You are viewing the SNL 2018 Archive Website. For the latest information, see the Current Website.

Poster A24, Thursday, August 16, 10:15 am – 12:00 pm, Room 2000AB

Neural oscillations as a brain signature of statistical learning?

Louisa Bogaerts1, Ayelet N. Landau1, Craig G. Richter2, Ram Frost1,2,3;1The Hebrew University of Jerusalem, Israel, 2Basque center on Cognition, Brain and Language, 3Haskins Laboratories

Statistical learning (SL), the ability to extract distributional properties of sensory input across time and space, is taken to be the main mechanism by which cognitive systems discover the underlying regularities of the environment. Since the seminal demonstration of Saffran and colleagues (1996) that infants are able to segment speech on the basis of transitional probabilities, a large number of studies have demonstrated that people often display a remarkable sensitivity to the co-occurrence of stimuli in an input stream. This was shown across a range of stimuli, for newborns as well as adults, and across sensory modalities. By highlighting experience-based principles for detecting regularities SL research has offered a new perspective on how language regularities are acquired. Recent neuroimaging studies have associated SL with domain-general regions responsible for binding temporal and spatial contingencies in different modalities (hippocampus, medial temporal lobe), as well as with domain-specific visual and auditory cortical networks. Here we aim to go beyond the neurobiological “where” of SL and use electro-encephalography (EEG), which can reveal oscillatory activity. Neuronal oscillations reflect rhythmic fluctuations in the inhibition/excitation balance of neuronal populations and have been proposed to be instrumental in accounting for sensory processing, attentional selection and memory formation. We present data (n=36) from a classical visual SL task. This task consists of a familiarization phase in which participants are repeatedly exposed (3 blocks of 18 repetitions each) to 8 triplets of shapes, embedded in a continuous stream. Subsequently, learning is assessed behaviorally through a set of two-alternative-forced-choice questions. The intriguing possibility we explore is that pre-stimulus neural oscillations in the Delta and/or Alpha-Beta range may provide a brain signature of the anticipation of the predictable stimuli in a sequence, and hence of regularity learning. Our behavioral results indicate that 65% of the tested participants had significant above-chance performance at the individual level. For those “learners” we found 1) increased Delta activity (2-4Hz) over frontal and centro-parietal electrodes in the 400ms to 100ms time window before the presentation of predictable shapes relative to unpredictable shapes and 2) a similar modulation of centro-parietal activity in the 11-14Hz frequency range, in the 150ms prior to stimulus onset. Importantly, looking at the entire sample, the relative difference in Delta power measured in the last learning block was highly correlated with behavioral learning outcomes. These findings will be discussed in the context of the believed functional relevance of neural oscillations in regularity learning and predictive processing. Revealing a spectral signature of learning holds the promise of offering an online learning measure providing critical insights regarding the mechanisms of SL.

Topic Area: Perception: Orthographic and Other Visual Processes