Gary Marcus, who seems to be gradually emerging as the Neil de Grasse Tyson of psychology, has a nice Times op ed today calling for greater scholarly focus on (and possibly funding for) efforts to not just compile but to make sense of the huge amounts of data which neuroscientists have been accumulating over the past decade or so. He notes, in particular, that the success of the field “ultimately rest(s) not just on the data to be collected but also on what can be done with those data once they are collected.”
He gives an indication of what he has in mind in observing that “we know that there must be some lawful relation between assemblies of neurons and the elements of thought, but we are currently at a loss to describe those laws. We don’t know, for example, whether our memories for individual words inhere in individual neurons or in sets of neurons, or in what way sets of neurons might underwrite our memories for words, if in fact they do.”
While this is surely well taken, it struck me as a bit unfortunate that the observation is not fleshed out by mentioning the field implicated in the “”memories of individual words” which assemblies of neurons will need to be reconciled with. That field is, of course, linguistics.
Gary, who is, at least for Facebook purposes, a friend, responded to me when I brought this to his attention that “linguistics actually was in the early draft, but cut in the savage but elegant trim to 800 words.”
The cut was understandable. However, one wonders what direction the piece would have gone in had the cut been restored and Gary able to develop it somewhat. Obviously, I don’t know, but I can suggest one sort of story he could have told which might have made the case for the relevance of the field.
It would begin with the observation that we all have a pretty good intuitive understanding of what a word is though if you ask most people, they would likely offer as a definition that a word is the group of letters separated by spaces when we write or read texts. But of course people don’t really believe that since they know that whose who either speak unwritten languages or who, for whatever reason, do not read or write are well aware of what linguistic segments are and are not words. Even those English speakers who have only heard or spoken, for example, the sentence
1)The dog is well trained.
know that the unit indicated by the letters “dog” is a word, whereas the last two letters “-ed” are not.
But that doesn’t mean there isn’t a grey area-and we intuitively understand this as well.
For example, consider the closely related sentence
2) The dog’s well trained.
Is “-‘s” a word? I don’t know what people would say, though I think it’s likely that those who provided a yes or no answer would be more or less evenly divided into two groups: lumpers designating “-‘s” as included within the single word “John’s” and splitters who would argue that “-‘s” remains a separate word.
Both sides have a basis for their conclusions.
The splitters might argue that we know that “-‘s” in 2) is a word since it can be moved to the front to produce the question
3) Is the dog well trained?
It would seem that only separate words are sufficiently self-contained linguistic objects so that they can be moved around in this way so on these grounds it seems reasonable to designate “-‘s” (what linguists call a clitic) as belonging to the mental category word.
But to that the lumpers would respond: yes, but notice that you can make the same argument with the “-ed” in the word “trained.”
Consider the sentence
4) Bill trained the dog.
When we change that to a question
5) Did Bill train the dog?
the same thing happens: you detach the “-ed” from the stem “train”, move it to the beginning of the sentence and then-so that it can function as a self-contained word-tack on a “d-” at the beginning to make “did”. We do almost exactly the same thing in converting in 2) to 3): the “-‘s” is altered to “is” and then it is moved to the beginning to make the question.
So, says the lumper, if you want to call “-‘s” a word, you have to call “-ed” a word too. But we know that “-ed” is not a word, it’s just a syllable (more precisely a past tense suffix), so by that logic, the lumper concludes the splitter is wrong to call “-s” a word.
To be honest, I don’t know who’s right-whether clitics such as “-s” should be construed as words-or “phonological words” as the linguistics refer to them (as distinct from the orthographic word which is an artifact of writing systems, as mentioned above). I do know that the question of what constitutes a word boundary is something which linguists have given considerable thought to having advanced what is known as the “prosodic hierarchy” for this purpose. Applying one formalism devised to describe the relevant facts, the boundaries between “Bill” and “trained” in 4), understood to be considerably more prominent than the boundary between “the” and “dog” with the former thereby represented with three as opposed to, for the latter, one pound sign (#) resulting in something like
I’m not going to continue with this discussion here since my objective in laying it out is not to provide an answer but rather to show in a simplified form the kinds of arguments which are made by linguistics-or more precisely one specific aspect of these arguments. Notice that nothing in what was just outlined required any sort of neurological data. This is so because, to repeat Gary’s observation, neuroscience as of yet has nothing to say as to “whether our memories for individual words inhere in individual neurons or in sets of neurons, or in what way sets of neurons might underwrite our memories for words, if in fact they do.”
