« previous post | next post »

Geoff Pullum's wonderful example of the perils of trying to avoid split infinitives: it can lead to a straightforwardedly ungrammatical result. Simplifying the original to make the point clearer, the entirely unproblematic

(1) Rockefeller has pledged $100 million to dramatically increase learning opportunities for Harvard undergraduates

was incorrected to the entirely problematic

(2) Rockefeller has pledged $100 million to increase dramatically learning opportunities for Harvard undergraduates.

I intend to post later on other aspects of the phenomenon, but here I want to say a few words about the "nonintervention constraint" violated in (2): in English, nothing can intervene between a verb and its direct object (unless the direct object is long, complex, or "heavy"). I want to say these few words because I think the constraint — which I'll refer to as *V+X+DO (yes, extraordinarily clunky, but I have at the moment no better alternative) — is a beautiful piece of English syntax.

Nothing I'm saying here is novel with me, by the way.

First background comment: *V+X+DO is a constraint on English specifically. In many languages, including a number of familiar European languages, not only are adverbials intervening between V and DO not prohibited, this is a preferred order for many types of adverbials, a fact that shows up in characteristic foreignisms in English, like "I saw yesterday Norman" and "I shook violently the box" from native speakers of German. 

Second background comment: English has a number of other nonintervention constraints, ranging from the very general (like this one) to the very construction-specific (as in quasi-serial verbs). There is almost surely no larger generalization covering them all (while allowing intervention for to + VP and for the oblique-object cases I'll take up in a moment).

Third background comment: the statement of the constraint above needs to be refined in several ways. In particular, it's a strong constraint when the intervening element X is an "integrated" adverbial, as in (2), but when the intervening material is a parenthetical, or "supplementary", modifier, as in (3), intervention is much easier:

(3) Rockefeller has pledged $100 million to increase — dramatically — learning opportunities for Harvard undergraduates.

Fourth background comment: to do the job the constraint is supposed to do, "direct object" in it must be understood as denoting a phrase and not a single word; otherwise, it would ban things like

(4) We confronted enormously large lizards. (cf.: *We confronted bravely large lizards.)

in which two words intervene between "confronted" and "lizards". These words are irrelevant, of course, because they're PART OF the direct object. This is another case where a proper understanding of how grammar works depends on seeing things in terms of constituent phrases rather than (as in many presentations of "traditional" or "school" grammar) single words — a point I make here every so often, most recently here.

Fifth, and final, background comment: like other constraints on specific syntactic constructions, *V+X+DO presents a puzzle in language acquisition: how do people acquire "negative knowledge", knowledge that certain instances of a configuration are NOT grammatical? (This is the general puzzle known as Baker's Paradox, discussed on Language Log here.)

Now to the main point, which is that *V+X+DO distinguishes elegantly between direct objects (the V is transitive) and oblique objects (OOs), marked by prepositions (so that the V is intransitive); only the former are ruled out by the constraint. English has verb pairs with similar meanings but syntax differing on exactly this point, for example:

(5a) DO: *We accepted yesterday the terms of the agreement. (cf.: We accepted the terms of the agreement yesterday.)

(5b) OO: We agreed yesterday to the terms of the agreement.

In fact, what is arguably the same verb can occur with both sorts of objects (and perhaps a subtle difference in the meanings of the constructions).  Again the contrast is striking:

(6a) DO: *I played for hours the piano. (cf.: I played the piano for hours.)

(6b) OO: I played for hours on the piano.

(7a) DO: *You must exit immediately the building. (cf.: You must exit the building immediately.)

(7b) OO: You must exit immediately from the building.

(8a) DO: *They fought valiantly the enemy. (cf.: They fought the enemy valiantly.)

(8b) OO: They fought valiantly against the enemy.

Such examples, and many others that could be listed, show that the constraint can't be stated in purely semantic terms. That's because whether a verb takes a DO or an OO or both (and if an OO, which preposition it selects) is to a large part a matter of convention, not deducible from the meaning of the verb; the details differ from dialect to dialect and style to style and change over time. (There are some subtleties here I'll look at in another posting.) They also give second-language learners a lot of trouble.

Final note: intransitive Vs that select Ps in their (oblique) objects are often treated in handbooks for second-language learners as lexical units with those prepositions, and the V+P pairs are often listed in specialized dictionaries along with verb-particle combinations (like give up 'abandon', as in "We gave up the fight"), as "complex verbs", "phrasal verbs", or "idiomatic verbal phrases". Putting the two phenomena together makes some practical sense: both require a good bit of memorization of V+X pairs. But they differ syntactically on a number of dimensions, and good reference grammars of English (CGEL, in particular) show how.




  1. john riemann soong said,

    May 2, 2008 @ 2:49 pm

    How and when in the history of English did this constraint develop?

    Is there a correlation between this constraint and whether a language is right-branching or left-branching?

  2. ST said,

    May 2, 2008 @ 3:43 pm

    What about indirect objects without prepositions? For example, "He gave her the bowl." Why doesn't that violate the constraint? "The bowl" is the direct object and "her" intervenes between it and the verb.

  3. Josh Millard said,

    May 2, 2008 @ 4:03 pm

    I have some rusty armchair recollection of the notion of "gave" there as a 'bitransitive' verb, effectively taking two objects; "her" is not intervening the object of the verb "gave', it's necessarily occupying the role as the first object, with "the bowl" the second object noun-phrase.

