The alignment of deaf children's gaze with parent signed input

Background

Children acquiring American Sign Language (ASL) must learn to recognize when to direct gaze to their parents in order to perceive linguistic input (Lieberman et al., 2014). Additionally, parents need to monitor children’s attention and time their input to coincide with child gaze (Swisher, 1999). In the current study, we asked how children’s perception of parent signs changes across development. We predicted that as children gain more experience interacting in ASL, there would be an increase in the proportion of parent sign tokens that are perceived by the child. We further hypothesized that parents might change the degree to which they scaffold children’s attention over time, such that parents of younger children wait for child gaze before beginning to sign, but parents of older children begin signing with the assumption that children will shift gaze to perceive the sign shortly after sign onset. 

Methods

We analyzed naturalistic play sessions between deaf children (n=21) aged 9-60 months (M = 34 months) and their deaf (n=13) or hearing (n=8) parents who used ASL. Sessions were videotaped and coded in ELAN for all ASL signs and child eye gaze. For each parent sign token, we determined if the sign was seen, the proportion of overlap between the sign and child gaze, and the timing of child gaze relative to parent sign onset and offset. 

Results

Across all participants, 43% of sign tokens (n = 5629) were at least partially perceived, and of these, 65% were fully perceived. Among the perceived tokens, the child was looking to the parent at sign onset but looked away before completion 17% of the time, and the child shifted gaze to the parent after sign onset 12% of the time. The remaining 6% of signs were seen only in the middle of the sign. The amount of overlap in parent signs and child gaze increased with age, but contrary to our predictions, the proportion of signs perceived only at sign onset or offset were consistent across age. We then ran a logistic regression model to predict the odds that a sign would be seen by the child, with age, parent hearing status, and number of sign tokens produced in the session as predictor variables. As expected, increasing child age improved the odds of a sign being seen (Estimate = 0.03, SE = 0.003, p < 0.001), as well as the number of tokens produced by the parents (Estimate = -0.003, SE = 0.0003, p < 0.001).