approve (of)

« previous post | next post »

William Safire has taken up (in his column in the NYT Magazine of 11 May) the knotty question of whether political candidates should say they approve some message or approve of it. This caught my eye because I've been thinking recently about "diathesis alternations" in general (see here and here), and in particular about alternations in English between direct objects (no preposition) and oblique objects (marked by a preposition). I have an unfortunately large file of cases, in most of which the oblique, intransitive, construction is historically older (with the direct, transitive, construction a more recent innovation). But for approve, the oblique variant is the innovation; the first OED dates are ca. 1380 for the direct, 1658 for the oblique (with P on at first, later supplanted by of).

Safire tells us to lose the of. There are at least five things worth commenting on here.

To start with, why is Safire considering the question at all? Well, it began with e-mail from a reader who noticed that the candidates are saying approve this message, though his "Boston-school-systems teaching places of in front of this". Safire replied that the 2002 Bipartsan Campaign Reform Act ("McCain-Feingold") stipulates this wording. But even if the law were not involved, I think he'd have taken up the question.

That's because he's looking at two variants, approve of and plain approve, which in this context probably seem to him to be semantically equivalent. But, as I've noted many times, usage writers are reluctant to allow "free variation"; to their minds, there should always be a "correct", or at least a favored, choice, and there should be a justification for this choice. I think of this as the One Right Way principle, and I think that as a general principle it's a rotten idea.

[Stop before you begin writing a comment! I'm only saying that One Right Way is a rotten idea AS A GENERAL PRINCIPLE; instead, there should be room for variation from person to person and occasion to occasion, and people should be free to make some choices for reasons they can't articulate. I am NOT saying that all variants are equally good, all the time; it's often the case that in particular contexts, one variant is superior to another.

I'm a bit annoyed to have to make this disclaimer, but if I don't, I will surely be pilloried by critics. Whenever I denounce some extreme position (like One Right Way), a certain number of readers will take me to be advocating some equally extreme, and absurd, contrary position. So my postings are accumulating more provisos, disclaimers, restatements, explanations, and the like. They are becoming less fun to write, and probably less fun to read.]

In any case, instead of saying that, well, different people have different preferences, but both variants are standard, Safire elects to choose a winner.

That brings me to point two. Though Safire seems to think of the direct and oblique variants as semantically equivalent in the campaign-ad context, they are certainly not so in general. Whatever their history (which involves considerable overlap in usage) direct approve and oblique approve of are now clearly differentiated in many contexts, as (in the OED's glosses, quoted by Safire) 'confirm authoritatively' and 'pronounce to be good', respectively (in the MWDEU account, 'sanction officially' and 'take a favorable view'). Compare:

I didn't approve that application for retirement.

I don't approve of homosexuality.

Neither variant can be traded for the other without change of meaning (or, in the second case, grammaticality). Yes, they can overlap, but they're far from free variants. (There's more discussion in MWDEU.)

The remaining points have to do with Safire's justifications for his choice of direct approve. In the order they appear in his column:

Historical priority: direct approve is the older — the original — variant.

Stylistic association/connotation: oblique approve of "sounds elitist, out of touch, like a blessing by the Charles River Gang."

Omit Needless Words (in an especially hyperbolic version): "If you can cut earmarks out of the budget deficit, you can cut the needless word out of the sign off, and it saves a fraction of a second, which–multiplied by thousands of TV spots–adds up to real money."

[Ok, on the last item, I'll grant that this is intentionally, and humorously, hyperbolic — but it looks just like the advice literature on "excrescent of" in various constructions, in particular in exceptional degree marking (too big of a dog and the like).]

So, point three, on historical priority: the idea that an "original" meaning-form pairing must now be the "correct" one is very silly, a piece of mystic "originalism" akin to the Etymological Fallacy (on which Language Loggers have written many times). Things change, and over the years they can change dramatically. New variants arise and compete with older ones. Sometimes the older variant vanishes or survives only in certain restricted contexts, sometimes the newer variant goes out of fashion, sometimes they end up in stable variation (usually differentiated in some way: in semantics, discourse function, style, or sociolinguistic value).

Things mean now what they mean NOW. It counts for nothing that direct approve got in first (especially when we're talking about events from hundreds of years ago). What's now is now.

