Segmentation through Learning in a Cultured Neuronal Network in Response to Repetitive Stimulation

Abstract

In this study, we hypothesized that the segmentation of continuous stimuli in cultured neuronal networks would emerge as an increase in the reproducibility of evoked response patterns to a specific repeated sequence of input stimulation. We examined whether cultured neuronal networks could segment a specific stimulus pattern from random continuous stimulation. This was evaluated both before and after conditioning with a specific stimulus pattern. We defined stimulation applied from three spatially distinct electrodes (A, B, and C) as“ letters. ”Based on combinations of these“ letters, ”we constructed a“ target stimulus pattern ”consisting of a single stimulation (C), a doublet stimulation (BC), or a triplet stimulation (ABC). These were repeatedly presented to the cultured neuronal networks as conditioning stimuli. Before and after conditioning, we applied a test stimulus sequence in which all possible singlets (e.g., A, B), doublets (e.g., AB, CC), and triplets (e.g., ACC, BBB) composed of A, B, and C occurred with equal frequency. We then analyzed the reproducibility of all evoked response patterns appearing in the test sequence. In several samples, we observed that only the evoked response patterns corresponding to the target stimulus patterns became selectively more reproducible. These findings suggest that the segmentation of continuous stimuli―a fundamental ability underlying language acquisition―can be implemented in a cultured neuronal network.