March 31, 2007

Cut 'n' paste


Today's New York Times has a heartbreaking story (beginning on p. 1 and continuing on p. 8), "Iraqi Widow Saves Her Home, but Victory Is Brief", with three photos on p. 1 and three more on p. 8.  The third photo is of the widow's granddaughters; the first sentence of its caption summarizes the whole story:

THE DAY AFTER Ms. Saadoun's phone call to a nearby military base kept her family from being falsely evicted, but she was killed the next day at a market.  Her granddaughters were among those in mourning.

As you can see, this sentence has been mangled in editing, probably as a result of shifting from one version of the sentence to another, giving something that looks like an on-line sentence blend but is subtly different from it.


Here are what I take to be the alternatives the caption writer considered:

(1) The day after Ms. Saadoun's phone call to a nearby military base kept her family from being falsely evicted, she was killed at a market.

(2) Ms. Saadoun's phone call to a nearby military base kept her family from being falsely evicted, but she was killed the next day at a market.

The first of these has an initial adverbial subordinate clause followed by the main clause; the second is a coordination of two sentences, with contrastive but linking them.  Both present the two events in the order they happened, but in (1) the material locating the second event is in the initial modifier, while in (2) it's in the second conjunct.  If I had to choose, I'd go for (2), because it represents the temporal information more directly and because it conveys contrast explicitly rather than implicitly.  My guess, in fact, is that the caption writer started with (1) and then decided to revise it to (2), and didn't quite finish the job.

On this account the problem was introduced in the editorial process, rather than in the original writing.  It's what I'll call a "cut 'n' paste error" (though actual cutting and pasting don't have to be involved).

(I went on-line to see if the cut-'n'-paste error had been fixed, only to discover that the photos -- actually, a slightly different set -- had been organized into a slide show, with the granddaughters last, and with the captions rewritten to tell things event by event, so there's nothing that corresponds to the problematic sentence in the paper version.  In the paper version, the photo of the granddaughters is right above the story, so it would have made sense to try to summarize the narrative there.)

At first glance, the caption sentence looks a lot like a syntactic blend, an on-line error that results from the speaker or writer entertaining two formulations and producing something with parts of each.  In the common "splice blend", the first part of one of the competitors is spliced together with the second part of the other; there is usually a overlap portion in the middle, a piece shared by the two competitors.  So, if you're entertaining the alternative idioms

blown out of proportion
taken out of context

(which share out of), you can end up producing either of the following (from a 2005 discussion on the American Dialect Society mailing list):

[ blown [ out of ] context ]
[ taken [ out of ] proportion ]

(where the red brackets enclose the contribution from the red idiom, and the blue brackets the contribution from the blue idiom).  The caption sentence can be analyzed in this fashion.  I'll treat (1) as the red sentence and (2) as the blue, and put the three portions of the example (initial red portion, medial overlap portion, final blue portion) on separate lines:

[ The day after

[ Ms. Saadoun's phone call to a nearby military base kept her family from being falsely evicted, ]

but she was killed the next day at a market ].

But the analogy isn't exact.  That final portion isn't all blue: only but and the next day are blue; she was killed and at a market are shared between blue and red.  The caption sentence has EVERYTHING from sentences (1) and (2) in it, unlike usual examples of inadvertent blends, which have only part of each.

In addition, sentences (1) and (2) are pretty long.  It's hard to imagine that the writer was entertaining both (1) and (2), in full, at the beginning of the writing process.  The contributors to syntactic blends are almost always short expressions.  In fact, one of the contributors is almost always a fixed expression of some sort, and very often both are (as in the proportion/context examples above); it's plausible to assume that the speaker or writer is retrieving one or both of them as wholes, rather than composing them from scratch (as would surely be the case for (1) and (2)).

So a blend analysis isn't very plausible.  There is, however, another way in which the caption sentence could have arisen as a on-line production error (rather than an editing glitch): through a shift from one structure to another during production, a kind of anacoluthon.  The speaker or writer loses track of an original plan and embarks on another, perhaps building on material most recently produced.  In the case at hand, the writer might have lost track of the (1) plan during that long middle portion and shifted to the (2) plan.  Such shifts certainly occur, especially in speech; some speakers (including several U.S. presidents) are famous for their wandering from one structure to another in long, rambling sentences.  They seem to be much less common in the writing of practiced writers, but we can't exclude the possibility in the case of the caption sentence.

To sum up the discussion so far: we're faced with an expression that is in some way a combo, a combination, of two others.  The question is how the combo arose, and I've suggested three possibilities in this case: as an inadvertent blend (resulting from competition between plans), via inadvertent anacoluthon (resulting from a shift in plans), or as a cut 'n' paste error (also inadvertent, but arising in the editing process).  It's impossible to say for sure why this single event occurred, but it is possible to argue that some accounts are more plausible than others.  (For other cases, the plausibilities might be quite different.)

Two larger lessons here.  The first is that a configuration -- like combos -- can arise in many different ways; we'll need to distinguish many types of phenomena.  (This lesson is familiar from syntax.  There are, for example, a great many ways in which clauses or clause-like expressions can be missing a verb in English; we need to posit a rather large number of constructions -- with rather different syntactic details, semantics, sociolinguistic statuses, and discourse functions -- that allow for the omission of a verb.)  The second lesson is that without further information, single occurrences can almost never be assigned with complete assurance to a type; isolated events are generally inscrutable.

Suppose it's been reported that someone said "I didn't see nobody" (with the prosody of "I didn't see Peabody").  What do we make of this?  Well, it could be an inadvertent blend (of "I didn't see anybody" and "I saw nobody"); or just a run-of-the-mill negative sentence from a speaker of a variety with multiple negation; or a joking quotation of multiple negation from a speaker whose variety doesn't otherwise have it.  No doubt other possibilities could be imagined.  In any case, there's more than one way this expression could arise, and to decide which type to assign it to we have to know more about the speaker and the context of speech.

In fact, combos can arise in many other ways than the three I talked about above.  These three are all inadvertent errors, but there are also deliberate combos: for example, portmanteaus like smog and Brangelina and chillaxing (chilling + relaxing) and many others; deliberate telescopings like filk for fresh milk (reported to me by Eric Lee, from his Stanford household); and deliberate "piling on" on of material for emphasis, as in little small and small little.  And there are still other types, which I'll forbear discussing here so as to keep this posting to a reasonable size.  I've started using "combo" (as above) as an umbrella term covering all these phenomena; but it's merely a label for a set of problems, rather than an analysis of anything.

zwicky at-sign csli period stanford period edu

Posted by Arnold Zwicky at March 31, 2007 01:56 PM