« previous post | next post »

From the annals of incorrection: cases where, because of some structural similarity between constructions C1 and C2, some people see C1 (incorrectly) as an instance of C2, where C2 is believed (incorrectly) to be non-standard (or defective in some other way), so that these people avoid C2 and replace it by something else. The proscription against C2 has then CONTAMINATED the innocent C1. On to cases.

Act 1. Stranded to: C1 is "stranded to" (Verb Phrase Ellipsis licensing the omission of a VP complement of infinitival to, as in I don't want to) and C2 is "stranded P" (where fronting of the NP object of a preposition leaves that preposition without an immediately following NP, as in What do you attribute that to? and I don't know what to attribute that to). If Dryden's Rule — No Stranded Ps! — is in force for you (against the practice of English writers for centuries), and if you think (inaccurately) that infinitival to is a P, then you'll hate stranded to and will insist on using an explicit VP (e.g., I don't want to do it) instead of VPE. This would be foolish, since in current educated usage, stranded to is so uncontroversially normal that no one even notices it. (The history is more complicated, and I hope to write about it here, but the current state of affairs is very clear.)

This is type 1 incorrection: a correct variant, falsely believed to be incorrect, is replaced by some other correct variant. (In type 2 incorrection, the replacement is less satisfactory than the original, but still acceptable. In type 3 incorrection, the replacement is itself incorrect, and we have a type of hypercorrection.)

Anecdote: when I first began studying stranded to, in the late 70s, I devised a questionnaire asking for judgments on stranded to in different syntactic contexts (a paper reporting on the early part of the project can be viewed here). I administered the questionnaire to several groups, including an entire Ling 1 class at Stanford. As is usual in such studies, the materials included a few sentences that, from pre-testing, I believed to be unquestionably acceptable and some I believed to be unquestionably unacceptable. I then used bizarre responses to exclude some subjects from the analysis in a principled way. (Subjects sometimes give random responses, just to get the "test" over with. Some subjects are simply inattentive. Others don't really understand the instructions. And, alas, still others invert the scale of judgments.)

I got one stunning set of judgments — all as negative as could be — with a scrawled comment that went something like "LOSER! Don't you know you can't end sentences with a preposition!"

I have no idea how many people avoid stranded to because they subscribe to Dryden's proscription against stranded P, but they are encouraged in this folly by the practices of most dictionaries (and other reference works that use dictionaries as sources). The thing is that most English dictionaries have their entries organized on etymological grounds. So they'll have at least four noun pen entries, for pen 'writing implement', pen 'enclosure for animals' (with a number of extended senses), pen 'prison', and pen 'female swan'; there are four different etymologies here, and, indeed most speakers of English think that these are "different words" that happen to be pronounced the same.

But the dictionaries will also put infinitival to into same entry as all those uses of to as a preposition, and the "part of speech" for that entry will probably be prep. — because infinitival to originated as a use of the preposition to in combination with a nominalized verb as object. (This is the treatment in the OED and in AHD4. NOAD2 puts the two into a single entry, but has subentries with different part-of-speech labels, "preposition" and "infinitive marker".)

The problem is that infinitival to is no longer any kind of preposition, and a rational dictionary should distinguish the entries (while noting that one of the items is historically derived from the other).

Meanwhile, kids are being taught (a) that to is a preposition, and (b) (Dryden's Rule) that prepositions must not be stranded. Both clauses are wrong, but in combination they are a disaster: Dryden's Rule contaminates the innocent stranded to.

Act 2. On to split infinitives, and the first sort of contamination of other constructions by No Split Infinitives (NSI). What these first cases have is an adverbial intervening between words WITHIN an infinitival VP, though not actually between to and the head V of its complement. [Digression. The usual account of split infinitives is that an adverbial intervenes between to and "its verb", and in strictly linear terms, this is correct (to + boldly + go). But in fact the adverbial is merely the first constituent of the VP complement of to; to and its VP complement (boldly go in this example) are immediately adjacent.]

An example from Thomas Jefferson:

A morsel of history is a thing so rare as to be always valuable.

This quotation turned up as a random .sig quote for Jess Anderson on soc.motss five years ago. Anderson wrote about it:

My residual inner prescriptivist always wants to change that to "as always to be". 

