Of Minds and Language (9 page)

P
ARTICIPANT
: I have a sort of exploratory question about the relationship of symbolic items that enter into Merge and content. One of our recent graduates wrote a dissertation on generics and he came to a conclusion where he basically just supposes a GEN operator and finds variables, and then that points him to a generalization. And while I'm sympathetic to that sort of approach, I'm not sure it is a strategy for studying mental content and its relationship to language in this way, because it sort of seems like, well, you try to work it out in a more conventional generative semantics way, but after a while you think, well, I can't really get this to work out, so let's just invent a new operator and say, hey, there's this mystery box in the brain that takes care of it. So while I think it is great to come up with answers like that, I'm just wondering about the research value of this and how to make this a little more solid.

C
HOMSKY
: Without going into that particular work, I think there is a question one has to ask about these things, and that is whether they are actually answers, or whether they are simply reformulations of questions. I mean, you have a certain phenomenon that is puzzling. You can sometimes kind of reformulate that phenomenon in technical terms, introducing certain assumptions about the nature of the mechanisms and so on. But then, the question you always have to ask yourself is whether your explanation is of the same order of complexity as the description of the phenomena. And I think it often turns out that it is. It often turns out that the explanation is approximately of the same order of complexity as saying here is what the phenomena are, in which case it is not an answer. It may be useful. Maybe it is useful to reformulate the question that way, and maybe that carries you on to some next stage, but it is a question you always have to be very aware of. Take things like work trying to explain ECP,
³⁷
or the
that
-trace phenomena or what have you. Possibly you get things which you could call explanations, but when you look at them properly, it turns out they are not really explanations; they are reformulations because you are introducing assumptions for which you have no reasons other than the fact that they help to account for this phenomenon. And insofar as that is true, you are restating the phenomenon in an organized way. Now again, that could be a useful step, because maybe this organized way of restating it leads to
suggestions about how to get a real explanation. But my suspicion about this case is kind of like that. Like where did that operator come from? Is it anything other than just a restatement of the data that we are trying to somehow find an account of? In that case, it is not an answer, though perhaps a useful step towards one. I think it is a question that always should be asked.

Cedric Boeckx

I wanted to discuss an issue that speaks to both linguists and non-linguists, and what I am going to try to do is first of all phrase a series of very general questions and then take one specific example, Merge (the most basic kind of example that I can take from the linguistic literature), in order to address particular questions of evolution with regard to that process.

To begin, let me just give you the context of my presentation. It is basically the biolinguistic perspective that Chomsky defined very well in the eighties by enumerating a series of questions that I think ought to be on everybody's agenda. The questions are as follows:

Part of what I would like to do in this paper is briefly establish a parallelism between a question that we have understood fairly well in the linguistic literature, namely the developmental question (2) and its analogue or cousin in the sense of evolution.

Another thing that Chomsky did that was very useful was to trace historical antecedents for these questions and give them names. So, for example, (1) is called Humboldt's Problem, and (2) is Plato's Problem, and that is the one that we are all very familiar with. Question (3) is Descartes's Problem, in many ways still a
mystery. Question (4), interestingly enough, is not easy to name. It is about the brain–mind connection, and very few people have had good intuitions as to how to go about solving that mystery. You could call it Broca's Problem or Gall's Problem, but it is very difficult to find insightful antecedents for this issue. I think there is a lesson to be learned from the fact that we cannot really name that question, despite the fact that nowadays question (4) is taken in many circles to be the one on which the future of linguistics depends. By contrast, problem (5) is very easy to name, and although no one has applied this name to my knowledge, it can easily be called Darwin's Problem. Just like Humboldt, Descartes, and to some extent Plato, Darwin was very much interested in language, and in fact if you read
The Descent of Man
, there are very interesting reflections on language. Interestingly, Darwin establishes connections between our “language instinct” (that is where the term comes from) and the abilities that for example birds display when they sing. I think if we actually read those chapters in Darwin, we would not be misled by some of the recent heat on songbirds. Darwin was ahead of his time in that context as well.

The questions that Chomsky raised defining the biolinguistic literature find very obvious correspondences with those that Tinbergen put forth in 1963 in a famous paper called “On Aims and Methods of Ethology.” These are the questions:

i. What stimulates the animal to respond with the behavior it displays, and what are the response mechanisms?

ii. How does an organism develop as the individual matures?

iii. Why is the behavior necessary for the animal's success, and how does evolution act on that behavior?

iv. How has the particular behavior evolved through time?

You can see that if you decompose those questions and rephrase them, inserting language in them, you get exactly the same set of questions that Chomsky put on the agenda. When Tinbergen put forth those four questions for ethology, he was very much under the influence of Ernst Mayr, and Dobzhansky's (1973) assertion that nothing makes sense, except in the light of evolution – Darwinian evolution, that is.

