September 11, 2007

What's it all about?

Some time ago, Mark Liberman came across a peeve about at about in expressions like at about 10:30, and countered that there was nothing wrong with it -- it means 'at approximately', which is neither incoherent nor redundant -- and is attested in the writing of eminent authors over the centuries.  Garner's Modern American Usage finds no fault with it, nor does Merriam-Webster's Dictionary of English Usage.  Yet, as MWDEU notes, a long list of manuals condem the usage -- a fact that itself calls out for some explanation.  But first, some words from Zippy the Pinhead about about:

(MWDEU covers most of the territory I'm about to discuss, citing especially Bergen and Cornelia Evans, A Dictionary of Contemporary American Usage (1957), which -- unlike most of the other manuals -- is right on target.)

The use of about in at about is approximative, roughly as in the last panel of the Zippy strip.  The uses in the first two panels are ordinary uses of about as a preposition (P) taking NP objects -- what I'll refer to as an OPERATOR use of Ps.  (The third panel has hard-to-classify idiomatic uses.)  But in the last panel, about is functioning as an ADVERBIAL, in this case modifying predicate adjectives.  A number of Ps -- about, around, under, over, for instance -- have adverbial uses, often more than one kind of adverbial use; all four of the Ps I just listed can serve as modifiers of numerical or time expressions of one kind or another, as in:

(1) About/Around/Under/Over ten people came to the party.

(2) We'll leave in/for/after about/around ten minutes.

(3) We'll stay here until about/around 10 o'clock.

(Different Ps have somewhat different syntax.)

Now, if you fail to appreciate the difference between operator Ps and adverbial Ps, and take all Ps to be operators, you'll think that at 10:30 has one operator P, conveying a (relatively) exact time, and that about 10:30 has another operator P, conveying an approximate time.  As a result, you'll find at about 10:30 to be incoherent, as the griper Mark Liberman cited did. 

The mistake in the reasoning here almost surely stems from a failure to distinguish syntactic CATEGORIES (like P) from syntactic FUNCTIONS (like operator and adverbial).  We've commented here many times on the confusions and misunderstandings that result from not distinguishing category and function.  To appreciate the syntax of English possessives (like Mary's in Mary's father), for example, you need to recognize that they are NPs (and not adjectives or adjective phrases), but NPs functioning as determiners (a type of noun modifier, distinct from adjectivals) rather than as arguments (subjects, direct objects, objects of prepositions, etc.).  In fact, the category/function distinction plays an important role in the next chapter of the at about story.

The most common critique of at about seems to be that it's redundant (rather than incoherent); the at should be omitted because it's redundant.  You get to this conclusion in five steps:

(a) the observation that the at in at about is omissible, in the sense that the versions with and without the at are both grammatical and don't differ significantly in meaning;

(b) the claim that, in general, elements that are omissible in this sense are redundant, meaning that they repeat information;

(c) the claim that redundancy, in this sense, is a bad thing; hence

(d) the conclusion that omissible elements should (or must) be omitted;

(e) in particular, the at of at about should (or must) be omitted.

(The careful reader will have noticed that step (d) is a special case of the famous Omit Needless Words principle.)

I would reject both claims (b) and (c) as general principles, but it's (b) I want to focus on here, because thinking about at about and similar expressions in terms of the superficial notion of omissibility in (a) and (b) leads people away from asking about the syntax of the expressions at issue.  Let's do that now.  I'll start with something reasonably simple, at 10:30 in

(4a) We met at 10:30.

This is pretty clearly a PP, with head P at and object NP 10:30, the whole thing functioning as an adverbial (which is what PPs mostly do).  The only complexity is the nature of the object 10:30.  Time Ps take objects denoting, among other things, locations in time (at / before / after noon), and when they do, the objects take adverbial rather than adjectival modifiers -- nearly, approximately, exactly, almost -- the same types of adverbials that modify numerical expressions (nearly / approximately / exactly / almost ten people).  (There are many complexities here about which combinations of P, adverbial, and object occur.)

On to at about 10:30, as in

(4b) We met at about 10:30.

This is just like at 10:30 in (4b) -- a PP functioning as an adverbial (of location in time) -- except that the object about 10:30 of at contains the adverbial-P modifier about.  It's entirely parallel to We met at approximately 10:30.

Now, finally, about 10:30, as in

(4c) We met about 10:30.