Backtrack to the think/thing thing. It seems pretty clear that have another think coming was the earlier variant, but these days nobody should care.  Both variants are well-attested and should be treated as alternatives in the (informal) standard language.

On to point four, nuanced stylistic/connotational differences between direct approve and oblique approve of. Here Safire hears an "elitist" tone in the oblique variant. I don't hear that, but I do hear a subtle difference between the two variants which is related to the difference Safire hears: the oblique variant is more "distanced".

This is a general property of obliques vs. directs (long noted by "functionalist" linguists), and it represents a kind of ICONICITY between syntax and meaning: direct objects are more closely tied syntactically to their verbs (recall my discussion of nonintervention), and so tend to convey close association between the event denoted by the verb and the participant denoted by the object, whereas oblique objects tend to convey more distant associations. Compare:

I played the piano for hours. [direct]

I played on the piano for hours. [oblique]

(Many other P~zero pairs show similarly subtle distinctions.)

My interpretation of what Safire hears in direct approve vs. oblique approve of is that he's taking distance to be social distance.

Finally, point five, ONW (to which I return with regrettable frequency on Language Log). 

[Ok, another disclaimer. Strunk (1918) managed to fold entirely reasonable warnings about wordiness and excess verbiage — what writing teacher has not issued such warnings? — in with proscriptions against specific usages that he disliked or judged (whether correctly or not) to be non-standard. This is a terrible idea. The justificatory use of ONW for specific usage proscriptions is simply indefensible, as far as I can see.

And it undercuts the genuinely valuable advice about wordiness and excess verbiage.

In any case, I am NOT saying that extra words are generally a good thing. I'm talking about the fine details of specific syntactic constructions. Saving an of in approve of is simply not a question of avoiding wordiness; this is just a slam against a specific construction that Safire doesn't prefer. (Note: I am NOT saying that approve of is necessarily standard just because some people use it. In fact, I think it IS standard, but I think that because of who uses it in what contexts. ONW, or that matter IANW (Include All Necessary Words), has nothing to do with this judgment on my part.)]

Appealing to ONW in cases of direct/oblique alternations is stepping into dangerous territory, because different cases work in different directions. Take the verb protest. There's a significant difference between American and British usage here. American speakers generally allow direct protest with NP objects as well as oblique (with the P against):

They protested against the war. [oblique]

They protested the war. [direct]

But the direct use — the one that saves a word — is judged unnatural or straightforwardly ungrammatical by British speakers. Some reports on the situation:

Garner's Modern American Usage (p. 649) says that against may be omitted in American English, but not in British English.

Burchfield's New Fowler's (p. 635) says that transitive protest is accepted in American English but is "far from natural" in British English.

MWDEU (p. 784) reports that the transitive is as common in the U.S. as the intransitive, but that British English still normally uses against.

The OED (draft revision Dec. 2007) says of transitive protest: "Chiefly U.S. To object to (an action or event); to challenge or contest; (also) to make the subject of a public protest or demonstration."

And Jan Freeman reported last year in a Boston Globe column on the scorn heaped on transitive protest (and battle and appeal) by British readers.

Meanwhile, British news sources (as reported by Alison Murie on ADS-L back in January) have innovated a direct use of agree (for agree to) that's just not acceptable to American ears:

Many countries hope that Bali will agree a two-year roadmap to negotiate a successor to the Kyoto Protocol, the main UN plan for fighting warming until 2012 … (link)

(I have examples from print journalism and from radio reporting, all from British sources. These are not errors.)

There are many other P~zero alternations discussed in the usage literature, in ADS-L postings, and even here on Language Log. The status of the variants (as standard, marginal, or non-standard) differs from case to case, and the situation is different from dialect to dialect. There is a general tendency for speakers to innovate direct uses, but many of these innovations are, at least at the moment, marginal (direct depart in depart San Francisco) or decidedly non-standard (direct abscond in abscond the money), so you wouldn't want to appeal to ONW in describing them; if these are covered in the manuals, it's IANW that's appealed to, telling you not to omit the necessary P. Obviously, neither ONW nor IANW is of any real use in these situations.

To sum up: the law's the law, so political candidates should use direct approve in their messages (though it's hard to imagine anyone being taken to court for using oblique approve of). And Safire's entitled to his preference for direct approve in this context. But his arguments in favor of this preference are not cogent. Oblique approve of is also standard, and in fact there's no reason to have to pick one of the variants as the correct one.




