October 18, 2007

Go go go!

A couple days ago Mark Liberman followed up on this question from Thomas Mills Hinkle:

Just now, my wife asked the following "Would you mind go checking on the laundry?" The (to me) error "go checking" made me think: what is the deal with go X constructions in English?

I have a speculation about this example, which relates it to the "go V" construction but also treats the verb mind as crucial in the story.

But first a pointer to an essential piece of literature on the subject, our own Geoff Pullum's 1990 article "Constraints on intransitive serial-verb constructions in modern colloquial English" (Ohio State University WPL 39.218-39).

Geoff's article does four things.

One thing it does is distinguish the construction -- which I'll refer to here as QSV (for "quasi-serial verb": I'll go see who's at the door) -- from a number of other constructions that can have the verb GO in them and share some syntactic or semantic properties with QSV, in particular the following (the names are mine, not Geoff's):

Hendiadys, with and: I'll go and see who's at the door;

GoPurp, a purposive construction with a marked infinitive: I'll go to see who's at the door;

AdvIng, "adverbial -ing":  I'll go fishing with you tomorrow.

(There are more constructions with GO in them, but these are the most similar to QSV.)

Another thing Geoff's article does is survey the peculiarities of QSV.  Most speakers allow some motion verbs other than GO as the first verb in QSV: COME and maybe RUN or HURRY.  And for most speakers, QSV is subject to the two conditions in the passage below (from a 2003 abstract of mine that adds some observations to Geoff's):

... (1) the Inflection Condition, which requires that both verbs be in a form identical to the base form (either the base form itself, or the present non-3rd-sg form, as in I go get some wine whenever I can, vs. *She goes get(s) some wine...); (2) the Intervention Condition, which disallows a dependent of  the first verb between the two verbs (*Go out get some wine); (3) a strong preference for face-to-face conversation (so that searching the standard corpora nets very few examples); and (4) a newly discovered statistical asymmetry (observed by searching a database of film scripts), involving a very strong preference (when QSV examples are compared to all relevant occurrences of go and come) for base forms over unmarked presents and, within the base forms, for the imperative over all other uses (with modals, with infinitival to, and in all other contexts).

(Hendiadys, GoPurp, and AdvIng are subject to neither the Inflection Condition nor the Intervention Condition.)

The Hinkle example violates the Inflection Condition: go looks like a base form, but checking certainly is not one; it's a "gerund participle" form ("form N", as I'll call it here), with suffix -ing.  The Hinkle example also violates a condition on complements of the verb mind, which have to be nominal (as in Would you mind a helpful suggestion?); VP complements are possible, but they are in fact nominal gerunds, in form N: Would you mind my suggesting an alternative?  Would you mind seeing who's at the door?  Fixing the complement of mind in the Hinkle example (where it's in the base form) only makes things worse: *Would you mind going checking on the laundry? violates the Inflection Condition twice.  (*Would you mind going check on the laundry? still violates it once, and *Would you mind go check on the laundry? satisfies the Inflection Condition but violates the condition on complements of mind.)

So the Hinkle example looks like a blend -- of QSV (Would you go check on the laundry?) and mind with a form-N complement (Would you mind checking on the laundry?), yielding a complement of the form: go + N-form VP (with semantics distinct from the AdvIng construction).  It could have occurred as an inadvertent speech error.  But, as Mark pointed out in his posting, it looks like some people have combined these features into an actual construction, not (so far as I know) previously reported.  (This sort of thing can happen.  The GoToGo construction -- I'm going home and take a nap -- that I've posted about here several times presumably originated in a telescoping or blending, but it's now just part of some speakers' linguistic systems.)  The question is what the details of the construction are like.  I'll save that discussion for an appendix to this posting, and return now to Geoff's article.

A third thing that Geoff does in this article is go through the previous history of linguists' treatments of QSV and the conditions on it, going back to the 1960s.  The sad lesson here is that the phenomenon was discovered again and again, and people wrote about it almost entirely without citing any of the earlier literature.

Finally, Geoff reports on some judgments he and I collected (way back when) about QSV, judgments indicating that there is a core pattern, with some variation.  But the full envelope of variation appears to include almost every logically possible system.  Mark has now searched for some of the minority variants and found several of them attested.

This general configuration is of considerable interest: people base their systems on what they hear, and in the case of QSV, they are mostly -- but not entirely -- reluctant to go past the evidence they have, in which the Inflection Condition holds.  What  they do here is UNLIKE what people do in many other situations: ok, you haven't heard a passive with a particular verb, but you press on and produce it, and no one bats an eye.  You go beyond the evidence you have.

