Maribel Romero: Mood Selection in Romance Complememt Clauses

Attittude verbs in Spanish and, more generally, in Romance, select for a specific mood in their complement clause. With only a few exceptions in some Romance languages, the classification goes as follows. Verbs selecting indicative mood include doxastic verbs ('think', 'believe', 'know'), verbs of communication (‘say, ‘announce’), verbs of certainty (‘be certain that…’), commitment (‘promise’), fiction (‘dream’), mental judgment (‘guess’, ‘understand’) and perception verbs (‘see that…’, ‘hear that…’). This is illustrated in (1), where a positive sentence with ‘believe’ / ‘know’ is grammatical if its complement clause appears in the indicative and ungrammatical if it appears in the subjunctive (Villalta 2008). Verbs selecting subjunctive mood are desire verbs (‘want’, ‘prefer’, ‘wish’), factive-emotive verbs (‘regret’, ‘be glad that…’), modal verbs (‘be possible’, ‘be necessary’), verbs of doubt (‘to doubt’), directive verbs (‘order’, advise’) and causative verbs (‘to make somebody do something’, ‘to manage’). This is illustrated in (2), where a subjunctive complement clause is grammatical and an indicative complement clause leads to ungrammaticality (Villalta 2008).

Researchers have tried to give a semantic analysis of mood selection in Romance, taking as the critical factor the intrinsic semantic characterization of the embedding verbs. The traditional idea relates mood selection in complement clauses to mood choice in conditionals: indicative is the “realis” mood and thus signals that we are quantifying over worlds close to the actual one (cf. the indicative conditional in (3a)), whereas subjunctive is the “irrealis” mood and signals that we are quantifying over worlds that are already known to be impossible or at least are more distant from the actual one (cf. the subjunctive conditional in (3b)). Several problems have been raised against this idea, and a substantial body of literature has produced new semantic analyses in the recent years, such as Farkas (1992), Quer (1998, 2009), Giannakidou (1995), Giorgi and Pianesi (1997), and Villalta (2008).

The goal of the present paper is, first, to present problems remaining in some of the most recent analyses, namely in Giorgi and Pianesi (1997) and Villalta (2008), and, second, to tentatively advance a way of implementing the traditional idea that circumvents these new problems.