  1. Josh Millard said,

    May 15, 2008 @ 4:22 pm

    I'm a bit annoyed to have to make this disclaimer, but if I don't, I will surely be pilloried by critics.

    Have you considered an enumerated FAQ? Anything that has required clarification half a dozen times (or whatever threshold fits) goes on the list, and in lieu of a tiring disclaimer you can simple drop number hyperlinks into an article as necessary ("[see #4 & #21]").

    Special item "#mu" on said list could be a boilerplate explanation reserved for replies to correspondence from a reader who did not apparently in fact see the suggested items. This one could recurse nicely for the superlatively inattentive-yet-insistent correspondent.

  2. Craig Russell said,

    May 15, 2008 @ 5:12 pm

    A nicely written and intelligent discussion of the question and the larger issues. As for the actual difference in meaning between the two expressions, for what it's worth, in my own usage and understanding, there is a pretty clear-cut distinction between the two: you approve things that you have control over, and you approve of things that you don't.

    So I might approve of a political candidate's commercial, but only the candidate himself has the ability to approve it. If an editor doesn't approve a story, the story doesn't go to print; if the reader doesn't approve of a story, the story still gets printed. Et cetera.

    Is this distinction between the two unique to me?

  3. Craig Russell said,

    May 15, 2008 @ 5:13 pm

    Sorry–on closer reading of the post, I see that you say pretty much the same thing.

  4. Got Medieval said,

    May 15, 2008 @ 5:50 pm

    Maybe McCain and Feingold or their legislation-drafting aides were trying to thread a semantic needle themselves. The expected construction would be "I approved this message," in the past tense, but then that would allow a politician to more readily say, "Well, sure, I approved it. I signed the authorization, but I sign off on a lot of things. I certainly don't approve of the sentiments expressed in the ad." By forcing the candidates to say, "I approve" they're forcing them to assert their ongoing approval of the message.

    A LexisNexis Congressional search might clear up what the congressmen thought they meant when they approved the legislation.

  5. Tim Silverman said,

    May 15, 2008 @ 5:53 pm

    I think transitive "agree" means something more like "agree on" than "agree to". You agree to something (e.g. a proposition) that already exists and is presented to you; but the agreement that you agree (transitively) comes into existence in the process of being agreed. That's how it seems to be for me, anyway.

  6. Peter Howard said,

    May 15, 2008 @ 6:08 pm

    Interesting that Safire thinks the oblique version "sounds elitist" and yet the OED, MWDEU, Craig Russell (and me, FWIW) all think that the direct version tends to imply some control, authority or officialdom. Which would suggest that the direct version is the elitist one. Arnold's "distanced" description makes more sense to me; if you don't have control etc, then you are in a sense more distanced.

    I was also interested that direct "agree" is "just not acceptable to American ears." As a BrE speaker, I'd noticed this innovation and assumed it to be a US import. But I often get this wrong. I once used the word "candyfloss" in a poem and was astonished when a whole bunch of Americans asked me what it meant.

  7. mike said,

    May 15, 2008 @ 6:11 pm

    Probably elsewhere you have addressed "graduated high school" from "graduated from high school."

    As for the disclaimers, you might enjoy, in a misery-loves-company sort of way, the ever-more-numerous accretions that Raymond Chen, who writes about Windows programming, has been obliged to include in his blog posts. Here's an example of an extended "Nitpicker's corner":

    And an example "Pre-emptive comment" that features a named heckler, no less:

    One more:

  8. Arnold Zwicky said,

    May 15, 2008 @ 6:41 pm

    To Peter Howard, who said:

    I was also interested that direct "agree" is "just not acceptable to American ears." As a BrE speaker, I'd noticed this innovation and assumed it to be a US import.

    Wonderful. Confronted with usages that are new to them, people tend to consign them to the Other (for some sense of Other). I have tons of examples. The most entertaining ones are cases in which people from group A assign the variant to group B, *and* vice versa. It's what "other people" say.

