Lorna Morrow, Patrick Sturt, Anthony Sanford & Linda Moxey
University of Glasgow
In this research, we use anaphor resolution to explore the mental representation of singular and plural entities during language processing.
Albrecht & Clifton (1998) reported processing difficulty when a singular pronoun refers to an antecedent embedded inside a conjoined noun phrase (as in 1):
|
(1) |
John and Mary painted the room. He really liked the colour. |
Albrecht & Clifton offer a syntax-based account: the conjunction cost is due to the difficulty of splitting apart the conjoined constituent at a syntactic level, in order to access one of the conjuncts. According to an alternative scenario-based account (Sanford & Moxey, 1995), the conjunction cost arises in (1) because of the difficulty of individuating the discourse roles of John and Mary in the scenario. Notice that both of these accounts predict difficulty for singular pronouns where the intended referent is individuated in some way from an alternative (either in syntax or in discourse). These accounts may also be extended to make the opposite prediction for plural pronouns, where processing should be facilitated when two intended referents are not individuated.
To explore these issues, participants' eye-movements were monitored while they read short passages like (2):
|
(2) |
a. |
John and Mary painted the room. He/They really liked the colour. |
|
|
b. |
John painted the room with Mary. He/They really liked the colour. |
|
|
c. |
John painted the room for Mary. He/They really liked the colour. |
In (2a), John and Mary appear in a single conjoined constituent, while in (2b) and (2c) they are split into two constituents. Thus, the syntax-based account predicts that a singular pronoun should be hard to process in (2a), but easier in (2b) and (2c).
In (2a) and (2b), John and Mary are both "agents" of painting (and are thus similar in their discourse roles), while in (2c), John and Mary have different roles. Thus, the scenario-based account predicts difficulty for singular reference in (2a) and (2b) in comparison with (2c). The reverse predictions hold for plural pronouns for both models.
The pattern of results suggests that both accounts are correct; for the singular conditions, first-pass reading times in the adverb region ("really") were longer for (2a) than (2b), and longer for (2b) than (2c) (though only the contrast between (2a) and (2c) reached significance). For the plural conditions, first-pass reading times were longer in the final region ("the colour") for (2c) than (2b), and longer for (2b) than (2a) (though again, only the contrast between (2a) and (2c) reached significance). Similar results were found for regression-path and right-bounded reading times. The difference in the location of this effect (adverb vs. end of sentence) suggests that plural reference is delayed compared with singular reference.
A follow-up experiment is currently being run using pronouns in object position.