Curiosity-based learning: A unique EEG signature of intermediate perceptual complexity

According to curiosity-based models, learning occurs at intermediate levels of
complexity, relative to existing mental representations. Low-level complexity does not
stimulate significant learning, as similar information is already mentally represented,
whereas overly complex information is difficult to integrate into existing
representations. However, little is known about the brain mechanisms underlying
preference for intermediate levels of complexity. The aim of this study was to address
this question by investigating EEG correlates of perceptual complexity. We quantified
complexity by the perceptual difference between category exemplars. The stimulus set
consisted of nineteen colour images of novel animals, where attributes such as posture (stooping to standing) and size of wing were varied incrementally. Images were presented successively, differing by either one, three or seven positions along the continuum; corresponding to small, intermediate or large levels of difference,
respectively. The final sample included fifteen 9-month-old infants. Stimuli were
presented on the screen consecutively, while EEG was recorded at 128 electrodes.
Event-related potentials (ERPs) were calculated and compared for the three conditions
(small, intermediate or large difference). Visual ERPs differed between small vs.
intermediate and small vs. large differences as early as 60-130 ms after stimulus onset, indicating less effort to process an image if it was preceded by a similar one. Central, attention- and memory-related ERPs between 500-600 ms were larger for large differences compared to small and intermediate differences, indicating larger
attentional allocation when an image is very different from the previous one. More
importantly, central ERPs between 350-450 ms were larger for intermediate differences compared to small or large differences. This indicates distinct attentional and/or encoding mechanisms when two images, presented consecutively differed at an intermediate level. In conclusion, the brain responds differently to levels of complexity and we found an EEG signature that reflects the role of intermediate complexity in perceptual learning.