Wednesday, November 10, 2010

Comments on NLC 2010 #1

NLC starts tomorrow and after letting the 14 boxes of scientific programs age in my garage for a few weeks, I finally pulled one out and had a glance. One of the first abstracts I came across was one by Willems et al. (abs #8) titled, A functional role for the motor system in language understanding: Evidence from rTMS. Since the title includes the term "language understanding" I was hopeful that they assessed language understanding. My hopes were dashed when I read that their task was lexical decision. They argue that lexical decision is a "classical indicator of lexico-semantic processing" (note the terminology change: they did not claim it was a classical indicator of "language understanding"). I suspect that like syllable discrimination in the phonemic perception domain, lexical decision is a highly misleading indicator of what normally happens in "language understanding" because (i) we don't normally go around making lexical decisions (it may involve additional cognitive processes not normally involved in comprehension), (ii) you don't need to understand a word to make a lexical decision (think, "familiarity without knowing"), and (iii) lexical decision data usually comes in the form of RT data even though it is a classic signal detection type paradigm, and therefore the data are subject to bias.

Skepticism aside, what did they do and what did they find? They stimulated left or right premotor cortex while subjects made lexical decisions to manual (e.g., throw) or nonmanual (e.g., earn) verbs. Left but not right PM stimulation lead to faster RTs to manual verbs compared to nonmanual verbs.

They conclude,

This effect challenges the skeptical view that premotor activation during linguistic comprehension is epiphenomenal... These data demonstrate a functional role for premotor cortex in language understanding.

I think they've shown clearly that premotor cortex plays a role in lexical decision. The source of this effect remains unclear (does it affect the semantic representation or just bias, e.g., prime, the response) and more importantly, the relation between lexical decision (the measured behavior) and language understanding (the target behavior) is far from clear.

In short, they have done nothing to curb the skepticism regarding the role of premotor cortex in language understanding.


Ivan Vankov said...

I can not quite understand the third problem with the lexical decision task (about LD being a typical signal detection task). Would please comment on this.

I agree that the relation between LD and language understanding is unclear. Still, the result is a strong evidence in support of the hypothesis that language processing (if not understanding) involves the motor system.

Did the subject reply by pressing a button? Could it be that the effect is due to motor priming the response action?

Greg Hickok said...

Lexical decision is a binary decision in which the subject indicates whether the stimulus comprises a word or not. This is structurally identical to detecting whether a signal (a word) is present or not (a non word). Without using SDT analyses, it is impossible to know the extent to which your measures reflect discrimination/detection ability versus bias.

GamesWithWords said...

This seems like one of those cases in which people just talk past each other. Did you mention this to the authors or ask them anything about it? My prediction is that they simply wouldn't understand your question. It's a funny phenomenon that many of the people studying language study lexical decision or noun-learning or something very accessible like that and think that's all that language is. Such researchers often feel that those of us coming from more of a language background are obtuse: why can't we see that their study of lexical decision for nouns explains everything there is to know about language? Some of the rest of us who study more classic linguistic phenomena like phrase structure, linking rules, etc., wonder exactly what they think they're studying. Everyone finds the other side insufferable. I think you can see some of that in recent arguments on my blog.

Unknown said...

(sorry, ignore/delete previously attempted comment)

Can you recommend any studies that test this hypothesis without the confounds of a lexical decision task or signal detection bias?

Greg Hickok said...

If you want to assess the role of (pre)motor cortex in language comprehension, then just use a comprehension test.

Ivan Vankov said...

I could not quite understand how the existence of a response bias questions the validity of reaction time data which are derived from correct positive responses. Is there any reason to believe that the bias is confounded with the experimental manipulation? If not, the effect of the bias should be uniformly distributed across conditions.

Greg Hickok said...

Well you'd hope that bias is equally distributed across conditions, but why not just do the calculation?

The thing is, subjects are clever. If they pick up on the fact that their motor system is being stimulated and they notice that they are hearing lots of words with meanings related movement, they might make a connection and start responding faster when meanings are motor related. I know it is a stupid explanation, but we did learn some lessons from Clever Hans.

More to the point though, the fact that stimulation of the motor system causes a modulation of response to motor-related words does not imply that the motor system is part of the semantic or lexical representation of those words. It could simply be modulating the *real* representation of the words via associative priming.