This is something of a surprise; about 10:30 is, according to everything I've said so far, a NP, just like about 10:30 in (4b), but here it's functioning on its own, without an operator P, as an adverbial (of location in time).  It's what we call in the syntax trade a BARE NP ADVERBIAL; its category is NP, but its function is adverbial.  In English, most NPs cannot serve as bare NP adverbials (exactly 10:30, for example, cannot), but some can, and a few of these alternate with P-marked variants, as in

(5a) We met Sunday.  [bare]
(5b) We met on Sunday.  [P-marked]

About 10:30 (and about in combination with many other time expressions) is like Sunday in allowing a bare variant.

There's a huge literature on bare NP adverbials in English; it's a complex topic, with lots of fascinating wrinkles.  But for my purposes here, it's enough to point out that what's notable about at about vs. plain about is not that there's a P-marked variant, but that there's a bare variant; "omissibility" is the special case.

[Addendum 9/12/07: Eli Morris-Heft reports that (4c) is completely unacceptable to him, which is to say that about 10:30 and the like are not available for him as bare NP adverbials in VP-final position, or perhaps in general.  There's plenty of variability in the sets of bare NP adverbials that individual speakers have; this might be just another data point.  Clearly, a great many speakers have no problem with bare NP adverbials in about -- or otherwise manuals wouldn't be recommending them as replacements for P-marked adverbials.]

And there's no redundancy in the P-marked variant.  The P makes explicit the relationship between the time denoted by its NP object and the time of the situation denoted by the clause the PP modifies.  In the bare variant, in contrast, this relationship is merely implicit; in a sense, the bare variant is underinformative (rather than the P-marked variant being redundant).  The relationship between the P-marked and bare variants is then parallel to many other cases of explicit vs. implicit marking -- for example, (explicit) that-marked complement clauses (I realize that pigs can't fly) vs. (implicit) unmarked complement clauses (I realize pigs can't fly).

What's gone wrong in so many advice manuals is that they've focused on omissibility (in sequences of Ps, in particular), treating this mechanically as a test for redundancy, without an appreciation of the syntax and semantics involved. 

Here's one treatment of at about from this literature -- in Roy Copperud's American Usage and Style (1980), where on p. 301, in a subsection of the entry on "piled-up prepositions", we are told:

Single preposition also sometimes superfluous:
- of: omit in "A low temperature (of) near 45 degrees"
- from: omit in "received (from) two to four inches of snow"
- at about: omit at from "(at) about 9:00"

(I've left in the of and from cases as extra entertainment for the reader.)

Note, first, that the usage in question is implicitly referred to higher-level principles -- piling up prepositions is, in general, a bad thing, and omissible material should be omitted (because it's "superfluous") -- though in fact the actual advice consists of a list of very specific cases, which don't hang together.  No one actually proposes that sequences of Ps are in general a bad thing (instead, particular sequences are proscribed), and no one actually insists that omissible material should always be omitted (instead, omission is prescribed in certain specific constructions).  You could spend hours collecting examples of perfectly impeccable sequences of Ps, of several different kinds (Sandy took the box from under the table; Terry walked out of the house; etc.).  And to insist on omission wherever possible would be to insist, among other things, that explicit marking should never be used when implicit marking is available; in particular, bare adverbials would have to be used instead of P-marked adverbials whenever both are available: (5b) out, (5a) in.

Second, though the principles appealed to in the manuals are over-general, sometimes absurdly so, the actual recommendations are often over-particularistic, focusing on a few cases while disregarding other entirely parallel ones.  What Copperud and other handbooks say about at about should carry over almost entirely to at around, though the manuals don't mention around.  The same holds of adverbial about in the object of Ps other than at which are "omissible" on occasion, as in the following pairs:

(6) On about Sunday, things will get worse.  /  About Sunday, things will get worse.

(7) In about June, things will get worse.  /  About June, things will get worse.

(8) I waited for about two hours.  /  I waited about two hours.

Third, these very particularistic recommendations are mostly framed in terms of linear strings -- the sequence X Y is to be avoided when one of them is omissible -- though they should really be framed in terms of structures or constructions, and (more important) the notion of "omissible" isn't made explicit.  In the case of at about, some occurrences are totally irrelevant: the person I yelled at about the failures, where at and about belong to different, parallel, constituents.  Some fail one or the other clause in the definition of "omissible", even though the structures are more or less of the right sort: I aimed at about ten targets, where the version without at is not grammatical (unless you're someone who can aim targets at things), because the at of aim at expresses goal rather than location; At about the corner, I fell down and At about these rates, indebtedness will decline in ten years, where the at is not omissible even though it expresses (metaphorical) location; and I yelled at about ten kids, where the version without at is grammatical, but is not even approximately a paraphrase of the version with at (goal rather than location again).

