A Matter of Ambiguity? Using Eye Movements to Examine Collective Vs. Distributive Interpretations of Plural Sets
Christine Boylan, Dimka Atanassov, Florian Schwarz, John Trueswell
Previous Work
Frazier et al. (1999) used eye-tracking reading times to compare processing loads of sentences that were explicitly distributive (involving the adverb each), explicitly collective (involving the adverb together), and locally indeterminate at the predicate.
Having found evidence for increased processing load associated with distributive sentences, they concluded that the processor initially pursues a collective reading, and thus the distributive / collective distinction was one of ambiguity and not vagueness.
However, an increased processing load for the distributive reading might be expected regardless of whether the underlying representation is vague or ambiguous
processor must stipulate a distributive operator D (the spell-out of which is each) to interpret a distributive meaning, which may incur processing delays
Moreover, increased reading times at the point of disambiguation preclude conclusions about exactly when listeners may have committed to a distributive reading during the processing the underdetermined predicate.
Introduction
eye movements of listeners were recorded to investigate the representation of collective vs. distributive interpretations of plural subjects in light of the Minimal Semantic Commitment (MSC) hypothesis
The crucial difference between these two proposed representation types is that an ambiguous representation forces a decision about an interpretation, while a vague representation tolerates unspecified features.
Given the prediction that an ambiguous item will prompt the processor to converge on one interpretation even in the absence of disambiguating information, we tested whether sentences underdetermined for collectivity / distributivity would nonetheless cause listeners to converge on a single interpretation.
Rather than relying on processing times to infer representational commitments, we employed the visual world paradigm to track which representations subjects considered over the course of hearing a sentence
Method
The eye movements of 24 participants were recorded as they listened to explicit/indeterminate collective/distributive sentences while they considered collective and distributive acts depicted on a computer screen
An earlier switch in gaze to one of the two images would indicate a processing preference for one interpretation over the other. Results and Discussion
Explicitly collective sentences prompted looks to the collective scenario at the point of disambiguation (i.e. together) and explicitly distributive sentences (using each) prompted looks to the distributive scenario (Fig. 2a)
Crucially, however, the indeterminate, nulldisambiguator sentences patterned along the same trajectory as together sentences: the predicate alone prompted looks to the collective, prior to hearing the final word of the sentence
We also compared the proportion of looks averaged across two time windows: an 800- ms interval before the onset of the predicate and an 800-ms time window following the predicate onset (Fig. 2b)
Since the distinction between ambiguity and vagueness lies at the interface of semantics and pragmatics, it is particularly important that we find a psychometric realization of the difference between these two types of representations.
Here we presented a method by which to investigate the time courses of these representations as they relate to distributivity, and we suggest this method may further contribute to the study of the semantics-pragmatics interface at large.
Results
In an ANOVA of proportion of looks to collective and distributive scenes, we found significant interactions between disambiguator and time window.
In a targeted analysis of disambiguator effects in each time window, we found significant differences between together and each sentences and the null and each sentences after the predicate onset but not before
Moreover, the together sentences did not significantly differ from the null form.
Thus, despite a lack of explicit disambiguating information, the indeterminate, nulldisambiguator sentences nonetheless prompted looks to the collective scenario, which was reliably different from the time course of distributive-directed each sentences.
This provides evidence that the listener has committed to the collective interpretation in the absence of disambiguating information.
This is consistent with a theory that treats the collective / distributive distinction as ambiguous rather than vague.
The results also indicate that this processing commitment is essentially immediate; i.e., as soon as listeners begin hearing the ambiguous predicate, they show a preference for the collective interpretation.
