Contrasting the First-Mention and Subject-Preference Accounts

A visual-world eye-tracking experiment investigated the influence of order of mention and grammatical role on resolution of ambiguous pronouns in Finnish. According to the first-mention account, general cognitive structure-building processes make the firstmentioned noun phrase the preferred antecedent of an ambiguous pronoun. According to the subject-preference account, the preferred antecedent is the grammatical subject of the preceding clause or sentence. Participants listened to sentences in either subject-verb-object or object-verb-subject order; each was followed by a sentence containing an ambiguous pronoun that referred to either the subject or the object. Participants’ eye movements were monitored while they looked at pictures representing the two possible antecedents of each pronoun. Analyses of the fixations on the pictures showed that listeners used both order-of-mention and grammatical-role information to resolve ambiguous pronouns. An essential feature of understanding texts is the ability to resolve co-reference relations. However, such relations, especially between pronouns and their antecedents, are often ambiguous. For example, when hearing the following pair of sentences, how do listeners determine to which person the pronoun he refers? (1) Tony Blair shook hands with George Bush in the White House. He wanted to discuss the situation in Iraq. Various factors have been shown to affect pronoun resolution at some stage during the comprehension process (e.g., Garnham, 2001). Considerable attention has been paid to heuristic strategies, especially to whether co-reference resolution is guided by language-independent, general cognitive processes or by specific linguistic factors. On the one hand, Gernsbacher (Gernsbacher & Hargreaves, 1988; Gernsbacher, Hargreaves, & Beeman, 1989) has proposed that the first-mentioned noun phrase in a sentence has a privileged status over other potential antecedents. According to the structure-building framework, the first-mentioned entity forms a foundation onto which further information is mapped when the mental representation of a sentence is built. As a result, the first-mentioned entity is the preferred antecedent of an ambiguous pronoun. The first-mention advantage is attributed to general cognitive processes that are independent of linguistic factors, such as the antecedent’s grammatical role. Evidence for the first-mention preference comes mainly from probe recognition experiments. Gernsbacher and Hargreaves (1988) investigated sentences containing two noun phrases and showed that probe recognition was faster when the probe word was the first-mentioned noun phrase. Moreover, Gernsbacher and Hargreaves (1988) and Carreiras, Gernsbacher, and Villa (1995) observed this advantage even when the first-mentioned noun phrase was not a grammatical subject. However, neither study investigated the processing of pronouns. Furthermore, Gordon, Hendrick, and Foster (2000) have argued that probe recognition reflects the use of special strategies specific to probe recognition rather than processes of language comprehension. Thus, it is uncertain whether the observed first-mention advantage generalizes to on-line pronoun resolution. On the other hand, preferences in pronoun resolution have been attributed to specific linguistic factors, especially to grammatical-role information. First, the subject-preference account claims that the preferred antecedent of an ambiguous pronoun is the grammatical subject of the preceding clause (Crawley, Stevenson, & Kleinman, 1990; Frederiksen, 1981). Frederiksen found that reading times for sentences beginning with a pronoun were faster when the pronoun referred to the subject of the preceding sentence rather than the object. Crawley et al. observed a similar preference. However, in both Address correspondence to Juhani Järvikivi, Department of Psychology, University of Turku, Assistentinkatu 7, FIN-20014 Turku, Finland; e-mail: [email protected]. PSYCHOLOGICAL SCIENCE 260 Volume 16—Number 4 Copyright r 2005 American Psychological Society studies, the subject was also the first noun phrase in the clause (as is usually the case in English), so the preference may also have been due to a first-mention advantage. Second, the proponents of the parallelism account argue that a pronoun is preferentially interpreted as co-referent with a noun phrase that has the same grammatical role (e.g., Sheldon, 1974; Smyth, 1994). Smyth (1994); Stevenson, Nelson, and Stenning (1995); and Chambers and Smyth (1998) observed that readers preferred antecedents that had the same grammatical role as the pronoun. The preferred antecedent for an object pronoun was the (second-mentioned) object in the preceding clause, a finding that cast doubt on the generality of the firstmention account. However, in all these studies, the clauses containing the pronoun and the potential antecedents were semantically almost identical, raising the possibility that semantic factors contributed to the preferences. In a language like English, it is very difficult to distinguish among the accounts—and, in particular, between the firstmention and subject-preference accounts—because the grammatical subject is usually also the first-mentioned referent. Thus, all of the accounts we have mentioned predict that Blair is the preferred antecedent in example (1). Therefore, in the present study, we used Finnish, a language with free word order and a gender-neutral third-person singular pronoun, hän (‘‘he/ she’’). In Finnish, grammatical roles are indicated through morphosyntactic marking, with the subject typically in nominative and the object in partitive, as in example (2): (2) Tony Blair kätteli George Bushia valkoisessa talossa. Hän halusi keskustella Irakin tilanteesta. Tony Blair (subject) shook hands with George Bush (object) in the White House. He wanted to discuss the situation in Iraq. In this sentence, the subject noun phrase, Tony Blair, is in the nominative singular case, and the object noun phrase, George Bush, is in the partitive singular, indicated by the suffix -ia. Kätteli is a third-person singular past-tense form of the verb kätellä (‘‘to shake hands with’’), which, unlike in English, is a simplex verb taking a direct (partitive) object. In (2), the word order of the first sentence is subject-verbobject (SVO). However, in Finnish, the word order can be reversed (OVS) without any change in the inflectional marking, as