Infants’ processing of distractors is associated with group-coherent modulation of occipital gamma bursts

We looked at active learning in 10-month-old infants, more specifically the dynamics and neural correlates of their processing of distracting stimuli.

Infants brain activity was recorded using electroencephalography (EEG) while they watched stimuli on a screen. Stimuli consisted of 1) a complex ongoing stimulus (40s-cartoon video, repeated 10 times); 2) a simple distracting stimulus (100ms-checkerboard, flashed 128 times). This paradigm being quite repetitive, infants tended to disengage from it at different points, which provided a good basis to explore individual differences in engagement. To account for such differences, the data was split into 3 stages (early, middle, late) of even trial length based on how long the participant remained on the task. The mixed nature of the stimulus set allowed to study both how infants learned about complex information and also how distractible they were throughout the session. The present study focuses on the second aspect.

N=31 infants were included in the analysis. Their EEG signals were pre-processed (bad electrodes interpolation, filtering, epoching around distractors, baselining with pre-distractor segments) and for each stage, individuals’ average time-resolved gamma power (55-100Hz) was derived over a set of occipital electrodes. A group average per stage was also derived at every time-frequency point. The distribution of the post-stimulus gamma power at each stage were compared (Kolmogorov-Smirnov tests, Bonferroni corrected) and effect sizes derived (Cohen’s d) for 1) individuals average time-frequency data, 2) pooled data: time-frequency values from (1) pooled together, 3) the group average time-frequency data.

We found that at the individual level (1), participants’ relative occipital gamma power was distributed significantly differently between stages (p<.001 for all comparisons), with small effect sizes on average (mean absolute Cohen’s d across participants and stages=0.346, std=0.246), suggesting that participants’ processing of the distractors changed throughout the study. However, these effects were unsystematic, revealing only marginal modulations (p<.001, Cohen’s d=-0.099, -0.087 and 0.181, comparing stages 1 -2, 2-3, and 1-3, respectively) in the pooled data (2). By contrast, the group average time-frequency landscape (3) revealed a significant gradual decrease in power between stages, with strong effects (p<.001, Cohen’s d= 0.381, 0.3853 and 0.633). Because the averaged time-frequency representation is influenced by the summing or cancelling of time-frequency components across subjects, this result suggests that there is a coherent, systematic expression of gamma bursts in the first stage, which diverges in time towards stage three. Since this effect is not seen in individuals nor in the pooled data, it cannot be explained merely by a global decrease in gamma bursts as the experiment unfolds. Rather, we show that gamma bursting is more systematically localized and coherent across participants at the beginning than at the end of the experiment. Further analyses will relate these measures with individual differences in attention/distractibility as measured with the ECBQ questionnaire.

This work unveils an important role of gamma-band occipital activity in infants individualised learning. While gamma-band activity is rarely investigated in infants, our results suggest that it holds important information on how visual stimuli are processed at the individual and the group level.