The effects of contextualization on fluency and language in three groups of children

A well-known feature of fluency disorders is that frequency and severity of stuttering are influenced by language formulation demand. Previous research has shown that as utterances become longer and more syntactically complex, children who stutter (CWS) exhibit greater frequency of stuttering. Research examining the narrative abilities of CWS has implicated language ability as one component of the stuttering problem for a possible subgroup of CWS. The influence of language on stuttering has even been discussed as a contributing factor in a developmental explanation of stuttering, as well as in maintaining the fluency disorder. In this study, language formulation demand was manipulated by eliciting 4 discourse samples from 35 children. Twelve CWS were matched for age and sex with 12 children with normally-developing communication skills (CNC), and with 11 children diagnosed with language impairment (CLI). The 4 samples were elicited by asking participants to generate 2 scripts for making a recipe, and retell 2 stories. One script generation and one narrative were contextualized, with objects or pictures immediately available; the other script and narrative were decontextualized. It was hypothesized that for CWS, decontextualization would result in increased frequency of stuttering-type and normal-type disfluencies, and increased maze behavior (i.e., false starts and revisions). Additionally, it was expected that discourse complexity, episode completion, and semantic complexity would be reduced in the decontextualized conditions. Similar patterns of increased normal-type disfluency and maze behavior and decreased discourse and semantic complexity were anticipated for CNC and CLI. Results show contextualization to be an aspect of language formulation demand that influences both fluency and linguistic complexity. For CWS, a significant reduction in frequency of stuttering was noted in the contextualized recipe condition. Across conditions, frequency of maze behavior was significantly greater than frequencies of stuttering- and normal-type disfluencies. The difference between frequency of stuttering- and normal-type disfluencies across conditions was nonsignificant. For all 3 groups, both decontextualized conditions produced greater frequencies of normal-type disfluencies and maze behavior, as well as reduced discourse complexity, episode completion, and semantic complexity