Of Minds and Language (67 page)

R
IZZI
: A small technical question about what the P600 effect really reveals, what kind of brain computation it expresses. You made a remark in passing, if I caught it correctly, according to which in a certain task, if the task was simply passive listening, you would not see a P600 effect. Does that mean that you see a P600 only when there is some kind of metalinguistic task, or not? Because of course that would lead to other different conclusions about what the effect really indexes.

F
RIEDERICI
: What I can say is that the P600 is a controlled process, so for example we have done an experiment where we had just these simple syntactic violation errors, and either there were 20 percent of the sentences that were incorrect vs. 80 percent correct, or the other way around (Hahne and Friederici 1999). What subjects had to do here was they had to judge grammaticality. So maybe not surprisingly, when you have 20 percent incorrect sentences in the experimental set, you see the ELAN and a nice P600. When you reverse the proportion of correct and incorrect sentences, you see the ELAN which is not even influenced in amplitude or anything by this variation. However, when you have 80 percent incorrect sentences you don't see a P600, I think – this at least was our explanation. The system would not go into the revision process any more, even though at the end of each sentence subjects had to do a grammaticality judgment task. We also see that depending on what task we
use, whether we have a probe-verification task or a grammaticality judgment, the amplitude of the P600 varies as a function of that. It is larger for grammaticality judgment, as you suggested, and not so large for some other task where you do not have to process the entire syntactic structure.

F
ODOR
: I'm interested in how much alike we all are in these respects and how much variability there is both in the location and temporal scale, because in the old days when there were only lesion studies as the data, we were taught that left-handed people had half their language in one hemisphere and half in the other, and so forth. So I'd like to know how tidy the LH–RH separation is and the time of the responses.

F
RIEDERICI
: I think with respect to the groups we investigated, I cannot say anything with respect to this issue as we have only looked at right-handed subjects so far. We've looked at left-handed subjects in one single fMRI experiment. In this study we also did a dichotic listening experiment on these subjects in order to figure out whether they really had the “crossed” hemisphere indication. The fMRI data revealed that only about half of the left-handers have a language dominance in the other hemisphere, that is the right hemisphere. Dominance classification based on dichotic listening worked much better than the usual handedness questionnaire (Hund-Georgiadis et al. 2002). Just looking at handedness would thus not be enough; you always have to do additional tests, and that then gives you more variables you have to consider in order to do a well-controlled experiment on language dominance. With respect to the timing of brain response, we haven't really looked at individual differences for the P600, but we did so with respect to ELAN with the MEG experiment, and I was really surprised to find that the peak of this early effect was not more than 25 ms apart between the subjects we have been looking at (Friederici et al. 2000).

F
ODOR
: I see. So one issue is more and less advanced language skills. I mean there are scales on which you can rank people, but my real interest is actually when the syntax is over in the RH, is it crowding out the prosody? Are these trying to occupy the same space?

F
RIEDERICI
: As I said, we don't have data on right-dominant subjects so I cannot answer this question. But you are right in raising the issue about the RH involvement in general. In our crucial experiment concerning the prosody–syntax interaction conducted in right-handers, we were looking at the brain's reaction to an element that is not directly at the point of the critical prosodic information. You first have the intonational break and we are looking
at the verb that comes two words after it.
⁷
So these data could mean that it needs some time for the prosodic information to influence syntactic processes. It is very difficult to find material where you can show the exact timing of the interaction of information types within these sentence structures. What we have done, therefore, is to look for material with a counterpart of the local phrase structure violation in the prosodic domain. What we have been doing is the following. In a violation sentence like the ones we were using before,
⁸
the prediction at the preposition, which is case-marked, could be two-fold. One is syntactic, where you predict a certain word category, but the other one would be a combination of syntactic and prosodic information. Because you know as a German speaker that the main verb should come at the end of the sentence, you predict that the next element after the preposition should not have a sentence-final prosody. In crossing these information types fully in a two-by-two design, we find that independent of the syntactic violation, the wrong prosodic intonation that an element has taken elicits an early right anterior negativity indicating RH processes. Moreover we find an interaction between prosody and syntax even for the combined violation condition (Eckstein and Friederici 2006).
⁹

L
AKA
: I was curious about the patients that you looked at. Outside of experimental conditions, are these CC patients people who show any symptoms of lack of integration of prosody and syntax? I was trying to recall these famous patients who had the CC cut surgically and could not recall any symptoms of this sort, and I was curious as to whether the patients you looked at showed any signs of this lack of integration.

F
RIEDERICI
: Well, you have to test these patients very carefully. Gross testing or coarse testing would not show that, because they always are able to compensate. I didn't go into detail concerning the anterior portion of the CC, which connects the two frontal cortices. For all our syntactic and prosodic studies in normals, we have seen activation also in the frontal operculum of the two hemispheres. For the moment we do not have a really good idea of how the two anterior and the two posterior portions contribute to the observed effect of the N400 in normals. However, you may remember that the N400 was reduced in those patients who had lesions in the anterior portion of the CC, but only reduced in the second part of the N400. The first part peaked well, but then the effect flattened out. I think that also the anterior regions (these are the frontal
operculum of the RH and the LH) talk to each other but in a secondary process. I think in the N400, at least this is what you would conclude from the data here, there is an overlapping of two processes. Thus those patients with lesions in the posterior portion are perhaps able to compensate in behavioral tasks based on the anterior portion of the CC.

P
ARTICIPANT
: Just a quick follow-up from what Janet Fodor was saying: the connection with syntax and prosody. What are your thoughts on the processing studies that have been done with sign language? One of the big issues used to be the use of the RH, but a British group seems to have managed to discard the RH effect for sign language processing. Do you have any thoughts on that?

