Statistical learning of a sequence of emotional faces in 12-month-old infants

• In Event: Poster Session 02
  In Poster Session: PS 02 Topic - Attention, Learning, Memory

Abstract

Previous research suggests that infants are capable of extracting statistical regularities from a continuous stream of elements, an ability that helps them making sense of the surrounding world. This capacity, referred to as statistical learning (SL), is thought to appear early on in development (Bulf et al., 2011; Teinonen et al., 2009), and to support infants’ learning in many different domains. Indeed, infants have been shown to extract statistical regularities from sequences of words (Saffran et al., 1996; Teinonen et al., 2009), sounds (Hannon & Johnson, 2005; Saffran et al., 1999), events (Roseberry et al., 2011), actions (Monroy et al., 2017, 2019) and many other types of stimuli. However, no research so far has examined infants’ SL when presented with a sequence of emotional faces. This ability may be crucial for infants’ understanding of social interactions, in which facial expressions of emotions often succeed one another following specific patterns, reflecting the mental states of the others.
The current study aimed to fill this gap by investigating 12-month-olds’ ability to extract statistical information from a sequence of emotional faces containing salient social cues. Infants were presented with videos of 2 actresses expressing the same facial emotion, and subsequently turning towards each other while maintaining the expression at its peak. Infants were first familiarized with a sequence of 8 videos, each video displaying a different emotion (i.e., anger, happiness, fear, sadness, surprise, amusement, disgust, exasperation). During the familiarization phase, four fixed pairs of videos were presented so that the transitional probabilities were of 1.0 within each pair (units) and 0.5 between pairs (part-units) (Figure 1). During the test phase, infants were shown units in alternation with part-units. We expected 12-month-olds to be able to extract the statistical information embedded in the sequence to which they were familiarized, showing a preference (i.e., longer looking times) for the low predictable transitions (i.e., part-units).
A repeated-measures ANOVA was performed on looking times at test, with Trial type (units, part-units) as within-subjects factor and Test trial order (familiar first, novel first) as a between-participants factor. The analysis revealed a main effect of Trial type, p=0.029 but no main effect or interaction with Trial order. A post-hoc paired-samples t-test showed that infants looked significantly longer to part-units (M = 58.3 s; SD = 27.6) than units (M =49.9 s; SD = 25.5); p = .03 (Figure 2). These results suggest that 12-month-old infants were able to extract the statistical information embedded in the sequence of emotional expressions within a salient social context.

Authors

• Julia Mermier, University of Milano-Bicocca
  Presenting Author

• Ermanno Quadrelli, University of Milan-Bicocca
  Non-Presenting Author

• Chiara Turati, Univesrità di Milano-Bicocca
  Non-Presenting Author

• Hermann Bulf, University of Milano-Bicocca
  Non-Presenting Author