The topic of the go get (or go + V) construction in English (Let’s go see what’s happening) has now come up (in comments) in connection with V + V compounds (sleep walk, stir-fry). I’ve spent years probing this construction, mostly in collaboration with Geoff Pullum, so I thought this would be a good time to reproduce a summary I wrote in 2000 for the Usenet group sci.lang (with editorial amendments and additions over the years).
Note: This is going to be long and pretty technical. And you should first read (at least) my 2007 Language Log posting “Go, go, go!”, which will put things in context and set aside a collection of issues off the main topic.
The Quasi-Serial Verb (QSV) construction in English, illustrated by
(1) Go see who’s at the door!
(2) I’ll come look at your homework in a minute.
(3) I want you to go check the gauges.
involves a head verb (V1) — GO or COME or, for some speakers, RUN or HURRY, and possibly others — in combination with an immediately following VP, the head verb of which (SEE, LOOK, CHECK in (1)-(3), respectively) I’ll refer to as V2. QSV is subject to quite a lot of variation from person to person; my discussion of this variation here is based on Pullum (1990: “Constraints on intransitive quasi-serial verb constructions in modern colloquial English”. OSU WPL 39.218-39.).
First, there are speakers who do not have QSV; such speakers reject (1)-(3) and all similar examples as ungrammatical. (Some older British speakers give these judgments, characterizing examples like (1)-(3) as barbaric Americanisms.)
SYSTEM 1: no QSV
Pullum gives citations, from Bartlett’s and the OED, from the 14th century on, to which many further examples could be added from the great age of hymnody:
“Come sound his praise abroad” (Isaac Watts, 1704)
“Come let us join our friends above” (Charles Wesley, 1759)
“Come tell of your ship and what is her name” (anonymous American, 19th century)
But Pullum notes that lexicographers and grammarians alike characterize the construction as “archaic” and/or “dialectal”. Presumably, the construction vanished from colloquial British English by the end of the 19th century and waned in American English, but then experienced an American revival during the 20th century; several people have suggested to me that the rise of the modern construction might have had its origin in a calque on a Yiddish construction (though I can’t see any evidence at all in favor of the idea, beyond the inclination of some commentators to treat what they see as Americanisms as due to Yiddish influence)
Perhaps modern QSV did arise in the U.S.; if so, it has now spread rapidly throughout the English-speaking world.
My hypothesis about its origin (a hypothesis still to be tested by the study of texts) is that the source is coordinate imperatives like
(4) Go, see who’s at the door!
(5) Come, look at your homework!
(The hymn texts sometimes have a comma, sometimes not: “Come, let us raise our voices high” (anonymous American, 19th century), “Come, let our hearts and voices join” (Alabaman O.H. Handley, 1959), “Come, let us join with one accord” (Georgian H.M. McGraw, 1935; sometimes printed without the comma).)
All that’s needed to get from sentences like (4) and (5) to an example like (1) is for the two imperative verbs to be prosodically united, as they surely are in the hymn texts. Then (1) can be reanalyzed (by listeners) as a verb-complement, V+VP, construction — a small construction or an idiom with an open slot. At this earliest stage, the idiom would be restricted to occurrence in the imperative. (It’s not uncommon for an idiom to be restricted to occurring within a certain other construction.)
SYSTEM 2: QSV limited to imperative sentences
A speaker with this system would accept (1) but reject (2) and (3). I have not (yet) encountered a system 2 speaker. But if my hypothesis about origins is correct, then there should be evidence that there were such speakers; a careful text search should yield examples of QSV in imperatives only, before it appears in other contexts.
Almost all of the early examples of QSV that I’ve already mentioned — from Bartlett’s, the OED, and the hymn texts — are in imperatives.
The next stage depends on the fact that imperative sentences in English use the BSE (base) form of the verb. A natural reanalysis would take the crucial fact about examples like (1) to be the appearance of a BSE form, rather than the specifically imperative use of BSE The result would be an extension of QSV to other uses of BSE, the two most frequent of which are (a) in the complement of a modal verb, as in (2), and (b) as the complement of the infinitive marker TO, as in (3).