  9. Elise Kramer said,

    May 15, 2008 @ 6:48 pm

    In addition to differing in terms of the action/position they describe, "approve" and "approve of" also differ in aspect: "approve" seems to describe a single action that has been completed, whereas "approve of" seems to describe a continuous mental state. "I approve this message" therefore seems a bit strange to me, for reasons I can't quite pinpoint — perhaps it's that the multiplied and displaced quality of the advertisement doesn't quite jibe with the speech act of approval. It's like the candidate is constantly re-approving the message every time it plays, but he technically only approved it once when recording the commercial. I suppose this has to do with some kind of need I have for a performative speech act to be physically reinscribed every time it is instantiated.

    "My name is X, and I approved this message" seems to me to make the most sense. "Approve of" also makes sense to me, but it has less official force to it (both because it doesn't sound legally binding and because it isn't a performative).

    I just discussed this with my partner, however, and he thinks "I approve this message" sounds absolutely fine. Some light argument revealed that our disagreement hinges on our different understandings of how the "I approve this message" clip is meant to be understood as spatiotemporally situated. Whereas I see the "I'm X and I approve this message" clip as a new speech act conferring approval upon each airing of the commercial, he sees it as visual evidence that at one point the candidate approved the commercial — as though his aides had just shown him the commercial, he had said "I approve this commercial!" and they had just happened to be taping at that moment. Whereas I see the "I approve this message" clip as spatiotemporally dislocated just like the rest of the ad, my partner sees it as a separate, "anchored" accompaniment. So to some extent this seems to be in the eye of the beholder.

  10. Ellen K. said,

    May 15, 2008 @ 7:07 pm

    I think "approve of" doesn't even make sense in this context. If someone made an ad, and the politician later looked at it, he could approve of it. But in my head, he can't approve of it before it's finished. And it's not finished when he's saying that line. Like, you can approve of things that exists, or general ideas, but not future entities. You can approve of the idea of something… because the idea exists now.

  11. Bob Moore said,

    May 15, 2008 @ 8:26 pm

    "Approve" and "approve of" certainly do not mean the same thing. When I travel on business, I frequently have nice dinners that go right up to the limit of the expense reimbursement policy of my employer. Some people no doubt consider this extravagant, and I am certain that some of my expenses have been approved by administrators who did not approve of them.

    Elise Kramer's observations about "approve" and "approve of" differing in aspect are on the mark. In the verb typology of Zeno Vendler, "approve" expresses an achievement, and "approve of" expresses a state. Typically in English, only stative verbs can be used in the simple present to say that something holds at the time of utterance with no special aspectual character, so "I approve this message," is likely doing something else.

    I believe what is going on is that "I approve this message" is a performative in J.L. Austin's sense. The utterance of "I approve this message" is the act of approval itself. A large number of speech-act verbs are used in this way; e.g., "I now pronounce you man and wife," "I declare the meeting adjourned," "We request the pleasure of your company," etc.

  12. rootlesscosmo said,

    May 15, 2008 @ 9:00 pm

    direct depart in depart San Francisco

    I would suggest that direct "depart" and "arrive" are railroad timetable usage that entered general US English some time in the 19th century. "Train 98 departs San Francisco 8 am [note that "at" is also omitted], arrives San Luis Obispo 2 pm."

    We on the railroad always personalized trains (technically "schedules") as "he": "When does ninety-eight get here?" "He's outa the City at eight, due in San Luis at two."

  13. Mark Liberman said,

    May 15, 2008 @ 9:29 pm

    I would suggest that direct "depart" and "arrive" are railroad timetable usage that entered general US English some time in the 19th century.

    Maybe timetable usage played a role, but depart with the place departed from as direct object has been around for a while, according to the OED:

    1536 in W. H. Turner Select. Rec. Oxford 138 Nicholas Hore paid for the wine and departed their company. 1548 HALL Chron. 114 All the Welshemen were depart the toune. 1597 HOOKER Eccl. Pol. V. i. (1611) 186 The soules of men departing this life. 1647 N. BACON Disc. Govt. Eng. I. lix. (1739) 112 No Clergyman or other may depart the Realm, without the King's Licence. 1712 ADDISON Spect. No. 517 §1 Sir Roger de Coverley is dead. He departed this life at his house in the country. 1734 tr. Rollin's Anc. Hist. (1827) II. II. 126 Jugurtha was commanded to depart Italy. 1839 KEIGHTLEY Hist. Eng. II. 33 The clergy were ordered to depart the kingdom. 1861 DICKENS Gt. Expect. xxxiv, Mrs. J. Gargery had departed this life on Monday last.

