On sentences

Someone linked this, and I couldn't help thinking "Seriously, someone got paid for pointing out that sentences have a subject and a bunch of stuff that isn't the subject, except when they don't?" She also mocks a couple of bad spoken sentences but they would have made more sense if she had correctly analysed sentences in the first place.

If she had used the terminology "topic-comment", her article would have been greatly improved, although still deficient. English sentences largely consist of a thing you are talking about, and what you are saying about it. In the spoken examples, what you have are a set of utterances that have different topics, which could be analysed as separate sentences, rather than components of the same sentence. In speech, more so than in writing, we often use what you might call topic chaining, where part of one topic's comment becomes itself a topic.

The old-fashioned notion of "subject and predicate" is more closely defined by calling it "topic and comment", allowing the terminology "subject" and "predicate" to refer to building blocks of a sentence (generally, the grammatical subject and the verb -- you might consider an adverb that qualifies the verb to be part of the predicate, or you might equally consider it along with any objects of the verb to be arguments of the predicate).

Just so we're clear, look at the sentence: "I gave the dog a bone."

Here the topic of the sentence is "I". The sentence is about me. The rest of the sentence tells the reader what I want to say about the topic "I". This idea quite nicely captures the commonality between comments about actions and comments about states, yet does not help us understand how they are different.

So we understand that:

I gave the dog a bone
and
I had a rueful look

have something in common, but using a deeper description helps us see how they are different. In brief, one is about something I did; the other is about something about me. The difference is not clearcut, of course, and the skilful writer can use sentences that look like the first type to represent the second:

I smiled.

Here we are describing an action, yet at the same time we imply a state. How about:

I gave the dog an angry look

?

While these sentences seem formally the same, they are functionally very different. I'm not going to get into how you might formulate a predicate grammar that captures the difference, but it should be reasonably clear that it would be possible.

There was some conflict in linguistics in the 1960s between those who believed that syntax and semantics are formally distinct: that we generate syntactic structures and semantics is in some way an overlay, so that the grammatical rules and the semantic rules are applied on different levels, with grammar at a deeper level; and those who believed that semantic predicates generated grammar. The latter lost (they were a minority of dissenters) but I feel that their view had some merit. The former is reductionist: somehow the meaning of sentences can be derived from their building blocks in the same way that the actions of chemistry can be derived from the interaction of electrons, so that what on the surface seems richly meaningful is in fact emergent from a simple process, gaining complexity only because of the great range of possible interactions. You could argue that the difference between transformational grammar and generative semantics lay simply in a disagreement over what form the "atoms" of syntax and semantics took.

I feel that where transformational grammar failed somewhat was that whereas it was very good at explaining why a sentence would be considered well formed in a language, it could not explain -- didn't try to -- why some sentences seem "better" to us in a qualitative sense. A couple of things are certainly true about sentences in English: we feel information is conveyed more succinctly by sentences where the verbs carry more of the semantic load than verbs, and we feel that even more so where the verbs describe actions, rather than states. Furthermore, I'd argue that we feel that sentences are better when their semantic load (the amount of meaning they contain, if we understand a sentence's meaning to be roughly the same as its information content) is spread over fewer rather than more words.

Each predicate is in a sense a peg on which we hang the information contained in its arguments, and we tend to feel that the more information is provided by the predicate, and the less by the arguments, the better the predicate-argument complex is at conveying information, or at least, that the risk is that arguments will not add much to the information conveyed.

Consider:

I broke the glass.

Here, usually, more information is conveyed by the verb than the verb's argument (unless you are answering the question "What did you break?", having said, perhaps, "I broke something".) Information can be understood as material the reader did not know, and I think it's simple enough to argue that there is more new in the fact you broke something than in what it was, if only because the fact of breaking is prior to the fact of what was broken.

However, consider:

I smiled sweetly.

Here you could argue that the underlying state that this sentence describes is more new than the fact that I smiled. "Sweetly" conveys a great deal, none of which we actually express.

Look again at:

I gave the dog an angry look.

Can we argue that the fact that I am angry (with the dog, presumably) has more information content than the fact that I gave the dog a look? I think so and I do think that a proper analysis of sentences needs to explain this. Time and space forbid me from trying (let alone that I have no idea how I would go about it without giving it some serious thought) but I think that at least having the notion that parts of a sentence are required to contribute to the semantic load of the sentence or face expulsion is important to a writer.

posted by Dr Zen @ 6:20 pm