SYSTEM 3: QSV limited to V1 in constructions requiring BSE
(It’s not uncommon for lexical items or idioms to be restricted to occurring in certain inflectional forms. The English modals, and the obligative BE of You are to leave this instant, are limited to occurrence in finite forms.)
There are speakers with this system. They accept (1)-(3) (and also examples with less frequent uses of BSE than these three), but reject examples like
( 6) I go/come water the plants whenever I get a chance.
System 3 is a minority option in present-day English, however. Most speakers accept examples like (6), with present tense go and come, as well as (1)-(3). How could this come about?
Here I need to uncover a subtlety in what it means to say that a word W in a sentence is in some specified form F. The first half of the answer is that the syntax assigns the feature F to W. But then we need to check that the morphology associates F with the phonological content of W. Being featurally an instance of F entails being formally an instance of F.
The subtlety is that the converse doesn’t hold: being formally an instance of F doesn’t entail being featurally an instance of F. For most verbs in English, all (featurally) present tense (PRS) forms except the 3sg (e.g. /go/ for the verb GO) are formally instances of BSE. Formal identity can be detached from featural identity.
In system 3, it’s featural identity that’s relevant. Now suppose we split the condition of system 3 into two conditions, detaching formal identity in the process:
SYSTEM 4: V1 in an instance of QSV
(a) must be featurally (and therefore also formally) an instance of the inflectional category required by its syntactic context; and
(b) must be formally an instance of BSE
Condition (a) is not, of course, a clause specific to QSV. It’s merely part of what it means for a sentence to be syntactically well-formed. Condition (b) is what Pullum calls the Inflection Condition, hereafter the IC.)
(As Pullum notes, the IC is shared by the try and V construction in English, but not by the go and V construction.)
In examples (1)-(3), the syntactic context requires BSE, so that the IC is automatically satisfied if condition (a) is. But in example (6), the fact that condition (a) is satisfied — go/come are featurally instances of the 1sg PRS form required for subject-verb agreement — doesn’t entail that the IC is also satisfied. Now, in (6) the IC is indeed (separately) satisfied; (6) is predicted to be ungrammatical in system 3 but grammatical in system 4. On the other hand, examples that satisfy (a) but not the IC are ungrammatical in system 4 (as well as in system 3):
(7) *She goes/comes water(s) the plants whenever she has a chance.
(8) *I went/came water(ed) the plants whenever I had a chance.
Examples like (7) and (8) point to an indeterminacy in the analysis of (1)-(3) and (6), namely an indeterminacy in the conditions on V2. V2 in (1)-(3) and (6), like V1 there, looks like the BSE form of its verb, but the conditions on V2 don’t come from the surrounding syntactic context; they are matters of QSV itself. And there are three (slightly) different grammars that would, equally well, predict the V2s that we see in (1)-(3) and (6):
grammar (A): V2 shares the features of V1
grammar (B): V1 governs the BSE form on V2
grammar (C): V2 shares the features of V1, and – the IC — V2 is formally an instance of BSE
Nothing we have seen so far would provide a way of deciding among these analyses. Different speakers could have different grammars, and we’d have no way of telling who has which one. (This is not a problem for the analytic framework, just a limitation on what linguists can know about idiolects.)
There are speakers for whom the analyses could be distinguished. Suppose that a speaker lacked system 4’s IC on V1; such a speaker would be extending the construction from its original BSE or BSE-like V1s to the full set of V1 forms. Then we could see evidence of either grammar (A) or grammar (B) (I assume there would be no reason to posit an IC on V2 if you don’t impose it on V1, so that grammar (C) is irrelevant here.) This gives us two further systems.
SYSTEM 5: V1 unconstrained, V2 subject to (A)
In system 5, the following examples would be grammatical:
(7′) She goes/comes waters the plants whenever she has a chance.
(8′) She went/came watered the plants whenever she had a chance.
This system is amply attested, though it’s a minority system. One example:
Jimmy Goes Asks the Barbershop About Obama Jokes (link)
(This is the connection to the examples of double inflection in my “Sleepwalking” posting.)
SYSTEM 6: V1 unconstrained, V2 subject to (B)
In system 6, the following examples are grammatical:
(7″) She goes/comes water the plants whenever she has a chance.
