Clausal backgrounding and pronominal reference:

A functionalist approach to c-command

C. Harris and E. Bates

2002

Pronouns normally follow their referents (e.g., As Joel walked to school, he thought how nice it would be if he could use a jetpack.). Occasionally however, the pronoun will precede its referent (e.g. Because he walked up the steep hill, Joel was tired when he got to school.). Taken together, examples such as these have lead linguists to posit that syntactic structure determines coreference patterns (the noun that a pronoun can refer to). That is, once a sentence has been diagramed, a pronoun cannot refer to a noun that is under it in the diagram (i.e. if it is c-commanded(?)). C-command has been studied extensively and many predictions from this assumption have been shown to be true experimentally.

An alternative view is that by backgrounding a clause, the normal anaphoric nature of a pronoun can be bypassed. By putting a pronoun in a subordinate (or backgrounded) phrase, context is provided for the main concept to be communicated subsequently.

Exp I

Participants were asked to determine who was performing the action in the initial clause. The choices consisted of the person mentioned explicitly in the latter portion of the sentence, someone else, either of those, and the final choice was that the sentence was ungrammatical. The stimuli differed in initial clause type. The three types of these were simple past tense, a subordinate clause, and progressive aspect (either past progressive or pluperfect tense).

Results: All conditions were considered grammatically acceptable. Additionally most were scored as 'coreferent permitted'; although the subordinate clause condition was rated as being permitted more than the progressive aspect, which in turn was more permitted than the simple past tense.

Exp II

The extent of backgrounding in the three conditions studied above was assessed in this study utilizing a story-continuation method. In this experiment, participants read a sentence in which two individuals were mentioned by name. They were then asked to make up a subsequent sentence as a continuation of the sentence they had just read. The researchers expected the information in the main clause would be featured most prominently in the continuations.

Results: The first person mentioned was featured in the continuations in all conditions other than the subordinate condition. The researchers attributed this to a 'advantage of first mention' effect. The main variable of interest for them however (the frequency of mentioning the second person from the stimulus sentence) paralleled the results of experiment I with more second person mentionings in the subordinate condition followed by the progressive aspect condition and finally the simple past tense condition. The researches concluded that the progressive aspect does serve a backgrounding function, and therefore further research on pronominal reference was warranted.

Exp III

The researchers were curious as to why the sentences in experiment I were rated as being permitted, and thought it was perhaps due to the lack of an explicitly mentioned second entity. They designed a 2 (main vs. subordinate) X 2 (simple past tense vs. aspect) X 2 (new name vs. repeated name) design. Participants were again asked to determine who was being referred to by the pronoun in the sentence but in this case, the either choice was omitted and the ungrammatical choice was divided into an 'ungrammatical' and an 'odd' choice. Raters could choose either or both of these but were forced to make a decision on the pronoun.

Results: Again the subordinate clause condition served as a more effective backgrounding device than the progressive aspect. The main clause, progressive aspect condition was the primary manipulation of interest to the researchers. Coreference was allowed more often in the progressive aspect condition than in the simple past tense condition for both the repeated name and new name conditions (although the effect was larger in the repeated name condition).

Exp IV

In experiment III, 20% of raters allowed for coreference in sentences that should have been blocked by c-command (past research found acceptance rates that were much lower). A pilot study ruled out the plausibility of an instructional bias effect, so the researchers varied the subordinating conjunction (when vs. after). Semantic plausibility was also varied by manipulating the private mental aspect of the verb in the main phrase (noticing something) and a more public communicative verb (stating something). This resulted in a 3 X 2 design with the three levels being sentence type (when- simple past tense, after- simple past tense, and when- progressive aspect) and the two conditions of verb type (private knowledge, public knowledge). Once more, participants were asked to determine to whom the pronoun was a referent to.

Results: As expected, the intrasentential referent was selected more often in the subordinate clause condition. Also as expected, coreference was accepted more frequently if when was used as the conjunction than if after was used. The researchers were surprised to find that the main clause verb manipulation (private vs. public) only had an effect on coreference ratings in the when-simple past condition.

Overall, these studies show that although aspect backgrounding can function to bypass normal pronoun use, subordinate clauses work better. Aspect facilitates coreference judgments best in only select types of sentence structures.

- in isolated sentences (exp I)

- when there is a repeated name penalty (exp II)

- when there is private mental knowledge and if a when clause is used as the conjunction. (exp IV)