Some children with autism spectrum disorders (ASD) may not produce speech because of speech-motor deficits, which may in turn produce atypical or developmentally young vocalization (Rogers & Pennington, 1991). In response to challenges in measuring these deficits in children with ASD, some scientists measure the complexity of non-word and word vocalizations as a proxy for speech-motor (dis)ability. Most methods of analyzing vocalizations involve either perceptually-guided spectral analysis or behavioral coding of recorded interactions (Singh & Singh, 2008; Paul, et al, 2011). These are extremely time-consuming procedures, and consequently brief vocal samples (e.g., 10 minutes), are analyzed resulting in relatively unstable estimates of vocal complexity (Yoder & Symons, 2010). Day-long (12 hours) samples of child vocalizations through small digital audio recorders and fully automatic computer analysis can now be used to derive a vocal age equivalency score (Oller et al., 2010). The across-occasion stability and association of this index with expressive language in children with ASD and children with typical development has not yet been tested.
Objectives:
To determine how many day-long sessions are required to derive a stable vocal age equivalency score in (a) children with autism spectrum disorders (ASD), and (b) children who are typically developing (TD).
To determine the extent to which the vocal age equivalency score is associated with expressive language in children with ASD and children who are TD.
Methods:
Forty-two children with ASD and 30 children who were TD wore a small digital recorder (LENA) for three days (occasions) within a 5-day period (i.e., 72 participants x 12 hours x 3 samples = 2592 hours of recording). Generalizability and decision studies were conducted on each participant group to determine how many sessions’ data would need to be averaged to estimate vocal age equivalency with an intraclass correlation coefficient (ICC) greater than .80 (our criterion for “across-occasion stability). Within 5 days of the vocal recordings, each child’s parent completed the expressive subscale of the Child Development Inventory (CDI-EL) and the Language Development Survey (LDS). The age equivalency and raw scores were derived from the CDI-EL and LSD, respectively.
Results:
In both groups, only one day-long session was necessary to derive a stable estimate of vocal age equivalency (ICC for children with ASD = .86; ICC for children with TD = .86). The vocal age equivalency from only the first recorded session was correlated with the CDI-EL and LDS at r = .70 and .75 in the ASD sample, respectively and at .73 and .68 in the TD sample, respectively.
Conclusions:
The validity of a measure cannot exceed the square root of its reliability (Nunnally, 1978). If the vocal age equivalency score is a measure of vocal complexity, it should correlate with expressive language (Cronbach & Meehl, 1955). To put the .68 - .75 correlations with expressive language in perspective, other measures of vocal complexity (e.g., consonant inventory, canonical babbling ratio) concurrently correlate with expressive language between .38 and .65 in ASD or at-risk-for-ASD participant samples.
See more of: Core Symptoms
See more of: Symptoms, Diagnosis & Phenotype