Semantic evaluation of syntactic structure

Lyn Frazier, Anne Cook, Maria Nella Carminati, and Keith Rayner
lyn@linguist.umass.edu
University of Massachusetts

Many studies have shown that perceivers initially take a postverbal NP to be a direct object, resulting in difficulty in 'early closure' sentences like (1a), especially if the postverbal NP is long and a plausible object for the verb (Frazier and Rayner, 1982, Pickering et al, 2000 and references therein)

(1)

a.

As John hunted the rabbit escaped through the woods

b.

As John hunted the rabbit it escaped through the woods.

In an eye movement study, sentences like those in (1) were compared to sentences where the subordinate clause verb phrase could not readily be interpreted as bounded because it appeared in the progressive, which denotes an ongoing activity.

(2)

a.

As John was hunting the rabbit escaped through the woods.

b.

As John was hunting the rabbit it escaped through the woods.

It was hypothesized that the perfective form in (1) semantically reinforces the late closure error in (1a): the predicate is more readily interpreted as a completed event when the predicate is bounded. The direct object serves to bound the predicate. Consequently it was expected that the difference between (2a) and (2b) in the disambiguating region would be smaller than the difference between (1a) and (1b). In other words, in (1a) the direct object analysis is semantically confirmed resulting in longer reanalysis times than in (2a) where the activity predicate is merely consistent with a direct object but does not particularly prefer to be bounded. Eye movement data supported the prediction.

The bounded predicate hypothesis makes the further prediction that the late closure error in (3a), relative to (3b), should be smaller than in (1a), relative to (1b).

(2)

a.

As John hunted frightened rabbits escaped through the woods.

b.

As John hunted frightened rabbits they escaped through the woods.

The bare plural object in (3) will not semantically confirm the direct object analysis of the initial clause because bare plurals do not readily bound the predicate. Preliminary results from a second eye movement study confirm the prediction.

In principle, one might try to model the results in an experience-based model in terms of decreasing the transitivity bias of each verb in the language when it occurs in the progressive or when it co-occurs with a bare plural NP. But this approach would fail to capture the systematicity of the phenomenon and would lead to incorrect expectations about where in the eye movement record the effects appear.