Thursday, December 13, 2012

Occam's razor and the point shavers

One of the points I've been mulling over from the Honda book is the assertion that Occam's razor (the principle of parsimony) would dictate that remembering two rules, say A & B, is harder and more complex than remembering the choice between those rules, A OR B. Thus, it would be more parsimonious to posit

1. John gave to Mary the book

as opposed to 2. John gave the book to Mary

ss the underlying structure of

3a. John gave the book to Mary and 3b. John gave Mary the book.

To get from #1 to the first sentence, you would delete "to"; to get to the second you'd move the prepositional phrase. Thus you have A OR B which Honda posits as parsimonious. However, to get from #2 to the first sentence, you'd do nothing, but to get to the second, you'd have to move the phrase forward AND delete the "to". Thus the second rule has more complexity?

At first my take on this was to wonder who said there's an underlying structure in the first place? Or, that the underlying structure itself has any order? When I was in graduate school we had this idea of marked/unmarked which didn't quite explain everything but we'd use it in a case like this: When the indirect object is right up against the verb, it must be marked, and it's marked by not having its preposition. OK, so it's like a sign of a specially marked indirect object.

Honda kind of flies through this without much soul-searching or discussion, but to me it brings up the heart of the problem: how do we explain the internal workings of our minds in the simplest way possible? We native speakers hear such sentences every day and have representations of these rules (transformations, allegedly) in our minds.

My students were of no help in the matter. They stared at me blankly when I told them that 3a. and 3b. were legitimate sentences in English; that we consider them both grammatical; that linguists would like to explain their relationship and explain the rules we native speakers use to get from 3a. to 3b., or from the underlying structure to each of them. They had no problem with the idea of "direct object" and "indirect object" which some had heard of, vaguely, at some time in their past. They really really didn't know which would be more parsimonious.

My second reaction is that any time you have A AND B, you have to possibility of bundling them, and saying something to the effect that, "any time you do this, you have to do these," putting A and B together, and remembering them together as basically one rule, one move. Remembering two rules AND remembering to choose one or the other but not both, seems to me more complex, and requires that you separate them out in your mind, and keep them separate. Oh bother!

What might be required here is a closer look at what the rule of parsimony really entails. I think that almost all linguists buy into the idea of Occam's razor to begin with; that's because, with language having as many rules as it does, everyone has to grab and remember them as well as possible, and what can't be remembered, doesn't survive. However you represent what is happening in our minds, it has to ring true with us; it has to make us say, "that's what I do, I've been doing that all along."

Now my students, as I said, were of no help here. They don't know what they've been doing all along, and they're not sure if "transformations" or "movement" is a fair representation of what is happening. Finally I challenged them: You tell me what is happening (I said). If you don't like Honda's explanation, find a better one! (silence again).

Two issues came up repeatedly. One was Chomsky's claim of hierarchical structure of trees. Thus we have trees which represent the underlying structure of the sentences we produce, and moving a prepositional phrase such as "to Mary" to the end of a sentence would allegedly move it to a higher node, thus would align with a generalization that all transformation moves things to a higher node (I have come to align this general rule with the rule about kittens and Christmas ornaments) - but, I don't really agree with this for several reasons. First, it seems that maintaining this theory would cause trouble with other structures, most notably passive (John kissed Mary becomes Mary was kissed by John) - in which you would have to go out of your way to posit underlying structures with nodes that were somehow lower than places where these structures would end up. But second, what justification is there for calling the whole picture hierarchical? None. I don't feel, deep in the quiet center of my mind, the need to take any language I hear and construct my understanding of it in a hierarchical way.

Now mind you, we construct most things in heirarchical ways. When we see a group of people, we try to figure out who their leader is. When we see trees, we say, it's a kind of elm, and elm is a kind of shade tree, etc. People have constructed hierarchies and have constructed hierarchies in our understanding of the universe, especially for other things we don't understand. But that doesn't mean that nature constructed them. Or that we, in our inner minds, constructed what we heard as hierarchical in nature.

So I would not be surprised that people believe that the structure of language is hierarchical in nature, and that they believe Chomsky when he claims that such hierarchy-based transformations make up what we do and understand as language; Chomsky has had a good run with this. But my question remains: what is parsimonious? What is the way that most people will remember this stuff?

No comments:

Post a Comment