As designated Page Six for the Institute, I guess it’s my duty to report on some “state of the field” discussions that I’ve been privy to this week. Since it seems like this is one of the inevitable and special things that happens at these kinds of events. I’ll start with a wrapup of Joan Bresnan’s provocative (or not, depending on who you are - but she meant it to be provocative) lecture on Tuesday night, “Empirical Foundations of Syntax.”
Bresnan began the talk with about a 10-minute homage to Bill Labov, from whose 1975 article “Empirical Foundations of Linguistics” she fashioned her own title. In that piece, Labov critiques the common practice of introspection among syntacticians (”I wonder if this is grammatical…well, it sounds kind of bad…let’s just give it a star, shall we?”). In particular she takes up the criticism of Chomsky’s (from 1957’s Syntactic Structures) suggestion to “let the grammar decide” when in doubt about grammaticality: “Whose grammar?” she asked. Dealing with introspection led to a bunch of mismatches between what syntacticians had been calling grammatical and what was actually found in naturalistic data (she used examples of resumptive pronouns and relative clause extraposition and something else…). That is, people actually do produce stuff that syntax claims is “ungrammatical.”
As retold by Bresnan, Labov’s admonition to the field, to mitigate the disastrous contrast between what actually happens and what gets described, was to take on an empirical approach to data and analysis. And then she put up this slide that said:
Most sociolinguists listened to Labov. Most syntacticians did not.
–Dramatic pause while you digest that dramatic claim–
For the rest of the talk Bresnan presented lots of data from her own and others’ research, employing mainly corpus-based and experimental methods for doing syntactic research. What they’re looking at is now-totally-hot probabilistic grammar. Her results show that you can make a model of usage based on corpora, and the model can predict with good accuracy what constructional variant will occur given a set of grammatical features (for items where there’s more than option, such as dative alternation between double-NP and PP). AND that people tend to judge things as acceptable or not in line with those same probabilistic patterns as the model does. Which is to say, people seem to be judging grammaticality not just on some internalized knowledge of grammatical structure, but also on patterns of actual use.
Then she raised the question of whether speakers have implicit knowledge of statistical patterns, presenting it as something that was a “really CRAZY” suggestion. Such knowledge would give speakers the ability to predict what others are going to say, explains apparent inconsistencies in their grammaticality judgments, and varies across dialects and varieties, just as grammar does. [I can't tell from my notes whether that sentence is paraphrased or verbatim from Bresnan, so apologies if I took it verbatim without quote marks.] This doesn’t mean there’s no grammar, as she was sure to point out in the Q&A section. It just means that there’s more than one mechanism working to construct people’s linguistic knowledge - and (at least) one of those mechanisms is usage-based, i.e. frequency-based.
OK. There are a lot of interesting things about the talk, either about its content or its manner or presentation or its implications. First of all, a syntactician spent most of her talk lauding Labov, a sociolinguist, for his approach to linguistics. This is, actually, kind of crazy: the two subfields haven’t historically been on awesome terms, and sociolinguists often feel like their emphasis on empiricism, actual usage, and speaker-influenced (contra linguist-influenced) models, gets very wrongfully overlooked by the mainstream/theoretical parts of the field. So on the one hand, it’s nice to see sociolinguistics (finally) getting a claim to best practice.
On the other hand, I was sitting next to another sociolinguist at the time, and we basically just passed a “Well, duh” look at each other back and forth during the talk. Fundamentally, what Bresnan was saying is that variation exists, that it’s somewhat predictable (variable rules, anyone?), and that we should seek to account for it in our theories/models of language. YES. Yes, this is true. Anyone who’s ever dealt with dialects (”nonstandard” or whatever) deals with the fact that people produce stuff that isn’t supposed to be possible, or that isn’t supposed to be normal, for a particular language’s (= individual’s) grammar; especially if you’re trying to look at variation within Chomsky-centric generative theories (I’m thinking immediately of Alison Henry here). And that their realizations are often conditioned phonologically, socially, or grammatically. So it’s frustrating, really - like, really really frustrating - that while Bresnan, one of the most prolific and respected and brilliant syntacticians, is recognizing sociolinguistics for its contributions to theory, she doesn’t cite a single actual sociolinguistic finding or analysis in the process. AND that it’s taken this long for the recognition to happen at all.
Meanwhile, on my right foot is the question of Labov and sociolinguistics. After his lecture (Divergence) a few weeks ago, several sociolinguists were kind of shocked at a comment he made towards the end of his talk: it had something to do with there being no social meaning to the difference between North and Midland spread (or not) of the Northern Cities Shift vowels. He said that NCS was a purely structural change, not affected by social patterns. Or that’s what I got out of it - I’m still not sure I understand his claim entirely. This is not a very sociolinguisticky thing to say, at least not in this day and age. In fact, there’s kind of a whole post-Labov school (I’m thinking primarily of people like Eckert, Mendoza-Denton, Bucholtz, Alim, Moore, etc. — if anyone grossly objects to this term, speak up — perhaps you might like “third wave variation” better?) that is more interested in emergent social meaning of linguistic practice than the numerical distribution of linguistic variables. This maintains a focus on interaction and statistical patterns but situates the locus of social differentiation in practice; not the locus of variation in social categories. [I don't know if that made any sense. I'm still but a wee linguist, and if anyone with more credentials is reading this, please add some sense...my brain is mushy mushy mdajfldjaofdaoiaoieewoy....]
The buzz goes like this: Bresnan is citing Labov, which at first it’s like, “Whoohoo! Sociolinguists represent!” Except then you think about it and you’re all, “But wait, are Labov’s goals/methods nowadays still representative of ’sociolinguistics’? Didn’t he just say that social stuff sort of doesn’t matter?” [update: I didn't mean this to sound like I was conflating a bunch of different sociolinguistics into one big sociolinguistics. There are a bunch of different sociolinguisticses. I don't want to be misinterpreted, and I think that all of the discussion here at Stanford about Labov in various classes and with various people has really got me flummoxed and kind of confused and torn.] And then it’s like, “Wait, maybe they just like us for our use of numbers.” And meanwhile other people are like, “Numbers aren’t the whole picture! We need to stop focusing on numbers so much!” And then you go, “Did she really just call discourse markers disfluencies? My, we still have a long way to go.”
Anyway. It was interesting. [on the title]