F
RIEDERICI
: Well, we have thoughts and we have data. Prosodic information is very much encoded in mime and facial gestures, so if you are able to separate those out in an fMRI experiment, you should see a very similar distribution as for normal language processing. I mean, forget about the visual cortices, because the information has to go through that in the first place, but then when you only have the subjects looking exclusively at the hand movements, that is not enough, as prosody is very often signaled by eyebrows and other facial gestures. I think I know of no study that has very nicely separated those two aspects, but it is a nice idea to do that.

If there are no other questions, I would like to thank the organizers for holding this wonderful conference and for inviting me to give this talk to you. Thank you.

Noam Chomsky

First of all, I'm here over my own strong objections. When I saw the program I wrote to Massimo and said that I'm not the right person to do this so somebody else ought to, and I suggested that he ought to because he's the one person who covers all of these topics and I don't. But he's very persuasive, so I fell for it, and that just made it even worse. He said I should go on as long as I liked. My children used to have a line; if they asked a question they used to say, “Please, just the five-minute lecture.” So I'll just go on until you shut me up.

I've tried to think a little bit about how to organize some comments. An awful lot of fascinating material has been presented here, some of which I understood, some of which I didn't. What I'll try to do is pick out some points that come to mind, starting from the most general to the more specific, and expressing an apology in advance to everyone whose work I misrepresent. I'll try to do as little of that as possible.

The most general point was a significant one of Jim Higginbotham's presentation (see pages 142–154) which actually carries an important lesson. Namely, that if you look back in history, you find that they were often recovering ground that had been partially attained and understood. And that's true; it came up in the discussion that generative grammar goes back 2,500 years. It didn't have to be discovered in the late 1940s; it goes back to the Paninean tradition which developed for centuries.
¹
And Panini himself was the result of a long, mostly unknown prehistory. And the same is true case after case; the same sort of thing happens in the sciences all the time.

So in biology, Mark Hauser gave a talk here on the illusion of variety (see
Chapter 19
), which was the position of Geoffroy in the famous Cuvier– Geoffroy debate.
²
Geoffroy was thought to have been demolished in the debate,
but it's coming back that he was in a deep sense right and that rational morphology, which had been derided for centuries, is somehow right.

This sort of thing continually happens, and the lesson is that science is a kind of hill-climbing. But you can get caught on the wrong hill, and you have to know that you should go down and start somewhere else and then go up; and often you find that people were higher up before you somewhere else. It's very easy to just get caught up in one's own conviction that the interesting line of work, which is raising technical problems and is fun to solve and so on, is really the answer, when it could very well be a sidetrack. We have all too many examples of that in the past – the recent past and the more remote past – and you have to constantly keep in mind the importance of knowing and remembering what happened, and keep an open mind about whether maybe those guys weren't so dumb as they looked, and that it's worth doing.

I'll just give one personal anecdote that came to mind in that discussion. Around 1960, a very famous and accomplished historian of Classical Greek Philosophy, Gregory Vlastos, came to Boston to give a talk on one of Plato's dialogues, the Meno.
³
And a couple of us from the research lab in electronics decided it would be fun to go – I can't remember exactly who, I think it was Jerry Fodor and maybe Julius Moravcsik, who was a visiting philosopher. So we went over and when Vlastos opened the talk, which was for philosophers, he was very apologetic for talking about all this stuff that we know is wrong and has been disproved. But he gave a serious talk about a serious philosopher, and after the discussion at the end we went up and talked to him, and it took him a while to figure out that we actually thought that Plato's argument was right and that it wasn't a crazy stupid thing that had been disproved. And when he finally realized, he got very excited and we went off and had lunch together and had a great discussion. He believed it, but it simply wouldn't have occurred to him that anyone in those days – in the 1950s or 60s – could possibly pay attention to this old, boring, dead stuff, which in fact turns out I think to be fundamentally right. It's come back in another form, but that can happen very often. And it's worth remembering.

The kind of work that Marc Hauser and Chris Cherniak were talking about in this conference, and the other work that's been referred to, is reconstructing and recovering consciously ideas that had been discredited. We were under the illusion that variability is limitless, the same illusion that resulted from anthropological linguistics.

I think that the broadest issues that arose in the discussion were the questions about the prerequisites for experience. So, to quote Rochel Gelman – quotes
here mean whatever I jotted down, probably wrongly – she made a distinction between core domains and other domains as a prerequisite for experience (see
Chapter 15
). The core domains involve a high-level, abstract conceptual framework, and mental structures actively engaging with the environment from the start. These core domains have several properties: they're reflexive, they're quick, they converge, they're common among people. They have many of the properties of Jerry Fodor's modules,
⁴
except that these are a different kind of module; these are acquisition or growth modules, not processing modules. It's not an unrelated notion but they are conceptually distinct. These are the rough properties of the core domains and they also involve from the start high-level abstract conceptual structures, not just picking out sensations and so on.

Randy Gallistel gave an amplification of this by providing a Kantian framework of foundational elements in terms of space, time, number, and so on (see
Chapter 4
). In ethological terms, the core domains with these foundational terms are what provided the Umwelt, the world of experience of the organism, which differs for each organism but is some kind of a complex world, and that's the kind of world that you're presented with. If we go back a century before Kant, there was rich discussion of these topics by mostly British philosophers, the neo-Platonists and the British empiricists, who talked about what they called cognoscitive powers. Somehow the organism has rich cognoscitive powers – they were only talking about humans – and these involved gestalt properties, causality, intention, and lots of others.