Neural Correlates of Implicit Learning In Young Children with ASD

Background: Communication impairment is a core deficit in Autism Spectrum Disorders (ASD), yet language outcomes vary tremendously, with approximately 50% of children remaining non-verbal. Part of the difficulty in predicting language outcome lies in the limitation of standardized assessments to capture specific areas of strength and impairment and, importantly, to characterize neural mechanisms underlying behavior. One cognitive domain critical for language function is statistical (implicit) learning, or the extraction of patterns from probabilistic cues. This domain can be very difficult to characterize behaviorally in young children, but it can be investigated using event related potentials (ERPs), as has been done in several infant studies.

Objectives: (1) To characterize the neural correlates of implicit learning in young children with ASD enrolled in an intensive, standardized, 3-month intervention program and to then investigate the association between language function and implicit learning. (2) To characterize change in implicit learning after intensive intervention.

Methods: Children were recruited through the UCLA Early Childhood Partial Hospitalization Program (ECPHP). ECPHP ascribes to an informed, eclectic treatment approach consisting of both adult and child directed instructional techniques. Children receive 30 hours/week of intervention for approximately 12 weeks. Children entering ECPHP over the past 6 months were investigated with an ERP paradigm of implicit learning. In this modified oddball paradigm, children were first exposed to a continuous sequence of shapes organized into three pairs repeated 10 times each. In the test phase, children were exposed to learned pairs (80%) and unexpected pairs (20%). The principal dependant variable was the frontocentral P300, which occurs between 210-350 msec after stimulus onset. EEG recording was accomplished using a 128 Hydrocel Geodesic Sensor Net System (EGI Inc), using NetStation software. Data were amplified and filtered (0.3-30 hz), with a sampling rate of 250 hz, and digitized using a 12 bit National Instruments Board. ERP data were edited using NetStation 4.4.

Results: 15 children were investigated, with 13/15 providing adequate EEG data.  Ages ranged from 3-5, with 1 female. 3/13 were nonverbal. As a group, children showed evidence of statistical learning, as shown by differences in P300 amplitude between the expected and unexpected conditions. In preliminary analyses, the nonverbal group (n=3) showed less differentiation between the two conditions than the verbal group (n=10), suggesting an association between impairment in language and statistical learning. No hemispheric differences were seen. In the three children who were followed longitudinally, there was evidence of a larger P300 to the unexpected stimulus after intervention.

Conclusions: This is one of the only studies investigating neural correlates of implicit learning in young children with ASD. These preliminary findings not only demonstrate that implicit learning can be characterized with ERPs in this population, but they also suggest an association between implicit learning and language function. As we continue this study and increase our sample size, we will be able to create models that include both behavioral and electrophysiological measures to better understand pathways to language impairment in ASD, with the ultimate goal of defining predictors of language outcome in this population.