A break in parental interaction does not affect the temporal dependency of infant social engagement, but disrupts non-social engagement

Infant looking patterns during interaction offer an early window into social and nonsocial engagement. Recent evidence indicates that infant looks exhibit temporal dependency—one look duration predicts the next look duration. It is unknown, however, whether temporal dependency emerges as infants structure their own looking or whether it is influenced by interaction. We examined whether a perturbation of social interaction affected temporal dependency. Using the Face-to-Face/Still-Face procedure, we compared temporal dependency during parental interaction (the Face-to-Face & Reunion episodes) to parental non-responsiveness (the Still-Face episode). Overall, the durations of successive infant looks were predictable; past behavior constrained current behavior. The duration of one look at the parent (Face Look) predicted the duration of the next Face Look. Likewise, the duration of a look at any place that was not the parent’s face (Away Look) predicted the duration of the next Away Look. The temporal dependency of Face Looks (social engagement) was unaffected by the Still-Face perturbation, but the temporal dependency of Away Looks (nonsocial engagement) declined during the Still-Face. Infant temporal structuring of engagement during social looking is not dependent on parental interaction while the disruption of interaction affects infants’ structuring of their own non-social engagement.

The final model (Model 5) accounted for 3.46% of the variance in the empty model (PVAF;24 In this final model, the previous look duration and the look duration two previous were predictive of a given look duration. Temporal dependency was indicated in the FFSF for Face Look durations, 10 = .11, SE ( ) = .02, t (108) = 6.98, p < .001. The results are summarized in Table 3; the prediction of Face Look durations is depicted in Figure 2.
differences between infants from that mean. 10 is the mean change in Face Look durations for every one second change in the previous Face Look duration, and r 1i represents random differences between infants in that change. 20 is the mean slope of change in Face Look durations for every one second change in the Face Look duration two previous, and r 2i represents random differences between infants in that change. 10 and 20 index temporal dependency. is unexplained residual variance in for the jth Face Look duration for the ith infant. Additionally, the covariances of all random effect terms (r 0i -r 3i ) are included in this model

Summary of Away Look modeling
The FFSF Away Look model was constructed by first generating a base or empty model (Model 1), using the mean look and random differences between infants in that mean look. This model showed non-trivial dependency between individuals, ICC = .88, the rationale for using a multi-level modeling approach. Next, a model was constructed including both previous look duration as a predictor and random differences between infants in the strength of that prediction (Model 2), which was a significant improvement over Model 1, χ 2 (3, J = 109) = 441.28, p < .001.
Next, a model was constructed which added the effect of the FFSF protocol, with two dichotomous variables encapsulating the differences between the Still-Face episode and episodes with interaction (Face-to-Face and Reunion), and the differences between interaction before (Face-to-Face) and after (Reunion) perturbation of the interaction (Model 3). These also included terms for the random differences between infants in the predictive strength of each episode-based prediction. This was a significant improvement over Model 2, χ 2 (4, J = 109) = 241.21, p < .001. Next, we constructed a model which added a second previous Away Look duration as a predictor, as well as random differences between infants in that prediction (Model 4). This was a significant improvement over Model 3, χ 2 (6, J = 109) = 224.82, p < .001. A model incorporating two interaction terms was next tested, one between the previous look duration and the Still-Face contrast, and the other with the previous look duration and the Face-to-Face vs.
Supplementary Information 6 was a smaller effect of temporal dependency compared to the Face-to-Face and Reunion episodes. Temporal dependency was found in the FFSF for successive looks away, 10 = .05, SE ( ) = .02, t (108) = 2.94, p < .01. The results are summarized in Table 3; the prediction of Away Look durations is depicted in Figure 3.
In summary, the equation for the final model of Away Look was: Where represents the duration in seconds of successive Away Looks, the jth in a series of looks for infant i. 00 represents the intercept of Away Look durations, r 0i represents random differences between infants from that intercept. 10 is the slope of change in Away Look durations for every one second change in the previous Away Look duration, r 1i represents random differences between infants in that change. 10 indexes temporal dependency. 20 is the slope of change in Away Look durations due to parental Still-Face relative to the Face-to-Face and Reunion, r 2i represents random differences between infants in that change. 30 is the slope of change in Away Look durations for every one second change in the previous Away Look duration during parental Still-Face relative to the Face-to-Face and Reunion, r 3i represents random differences between infants in that change. is unexplained residual variance in for the jth Away Look duration for the ith infant. Additionally, the covariances of all random effect terms (r 0i -r 3i ) are included in this model.