Imitation In a Large Cohort of Preschoolers with ASD: Measurement Structure and Correlates

Background:  Children with autistic spectrum disorders (ASD) show impaired imitation of actions (Rogers & Williams, 2006). Indeed, poor elicited imitation is an early predictor of ASD in high risk infants (Zwaigenbaum et al., 2005). Most detailed investigations of imitation in ASD have used relatively small and/or selected samples, and few imitation tasks, often of 1 or 2 types (e.g., actions with objects, conventional gestures). These tasks may be insufficient to describe the growth of imitation skills in the preschool years.

Objectives:  The present study examines (1) the structure of a novel multi-dimensional elicited imitation measure, and (2) the correlates of imitative performance in a large representative sample of preschoolers with ASD.

Methods:  The study draws upon baseline data (within 4 months of diagnosis) from a longitudinal study of preschoolers with ASD (n=361; 83% males; mean age=40.8, SD=9.2 months). The Multidimensional Imitation Assessment (MIA; Lowe-Pearce & Smith, 2005) contains 48 items organized into 11 a priori domains. In this study, performance was scored in vivo on a 4-point scale for accuracy (0 = no response 1 = emerging response / not imitation; 2 = partial / inaccurate imitation; 3 = exact imitation). Age equivalent (AE) scores were used for measures of intellectual (Merrill-Palmer-Revised Scales, Developmental Index), and language abilities (Preschool Language Scales 4^th ed., Auditory Comprehension and Expressive Language). Other measures were of adaptive behaviour (Vineland Scales of Adaptive Behaviour, 2^nd ed, Adaptive Behaviour Composite), initiating and responding to joint attention (JA; Early Social Communication Scales) and autism symptoms (Social Responsiveness Scale Total score).

Results:  A principal components analysis (PCA) with varimax rotation on the 48 MIA item scores yielded a 3-factor solution explaining 76% of the variance. Factor 1 - Actions With Objects, accounted for 68.9% of variance; Factor 2 - Vocalizations/Words with Objects, for an additional 4.1%, and Factor 3 – Actions Without Objects, for a further 2.9%. Mean accuracy scores for each factor increased across age groups (2-, 3-, 4 –year-olds), all p’s <.001. MIA Total accuracy scores were significantly positively correlated with all ability and JA measures (r’s from .18 to .63; all p’s<.001). However, relationships with the Actions Without Objects factor appeared to be somewhat more heterogeneous, warranting further exploration. Specifically, for 2-year-olds, significant correlations were seen between both cognitive and language abilities and imitation accuracy, for both Actions With Objects and Vocalizations/Words With Objects (r’s from .20 - .50, all p’s <.02). However, for Actions Without Objects, only language scores were significantly correlated with imitation accuracy (r’s = .21 and .26, p’s <.02 and .003 for EL and AC). Relationships were non-significant between autism symptoms and all imitation factor scores.

Conclusions: The MIA provides an index of age-related improvements in multiple aspects of imitation accuracy. Furthermore, associations with measures previously linked to imitation show evidence of criterion-related validity for the MIA. We conclude that the three empirically derived factors of the MIA provide useful indices for assessing longitudinal change in imitation in preschool children with ASD.