Noncompositional

During this year’s annual meeting of the LSA at Albuquerque, Sandra Thompson (ST) gave a plenary address entitled What are clauses for? Understanding grammar in terms of social action.. In discussion of the talk along with various people in the days following, I found that most people either disagreed with the premise or conclusion of the talk, or were unimpressed or confused regarding basically all aspects of the talk: the premise, the methodology, the argumentation, and the conclusion (if there was such). (And for those who didn’t happen to see the talk, my attempts at summarization fared no better). So I decided to write up my summary of the talk, what I think about it, and all that good blogging stuff.

For this summary, I am going off what I remember from the talk, the brief notes I took during it, and a paper by Sandra Thompson and Elizabeth Couper-Kuhlen in Discourse Studies (vol 7, 2005) called The clause as a locus of grammar and interaction. The talk followed the paper rather closely, though the former had more data and an additional section on adjacency pairs (though she may have called them “conversational doublets” or something else like that). The paper also compares English data to Japanese. I unfortunately do not have access to the abstract, and the slides are not available online AFAICS (…can see).

The basic goal of the talk, as I saw it, was to introduce a paradigm (if you will) in linguistics called Interactional Linguistics. It draws on findings and intuitions from conversational analysis, discourse analysis/pragmatics, interaction studies, and possibly also construction grammar and evolutionary and connectionist approaches to language development and use. Interactional linguistics is not, as the paper says, a separate level of linguistic analysis nor is it a module in the grammar:

When we talk about language in interaction, then, we are not so concerned with staking out a new ‘subdiscipline’ as we are with discovering the nature of grammar seen as social action and interaction. (p. 483)

On this (new) view, each unit of traditional language analysis, from phonemes to clauses and beyond, are subject to a new sort of methodology that centers around interaction - how are these units used in interaction, and indeed how might they arise from interaction? In the talk, ST focuses on clauses because they, she claims, are of the utmost importance in linguistic interaction and in the cognitive organization of language-as-tool. The clause, in turn, is dominated by the predicate, which is associated with a set of phrases, or arguments. The paper states:

[The predicate] is the element that enables recipients ["hearers"] to know what social action is being carried out by a given utterance in a given [...] context. (485)

This view is reflected in the talk, which claimed that (at least) the following two aspects of interaction are key to understanding the nature of the clause: projectability and intersubjectivity. Projectability involves the notion of expectation, and in particular expectations that conversants have about the course of the conversation. The ability to project the trajectory of a clause (i.e., predict what the current speaker is doing, when they will finish, and what they will have accomplished) is crucial to manage turn-taking and joint-utterance productions (finishing what your partner has started to say).

ST gave several examples of English turn-taking and joint-production, and the paper gives some more, as well as several Japanese examples of each. ST noted that each of these phenomena is oriented towards the completed clause. That is, proper (expected, non-intrusive) turn-taking occurs most often at the completion of a (prosodically-closed) clause, and joint-production is either of entire clauses or of the last parts of clauses (the situation is more complex in Japanese). Both of these would require the ability to project the occurrence and trajectory of the clause. The paper also points out that cases of speaker overlap in English are most frequently observed with what I might call “extra-clausal” material (on analogy with extra-metrical segments). That is, tag questions, turn-extensions (like “Have you got Seacliff’s phone number? … by any chance“), and other such material. It is rarely co-incident with the last lexical item of the previous turn (though it may slightly overlap with the last part of the last word). What ST/ECK claim is that this is evidence that at the level of interaction, the clause is the most important linguistic unit: the locus, if you will.

ST went on to speak intersubjectivity. These are those parts of the clause that indicate speaker stance, evidentiality, epistemicity, and evaluation. I have to admit that I didn’t exactly see the connection of this part of the talk to the previous part, so I didn’t take many notes as I listened. I will repeat one point made regarding the concept of subordination and complementation. ST brought up examples of I (don’t) think as examples of evidentiality markers, and claimed that analyzing sentences with these strings as involving evidential markers and focused clauses (that is, [[I think] [he's coming]]) may capture more linguistic and/or interactional generalizations than the traditional analysis (i.e., [[I] [think he's coming]]). To me, it seemed more like a point for Interactional Linguistics, and not so much a clause-related point. Maybe I missed something.

Finally, ST brought up a series of around a dozen of adjacency pairs, or two-speaker give-and-take interactions. Examples include request-acceptance, request-refusal, question-answer, invitation-rejection, assertion-confirmation, and so on. Each of these use, not surprisingly, clauses. The take-home point from this part, which unfortunately ST did not make (clearly), is that given such interactions must take place regardless the type of interactional tools available, when language is chosen, inevitably the clause is also chosen. This is the sense in which it is the locus of interaction.

(Of course often such interactions do not involve clauses, but simply single-word or single-phrase utterances. These, ST said, does not weaken the claim that clauses are of primary importance, because the interpretation of such utterances is completely dependent on nearby clauses. She also claimed that the notion of “deletion” of lexical material in, e.g., VP ellipsis, comes from the sense that the clause, with all of its parts, is primary. There is no need, she said, to propose that something is actually deleted: we must merely realize that such “incomplete” utterances are only used when there is an available clause with respect to which the fragment can be interpreted. She did not provide a full story of “missing constituents,” probably because she doesn’t (yet) have one, and the paper only mentions these in a footnote.

If you’ve followed so far, you’re probably thinking: uh…duh. Of course we use clauses. They are what the grammar produces (that is, complete sentences). The grammar, independent of conversational principles, provides us with certain fixed linguistic structures, and so we use these, and not others, in our interaction. As to this, ST was basically silent. She addressed the point at the beginning, making reference to the notion of the “emancipation” of generalizations. In other words, the grammatical forms we have here today are the frozen formulas that language users long long ago created, motivated by communicative needs. In other words, today’s syntax is yesterday’s pragmatics. Of course, here is where the concepts of emergent grammar and theories of language evolution probably come into play, and where my commentary becomes ignorant and probably not worth reading. But that’s where I think she would go.

Finally, the question period opened with Chris Potts bringing in everyone’s favorite distinction after descriptivism/prescriptivism. That’s right: description versus explanation. He agreed that interaction studies provides a crucial part of linguistic analysis, but pointed out that you can’t just do without traditional grammar. For instance, even if you’ve got [I think] [he's coming], you still need to deal with questions of formal properties, like (here I’m filling in what he left unsaid) subordination, island phenomena, c-command, agreement, and so on. ST agreed with this stance. I felt compelled to point out the potential for speakers to sabotage standard grammar and reanalyze strings like “I think” as separate units, pointing out subject-auxiliary non-agreement in a sentence like I don’t think he knows, does he? (*do I?). Another scholar (don’t know her name) pointed out that some people with language-related deficiencies (due to neural damage) can, for instance, use “I think such and such,” but otherwise cannot use subordination, or even non-1sg forms with think.

So the goal of the talk was, in my mind, to introduce a new way of doing linguistic analysis that makes heavy use of what conversation analysis and other social-type linguists have been doing (aka the “fun part” of linguistics, or perhaps the “ooey, gooey, creamy center of language study” [or is that the "periphery"?]). For someone expecting this, the talk was manageable, since the extra bits could be filled in. For people expecting something else, or something “more,” the talk was probably a disappointment or just totally incomprehensible. Hopefully I made clearer what went on that Thursday evening.

And that’s all she wrote. I recommend the paper, if you can get it, and there’s also a bibliography of Interactional Linguistics literature (as well as literature directly cited in the talk) available from Sandra Thomson’s web page.