Visual Statistical Learning In Infants at Risk for ASD

Background: Language acquisition has been shown to involve the extraction of patterns from probabilistic cues in the auditory input.  Accordingly, infants have been shown to readily learn statistically defined patterns in auditory as well as visual domains, which presumably facilitates language development.  Infant siblings of children with autism spectrum disorders (ASD) are at increased risk for both ASD and language impairments.  Do these high-risk infants exhibit impairment in statistical learning?  No prior studies have investigated statistical learning in infants at risk for ASD.

Objectives: Here, we used an event-related potential (ERP) paradigm to examine cortical activity in infants at high-risk for developing ASD and low-risk, age matched controls as they participated in a visual statistical learning experiment adapted from the habituation paradigm described by Kirkham et al. (2002).  The aim was to examine group differences in statistical learning.

Methods: Infants were exposed to a continuous sequence of colored visual shapes which were organized into 3 pairs repeated 10 times each in a random order during a training phase. A test phase consisted of a maximum of 200 randomly ordered frequently occurring pairs.  80% of the pairs were learned in the training and 20% were unexpected.  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.  The principal dependent variable was the magnitude of the P300, an ERP component that reflects cognitive abilities such as attention and information processing. It is exhibited in response to unexpected stimuli.  To examine the neural correlates of visual statistical learning, we examined the peak amplitude of this component when infants were presented with a specific shape.  We compared P300 peak amplitudes of unexpected sequences, which were formed by the second shape of one learned pair and the first shape of another, among the high-risk and low-risk infants.

Results: Five high-risk infants and seven low-risk infants completed the training and a sequence of a minimum 20 pairs and a maximum of 140 pairs during the testing phase.  The data show that both groups learned the shape pairs in the training phase, as all exhibited the P300 to the unexpected sequences.  An interesting trend emerged showing that high-risk infants exhibit a larger P300 peak amplitude to the unexpected stimuli than do low-risk infants.

Conclusions: Our preliminary data suggest that high-risk infants and low-risk control infants may differ in neural markers of statistical learning.  The underlying etiology of the larger neural response needs to be further explored as our sample size increases. In addition, data collected from these infants at later time points can provide more insight into the development of this cognitive process and its potential implications for ASD diagnosis and language acquisition.