Date of Completion


Embargo Period



Autism, Language Acquisition, Nonadjacent Dependencies, Statistical Learning, Sleep, Developmental Psychology

Major Advisor

Letitia Naigles

Associate Advisor

Heather Bortfeld

Associate Advisor

Inge-Marie Eigsti

Field of Study



Doctor of Philosophy

Open Access

Open Access


In order to acquire some aspects of grammar, such as wh-questions and verb tense agreement, children must be able to learn nonadjacent dependencies. This type of learning has been demonstrated in both children and adults, but is reported to be difficult. The current study investigated whether children with autism (ASD) would show a similar pattern of nonadjacent dependency learning as seen in typically developing (TD) children, and whether variation in this ability would relate to language levels, sleep habits or characteristics of ASD.

Ten TD children (M age 21.04 months), and ten children with ASD (M age 43.86 months) were tested in their homes on two consecutive days. Statistical learning abilities were assessed using a visual fixation paradigm (e.g., Shi et al., 2006). The audiovisual stimuli were presented via a computer monitor and speaker. A camcorder positioned behind the monitor captured children’s direction of gaze. During the familiarization trial, children watched a silent video while an artificial grammar was presented for 20 minutes. In the test trial, children watched a 3.5-minute video of a moving checkerboard, while they heard familiar and unfamiliar audio strings interspersed.

Audio stimuli consisted of 3-item strings of nonce words (e.g. vot_kicey_jic) in which the first and third item always co-occurred, but the second item varied. Test stimuli were a subset of the training strings (i.e., familiar), and a subset of unfamiliar strings (i.e., from an alternate grammar). Average looking-times were calculated offline for familiar and unfamiliar trials, to determine if each child discriminated between trial types.

Neither group of children showed significant discrimination between familiar and unfamiliar audio trials. However, individual children’s looking patterns revealed a qualitative difference in looking preferences. Most of the TD children looked longer during unfamiliar audio trials, while most of the ASD group looked longer during the familiar trials. Children’s looking times were correlated with measures of language levels, sleep habits, and autism characteristics. Few measures significantly predicted looking times during the statistical learning task. However, the measures that predicted TD looking patterns were different from the measures that predicted ASD looking patterns. Implications for these findings are discussed.