Temporal Dependency and the Structure of Early Looking

Abstract

Although looking time is used to assess infant perceptual and cognitive processing, little is known about the temporal structure of infant looking. To shed light on this temporal structure, 127 three-month-olds were assessed in an infant-controlled habituation procedure and presented with a pre-recorded display of a woman addressing the infant using infant-directed speech. Previous individual look durations positively predicted subsequent look durations over a six look window, suggesting a temporal dependency between successive infant looks. The previous look duration continued to predict the subsequent look duration after accounting for habituation-linked declines in look duration, and when looks were separated by an inter-trial interval in which no stimulus was displayed. Individual differences in temporal dependency, the strength of associations between consecutive look durations, are distinct from individual differences in mean infant look duration. Nevertheless, infants with stronger temporal dependency had briefer mean look durations, a potential index of stimulus processing. Temporal dependency was evident not only between individual infant looks but between the durations of successive habituation trials (total looking within a trial). Finally, temporal dependency was evident in associations between the last look at the habituation stimulus and the first look at a novel test stimulus. Thus temporal dependency was evident across multiple timescales (individual looks and trials comprised of multiple individual looks) and persisted across conditions including brief periods of no stimulus presentation and changes from a familiar to novel stimulus. Associations between consecutive look durations over multiple timescales and stimuli suggest a temporal structure of infant attention that has been largely ignored in previous work on infant looking.

Fig 1. An illustration of the temporal dependency hypothesis.

Individual looks occur within trials of the habituation protocol. Individual look durations positively predict subsequent look durations. Previous trial durations predict subsequent trial durations.

Fig 2. Temporal dependency.

Individual look durations are predicted by previous individual look durations (A–C) and trial durations are predicted by the previous trial duration (D). Durations are displayed in seconds on a log 10 scale. A. The modeled effect of five previous look durations predicts the duration of the nth in a series of looks (see Table 3). B. The modeled effect of look count (habituation) and previous look duration predicts the duration of the nth in a series of looks (see Table 4). C. The modeled effect of look count (habituation) and the duration of the last look of a previous trial predicts the duration of the first look of the next trial (see Table 5). D. The modeled effect of trial count (habituation) and previous trial duration predict the duration of the nth in a series of trial durations (see Table 6).

Fig 3. Individual infants’ temporal dependency parameters were derived from the final multilevel model (Table 4).

A. Infants with shorter mean look length had higher levels of temporal dependency, r(125) = -.42, p <.001. B. Using a median split, short lookers exhibited stronger temporal dependency than long lookers, t(125) = 5.28, p <.001. Bars indicate standard errors of the mean.