The influence of communication sample length on reliability and convergent validity of vocal measures derived from the Communication Complexity Scale

Purpose: 

This study examines the effects of communication sample length on the reliability and convergent validity of six vocal measures for children with autism spectrum disorder (ASD) with minimal verbal skills. The results are expected to inform recommendations for the length of communication samples for clinical and research purposes.
 

Method: 

Participants included 31 children with ASD (24 boys and seven girls; Mage = 6;7 [years;months], SD = 17 months) with minimal verbal skills. We coded six vocal measures that focus on vocalizations and early word productions from the Communication Complexity Scale (CCS) scripted administration protocol. To evaluate reliability of different sample lengths, we calculated intraclass correlation coefficients between the full CCS sample and 1-, 3-, 7-, 10-, and 20-min samples. To examine convergent validity, we calculated correlations between the six vocal measures for each sample length.
 

Results: 

When coded from 10-min samples from the beginning of the CCS, all of the vocal measures exhibit adequate reliability with the full CCS sample. Some vocal measures exhibit adequate reliability with samples as short as 3 min. For convergent validity, all of the correlations between the vocal measures exceed .40 and are statistically significant for the 10-min samples except for some of the correlations with the proportion of communicative vocalizations. Similar results were found for other sample lengths.
 

Conclusion: 

Findings support coding 10-min segments from the CCS scripted administration protocol to evaluate the vocal development skills of children with ASD with minimal verbal skills.