Make it snappy: Predictability and speech rate effects on child response latency and social desirability judgements

Background: The amount of time it takes a conversation partner to respond is pragmatically meaningful. People infer that long response latencies are indicative of a reduced willingness to cooperate (1). Adults report lower rapport with conversation partners when response latencies are slower (2). Furthermore, among children, faster turn-taking is associated with higher ratings of peer social engagement (3), and children indicate a reduced desire to interact with speakers who exhibit slower conversational responding (4). However, generating fast responses involves high cognitive demand, requiring listeners to concurrently plan their own upcoming turn (5), and predict the content (what the utterance is about) and timing (when the utterance will end) of their partner’s current turn (6). It is therefore unsurprising that children’s response latencies remain much longer than adults’ into middle childhood (~1000-ms) (7), given that working memory capacity continues to develop throughout this time (8). Nevertheless, previous research has found that children as young as 3-5-years-old can make content-based predictions which facilitate early response planning (9). This therefore suggests that the cognitively demanding component of this process may be the need to make predictions about timing, rather than content. Indeed, response timing is far from straightforward, even for adults. On the one hand, some findings indicate that adults produce faster responses when hearing faster speech rates (10,11). This has been described as ‘entrainment’. In contrast, Roberts, Torreira & Levinson (12) found that faster speech (in a telephone corpus) was linked to longer response latencies, which could be the result of a higher cognitive demand of processing faster speech. Given their reduced capacity for cognitive load, it is possible that children might take longer to respond to faster than to slower speech rates. Conversely, given previous findings that children disprefer slower responders (4), they may entrain to the speech rate of their conversation partner, showing faster responses to faster speech rates. We carried out two studies with children aged 7-8-years (N=36 per study) to investigate the following questions: • RQ1: How do predictability and speech rate influence children’s response latencies? • RQ2: How does speech rate influence children’s social impressions of their conversation partner?

Method: Children were presented with a computer game which involved verbally answering yes/no questions from two virtual characters (‘Fast’ vs. ‘Slow’ speaker). We manipulated the content predictability of the questions (Predictable vs. Unpredictable) and the speech rate of the pre-recorded audio (Fast vs. Slow). In Study 1, the Fast speech rate was created by reducing the length of the original audio files by 10%, whilst the Slow speech rate represented a 50% increase. In contrast, in Study 2, both conditions evenly differed from baseline by 20%. To investigate RQ2, at the end of the game, participants provided social judgements about the Fast vs. Slow speakers on a scale of 0-100. In Study 1, participants rated each character’s friendliness, intelligence, and the extent to which they would enjoy chatting to them and would want to be friends with them. In Study 2, two additional statements were added, namely whether the speakers seemed fun / would be good at having a conversation. In both studies, participants were also asked to choose which character they would prefer to play the game with if they were going to repeat the task.

Results: RQ1: Response latencies Linear mixed effects models examined the effect of speech rate and content predictability on participants’ response latencies. In Study 1, Slow questions (M=798.65-ms) elicited earlier responses than Fast questions (M=1021.85-ms) (b= 221.83, p<.001). Participants also responded quicker to Predictable questions (M=836.56-ms) than Unpredictable questions (M=981.95-ms) (b= -101.96, p =.056). The same pattern of results was observed in Study 2, with participants responding earlier to Slow (M=712.54-ms) than Fast (M=879.28-ms) questions (b= 164.93, p<.001), and Predictable (M=707.62-ms) than Unpredictable questions (M=886.76ms) (b= -138.20, p < .05). There were no interaction effects (see Figure 1). RQ2: Social judgements Paired t-tests examined how speech rate influenced participants’ social judgements. In both studies participants rated the Fast speaker as significantly smarter than the Slow speaker (p <.05, d=.48). In Study 1, they also indicated that they would enjoy chatting with the Fast speaker (M=78.58) more than the Slow speaker (M=56.78) (p <.01, d=.67). Across both studies, no other comparisons were significant. In Study 1, when participants were asked who they would like to play alongside in another round of the game, they were significantly more likely to select the Fast speaker than the Slow speaker (p <.01). However, in Study 2, although the means were in this direction (Fast speaker= 56%), this was not significant (p =.505). Figure 1. Response Latencies by Predictability x Speech Rate

Condition: Conclusions Across both studies, we found that children responded earlier to Predictable questions than Unpredictable questions. Our novel finding is, however, that in both studies, children responded earlier to Slow questions than Fast questions. This is not in line with entrainment. However, it does align with certain findings from the adult literature (12,13), where it has been interpreted as an effect of cognitive load (5). Future research is needed to investigate whether cognitive load limits are the key, or whether slower speech rates may have facilitated more time for comprehension and response planning during the utterance. We also found evidence of social preferences favoring fast over slow speakers. In both studies, children perceived the Slow speaker to be significantly less clever than the Fast speaker. In Study 1, participants also indicated that they would enjoy interacting more with the Fast speaker than the Slow speaker, and they were significantly more likely to select the Fast speaker to play with again. That said, in Study 2, when the speech conditions differed equally from ‘baseline’, the Slow speaker was rated less unfavourably. Overall, whilst Slow speech rates were consistently associated with lower ratings of perceived intelligence, the extent to which children were willing to engage with the Slow speaker appeared to crucially depend on how slowly they spoke. Given the importance of speech latency in relation to social rapport (2,3), future research on this relationship is sorely needed.

References: 1. Roberts F, Francis AL. Identifying a temporal threshold of tolerance for silent gaps after requests. J Acoust Soc Am. 2013 Jun;133(6):EL471–7. 2. Templeton EM, Chang LJ, Reynolds EA, Cone LeBeaumont MD, Wheatley T. Fast response times signal social connection in conversation. Proc Natl Acad Sci. 2022 Jan 25;119(4):e2116915119. 3. Wilson M, Powell AR, Hernandez LS, Green E, Labahn C, Henderson H. Shyness, social engagement, and conversational response times in children’s dyadic interactions with an unfamiliar peer. Soc Dev. 2024 Aug;33(3):e12734. 4. McGuinness L, Abbot-Smith K, Gambi C. Seeing it in others versus doing it yourself: Social desirability judgements and conversation production data from autistic and nonautistic children. Autism. 2024 Nov 4;13623613241292172. 5. Levinson SC, Torreira F. Timing in Turn-Taking and Its Implications for Processing Models of Language. Front Psychol. 2015;(6). 6. Garrod S, Pickering MJ. The Use of Content and Timing to Predict Turn Transitions. Front Psychol. 2015;6:1–12. 7. Nguyen V, Versyp O, Cox C, Fusaroli R. A systematic review and Bayesian meta‐analysis of the development of turn taking in adult–child vocal interactions. Child Dev. 2022 Jul;93(4):1181–200. 8. Best JR, Miller PH. A Developmental Perspective on Executive Function. Child Dev. 2010 Nov;81(6):1641–60. 9. Lindsay L, Gambi C, Rabagliati H. Preschoolers Optimize the Timing of Their Conversational Turns Through Flexible Coordination of Language Comprehension and Production. Psychol Sci. 2019 Apr;30(4):504–15. 10. Wynn CJ, Borrie SA, Sellers TP. Speech Rate Entrainment in Children and Adults With and Without Autism Spectrum Disorder. Am J Speech Lang Pathol. 2018 Aug 6;27(3):965–74. 11. Corps RE, Gambi C, Pickering MJ. How do listeners time response articulation when answering questions? The role of speech rate. J Exp Psychol Learn Mem Cogn. 2020 Apr;46(4):781–802. 12. Roberts S, Torreira F, Levinson S. The Effects of Processing and Sequence Organization on the Timing of Turn Taking: A Corpus Study. In: Frontiers in Psychology. (6). 13. Hoogland D, White L, Knight S. Speech Rate and Turn-Transition Pause Duration in Dutch and English Spontaneous Question-Answer Sequences. Languages. 2023 Apr 22;8(2):115.