  14. Roger Levy said,

    May 15, 2008 @ 10:25 pm

    I think Bob Moore's example says it all :) "Approve" and "approve of" are far from meaning equivalent.

  15. Garrett Wollman said,

    May 15, 2008 @ 11:38 pm

    This discussion brings to mind a couple of notions from my experience in computing.

    In computer languages, where syntax and semantics are both more constrained than in natural languages, tokens which are mandatory in the syntax but serve no purpose (either syntactic or semantic) are referred to as "noisewords". There is a heritage of computer-language design based on the theory that valid statements in computer language ought to look at least something like human language. Of course, the converse is not true (yet, anyway) because the grammar of the computer language is much more restricted. This design idea is what gave us languages like COBOL, SQL, and HyperTalk (remember HyperTalk?) and is these days frowned upon — conciseness is considered more important than an imperfect notion of readability.

    An example for those who don't do this sort of thing: in SQL, a database-query language, the most fundamental statement in the language is "SELECT", which extracts data from the database (by convention, SQL keywords are written in upper case). It has numerous optional clauses, so the full grammar is quite complicated, but it could be summarized as "SELECT columns FROM tables WHERE conditions GROUP BY group-columns HAVING more-conditions ORDER BY columns". The "BY" in "GROUP BY" and "ORDER BY" is a noiseword; it serves no syntactic purpose (the grammar would be no more ambiguous without it), nor does it serve a semantic purpose (there is no other token allowed after "GROUP" or "ORDER" — so "BY" doesn't distinguish one kind of grouping or ordering from another — yet it is mandatory. Its only purpose in the language is to make statements read somewhat more like English sentences.

    One of the languages that has taken the notion of conciseness to an extreme is Perl. Perl's creator, Larry Wall, has some unusual linguistic ideas, some of which come from the same line of reasoning as gave us that useless "BY" token in SQL, but he ended up in a completely different region of design space. Perl's motto is "There's More Than One Way To Do It" (TMTOWTDI), and for almost anything you might want to do, there are often three or four ways to do it. Perl has a highly ambiguous grammar, although the parser resolves most ambiguities in the least-surprising way, reflecting the notion that programmers ought to be able to express themselves in different ways depending on personal preference and the requirements of the task at hand. For example: in most computer languages, there is a single way to express a conditional — indeed, it is so stylized that it's often referred to as the "IF … THEN" construct even when a given language doesn't use that syntax. Perl not only has an "if (condition) { statements }" statement, but it also has a "statement if condition" construct (because sometimes it's more natural to think of the condition as less important than the statement), and indeed complementary "unless" statements to go along with "if" (because sometimes it's more natural to think of the exception rather than the rule). Depending on the exact requirements, I could name two or three more ways to express the same semantics in Perl. Many language designers — particularly academics — decry this "profligacy" on the part of Perl (and languages inspired by it); they seem to be motivated by the same sort of "There's Exactly One Way To Do It" (TEOWTDI) philosophy as is espoused by many die-hard prescriptivist writers.

  16. John Cowan said,

    May 16, 2008 @ 12:11 am

    I had understood that "protest something" grates on BrE ears because BrE has a different role for the direct object: in Britain, convicts protest their innocence, whereas in America, they protest their conviction.

    However, none of the non-obsolete senses of PROTEST (v) in the current OED revision seem to back this up: sense 3a is "protest that" in the sense "state that", 4a is a technical use in commercial law, 6a is "protest against", 6b is "protest" meaning "utter a protest", with a direct or indirect quotation, 6c is "protest" in the sense "collectively dissent", 6d is the U.S. usage with a direct object, 6e is technical sports language, 6f is used of things to mean "creak under pressure".

    I've written to the editors to inquire, or enquire, as the case may be.

  17. Pink Pig said,

    May 16, 2008 @ 12:32 am

    I hesitate to quibble about such arcana, but while I agree (generally) with the principles stated here, I think this is a very bad example to cite. "To approve" (v.t.) and "to approve of" (v.i.) bear little relation to each other. The transitive form is a statement of an official act, e.g. "The legislature today approved a measure that gives umpteen zillion dollars to the Teatasters International Union of Peoria." The intransitive form is a statement of opinion, e.g. "I/He/She/We approve of the death penalty for ponces." I think the test is that you can't meaningfully say, "I approve X of Y", because the two meanings aren't the same. It's a different situation with "I agree with " and "I agree on ", because you can say "I agree with on ".