And poster David Fenton chimed in with:

I long ago decided that the faux-Latin rule against split infinitives is something to be blithely discarded. 

(soc.motssers have been bashing split infinitives for many years, but they've also had their defenders).

But Éamonn McManus and I pointed out that neither Jefferson's sentence nor Fenton's has a split infinitive in it: in linear terms, the adverbial comes between an auxiliary verb (be) and the head of its complement (valuable, discarded), not between to and the head of ITS complement; to and be are contiguous. But the adverbial does happen to be inside an infinitival VP. Dangerous territory, if you're fretful about NSI.

By the way, Anderson's proposed incorrection moved Jefferson's always — from a position that is possible, but stylistically marked these days, all the way to positioning before the to, though to my ear the most felicitous slotting of the adverbial would be — ta dah! — in the "split infinitive" position:

A morsel of history is a thing so rare as to always be valuable.

Occasionally, I've had people object to things like

… to have quickly established …

on similar grounds — in linear terms, an adverbial between an auxiliary (in this case, have) and the head of its complement, but once again within an infinitival VP.

In all these cases, NSI is contaminating a perfectly innocent construction allowing adverbials to occur in "post-Aux" position.

Act 3.  Four years ago, when Geoff Pullum and I discussed some of the cases from Act 2 in e-mail, I wondered if there were people who were edgy about adverbials in post-Aux position, even in VPs that weren't infinitival. I reported on students who were uneasy about things like

It began gently raining.


We made them rapidly run through the possibilities.

These have adverbials intervening between a V and the head of its complement, but this time the V (began, made) is not an auxiliary. Still, this positioning bears some similarity to split infinitives, so that NSI might contaminate these constructions too.

(Note: in some cases, you might prefer to have the adverbial following the verb it's associated with rather than preceding it. That's a matter of taste. The point is that both positions are available, and that's a good thing, because they are clearly different in their prosody and subtly different in their discourse functions.)

But could NSI contaminate everyday post-Aux placement of adverbials? In things like:

I have rapidly run through the alternatives.

Well, yes, even there. Ben Zimmer wrote me a few weeks ago to relay a message from a friend in a doctoral program in field F at university U, saying that:

… one of my committee members changed my sentence beginning "It can easily be shown that…" to "It easily can be shown that…"

I suspect that this is an extension of the "don't split an infinitive" non-rule, where you are not allowed to separate verb parts at all. argh!

google is on my side, 2.3 million to 17,000.

This is an incorrection of type 2, a disimprovement but not a disaster. With auxiliary verbs, the post-Aux position is the unmarked choice for many adverbials, but positioning before the Aux (in linear terms, between subject and predicate) is also possible, though somewhat marked, when the adverbial belongs, semantically, within the complement VP. (When it's functioning as an adverbial in the larger clause, post-Subj position is entirely normal. Yes, the facts are very complex indeed.)

For non-auxiliary constructions, things are a bit different: to my ear, gently began to rain is even more marked than easily can be shown (the gently example sounds to me like a poetic displacement); and similarly for rapidly make them run through the possibilities (where it's the running, not the making, that's supposed to be rapid). For auxiliary verbs, adverbs that belong, semantically, within the complement can be fronted before the auxiliary, but this is much harder for non-auxiliary (though nevertheless VP-complement-taking) verbs like begin and make.

The larger point is that even very ordinary constructions (like the one in can easily be shown) can be contaminated by NSI.

Act 4. In our old discussion of these things, Geoff Pullum issued a wry challenge:

… see if you can get a prohibitionist about split infinitives to understand that "I attributed it to really being interested in the subject matter" is NOT one.

This has the PREPOSITION to with a present-participial VP (a "nominal gerund") serving as its OBJECT — a VP really being interested in the subject matter that just happens to begin with an adverbial, the intensifier really.

[It's hell to try to talk about nominal gerunds without getting all technical. The problem is that in their external syntax, nominal gerunds act (almost entirely) like NPs, while in their internal syntax, they look (almost entirely) like (present-participial) VPs. They are classic cases of "mixed category" expressions (there are many others, in many languages), and there's a huge literature about them. Terminology is not the point here, but if you're going to talk about the phenomena, you need SOME labels.]

