Hebb's Vision: The Structural Underpinnings of Hebbian Assemblies

In 1949, Donald Hebb proposed that groups of neurons that activate stereotypically form the organizational building blocks of perception, cognition, and behavior. Finding the structural underpinning of such assemblies has been technically challenging, due to a lack of large-scale structure-activity maps. Here, we analyze this relation using a novel dataset that links in vivo optical physiology to connectivity using postmortem electron microscopy (EM). From the fluorescence traces, we extract neural assemblies from higher-order correlations in neural activity. Physiologically, we show that these assemblies exhibit properties consistent with Hebb's theory, including more reliable responses to repeated natural movie inputs than size-matched random ensembles and superior decoding of visual stimuli. Structurally, we find that neurons that participate in assemblies are significantly more integrated into the structural network than those that do not. Contrary to Hebb's original prediction, we do not observe a marked increase in the strength of monosynaptic excitatory connections between cells participating in the same assembly. However, we find significantly stronger indirect feed-forward inhibitory connections targeting cells in other assemblies. These results show that assemblies can be useful components of perception, and, surprisingly, they are delineated by mutual inhibition.