Testing the perception of a "categorical" rule:
Wanna experiment in syntax
This paper explores a new field of application for sociolinguistic methodology:
the empirical testing of aspects of linguistic theory. The notion of a
"grammatical utterance" within theoretical linguistics is
traditionally based upon the non-experimental methodology of gathering data
through introspection. While this approach was a necessary first step in
developing linguistic theory, a social science must develop an objective means
of gathering data. Labov (1975) convincingly demonstrated that introspective
judgments on grammaticality do not accurately reflect actual usage of language
in the speech community. In order to improve the objective accuracy of
grammaticality judgments, we developed a methodology which does not rely upon
intuition, but rather is grounded in empirically observable fact.
To demonstrate the methodology, we have selected one well-known phenomenon:
English wanna-contraction. Chomsky (1981:180-82) states that the
phonological reduction of want to to wanna is constrained
syntactically, as illustrated in (1-2):
- (1) They want [PRO to visit Paris]
- (1') They wanna visit Paris
- (2) Whoi do they want [ [ti] to visit Paris]?
- (2') *Who do they wanna visit Paris?
According to GB syntax, (2') is categorically prohibited because there is
illicit contraction over an NP-trace. While (2') intuitively seems
ungrammatical, there is no extant methodology with which to empirically support
this intuition.
To address this issue, we constructed and carried out an experiment to test
the GB claim. Our experiment is a perception task, where an informant's response
to a sentence indirectly but unambiguously indicates whether a sentence like
(2') is grammatical for that speaker. The task consists of listening to a short
story and answering the following question (posed with the contracted form to
half the informants and with the uncontracted form to the other half):
(3) Who do you {want to / wanna} help?
The context of the story allows for either an object-extraction
interpretation (as in (4)) or a subject-extraction interpretation (as in (5)):
(4) Object Which onei do you want to help ti ?
i.e., I want to help X.
(5) Subject Which onei do you want ti to help?
i.e., I want X to help.
We find the following distribution of responses:
| Question |
|
Object interpretation (a) |
|
|
Subject interpretation (b) |
Total N |
| want to |
8 |
20% |
< |
32 |
80% |
40 |
| wanna |
51 |
65% |
> |
27 |
35% |
78 |
chi-square = 20.01, p < .001
Our experimental methodology has allowed us to empirically support the GB
claim: there was a significant shift toward the object-extraction interpretation
when the stimulus question contained the contracted form. However, our results
also reveal that the prohibition against contracting over a trace is not
categorical: 35% of the responses to the contracted form did maintain the
subject-extraction interpretation (which is prohibited in GB syntax). This
finding suggests that syntactic rules may not be as categorical as the theory
assumes. Similar methodology should be applied to determine what other parts of
the theory are empirically supportable and which require revision, as well as
where variable rules best describe the language.
Return to list of Naomi's research
interests.
| 