Isaac Treves

Abstract

Current theories value top-down processing as central to the rich predictive abilities of the adult human brain. It remains unclear whether infants have top-down prediction, and if so, how and where it emerges in development. Brain imaging in infants is a crucial method for answering these questions. The work of Emberson et al. with functional near-infrared spectroscopy, (fNIRS), suggests global, top-down learning may be present in infancy1. In this experiment, we aim to isolate top-down neural activity only attributable to prefrontal involvement. We developed a sequence learning task which should require this prefrontal, top-down processing. The task has one ‘fixed’ condition with a learnable sequence and another ‘random’ condition with the same stimuli but no consistent sequence. We hypothesized prefrontal, top-down learning should cause attenuation of neural activity in the fixed condition, but not the random. We conducted the task with adults, allowing us to conduct fNIRS imaging with concurrent eye-tracking. Eye-tracking results confirmed that learning took place in the fixed condition. In addition, the fixed condition had significantly less activity than the random condition in the occipital and prefrontal cortices, in the deoxyhemoglobin signal. Yet the oxyhemoglobin signal and our block-by-block learning analyses showed no significant results. It is clear that there are limitations with our experiment—including rest period noise and mixing of attention and expectation effects. Whether fNIRS can be used to ask high-level questions about learning deserves serious consideration.