The problem is that the manuals give you generalizations -- high-level ones (Omit Needless Words) or specific ones (omit at in at about) -- and one or more instances of the generalizations, but nothing to indicate the limits of the generalizations.  In effect, they're saying

Follow this advice, unless that would be wrong.

and they're obliging the reader to try to divine their intentions from the examples they give.  Not very helpful.

Sometimes, I'd guess, the usage advisers are simply unaware of the complexities in their advice or don't understand the details of the constructions they're talking about.  Other times, I'd imagine, they're aware that more needs to be said but shrink from introducing technicalities, or (relying on the conceptual apparatus of "traditional grammar") they just don't have adequate vocabulary to convey those technicalities.  After all, how many people know about bare NP adverbials?

As I've said before, these nuggets of advice all start from specific events: somebody noticed a class of examples, judged them to be in some way imperfect, and then formulated a principle to appeal to in proscribing them for others -- a process that promotes both overspecificity (focusing only on the motivating examples) and overgeneralization (leaping to abstract characterizations of the perceived problem).  These proposals aren't seen as hypotheses about how the language works, but just as bits of advice about what to avoid in your language.  Then, in some cases, these ideas disseminate, as ideas do in communities.

Which brings me to my fourth point: as I've also said before, the advice manuals, unsurprisingly, tend to share opinions and attitudes.  Through their influence and the influence of editors and teachers, some usages get picked out for special opproprium, out of all proportion to their significance in the larger scheme of things (why should anyone care about the saving of the little word at in at about?), sometimes without reference to the practice of "better authors" (recall that at about has a long and distinguished history), and often without connection to very similar usages (at about gets bad press, at around and for about escape notice).  Ordinary people develop a prejudice -- a pet peeve -- against these usages, complete with viscerally unpleasant responses to them.  There are fashions in pet peeves, linguistic pet peeves included, as in other things.

Finally, a note about how I got interested in at about in the first place.  One of my current projects is to investigate instances of Omit Needless Words and Include All Necessary Words advice in the manuals.  (I'm now thinking of it as the OI! project: Omit!/Include!  Oi.)  ONW and IANW figure in usage advice in two different ways.  Sometimes it seems pretty clear that a usage is deprecated for social reasons -- because of the people who use it or the contexts in which it's used.  The primary objection is that the usage is non-standard, specific to some social group or region, innovative (or perceived to be so), informal, or mostly spoken rather than written, and this objection is bolstered by a SECONDARY appeal to some general principle, for instance ONW or IANW.  (The reasoning here goes both ways: in a widespread piece of language ideology, the standard, general, established, formal, written language is taken to be intrinsically good, so that variants that are non-standard, restricted, innovative, informal, or spoken are EXPECTED to be intrinsically defective in some way: sloppy, vague, redundant, illogical, etc.)

Examples: the various sorts of "intrusive" of -- notably, in off of and other combinations of prepositions with of (It fell off of the table), and in exceptional degree modification (That's too narrow of a topic for a paper) -- are judged by many to be non-standard, or at least innovative, informal, or spoken.  ONW is then appealed to as a backing for the advice that these usages are to be avoided.  Meanwhile, the determiner a couple without of (A couple people complained rather than A couple of people complained) is judged by some to be non-standard, or at least innovative, regional, informal, or spoken.  IANW is then appealed to as a backing for the advice that this use is to be avoided.

Sometimes, however, appeals to general principles (like ONW and IANW) lack any evident social basis; these are PRIMARY appeals.  The objection to at about as a violation of ONW seems to be of this sort, as does the objection to the then of if ... then as a violation of ONW.  (I hope to post again soon on the then case.)

My hypothesis in the OI! project is that secondary appeals to ONW and (especially) IANW considerably outnumber primary appeals.  To investigate this hypothesis, I've had an intern, Rachel Cristy, inventorying OI! appeals in a collection of manuals; it was Rachel who pointed me to at about as an example of a primary appeal to ONW that is, or at least was, surprisingly popular (perhaps as a result of a fashion in peeves).  (My thanks to the office of the Vice Provost for Undergraduate Education at Stanford for funding Rachel's internship.)  Meanwhile, I code and tabulate.

zwicky at-sign csli period stanford period edu

Posted by Arnold Zwicky at September 11, 2007 12:04 PM