An interesting piece of data came up the other day worth sharing. I’m trained to notice funny constructions, as are we all, like the time someone close to me said the sentence in (1):

(1) These bottles drink better.
(context: longnecks, as opposed to stubbies).

Usually discussions like this can’t go on phonoloblog, because they’re not phonological, like the funny coreference I heard from a broadcaster in (2):

(2) Nobody draws more comparisons to his_i father than Brett Hull_i.

Neither of these was spoken by a linguist, which is important, because sometimes linguists discuss certain constructions which are impossible (or “at best, marginal”) but by virtue of having discussed them suddenly start generating them. For example, lately I’ve seemed like I’m lots of things. Then there’s the matter of all those clauses that people don’t like it when I extract out of. I hear these things from other people too, non-linguists, but as you know, sometimes it’s a performance issue.

It’s all syntax too, usually not appropriate for phonoloblog, but don’t worry phonoloblokes, a good one came up the other day involving a syntactic claim that depends on phonological evidence.

Now, “want + to” and “going + to” can contract to “wanna” and “gonna” under certain conditions. The evidence for contraction is phonological: loss of the /t/ and reduction of the vowel in “to”. There’s a claim sometimes made that such contraction can occur only if the “to” is an infinitive particle of a particular kind of VP; I’m not sure who made the original observation, but several sources are cited below. What you call this VP depends on your framework, but in as neutral terms as I can imagine, it’s one whose implicit subject is the same as the subject of the matrix verb (i.e., the “want” or “go”).

Let me try to rephrase: first, if the “to” is a preposition, the claim is it cannot contract:

(3) *I’m gonna the store.

Second, if the “to” is an infinitive particle, the claim is that it cannot reduce if its subject is a wh-trace. I provide a set of uncontracted and contracted constructions in (4). In (4a-d), the empty subject of “win”, the lower verb, is the same as the overt subject of “want”, so contraction is allowed in both the declarative (4b) and wh-raised (4d) versions. But in (4e-h), “win” and “want” have different subjects. While contraction is uninterestingly impossible in the declarative (4f), the sticking point is the ungrammatical claim for the wh-construction (4h).

(4) a. I_i want e_i to win the game
b. I_i wan_ina win the game.
c. What_j do you_i want e_i to win t_j?
d. What_i do you wanna win t_i?

e. I want Jane to win the game.
f. no contraction of (4e) is possible
g. Who_i do you want t_i to win the game?
h. *Who_i do you wan_ina win the game?

I did a little googlage, using the search term [gonna empty category reduction VP complement], and the second hit was a posting of Radford (2000), in which these claims are made explicit, and in which some other references are cited, including Boskovic (1997) and Jaeggli (1980) among others, all dealing with (among other things) the ungrammaticality of sentences like (4h).

Radford’s data include the following:

(2) b. Who do you think’s been lying?

(3) a. Who do you wanna beat?
b. *Who do you wanna win the race?

The claim of ungrammaticality for (4h) and Radford’s (3b) is an arguing point about what kinds of empty categories can allow contraction, and what kinds of stipulations different theories need to make to account for the facts. Contraction is thus phonological evidence for a syntactic claim (a rarity, no?).

Months ago I heard a counterexample to the PP claim for (3). In an answer to “what are you doing for lunch?”, the reply was the sentence in (5). (“Chipotle” in this case is a chain of burrito restaurants.)

(5) I’m gonna Chipotle.

The person who noticed it remarked on how “Chipotle” had been verbed, but I remarked that perhaps the structure was prepositional. That is, it was contracted to the same structure in (3), but with “Chipotle” instead of “the store”. Impossible – you can’t do that! In this case the debate can’t be settled, since either parse is plausible. To boot, the utterer was a linguist who either liked to verb things or reduce PPs in “impossible” ways for the sake of irony.

Flash forward to the season finale of The Apprentice, a so-so reality show in which contestants vie to be a yes-man for Donald Trump. In an eye-splitting three hours we review the evidence to choose between Jennifer, the unflappable and attractive double Ivy League graduate, and (boy) Kelly, the robotic conflict resolver with an alternating past of West Point decorum and private enterprise entrepreneurialism. Cripes, no one knows what to do, so Donald goes through a painfully staged audience survey to elicit a nearly unanimous decision in Kelly’s favour. Anyway, a viewer watching this with me says the sentence in (6):

(6) Who do you wanna win, Kelly?

This was uttered with an intonation you might expect from “Where do you wanna go, Applebee’s?”. But in this case, there is a trace subject of “win” (actually, of “na win”), which is problematic. There is no alternative structure we could propose that would preclude t-contraction (e.g., where “who” has a trace as the complement of “win” – that’s not what was asked). Also, it was not produced by a linguist, so no chance of grammatical irony.

So what do you do? I don’t mean to pick on this group in particular, because debatable claims of syntactic grammaticality show up in other places, with other protagonists. Shoot, even phonologists do it. It just worries me when a claim of grammaticality – a premise to an argument – is false.

The internet could be a great corpus to look for funny sentence structures like (5) and (6), because even if they’re not ungrammatical, it’d be nice to get a feel for their frequency. Unfortunately, the lower-than-natural rate of contraction in text, the typo risks, and the linear search terms all reduce our ability to find other examples. What would be nice is a crawler that projects trees (including empty categories! ) for all the pages it reads. Boy, would that ever be nice.