[lingtalks] reminder TODAY: Florian Jaeger (Linguistics Colloquium)

[Klinton Bicknell]

I'm happy to announce the first linguistics colloquium of the year will be Florian Jaeger, speaking on "Probabilistic Syntactic Production." Florian is a new postdoc in the Psychology Department here at UCSD. For more information, see his website at http://www.stanford.edu/~tiflo/ .


:: Details ::

Date: Monday 2 October (this coming Monday!)
Time: 2-3:30
Location: AP&M 4301


:: Abstract ::

Probabilistic Syntactic Production
T. Florian Jaeger (Psychology Department, UCSD & ex-farmer)

While there is a considerable amount of evidence that syntactic comprehension makes use of probabilistic information (e.g. Clifton et al., 1984; Tanenhaus et al., 1985; Trueswell et al., 1993; MacDonald, 1994; etc.), comparable evidence for syntactic production is sparse. I present evidence from spontaneous speech showing that syntactic production is rather strongly affected by probabilistic information.

The evidence come from a well-studied case of syntactic reduction, optional that in certain types of complement clauses (1) and relative clauses (2), which can be produced with or without initial that (e.g. Bolinger, 1972; Thompson & Mulac, 1991; Ferreira & Dell, 2000; Race & MacDonald, 2003, Temperley, 2003; Ferreira, 2003, 2005; Roland et al., 2005):

(1a)  I bet [ that I am not doing this for you but only for myself]
(1b)  I bet [       I am not doing this for you but only for myself]

(2a)  I can't believe nobody noticed the mess [ that we made]
(2b)  I can't believe nobody noticed the mess [       we made]

[if I get data on reduction in subject-extracted RCs by Monday, I'll present those, too]

I show that speakers are the more likely to produce the reduced variants (1,2b), the more predictable the complement/relative clause is. This points to a probabilistic optimization of syntactic production (where grammar allows it). I discuss to what extent this optimization is due to production pressure alone or whether it possibly reflects speakers' consideration of comprehenders' difficulties (probabilistic audience design).

The results are also of interest to the ongoing debate whether probabilistic information is part of grammar (e.g. Gahl & Garnsey, 2004, 2006; Newmeyer,
2006): the predictability of a complement clause is known to influence its comprehension ( e.g. Garnsey et al, 1997). Given that I show the same predictability to influence production, the probabilistic information seems to be part of a system that is accessed by both production and comprehension mechanisms. Unless there is evidence to the contrary, it is not clear why this shared system should not be what is commonly referred to as `grammar'.