But there's to followed by an adverbial and then a non-finite verb form (being), so it's superficially similar to the split-infinitive configuration, and things are ripe for contamination.



  1. Bob Lieblich said,

    May 10, 2008 @ 4:25 pm

    Here's the first section of Title 31, United States Code, Section 3730, cut and pasted off the Web: "The Attorney General diligently shall investigate a violation under section 3729." Not a "to" in sight. I believe Fowler would have called this the product of someone who cares very much about SI's but does not know what one is. This sort of idiocy is especially prevalent in the writing of lawyers — well, American lawyers anyway. For the good of the profession (yes, I'm one of them), I do my best to extirpate such monstrosities. But it would take an Act of Congress to get rid of this one.

  2. Gary Kennedy said,

    May 11, 2008 @ 2:54 am

    I doubt your claim that most speakers of English think there are four different nouns "pen," since most of them will not know the "male swan" meaning at all. I would also have thought (except that you've just instructed me otherwise) that "enclosure for animals" and "enclosure for prisoners" are usages of the same word.

    [AMZ: 'female swan', now corrected in the text. For the other points, see below.]

  3. JimG said,

    May 11, 2008 @ 4:08 am

    Pen, as for animals, should be easy to distinguish as different from pen, shortened from penitentiary. Unless the animal usage evolved from the truncated human prison usage a longlonglong time ago. We can ignore any misspellings of Language Log's maternal soul, UPenn.

    Now, why should the adjective penitentiary have become the noun?

  4. Brett said,

    May 11, 2008 @ 10:24 am

    I suspect that "pen" for prison is neither simply derived from "pen" meaning an enclosure for animals nor simply a shortening of "penitentiary"; rather, it's a combination of both. It is a happy coincidence that by shortening "penitentiary," we get a form that is identical to a term for a place where something is confined. I doubt people would have adopted the shortened form if it didn't have this twist to it.

  5. Arnold Zwicky said,

    May 11, 2008 @ 10:44 am

    To Gary Kennedy: There's no winning on "pen". Every time I've cited this example and given only the three senses that most Americans know, I've gotten mail twitting me for omitting the 'female swan' sense (sometimes expressing surprise that I should be so ignorant).

    So, revise:

    … there are four different etymologies here, and, indeed most speakers of English think that these are "different words" that happen to be pronounced the same.


    … there are four different etymologies here, and, indeed, whichever of the four senses a particular speaker of English is familiar with (many do not know the 'female swan' sense, and some do not know the 'prison' sense), that speaker will think that these are "different words" that happen to be pronounced the same.

    At this point, someone will complain that I have left out the British regional word "pen" 'hill' (in the OED!). And someone else will chime in, with Gary Kennedy, that some people have collapsed the 'enclosure for animals' sense and the 'prison' sense into a general 'enclosure' sense (against the etymology, but ordinary people don't attend to etymologies and shouldn't be expected to). No doubt there is more variation.

    All this detail is fascinating, but beside the point (which is that people do sometimes distinguish "different words" that are pronounced the same, and that dictionaries generally have separate entries for words with different etymologies), and it would be a mistake to add more provisos and parentheses about the variation in the understandings of "pen".

  6. Sili said,

    May 11, 2008 @ 5:27 pm

    Personally, I find it rather disturbing that a uni student can consider "LOSER! Don't you know you can't end sentences with a preposition!" a good response. Kids these days mumble grumble. The lack of reflection is rather disturbing.

  7. Arnold Zwicky said,

    May 12, 2008 @ 10:08 am

    To Sili: kids those days, not these days. I did the questionnaire study in 1982 — 26 years ago.

  8. Jonathon said,

    May 12, 2008 @ 3:21 pm

    Bob Lieblich: I've previously encountered the idea that it's wrong (either flat-out ungrammatical or at least stylistically bad) to put anything between the pieces of a verb phrase. I can't remember if this was in an actual usage manual or if this was just advice that I'd heard in a writing class or something. Either way, I assume it comes from a logical extension of the SI prohibition.

  9. John Baker said,

    May 12, 2008 @ 3:26 pm

    Arnold, would not your analysis apply with equal force to instances where C2 is in fact nonstandard, but C1 is not actually an instance of C2 or otherwise defective in any way?

RSS feed for comments on this post