But for some constructions, people are conservative.  They (mostly) don't generalize to new cases.  You hear I go help them whenever I can, but you (well, most of you) don't go on to say She goes help(s) them whenever she can.  This is an instance of Baker's Paradox, which I discussed here about a year ago in connection with a constraint (for many speakers) on independent possessives that allows Let's meet at Sandy's but not Let's meet at mine (with mine not anaphoric).

Characteristically, in cases where most speakers resist generalization, having apparently learned an arbitrary constraint on some construction (like the Inflection Condition), there are some speakers who strike out anyway.  If you look hard enough, you can find a few people who don't adhere to the constraint; you find things like i called Melanie and went meet her in Moreauville (in Mark's posting), and more.  Sometimes the constraint applies in some dialects but not others; things like (non-anaphoric) Let's meet at mine are widely attested in British English.

A few words about the history of QSV.  Most people reflecting on the matter assume that QSV (go V) is just the "short form" of the "full form" Hendiadys (go and V); as Mark reported, this was Hinkley's own idea.  In general, when faced with a longer and a shorter variant with similar meaning, many people, linguists included, assume that the shorter variant is derived, historically and maybe synchronically as well, by truncation of the longer form.  This is often not true historically, and is very rarely, if ever, a satisfactory synchronic analysis.  In any case, Mark casts doubt on Hendiadys as the historical source of QSV (both have been in the language for a very long time).  The fact that Hendiadys is subject to neither the Inflection Constraint nor the Intervention Constraint casts further doubt on it as the source of QSV.

In my alternative story (in my QSV abstract),

... (sentence-initial) hortatory go and come with imperatives (Go, get some wine! Come, see how it's grown!) were reanalyzed as forming prosodic, syntactic, and semantic units with them; the resulting construction was then extended from the imperative to other uses of the base form, and then to homophonous finite forms (thus yielding the Inflection Condition). 

This is just a suggestion, but it does offer a possible account of the statistical asymmetries in modern QSV, which reflect the historical trajectory of the construction.

Appendix.  We started with examples of mind with a complement composed of go and an N-form VP, as in:

mind + [ go + [ checking on the laundry ] ]

You can also find examples with try:

If you guys like what you read, try go getting his books

Try go telling a fanatic Muslim something about Jews.

You should try go asking some Koreans.

and probably with some other head verbs as well (I've just started looking at such data; this is a very preliminary report).  What's striking about the examples Mark and I have collected so far is that the head verbs are all in their base form: in imperatives (try go getting his books) or complements of modals (would you mind go getting your sister, should try go asking some Koreans).  Could this be a version of the Inflection Condition, applying here to the head verb?  If so we could hope to find some present non-3rd-sg examples, that is, with the verbs not in the base form but in a finite form identical to the base form -- something like

Whenever I need to know more about hangul, I try go asking some Koreans.

To complicate things still further, there are cases of go plus N-form VP "on their own" (not as a complement to a main verb like mind or try), cases that are not examples of AdvIng:

Don't go getting my hopes up.  [cf. Elton John's 1976 hit "Don't Go Breaking My Heart"]

How does one go getting it back?

Only a few examples so far, but they have base-form go.  Again, we could hope for some present non-3rd-sg examples, pointing to the Inflection Condition.  (It's possible that the two examples above illustrate two different constructions.)

To sum up:  there are apparently two constructions here, call them (arbitrarily) A and B. 

Construction A involves two constituents: the verb GO and an N-form VP.  On the evidence so far, the verb is constrained to its base form (or possibly it's subject to the Inflection Condition).

Construction B involves two constituents: a head V and a base-form complement in Construction A.  On the evidence so far, the verb is constrained to be in its base form (or possibly it's subject to the Inflection Condition.)  Also, on the evidence so far, the verbs that can head Construction B are verbs (like mind and try) that can occur with complements that are plain N-form VPs (I don't mind checking on the laundry, I tried asking some Koreans).  But which such verbs?

Lots of work still to be done here.

One final note: in searching for constructions with GO plus N-form VP, I turned up yet another one I hadn't noticed before, illustrated by

There you go getting obtuse again.

There you go, asking more questions.

This one obviously isn't restricted to the base form.  In fact, it isn't subject to the Inflection Constraint:

There he goes getting obtuse again.

It seems not to be relevant to the other constructions we've been looking at.

Posted by Arnold Zwicky at October 18, 2007 03:08 PM