The same time on the screen, a different outcome
In the United Kingdom and Singapore, 47 infants aged 8 to 10.5 months heard three artificial languages made from syllable triplets. Each language was paired with the same speaker, whose glasses were clear, partly shaded or fully dark. Time spent looking at the screen did not differ significantly between conditions. Statistically significant learning was detected only for the language presented with fully visible eyes, while the partial-gaze result was borderline. The EEG analysis included the 42 infants with usable recordings. The study was published in Nature Communications.[1]
The brain measurements show a subtler distinction. Directed adult-to-infant connections were detected in all three gaze conditions, but fully visible eyes strengthened connectivity relative to the no-gaze condition. The adult-to-infant connectivity pattern predicted learning better than the infant's within-brain connectivity and mediated the relationship between gaze and learning. Neural entrainment to the speech rhythm was significant only under full gaze, yet did not predict learning. The result therefore does not say that two brains connect in only one condition; it says that gaze changes the weight of connectivity already present.[1]
The limits of 47 behavioural results and 42 EEG datasets
The weight this result can carry should not be overstated. The behavioural sample comprised 47 infants and the EEG sample 42, limited to two cities and artificial languages presented on a screen. The learning effect under full gaze was small to moderate, while partial gaze was borderline at an adjusted p value of 0.058. Because the adult video and EEG were prerecorded, there was no real reciprocity, and the mediation analysis still requires causal validation in an independent experiment. The next test should vary gaze availability in a larger sample and during live, bidirectional interaction.[1]
One distinction remains important. Although infants looked at the screen across all three conditions, learning did not emerge equally; total visual attention neither differed between conditions nor predicted learning. This suggests that visible eyes may do more than prolong looking and may change the weight of neural connectivity running from speaker to infant. It would be wrong to turn that into a rule about screens. The narrower result is that a social cue can change which part of an unchanged speech stream is treated as worth learning. Whether real-world reciprocity strengthens that effect remains open.[1]