The Nominal Analysis of Children's Interpretations of Adjunct Pro Clauses

The hypothesis that children’s errors in interpreting adjunct PRO clauses are due to their use of a nominal structure was tested in two act-out experiments. Four- to six-year-old English-speaking children permitted a by phrase inside an adjunct clause containing an intransitive verb to be interpreted as agentive, and they gave such an interpretation for a construction for which a nominal analysis is not permitted in the adult grammar (when PRO clauses). This behavior can be accounted for if children have knowledge of general principles governing the interpretation of nominal and PRO constructions, and use a nominal analysis in interpreting adjunct clauses. In the second experiment, children distinguish between constructions that are unambiguously nominal and those that are ambiguous between a nominal and PRO structure, permitting agentive readings of by more frequently in the former case. This argues that adjunct PRO may be acquired by some children at a point during the preschool years. Overall the results fit a view of acquisition in which the language learner actively analyzes the input data, using knowledge of general grammatical principles, and is not narrowly bound by his or her current knowledge of the lexicon of the language.*

Introduction

In the adult grammar of English, the missing subject of a temporal adjunct (adverbial) clause is interpreted as coreferential with the main clause subject. Thus in each of the sentences in 1–3 the embedded subject, designated PRO in the manner of Chomsky 1981, is interpreted as Fred.

(1) Fred_i hugged Sue before PRO_i leaving the room.

(2) Fred_i was hugged by Sue before PRO_i leaving the room.

(3) Fred_i ran past Sue before PRO_i leaving the stadium.

The rule for interpreting the PRO subject of an adjunct clause follows from a simple condition on the interpretation of missing subjects that are obligatorily interpreted as referring to a noun phrase in a superordinate clause (missing subjects that are obligatorily controlled): the PRO subject must be c-commanded by the entity it refers to.¹ Adjunct clauses such as those in 1–3 are structurally attached outside the verb phrase; thus only the main clause subject c-commands the PRO subject of an adjunct clause (4)² and control by a direct object or the object of a preposition in the main clause VP is blocked.

(4) [_IP Fred [_IV hugged Sue] [_PP before [_IP PRO leaving the room]]]

Despite the fact that the obligatory control of the PRO subject of an adjunct clause by the superordinate subject is a fact of adult grammar that receives a simple explanation in linguistic analysis, numerous experiments using comprehension and judgment tasks³ have shown that children may be six years or older before they reliably follow the rule of making adjunct PRO refer to the subject of the main clause. The goal of this article is to compare competing explanations of children’s errors in interpreting adjunct PRO.

1. Children’s errors with adjunct PRO

The following patterns of performance have been observed in experiments testing children’s interpretation of adjunct PRO:

1. a. control by the main clause subject

2. b. control by the main clause (direct or prepositional) object

3. c. control by the subject or direct object

4. d. control by the agent of the main clause

5. e. free interpretation of the subordinate subject.

Pattern a is an adult response, but may arise as the result of a primitive interpretive strategy that links the subject of an embedded clause to the main clause subject, resulting in correct responses for sentences such as 1–3, but leading to error in interpreting, for example, the relative clause in sentences such as 5.

(5) The horse kicked the cow that jumped the fence.

The use of such a response strategy appears to be quite rare (for example, accounting for the behavior of only two out of sixty children in an act-out study by Hsu et al. 1985).

Pattern b was reported for some children in Hsu et al. 1985, and in act-out and judgment experiments reported in McDaniel et al. 1990...