Transparency Effects with Infinitive/Subjunctive Complements
It has been observed that infinitive/subjunctive clauses pattern alike regarding binding phenomena. In particular, infinitive/subjunctive complements are typically more transparent, showing clause-mate effects like anaphor binding and obviation. Here, we will look at another piece of evidence for the covert restructuring analysis: it can also account for transparency effects with infinitive/subjunctive complements.
Let us first look at anaphor-binding facts. There are languages in which an anaphor can refer to an antecedent across an infinitive or subjunctive clause, but not across an indicative clause. Icelandic is the best-studied language of this type (see, e.g., Anderson 1986; Manzini and Wexler 1987; Koster 1987; Manzini 2000).
(1)
As shown in (1b–c), when the reflexive pronoun sig is embedded in an infinitive or subjunctive clause, it can be bound by the matrix subject. However, as shown in (1a), when embedded in an indicative clause, it cannot be.
Let us next look at obviation facts. In languages like Catalan, Icelandic, Italian, and Spanish, the subject of a subjunctive clause is disjoint in reference from the matrix subject (see, e.g., Picallo 1985; Kempchinsky 1986; Manzini 2000). The following examples are from Catalan:
(2)
As shown in (34a), the coreference reading between the embedded and matrix subjects is possible when the embedded clause is indicative. When the embedded clause is subjunctive, however, the coreference reading is not possible, as shown in (2b).
The anaphor-binding and obviation facts show that infinitive/subjunctive clauses are more transparent than indicative clauses, though there are parametric variations among languages with respect to binding domains and the existence of long-distance reflexives that determine whether or not the transparency effects are observed in a given language (see, e.g., Manzini and Wexler 1987 for detailed discussion).
The covert restructuring analysis accounts for these transparency effects. Let us assume that TP constitutes a binding domain for reflexives in languages like Icelandic and pronouns in languages like Catalan. Let us also assume with Chomsky (1995), among others, that binding conditions apply at LF. Then, in the Icelandic example (1a), the antecedent Jo´n ‘John’ is outside the minimal TP that contains the reflexive pronoun sig ‘himself’; anaphor binding is not possible. In the Catalan example (2a), the antecedent il ‘he’, the matrix subject, is outside the minimal TP that contains the embedded subject pronoun il ‘he’; there is no obviation effect.
Let us next consider the Icelandic examples (1b–c) and the Catalan example (2b). Under the covert restructuring analysis, since the complement clause is infinitive or subjunctive in (1b–c) and (2b), the embedded AspP undergoes covert phrasal movement to the Spec of the matrix v. The relevant structures of (1b) and (2b) are represented in (3) and (4), respectively.
(3)
(4)
In (3), the antecedent Jo´n ‘John’ is within the minimal TP that contains the reflexive pronoun sig ‘himself’ in the covertly moved AspP in the Spec of the matrix v. Anaphor binding is possible, and hence (1b) is accept able. (1c) can be accounted for in the same way. In (4), since the ante cedent il ‘he’, the matrix subject, is within the minimal TP that contains the embedded subject pronoun il ‘he’ in the covertly moved AspP, they cannot be coreferential with each other. Hence, (2b) is deviant.
The covert restructuring analysis receives further support from a subject/nonsubject asymmetry with respect to anaphor-binding and obviation facts. Let us first look at anaphor binding. As shown in (5), in Icelandic, a reflexive in the subjunctive complement clause may not be interpreted as coreferential with a matrix nonsubject.
(5)
Note that the Icelandic reflexive is not subject oriented, as exemplified by (6).
(6)
Hence, the unacceptability of (5) is not due to any non-subject-oriented property of the Icelandic reflexive. Similar facts are found with obviation. In Catalan, an embedded subject and a matrix nonsubject can corefer, as shown in (7).
(7)
Under the covert restructuring analysis, during the derivation of (5), the embedded AspP, which contains the reflexive se´r ‘himself’, moves covertly to the Spec of the matrix v.
(8)
Since the matrix indirect object Jo´n ‘John’ is within the matrix VP, it does not c-command the reflexive sig ‘himself’ within the covertly moved AspP, which resides in the Spec of the matrix v. Hence, anaphor binding is not possible. Similarly, in (7), the indirect object Pere ‘Peter’, which is within the matrix VP, does not c-command the empty pronoun pro within the covertly moved AspP, which is in the Spec of the matrix v. Hence, the obviation effect is not observed.
