The comprehension and production of restrictive relative clauses

This project investigated the comprehension of relative clauses in English-speaking children and adults. In a series of 5 experiments across two studies we tested 300 participants, 183 of whom were children between the ages of 4 and 9.

Participants took part in a series of sentence-picture matching tasks while we tracked their eye movements in a visual world paradigm. This methodology yielded real-time online measures of comprehension as the sentences unfolded, and reaction time measures and accuracy measures of picture selection.  For the child participants we additionally took measures of working memory, of inhibition, and of syntactic skills.

Relative clauses are complex syntactic constructions whose comprehension requires the sophisticated integration of morpho-syntactic, semantic, and discourse-pragmatic information. Using online and offline methods we tried to disentangle the process from the outcome of the comprehension process, and we focused on the kind of semantic, discourse-pragmatic cues that guide the incremental process of sentence comprehension.

In the first set of experiments we focused on the well-attested comprehension asymmetry between subject relative clauses (The cow/the tractor that chased the deer) and object relative clauses (The cow/the tractor that the deer chased), and we investigated the extent to which the animacy of the head noun (animate: cow vs. inanimate: tractor) led participants to anticipate a subject or an object relative clause. We replicated the well attested subject/object asymmetry, and we found evidence for a novel visual surprisal effect in the interpretation of relative clauses with lexically inanimate nouns as head nouns. Using eye-tracking in the visual world paradigm allowed us to tease apart the role of lexical vs. perceptual animacy in the interpretation of relative clauses.

Overall, we did not find a predictive relationship between measures of memory, inhibition, syntactic skills, and relative clause comprehension.

In the second set of experiments we focused on the role of discourse in relative clause attachment. In the context of everyday language use, individual sentences are typically embedded in a larger discourse context as they are preceded by other information. Therefore, comprehending a sentence at any given point in time will depend, at least partly, on information included in previous sentences which has been in turn understood and retained in memory.

In three sentence-picture matching experiments during eye-tracking we manipulated the preamble before a target sentence. Participants listened to a sentence containing a main clause and a relative clause (The man saw the nurse with the child who was very tired) and then were asked a question (Who was very tired?) which implicitly required them to choose between the two nouns (nurse or child) as the head of the relative clause.

The relative clause in the task was structurally ambiguous as it could either be attached to the boy (low attachment, the preferred structural choice in English), or to the nurse (high attachment). We manipulated the discourse context preceding the target sentence in order to prime different attachment choices: it had been a very long day at school (bias towards low attachment to the boy), it had been a very long day at the hospital (bias towards high attachment to the nurse), it had been a very long day that Wednesday (control condition). In a second experiment we also increased the length of the preamble to further bias either a low or a high attachment interpretation. In neither experiment were we successful in manipulating the discourse context to overcome the very strong structural preference for a low attachment interpretation.

In a final experiment we manipulated a morpho-syntactic cue to test whether a categorical grammatical cue could over-ride the structural preference for low attachment. We manipulated the number (singular vs. plural) of the head noun and of the verb in the relative clause. So in a sentence like The man recognised the nurse with the boys who was very tired, we wanted to investigate whether the morphosyntactic cue of the number mismatch between the plural noun phrase the boys and the singular verb was in the relative clause would be strong enough to override the entrenched structural preference for low attachment.

The results from our child participants suggest that between the ages of five and eight, children develop their ability to make use of morpho-syntactic cues in complex sentences, but also develop their preference for minimal attachment, and that this preference interferes with their use of unambiguous and essential morphosyntactic cues. 

The findings of these studies reveal a complex interplay of sensitivity to semantic, discourse, and morphosyntactic knowledge in children’s sentence comprehension. By using both online and offline measures these findings make a contribution to our understanding of the process, and not just the outcome of sentence comprehension. 

Project Team: Ludovica Serratrice (Lead), Silke Brandt, Evan KiddElena LievenRoss Macdonald and Anna Theakston.

Start Date: September 2016

Duration: 3 years

(Work Package 10)