Evidence For Distributivity Effects in Comprehension

• Nikole D. Patson and Tessa Warren

Intro

• In the current paper, we introduce a new methodology for detecting whether a word in a sentence is conceptually represented as plural and use it to shed light on a debate about whether comprehenders interpret singular indefinite noun phrases within a distributed predicate as plural during on-line reading.
• self-paced reading on a sentence presented in one- and two-word chunk
• indicated that participants were slower to judge that one word was on the screen when the word was plural (e.g., cats) than when it was singular (e.g., cat)
• build different conceptual representations for distributed versus collective predicates, and interpret a singular indefinite noun phrase within a distributed predicate as plura

Results

• When the group was conceptually distributed (5a), participants incorrectly (in American English) used a plural verb (were) more often in their continuations than when the group was conceptually collective (5b). This indicates that participants were more likely to treat the gang as a plural when its individuals were more salient (5a) rather than when the group was the relevant referent (5b). This suggests that distributivity can make grammatically singular lexical items that have plural referents (e.g., gang, group) functionally plural during language production.
• This production and off-line comprehension work suggests that readers build different conceptual representations for collective and distributed predicates, and is consistent with the hypothesis that singular indefinite noun phrases within distributed predicates are often treated as conceptually plural.

Experiment 1

• Experiment 1 was conducted to test whether the Berent et al. (2005) methodology could be extended to sentences. It is possible that the added complexity involved in building and maintaining a sentence representation during the number judgment task, or the task demands of simultaneously carrying out self-paced reading and number judgments, might make participants less sensitive to interference than they were in Berent et al. (2005). Experiment 1 is also important because in order to use the paradigm to test ambiguous cases (as in Experiment 2), we must first establish that the paradigm works on simple, unambiguous sentences.
• The critical measure was the reaction time for the number judgment for correct number judgment trials only. There was a significant main effect of noun type such that ‘one’ responses were slower when the target word was plural
• Experiment 1 confirmed that even in sentential contexts, semantic plural information on a word interferes with singular number judgments.
• Specifically, if the distributing quantifier takes wide scope over the indefinite, and comprehenders build conceptual representations for distributed predicates that contain multiple exemplars of the referent introduced by the singular indefinite noun phrase, then one-word judgment times should be slower for indefinite noun phrases in distributed predicates than collective predicates.
• The experiment had a 2 × 2 within-participants design. The first factor was the quantifier type and was either distributed (a) or collective (b)
• They were asked to rate on a scale of 1 – 5 (where 1 was ‘definitely one’ and 5 was ‘definitely more than one’) whether the last word in the sentence referred to one or more than one object. Results indicated that the singular-marked distributed items were indeed biased toward a plural interpretation of the noun phrase. Participants rated the singular-marked distributed items as being closer to the ‘definitely more than one’ end of the scale
• than the singular-marked collective items

Experiment 2

• The results of Experiment 2 confirm the hypothesis that singular indefinite noun phrases in distributed predicates can indeed be treated as conceptually plural during reading
• There was no reliable effect of distributivity and no reliable difference between the plural-marked conditions, indicating that the difference in the singular-marked conditions was unlikely to be the result of one kind of predicate being more costly to compute than the other. These results indicate that in these items the distributing quantifier took wide scope over the indefinite
• indicate that conceptual plurality interferes with number judgments during sentence comprehension
• These findings (Filik et al., 2004; Paterson et al., 2008) indicate that comprehenders do not build conceptually plural referents on-line for indefinite noun phrases in distributed structures that off-line norming had indicated were likely to be interpreted as plural.