Neurology is no help in telling us anything about what a word is, let alone whether “-s”, or for that matter “the” or “dog” is. But does that mean we should abandon trying to come up with an answer for what we understand words to be and how they are arranged to form sentences until neurology has something to tell us? Surely not; it means rather that we need to look elsewhere for evidence, namely in the sort of data the lumper and splitter were appealing to in their argument: our intuitions with respect to linguistic form. To take their examples, we know intuitively that certain sentences including 1)-5) are all more or less unproblematically acceptable in English, and based on this, they were correct in assuming that the arguments they were offering were supported by these sentences.
Furthermore, they were right in proposing that 5) can be derived from 4) and 3) from 1) or 2) according to the movement operations outlined, though it’s important to keep in mind that this is only the initial stage of the inquiry. To return to their argument as an indication of how it can be extended, as we saw in 4) we know that you can form a question by moving the last syllable of a past tense verb like “trained” to the front of a sentence, after altering it to “did”. However, it will be seen that you cannot apply the same operation to passive sentences such as
7) The dog was trained by Bill.
Here moving the “-ed” produces the sentence
8) Did the dog was train by Bill.*
which is clearly impossible (as indicated by the asterisk). It follows that the computation deriving questions from declaratives is more complicated than the simple movement rule suggested by the splitter.
As it turns out, the problem of how one derives the grammatically correct question from 7) is hard-or, rather, I should say that it’s been relatively hard for linguistics to come up with the answer (though it’s very easy for us English speakers to do it!) But while tricky the problem has turned out to be by no means impossible, and linguists have made real progress identifying the mechanisms underlying passive constructions (in particular, the theory of theta roles interacting with Case assignment) which need to be posited to get the right answer.
I won’t discuss what these are except to reiterate the point that our intuitions with respect to language are both necessary and, as it has turned out so far, sufficient, to provide the data which goes into constructing this kind of account. In short, within a theory of syntax and neurological data is, at least so far, neither helpful nor relevant.
Now, as Gary notes, it is true that we will eventually hope to unify linguistic syntactic theory with neuroscience, just as, for example, genetics was unified with chemistry with the discovery of the DNA molecule. Prior to the unification of these two fields, genetics and chemistry were self-contained disciplines each achieving results based on theoretical frameworks each had devised for its own purposes. As it turned out, pretty much normal application of existing principles of chemistry was all that was required to explain the basic facts of genetics. But before this could occur, a substantial theoretical literature within genetics having to do with inheritance of specific traits, dominant and recessive genes, alleles, chromosomes etc. dictated the form of the answer which chemistry was required to provide by means of atoms, molecules, compounds, reactions etc. And Franklin, Crick and Watson were required to understand, at least to a sufficient degree, what these boundary conditions imposed by genetics were in order to devise the correct account within the terms fundamental to their own field.
As Gary suggests, much the same should be the case with linguistics and neuroscience: we know there are words, but beyond this we know there is movement of these elements in sentences. In fact, we know, according to linguistic theory that there are two types of movement covert and overt movement (one taking place before the sentence is uttered the other after). We also know that movement is constrained within certain configurations as can be seen when the linguistic hierarchy is represented as a tree structure derived from combining (i.e. merging) pairs of syntactic units-the basis of the computation we perform in assembling sentences.
As the linguistic account becomes more detailed and robust, it becomes on the one hand more intricate and requires a bit more work to assimilate. But neuroscientists should be enthusiastic about engaging it. For as linguistics develops, the questions which neuroscience can and should be able to derive the answers to become clearer. At this happens, the “troubles with brain science” which Gary cogently discusses, while surely never disappearing, will begin to recede into the background.