When the news hit the wires last Friday that Atlanta Falcons quarterback Michael Vick was pleading guilty to charges involving illegal dogfighting, the Reuters headline read:
Which of these three possible readings would you suppose is the one that the headline writer intended?
a) The verb say(s) functions as a more informal substitute for plead(s), since plead can be followed by the adjectival complement guilty to mean 'enter a guilty plea.'
b) The verb say(s) is quotative, with guilty as a direct quotation complement (or at least a partial one). This draws on our cultural understanding that saying "guilty" in court proceedings is elliptical for the declarative speech act "I am guilty" (i.e., "I am entering a guilty plea"). Such a reading would imply that the headline writer neglected to include quotation marks around the word guilty. (Compare the recent New York Post headline, "Cousin Vinny wiseguy says 'guilty' to go free.")
c) The verb say(s) is reportative, with guilty as an indirect quotation complement (or at least a partial one). In this type of ellipsis, "Vick says guilty" is journalistic shorthand not for "Vick says 'I am guilty'" but for "Vick says (that) he is guilty."
Attentive Language Log readers will recognize that the third reading is the most plausible one, given the source of the headline. As Arnold Zwicky explained a month ago, Reuters headlines are very often of the form "X say(s) C," where C is a complement clause missing a subject.
The headline that set the Language Log water cooler buzzing last month was: "Taliban say kill Korean hostage, set new deadline" (July 25). To help illuminate this construction, Barbara Partee and I dug up a large number of Reuters headlines taking the form "X say(s) find Y," as in "Scientists say find gene for child cancer syndrome" or "Statoil says found oil northwest of Shetlands." In such cases, the complement clause for say(s) is a finite VP with the subject omitted. If we were to restore the missing subject in each of these headlines, it would be a third-person pronoun coreferring with the antecedent X: "Scientists say (that) they find gene," "Statoil says (that) it found oil," and so forth.
Now it turns out that the predicates of these subjectless clauses don't necessarily have to take the form of a finite VP like "kill Korean hostage" or "find gene." Reuters headline writers are routinely writing headlines of the form Subject say(s) Predicative with the intended interpretation of Subject say(s) (that) Subject is/are Predicative. In ordinary English, such reported-speech constructions look like the following, with a predicate consisting of a finite form of be plus AP, PP, participial VP, or NP:
1. Joan says (that) she is enthusiastic about the team. (be + AP)
2. Joan says (that) she is on the board of directors. (be + PP)
3. Joan says (that) she is getting fed up with his shenanigans. (be + VP with V-ing head = Present Participle)
4. Joan says (that) she is protected against bankruptcy. (be + VP with V-ed/-en head = Passive Participle)
5. Joan says (that) she is a finalist for the big award. (be + NP)
In the peculiar register of Reuters headlinese, we would lose the subject of each complement clause (she) along with the copular verb be:
1a. Joan says enthusiastic about the team.
2a. Joan says on the board of directors.
3a. Joan says getting fed up with his shenanigans.
4a. Joan says protected against bankruptcy.
5a. Joan says a finalist for the big award.
Copula deletion is nothing unusual in elliptical headlinese, as in these recent examples from the Associated Press: "Gonzales a lesson in cronyism," "Wilson in good condition, hospital says," and "Idaho senator arrested in airport." To my eyes and ears, it's more than a little weird to lose the copula and the implied subject of the complement clause (even if the subject is just a pronominal placeholder). But that's just what the Reuters headline writers do on a regular basis. I went through the Reuters archive and collected five examples of each predicate type numbered above, beginning with the adjectival type exemplified by the Vick headline. Most of the types are easily adduced by looking at headlines for the month of August.
There's one additional type of predicate specific to headlinese: the infinitival "to VP," understood as a simple future or as shorthand for 'be ready/set/prepared to VP.' This construction is routinely used when reporting on corporations or other institutions that make projections for the future (earnings forecasts and the like). Not surprisingly, subjectless complements of this type abound in the Reuters archive, though whether these also involve copula deletion depends on how you interpret the tricky headline infinitival:
When I ran these Reuterisms past the keen eye of Arnold Zwicky, he wondered if it was possible for Reuters headlines to omit the subject in the complement clause for say without losing the copula. As Arnold notes, copula deletion is "customary in headlines, in fact, except when the head writer needs material to fill a line, but it isn't obligatory." As it turns out, copula deletion isn't obligatory even in the special Reuters case of Subject say(s) Predicative. Here are recent examples of the five predicate types with the copula intact but the subject missing:
Since the headline writers of newswires like Reuters are not concerned with the formal limitations of newspaper column widths, I doubt that the choice to keep or lose the copula is a matter of "filling a line." (It's generally not a concern for the online media outlets that reproduce the headlines, either.) Rather, it appears to be more of a stylistic choice, up to the judgment of a given headline writer or editor. And the Reuters practice of deleting the subject of the complement clause for say doesn't appear to be irrevocable "house style" either, since it's possible to find such headlines as "Bush, Karzai say they are aligned against Taliban" (Aug. 6) or "Freed Indian doctor says he is victim of conspiracy" (July 30). Still, the Subject say(s) Predicative flavor of ellipticality seems peculiar to the world of Reuterese. I wonder if it's enshrined in their style guides or if new staffers simply pick up this mannerism from their older colleagues.
All of this reminds me of another quirk of journalistic style: Time Magazine's much-maligned inverted syntax, famously ridiculed in 1936 by The New Yorker's Wolcott Gibbs: "Backward ran sentences until reeled the mind... Where it all will end, knows God!" Well, it did all end, just a few months ago. According to the New York Times, "the last remnants of Time’s signature syntax" were finally banished by editorial fiat in March, just 84 years after Henry Luce founded the magazine. Reuters, there's hope for you yet!
Posted by Benjamin Zimmer at August 28, 2007 01:09 AM