    This could well be outmoded or wrongheaded claptrap. But see also 'tell' ("he told her a story"), or sell ("he's selling her some cabbages"), etc. I need to read some more and newer books, regardless.

    Further, reworking that phrase to use a preposition:

    he gave the bowl to her
    *he gave to her the bowl"

  4. Karen said,

    May 2, 2008 @ 4:05 pm

    The bare (no preposition) indirect object's position is a remnant of case-marked English and is constrained to follow the verb. It's part of the complement, unlike adverbials, which the X here seems to represent.

  5. ST said,

    May 2, 2008 @ 4:09 pm

    For this rule, the direct object becomes "long and heavy" when it is followed by a colon and another thing or a list of things. For example, I've seen a lot of agreements written this way: "Defendant agrees to pay to the Plaintiff the following sum: one thousand dollars." I don't know if that's ungrammatical legalese, but I've gotten used to it and it seems natural to me.

  6. ST said,

    May 2, 2008 @ 7:30 pm

    Karen, sounds like a good explanation to me.

  7. Daniel Barkalow said,

    May 3, 2008 @ 1:06 am

    Some linguists have analyzed "he gave the ball to her"/"he gave her the ball" like "he sprayed paint on the wall"/"he sprayed the wall with paint". There are a number of similarities between the pairs of constructions, so this isn't too far-fetched. This would suggest that, in "he gave her the ball", the direct object is "her" (and not separated from the verb), and "the ball" is something strange (like a prepositional phrase with nothing as the preposition). In any case, the similarity with the variant with the other word order shouldn't be taken as particularly strong evidence of the structural relationship between the verb and the item given, since we have an example where a pair of variants give relationships with the noun phrases that are clearly different.

  8. Arnold Zwicky said,

    May 3, 2008 @ 11:16 am

    Amplifying a bit on Daniel Barkalow's comment: see my discussions of "diathesis alternations" in Language Log Classic, postings #5272 and 5400. The treatment of X in GIVE X Y as syntactically a direct object (with Y as some kind of oblique object, often called "second object") is widespread, approaching standard.

    Note: *syntactically* a direct object. But denoting the recipient in the act of giving — a participant role usually associated with a syntactic indirect object (as in GIVE Y TO X). So long as we're careful to distinguish syntactic functions from semantic roles, there's no problem here: there are default associations between the two, but in particular constructions, other associations are stipulated.

  9. bernarda said,

    May 3, 2008 @ 11:53 am

    In a previous thread you asked about grammar books. Well, one of my favorites is "Look it up" by Rudolph Flesch. I have an old copy that I bought at a second-hand shop and I don't know if it is in print. Now you just need to boldly go to look for it.

  10. Arnold Zwicky said,

    May 3, 2008 @ 3:02 pm

    To bernarda: I assume this comment is addressed to me. I don't know which thread you're talking about, or which posting of mine you're referring to. And I don't recall asking *about* grammar books. I do post a lot about grammar books, usage manuals, and the like; I have a gigantic collection of them, which I use in an extensive project on the advice literature about grammar, usage, and style.

    Flesch wrote a ton of books. I have The ABC of Style in my collection, but not Look It Up (though it's available used).

  11. David Marjanović said,

    May 3, 2008 @ 8:16 pm

    like “I saw yesterday Norman” and “I shook violently the box” from native speakers of German.

    German? Are you sure? As far as I remember, I never made this mistake, nor did anyone around me. In German, the adverb would go at the end of the sentence: ich sah N. gestern, ich schüttelte die Schachtel heftig — in the last example (but IMHO not the first), you could put the adverb directly behind the verb, but only in poetry, and even then it would be interpreted either as extra emphasis on the object or as an artificial rearrangement for the sake of rhyme and/or meter.

  12. john riemann soong said,

    May 3, 2008 @ 10:14 pm

    'Some linguists have analyzed “he gave the ball to her”/”he gave her the ball” like “he sprayed paint on the wall”/”he sprayed the wall with paint”.'

    DO and IO are syntactical concepts, not semantic concepts, right?

    I'm just remembering those assigned exercises in my freshman year of high school, and I had protested to my teacher how he had marked some of my analyses because I believed that semantics took a back seat to syntax, at least in grammar. Of course this gives us the concord / traditional agreement problem …

  13. john riemann soong said,

    May 3, 2008 @ 10:18 pm

    *semantic concord, pardon

  14. Bengt Skillen said,

    June 21, 2008 @ 10:35 pm

    In two disparate languages, French and Finnish, it is possible to place an adverbial immediately after the verb:

    English gloss: 'I sometimes eat candy.'

    French: 'Je mange quelquefois les bonbons.'
    Finnish: 'Syön joskus karkkia.'

    Literally in both languages: 'I eat sometimes candy.'

    I don't think positing an adverbial after the verb has anything to do with left- or right-branching rationale as French is definitely not left-branching while Finnish can often be, especially in academic or formal contexts. There is something else going on also as Finnish is a heavily inflected languge, both in terms of verbs and nouns, while French is much less inflected. Is this a sprachbund issue? I don't know. French and Finnish aren't spoken in adjacent areas, so it seems improbable.

RSS feed for comments on this post