semantics and pragmatics of to-contraction

Dick Hudson dick at LINGUISTICS.UCL.AC.UK
Fri Aug 3 18:42:33 UTC 2001

Dear Dan,
An even better account, in my opinion, refers to frequency. As Joan Bybee
has shown (in Phonology in the Lexicon, in Usage-based Models of Language,
ed. Michael Barlow and Suzanne Kemmer), frequently used forms tend to
shorten. This would explain why, e.g. "I('ve) got to ..." can shorten to
gotta, in contrast with the presumably less frequent "I got to ...". The
attraction of this explanation is that it avoids stipulating arbitrary
connections such as the one you suggest. But of course it may be wrong, in
which case its explanatory power dwindles to zero.

At 08:40 03/08/2001 -0500, you wrote:
>Comments on the following would be much appreciated.
>In the early years of trace theory it was claimed that traces received
>strong support from contraction facts. Recall that Chomsky proposed in 1976
>that traces
>are left behind by moved elements, as in (1):
>(1) Mortimeri was seen ti by Bill. (Mortimer moves from the object position
>to subject position in the passive, leaving behind the trace, ti.)
>It was then noticed that although contraction is generally allowed between
>verbs and infinitival to
>(and this is just part of a more general phenomenon: the cliticization of
>function words in English and other languages, a quite common process), as
>in (2) and (3), it is not allowed if a trace intervenes, as in (4) and (5):
>(2) a. Suzy promised [PRO to marry General Thade]. (PRO = understood subject
>of infinitive in Government and Binding and other theories)
> b. Suzy promistta marry General Thade.
>(3) a. I want to go to town.
> b. I wanna go to town.
>(4) a. Whoi do you want PRO to see ti?
> b. Whoi do you wanna see ti? (Contraction is OK across the PRO)
>(5) a. Whoi do you want ti to see Bill?
> b. *Whoi do you wanna see Bill?
> (5b) is ungrammatical according to this analysis because there can be no
>contraction across a trace. (Traces are unlike PRO in being marked by
>abstract, syntactic Case and being governed.)
>Pretty neat result. I certainly thought it was when I first saw it. But
>other linguists (especially Geoffrey Pullum and Paul Postal, the two
>principal critics of this proposal) were not so impressed. They brought up
>numerous facts which seemed problematic for the account. I
>considered the issue unresolved and hadn't really thought about it much in
>years. However, Keren Everett in unrelated work on the phonology of function
>words in the Carnegie Mellon University automatic speech recognition
>project, SPHINX, noted the following contrast, which made me rethink the
>(6) a. I got PRO to go. (meaning I have an obligation to go - notice that it
>isn't even clear this sounds right in the uncontracted form.)
> b. I gotta go. (Sounds much better than (6a) to me.)
>(7) a. I get PRO to go. (Indicating privilege.)
> b. *I getta go. (Sounds terrible to me and may be unattested in the CMU
>data base.)
>Chomsky's theory fails to predict the contrast between (6) and (7) because
>no traces are involved in either example.
>Note, too, the following additional contrast pointed out to me by George
>Lakoff (email August 02, 2001):
>(8) a. I got PRO to go! (Past tense of get, indicating privilege, meaning,
>for example, my boss let me go.)
> b. *I gotta go. (The contracted form can only indicate obligation, not
>So, we have the following contrasts:
>(9) a. I wanna go.
> b. *I getta go.
>(10) a. I gotta go. (obligation)
> b. *I gotta go. (privilege)
>Also, compare the following pairs (pointed out to me by Paul Postal email of
>(11) a. I ought PRO to go.
> b. I oughtta go.
>(12) a. I fought PRO to go.
> b. *I foughtta go.
>The trace theory of movement has nothing to say about the examples in
>(9)-(12). Therefore, it is too weak in this respect - it cannot handle all
>the facts. At the very least, we will need a separate account of other kinds
>of contraction.
>Let me tentatively propose a semantic account. Modal verbs involving
>intention of the matrix subject may contract (form a single unit with) their
>infinitival. Otherwise contraction is not allowed.
>Now, how might we account
>for the original contrast in (4) and (5)? In Role and Reference Grammar,
>WH-questions are only possible of focused material in English (probably
>universal). Objects (i.e. immediately postverbal arguments, since RRG does
>not recognize grammatical relations) are the unmarked focus position. For a
>subject to be questioned, however, it must be focused as well, but will
>require a marked structure. This 'marking' is indicated by the phonology, at
>least in English. Now reconsider (4) and (5), repeated here as (13) and
>(14), but without traces or PRO:
>(13) a. Who do you want [to see]?
> b. Who do you wanna see?
>No special marking is needed for object questioning because of unmarked
>English focus structure.
>(14) a. Who do you want [to see Bill]?
> b. *Who do you wan[na see Bill]?
>In (14), however, RRG would disallow the contraction because it eliminates
>the phonological salience indicating marked focus, a condition on
>questioning subjects in RRG, as observed above.
>Therefore, the RRG (semantic/pragmatic) analysis seems superior to the
>syntactic analysis.
>Dan Everett

Richard (= Dick) Hudson

Phonetics and Linguistics, University College London,
Gower Street, London WC1E  6BT.
+44(0)20 7679 3152; fax +44(0)20 7383 4108;

More information about the Funknet mailing list