    I don't think that this is analogous to the use of "of" to express the French partitive, as in your example "too big of a dog".

  18. Pink Pig said,

    May 16, 2008 @ 12:36 am

    Umm, that didn't come out right. Let me try again.

    It's a different situation with "I agree with <someone>" and "I agree on <something>", because you can say "I agree with <someone> on <something>".

  19. Anthony Hagan said,

    May 16, 2008 @ 2:31 am

    I'll agree with John Cowan: most usage of transitive "protest" grates with this forty-something speaker of Australian English too.

    However it is becoming more common here, e.g. in reputable media. I suspect I'm being left behind on this matter. (I must ask the opinion of my twenty-something sister in law…)

  20. Paul Carter said,

    May 16, 2008 @ 5:46 am

    I'm interested in the statement "the direct use [of ] — the one that saves a word — is judged unnatural or straightforwardly ungrammatical by British speakers" in a piece which challenges one particular speaker's judgements on a similar construction.

    The thing that bothers me is the question of how one decides when a judgement is a legitimate way of gathering data about grammaticality and when it is raving prescriptivism. Is it the case that grammaticality judgements ought to be backed up with actual usage data (e.g. the quotes on British versus American transitivity with and )? And if so, why do we bother with judgements at all? Why not go straight to the usage? Can we really say any more than "cat the sat mat on the" is absent from the data? If we're worried about so-called "performance errors" why do we read things like "I have examples from print journalism and from radio reporting, all from British sources. These are not errors"? How does one decide what is and what is not an error?

    I've been to enough talks by syntacticians which have descended into discussions about whether a particular construction is "bad" for a particular member of the audience to wonder seriously about this issue. I find it hard to work out why I should care whether a particular construction is "bad" for a syntactician – if you've thought about the hypothesis you should be excluded from the experiment, surely. And there's enough evidence out there to say that what people think they do in their language is often quite far from what they actually do – perhaps this applies less to syntacticians, but if we're asking the general public…

    Perhaps the nature of how we gather data is worth a piece on its own.

    NB I'm not challenging the particular judgements mentioned here: as a British speaker, "protest the war" sounds weird but "agree a deal" is quite normal to me. I don't know whether that's different in different parts of Britain.

    By the way, thanks for Language Log. It's beginning to teach me that syntax is not necessarily as unfathomable as I once thought.

  21. John Cowan said,

    May 16, 2008 @ 10:24 am

    The OED folks bounced the answer back to me this morning (my time): "protest one's innocence" is in fact lumped under 3a, not all of which is "protest that"; it also allows a direct object that represents a state, such as "innocence" or "excellence". The difference between this sense of "protest" and simple "state" is of course that "protest" involves the presupposition that the matter is under dispute.

  22. Coby Lubliner said,

    May 16, 2008 @ 11:34 am

    There is another British-American direct-oblique contrast: the use of "stop" or "prevent" + NP [AmE: + from] + gerund.

  23. Liz said,

    May 16, 2008 @ 1:49 pm

    The thing that makes the least sense about Safire's point #3 (Omit Needless Words), is that television ad time is not sold in fractions of seconds, and rarely in increments of less than 15 seconds at that. I understand he was being hyperbolic, but the idea of saving money by omitting one extra word from a TV ad doesn't even make sense hyperbolically.

  24. Janice Huth Byer said,

    May 18, 2008 @ 8:22 pm

    Safire's mistaken use of 'elitist' is probably defensive. While it may be presumptuous or impertinent to give approval outside your scope of legitimate authority, it's not elitist. But then, who wouldn't prefer to be called elitist, if only because, being inaccurate, it can honestly be denied? Consider these, if you would be so kind :)

    As CEO of my campaign, I approve this message.
    As a pundit, I approve of this message.

    As God, I approve homosexuality as an alterative for those inclined by Mother Nature. I don't approve the use of my name to persecute homosexuals.

    I approve of homosexuality for those inclined, not that it's any of my business.

    Could the distance we sense in the phrase "approve of" relate to its use in approving what is *beyond* the scope of authority?

    Note the word "of" is not needless. ONW rules, rocks!

RSS feed for comments on this post