Playing “Duck Duck Goose” With Neurons

Abstract

Reduced connectivity between sending and receiving neurons (i.e., synaptic depression) may facilitate change detection by reducing responses for recently viewed objects so new objects can be highlighted. In the experiment reported here, we investigated high-level change detection following semantic satiation, which is the loss of meaning following repetition of a word. A computer simulation of a word-reading neural network with synaptic depression identified key predictions of connectivity reduction. A dynamic-causal-modeling analysis of magnetoencephalography (MEG) responses collected during a category-matching task identified connectivity reduction between a cortical region related to orthography and a cortical region related to semantics as the cause of the reduced MEG response to a repeated word. As predicted, prior repetitions of a category-matching word presented immediately after the repeated word enhanced semantic novelty, as measured with the M400 component. These results demonstrate that a combination of neural-network modeling and connectivity analyses can reveal the manner in which connectivity fluctuations underlie cognitive functions.