Before reading further, consider the following newspaper headline, and make a mental note of what you think the article is about:
This headline appeared last week in the Calgary Herald. I eyeballed it without cognitive incident, and settled down to read what I thought would be one of those year-end, low-calorie servings of retrospective meta-journalism in which there is much musing about what the past year's news events all mean. Sort of a re-mastication of analyses about the lessons we might have gleaned from the Occupy Movement.
Actually, the article turned out to be news news, reporting on the following startling event:
Many Occupy L.A. protesters arrested during demonstrations in recent months are being offered a unique chance to avoid court trials: pay $355 to a private company for a lesson in free speech.
Los Angeles chief deputy city attorney William Carter said the city won't press charges against protesters who complete the educational program offered by American Justice Associates.
Later, the article explains:
Carter said the free-speech class will save the city money and teach protesters the nuances of the law.
Naturally, part of my surprise was due to the prospect of citizens being required to fork over money to a private business, which would teach them about the limits to their constitutional rights, in exchange for a reprieve from the legal repercussions of civil disobedience. But I was also surprised to find that, contrary to my breezy reading of the headline in which I'd parsed the phrase Occupy protesters as the subject of the verb offered in the active voice, I should have read offered as being in the passive voice, with the protesters being the recipients of free-speech lessons rather than the purveyors of those lessons.
When I showed my son the headline, asking him to imagine what the article was about, he proposed that someone was lecturing a group of Occupy protesters about free speech—in other words, he got the correct interpretation of the headline.
A garden-variety crash blossom, perhaps. But many of the crash blossoms that Language Log plucks from the press really do crash, some quite spectacularly, as in this recent (12/16/2011) example: "Bright sparks weather gala night power cut to party on."
More often than not, among the examples pressed between this blog's pages, something goes haywire in the reading of the headline itself, when it becomes apparent that an initial interpretation simply can't be right—either because it results in an ungrammatical pile of words rear-ending each other, or an absurd event ("Dog helps lightning strike Redruth mayor"; 10/27/2011) But the Calgary Herald headline had just sailed smoothly by, and had I not read the article, it wouldn't have gotten a second glance from me. And when I asked my son if he found the headline confusing in any way, he looked puzzled—his own reading of the headline, though different from mine, had also proceeded without incident and I had to explain the ambiguity in order for him to become aware of it.
Despite the ample opportunities for ambiguities in the English language, we hardly ever notice them unless something goes awry, and this happens surprisingly rarely. Not only that, but there's usually a congenial unanimity over how to interpret forms that, at least in principle, can bend in different directions. This headline appears to be a fine specimen of an ambiguous sentence that is perfectly balanced on a knife's edge of ambiguity, the kind of near-perfect linguistic Necker Cube that you don't see all that often in the wild.
But what is it that guides the parsing of a sentence in such a way that it's almost always interpreted with unanimity? And what's going on with this particular headline that allows it to be read so differently by two different people?
In fact, psycholinguists have painstakingly documented the various factors that contribute to one reading of an ambiguous phrase or sentence being favored over another. All things being equal, some structures seem to be preferred over others, whether due to their syntactic simplicity or their sheer frequency. For example, active readings tend to be preferred over passive ones, when both are possible. But this is mitigated by the specific verb in the sentence—some verbs rarely occur in the passive form, while others do quite often, and readers have implicit expectations based on the syntactic histories of individual verbs. The verb offered happens to occur roughly equally often in both forms. Then there are expectations about the likely thematic roles of particular noun phrases combined with particular verbs—that is, how likely is it that Occupy protesters will be the providers of lessons, rather than the recipients of lessons? Clearly, plausible scenarios can be constructed for both. Compare this headline, for example, to the structurally identical sentence: Homeless families offered shelter over the holidays. Homeless families are hardly in a position to be offering shelter, so one would expect that pretty much everyone would agree on how to read this sentence. In the normal course of parsing an ambiguous string of words, the mind weighs the evidence for each reading and settles on the winner swiftly and without fuss. Your conscious self is rarely aware of the deliberations of your internal parser.
The reason that most crash blossoms come to such a violent end is because the cues that guide the interpretation of the ambiguity tend to converge on a reading that happens to, well, eventually crash. But when it comes to the Calgary Herald headline, most of the evidence considered by the parser turns out to be rather inconclusive. The sentence could readily unfold either way. So why would my son and I be nudged towards very different readings? It may well come down to the wiggle room afforded by contextual expectations—perhaps my son had a slightly higher expectation that protesters would be subjected to lessons in civics from others. Possibly my own expectations were shaped by just having read several end-of-year retrospective articles. If I'd been in the mode of reading pure news reporting, my stylistic expectations might have been set differently. Notice, for example, that in order to get the active reading of the headline, the verb has to be interpreted as being in the past tense, which is a bit unusual for headlinese, where news events tend to be reported in the present tense. Normally the active version of our news headline might read as: Occupy protesters offer lesson in free speech. (For instance, a bit of news reporting on the same page was headlined: Researchers warn about too-loud toys.) But in the context of a retrospective piece of writing about lessons learned over the past year, the past tense would plausibly be used, even in a headline.
These are highly subtle considerations on which to stake an interpretation. Because ambiguities are so rarely poised between two equally viable readings, such subtleties tend not to cast a deciding vote very often, generally being dwarfed by more brutish parsing expectations. But in this case, they may well have played a role in tipping the reading one way for me, and another for my son.
Incidentally, the L.A. Times, where the article originated, has opted for the following entirely unambiguous headline (at least for its online version of the piece):
Some Occupy L.A. protesters may get a lesson in free speech.