Saying "that" in dialogue: The influence of accessibility and social factors on syntactic production

Abstract

Previous evidence suggests that when speakers produce sentences from memory or as picture descriptions, their choices of sentence structure are influenced by how easy it is to retrieve sentence material (accessibility). Three experiments assessed whether this pattern holds in naturalistic, interactive dialogue. Pairs of speakers took turns asking each other questions, the responses to which allowed mention of an optional "that" before either repeated (accessible) or unrepeated (inaccessible) material. Speakers' "that" mention was not sensitive to the repetition (accessibility) manipulation. Instead, "that" mention was sensitive to social factors: Speakers said "that" more when adopting another's perspective rather than one's own, and tended to say "that" more when attributing emotions to oneself rather than another. A fourth experiment confirmed that in a memory task, the original pattern is observed. These results suggest that "that" mention is sensitive to the cognitive forces that operate within a production task; in dialogue settings, social factors were especially influential.

Figure 1

Percentages of sentences produced with “that” as a function of main and embedded subject (“I” or “you”) in Experiment 1. Error bars report 95% confidence interval derived from subject ANOVA.

Figure 2

Percentages of sentences produced with “that” as a function of main and embedded subject (“I” or “you”) in Experiment 2. Error bars report 95% confidence interval derived from subject ANOVA.

Figure 3

Percentages of sentences produced with “that” as a function of main and embedded subject (“I” or “you”) in Experiment 3 (collapsed across whether the original sentence included or omitted the “that”). Error bars report 95% confidence interval derived from subject ANOVA.

Figure 4

Percentages of sentences produced with “that” as a function of main subject collapsed across embedded subject (top panel) and embedded subject collapsed across main subject (bottom panel) in Experiment 4. Error bars report 95% confidence interval derived from subject ANOVA.

Figure 5

Percentages of sentences produced with “that” (top left panel) and average certainty and unusualness (other three panels) as a function of main and embedded subject (“I” or “you”) in followup experiment. Error bars report 95% confidence interval derived from subject ANOVA.