(8″) She went/came water the plants whenever she had a chance.
Visser (An Historical Syntax of the English Language) reports that examples like They went look for him are to be found occasionally in American English. As here:
The houses were just about done. There were only 4 of them and each one of them sat on a two acre lot. They went look at it and, as I heard the story, decided to buy the one with the best view the same day. (link)
Returning now to majority idiolects, which have the IC on V1, there is a way, even there, to distinguish between grammars (A)-(C), at least for many speakers. The crucial evidence arises from the fact (which is, from the synchronic point of view, an accident) that the verb COME has a PSP come that is formally identical to its BSE. As a result, we can distinguish three variants of system 4 (all of them attested, though system 4C is by far the most common), according to which version of QSV they have in the perfect:
SYSTEM 4A: V1 constrained, V2 subject to (A)
(9A) I have come watered the plants whenever I could. [cf. system 5]
SYSTEM 4B: V1 constrained, V2 subject to (B)
(9B) I have come water the plants whenever I could. [cf. system 6]
[In addition, there are some speakers who have both system 4A and 4B -- who have alternative versions of the “same” construction, and accept both (9A) and (9B). Call this system 4A/B. (It is not unknown for a speaker to have distinct alternative versions of the “same” construction. So, for example, though Northern Midlands speakers in the U.S. usually have either My car needs fixing (the majority version) or My car needs fixed (the regionally restricted version), there are speakers who use them both and judge them both to be grammatical.)]
SYSTEM 4C: V1 constrained, V2 subject to (C)
In this system, perfect forms are possible only when V2 has a PSP that’s formally identical to its BSE, as a modest number of verbs do:
(9C) I have come put water on the plants whenever I could.
(10) *I have come water(ed) the plants whenever I could.
For what it’s worth, Pullum and I both have system 4C, which appears to be the majority idiolect .
A final complexity is that there are a fair number of speakers with yet another version of system 4, a version in which examples like (6) are grammatical, but examples like (9C) are not. For these speakers PRS come counts as formally identical to BSE come, but PSP come does not
What is at issue here, in brief, is the difference between formal identities that are systematic, that follow as consequences of generalizations about the morphology of the language, and those that are accidental. PRS come is systematically identical to the BSE (the identity holds for all verbs except BE), PSP come only accidentally so. There are then two versions of the IC: the one given above, merely requiring formal identity, and an IC’, more stringently requiring systematic identity.
SYSTEM 4C’: as above, but with IC’
This gives us eleven potentially different idiolects with respect to the QSV construction (1, 2, 3, 4, 5, 6, 4A, 4B, 4A/B, 4C, 4C’). System 1 lacks QSV. System 3 is an extension of system 2 from the imperative to base forms in general. All the systems 4x extend system 3 to base-like V1s; systems 5 and 6 extend system 3 further, to all possible inflectional categories for V1. Systems 4C and 4C’ extend system 4 to BSE-like V2s; systems 4A, 4B, and 4A/B extend system 4 further, to non-BSE-like V2s
Most of these eleven systems are attested, in the sense that the crucial forms have been elicited (details in Pullum 1990), Some of the systems are distinguishable only on the basis of crucial forms with a very very low probability of occurrence, like the perfects with come in (9A)-(9C).
The fact is that three types of contexts — the imperative, as in (1); the complement of a modal, as in (2); and the complement of infinitival TO, as in (3) — occur with high frequency, and one further context, the habitual simple present, as in (6), occurs with moderate frequency. For current speakers, these four contexts provide the raw data from which grammars must be induced, and crucial input is unlikely to be available for deciding among the various possibilities. Indeed, if you don’t hear very many examples like (6), or don’t appreciate the import of the examples you do hear, you might well stick, conservatively, to system 3. With more inputs like (6), and/or bolder extrapolations from inputs, you’ll end up with system 4 or higher.
Here ends the main part of the exposition. There are many further topics to be explored, including some evidence that QSVs are indeed compound-like (see Stephanie Shih, “Prosodic evidence for the lexical status of quasi-serial verbs”, 2009, here), despite their having many characteristics of verb + complement syntactic constructions (see Pullum & Zwicky, “Gerund participles and head-complement inflection conditions”, 1999, here).