexico-semantic information about action words depends on the integrity of the motor system" (p. 743). I beg to differ.

First, PD is NOT limited to the motor system. In fact, Boulenger et al. point out that "deficits in cognitive functions and subtle semantic language deficits have also been reported" (p. 744). It is impossible to know whether the failure to show priming effects is strictly a matter of motor dysfunction, or whether it stems from disruption of other functions supported by basal ganglia circuits. This is a point similar to one I raised in connection with ALS: just because a prominent symptom of a disease is motor, doesn't mean that the motor deficit is causing all the symptoms.

Second, depending on what you focus on in the reaction time data, the pattern of results could either support a verb processing deficit or a noun processing deficit. Have a look at the top "Patients OFF" panel in the graph above. While it is clear that nouns are priming and verbs are not, it is also the case that RTs to nouns are quite a bit slower than RTs to verbs in the control, unprimed condition (left side of graph). This is puzzling given that ON medication, the PD patients showed no RT difference to the same nouns vs. the same verbs. So one way to look at the result is that being off medication causes a selective deficit in noun processing relative to verb processing!

How do we reconcile these two interpretations? I don't know. It depends on which measure (raw recognition time vs. priming) is a better measure of "lexico-semantic" processing. Sometimes it helps to re-state the findings without all the interpretive baggage. Assuming that basal ganglia dysfunction is exaggerated when the PD patients are off Levodopa medication, the present study leads to the following conclusions:

1. Basal ganglia dysfunction reduces the masked-primer induced pre-activation of essential parts of the cerebral networks for verb (but not noun) processing. (This is a paraphrase of the underlying mechanism of masked priming as provided by the authors on page 744.)

2. Basal ganglia dysfunction slows the ability to recognize nouns relative to verbs in a lexical decision task.

Maybe priming-related pre-activation is a critical function of lexico-semantic networks, but it seems to me that slowed recognition is a bad thing as well, maybe even worse. Still, I don't know whether PD causes noun or verb problems (or both).

More generally, I'm beginning to wonder what lexical decision effects in these sorts of studies are actually telling us. On the one hand, it is possible to argue that lexical decision provides a highly sensitive measure of aspects of language processing, some of which are automatic and unconscious. In this sense, it seems like a good task. On the other hand, we don't normally walk around making lexical decisions on visually presented words. Does this task involve meta-linguistic processes that aren't normally involved in noun and verb processing? Is it a modality-specific (i.e., reading-related) effect? Note that modality-specific verb deficits have been reported (Hillis, et al. 2002).

So while the findings are certainly interesting, and add to the large literature demonstrating dissociations between noun and verb processing, the Boulenger et al. paper is not "compelling evidence" for motor involvement in action verb processing. We don't know that it is the motor system that is causing the problem, the results suggest the possibility of a selective noun deficit, and it is not clear what the task is measuring.


Boulenger, V., Mechtouff, L., Thobois, S., Broussolle, E., Jeannerrod, M., Nazir, T.A. (2008). Word processing in Parkinson's disease is impaired for action verbs but not for concrete nouns Neuropsychologia, 46 (2), 743-756 DOI: 10.1016/j.neuropsychologia.2007.10.007

Argye E. Hillis, Elizabeth Tuffiash, Alfonso Caramazza (2002). Modality-Specific Deterioration in Naming Verbs in Nonfluent Primary Progressive Aphasia Journal of Cognitive Neuroscience, 14 (7), 1099-1108 DOI: 10.1162/089892902320474544

No comments: