Objectives: The goal is to measure temporal dynamics of vocal adult-child interactions at multiple timescales in day-long recordings of preschool children with and without ASD using new large-scale automated analysis methods. Our hope is that this research will enrich our understanding of how the dynamics of interaction are altered by ASD and add to existing diagnostic and assessment indicators.
Methods: Data were from 438 day-long (12+ hours) longitudinal recordings from 26 children 16-48 months of age diagnosed with ASD and 78 age-, gender-, and SES-matched TD children. Recordings were processed using automatic speaker-labeling software. CR plots were then made for each recording. CR plots were square matrices (child on one axis, adult on the other) wherein each row/column pair represented an event where either the child or an adult vocalized. CR plots indicate temporal ordering between every child vocalization and every adult vocalization. The distribution of values perpendicular to the matrix diagonal (the "diagonal recurrence profile") represents leading-following tendencies at a range of timescales (Richardson et al., 2007). Amounts of adult-leading and child-leading were compared between ASD and TD recordings.
Results: Multiple regression controlling for child identity (random effect) and for child age, gender, and mother's education (fixed effects) revealed proportionally fewer immediate adult responses to child vocalizations in ASD recordings than in TD recordings (p < .001, beta = -.44). Additionally, for the ASD group the ratio of the number of child vocalizations following adult vocalizations within a 30-event lag window to the number of adult vocalizations following child vocalizations within the same size window was smaller than for the TD group (p < .001, beta = -.17). Restricting the analysis to child vocalizations that were identified as speech-like (cries and vegetative sounds excluded) or to non-speech-like vocalizations, similar patterns of results were observed.
Conclusions: Throughout the day, ASD and TD groups exhibit differences in dynamics of child-adult interaction. ASD children's vocalizations are less likely to be immediately followed by adult vocalizations, and the balance of child leading vs. following leans more toward the child following. This finding may relate to previous findings that children with ASD initiate interactions less often than TD children. Results also support CR analysis over day-long recordings as a potential method for assessing young children's interaction patterns in ASD.