In the realm of psychology or the mental properties of cognition, we are in an uncomfortable position because we have to deal with a big phenomenon called “evolutionary psychology,” which sort of reduces that question of Darwinian evolution to adaptation. However, if you talk to real biologists, they know that evolution is actually much richer than just adaptation. In particular, I think that we should bear in mind three things about evolution, which are valid for
everything including questions about the evolution of the language faculty. The three things are the three factors that for example Stephen Jay Gould identified in a wonderful book called
The Structure of Evolutionary Theory
(2002): first, of course, adaptation, but then there are two others that psychologists often forget, namely chance (accidents of various sorts), and then structural constraints (some of the things that fall into the laws of form, if you want: what Chomsky now calls “third-factor” effects). There is actually a good term that comes from Wallace Arthur (2004), namely “bias,” in the sense of “biased embryos,” meaning that embryos develop or evolve in some directions and not in others. So if you combine adaptation, bias, and chance, you get this ABC of evolutionary theory, which is worth bearing in mind, particularly in approaching questions on the evolution of language. In doing so, we should also recall some of the early results that Lenneberg put forth in his 1967 book on the biological foundations of language, where he was very much interested in questions concerning the brain–mind connection and the question of evolution.

I think we have made progress recently in linguistic theory that enables us to address those questions a little bit more precisely. In particular, it is well known to non-linguists who attend linguistics talks that the jargon is so developed that it is hard to start a conversation, let alone address questions that are of an interdisciplinary nature, much less design adequate experiments. But here I think that the minimalist program in particular has forced linguists to go more basic, that is to develop a series of questions and answers that to some extent may help us to talk to non-linguists and address those questions, in particular questions (4) and (5).

To continue with the fifth question, Darwin's Problem, I first want to note that in various ways it shares similarities to the way we approach Plato's Problem. As everyone knows, when talking about Plato's Problem, one has to mention poverty of stimulus and the fact that children face a formidable task that they have to solve within a very short window of time. The result in a very few years is uniform acquisition – very rapid, effortless, and so on and so forth. I think the only way to really answer Plato's Problem generally is to give a head start to the child and say that much of it (the ability to develop or acquire language) is actually innate and built in somehow, in the genome or elsewhere (epigenetics), but it is at least given, it does not come from the input the child receives. This way, you can make sense of the task that is being fulfilled and achieved within the very short window of time that we all encounter.

That is exactly the same problem as the issue of language evolution, because everyone who has thought about the evolution of language seems to agree that it also happened within a very short space of time. Just like in the context of Plato's Problem, it appears that human language as we know it developed very,
very rapidly; and it's uniform across the species (Homo sapiens). So the way we should try to address and solve that problem, given that short time frame, is to do exactly what we have done for Plato's Problem, namely to say that in large part you want to make the task “easy” – that is, you want to make sure that the thing that has to evolve is actually fairly simple. You also want to say that much of it is already in place when you start facing that problem. This brings us to the distinction, or the combination, of the language faculty in the broad sense (FLB) and in the narrow sense (FLN).
¹
The more you put into the FLB, the easier Darwin's Problem becomes. Just as we attribute a lot to the genome for his problem, so should we try to make sure that FLB contains quite a few things already, such that the thing that has to evolve is actually plausible as an organ subject to all the pressures of evolution.

I think that the FLB/FLN distinction becomes tractable or expressible especially in the context of the minimalist program, where you can begin to try to give some content in particular to FLN. And here I am building on work that Hauser, Chomsky, and Fitch did (Hauser et al. 2002, Fitch et al. 2005) by suggesting that one of the things that seems to be part of FLN is the operation Merge, which gives you this infinite recursive procedure that seems to be central to language. But here what I would like to do is suggest a slightly different take on the issue, or rather suggest a different way of defining Merge, that I think gives a slightly different program for linguists and non-linguists when addressing Darwin's Problem. Specifically, I think that there are some advantages in trying to decompose Merge a little bit further into more basic operations, to reveal not just the very general character of the operation, but also some of the specificity that gets into Merge to give you language and not just any recursive system.
²
In particular there is one thing that is quite clear about Merge and language: once you combine two units, X and Y, the output is not some new element Z, but either X or Y. So the hierarchical structure that we get in language is a very specific sort, namely it gives rise to so-called endocentric structures. That is the role of labels in syntax. So for example, when you put a verb and a noun together, what you get (typically, say, for the sake of concreteness) is a verb, and that verb, or that unit, acts as a verb for further combination. Now this, as far as I can tell, is very, very specific to language as a kind of hierarchical structure. If you look elsewhere in other systems of cognition (music, planning, kinship relations, etc.), you find a lot of evidence for hierarchical structuring of systems, possibly recursive ones, but as far as I can tell,
those hierarchical structures are not headed or endocentric in the same way that linguistic structures are. That, to my mind, is very specific to language, so while you find hierarchies everywhere, headed or endocentric hierarchies seem very central to language. And so of course they would be part of FLN.

As soon as you identify this as an interesting and unique property of language, the next question is how does that endocentricity come about? The brute force answer might be to say “Well, this is the way you define Merge.” But I think that there is a different, more interesting way of getting endocentricity that will actually raise other questions that people like Marc Hauser can address from an experimental perspective. For example, I have suggested (Boeckx 2006) that one way of getting endocentricity is by decomposing Merge into at least two operations. The very first operation is, say, a simple grouping procedure that puts X and Y together, and that presumably is very common across cognitive modules.
³
It is not very specific to language. Putting things together is presumably so basic an operation that it is, if not everywhere, at least in many systems. The next operation is selecting one of these two members and basically using that member as the next unit for recombination. For linguists, this is actually an operation that is well known. It is typically called a copying operation, where you take X and Y and then you, for example, retake X by copying it and recombine it with something else.