Associative memory of the mollusc Hermissenda crassicornis, previously correlated with changes of specific K+ currents, protein phosphorylation, and increased synthesis of mRNA and specific proteins, is here shown to be accompanied by macroscopic alteration in the structure of a single identified neuron, the medial type B photoreceptor cell. Four to five days after training, terminal arborizations of B cells iontophoretically injected with Ni2+ ions and then treated with rubeanic acid were measured with charge-coupled device (CCD)-digitized pseudocolor images of optical sections under "blind" conditions. Boundary volumes enclosing medial-type B-cell arborizations from classically conditioned animals were unequivocally reduced compared with volumes for naive animals or those trained with unpaired stimuli. Branch volume magnitude was correlated with input resistance of the medial type B-cell soma. Such associative learning-induced structural changes may share function with "synapse elimination" described in developmental contexts.Associative memory of the mollusc Hermissenda crassicornis-i.e., a remembered link between at least two discrete stimuli previously associated in time-has been shown to be accompanied by changes of membrane currents (1-4), changes in protein phosphorylation (5), and changes in synthesis of mRNA and of specific proteins (6, 7). These changes have been measured either in single identified neurons or in their immediate vicinity and can account for storage and recall of a Pavlovian conditioned response by the intact animal. Morphological changes of neurons, unlike those changes just mentioned, have never been directly related to associative memory of Hermissenda or any other species, although they have been observed to occur in a variety of developmental (8-13) and training (14)(15)(16)(17)(18) contexts. An inherent difficulty with the training experiments is to distinguish those training-induced modifications due to motor activity and sensory stimulation themselves from those that store memory for later recall. Such a distinction is possible with an associative paradigm that includes a control group receiving sensory stimuli not associated in time but with intensity and duration identical to those used to associatively train the animals.Here we report marked anatomic changes in a single identified neuron (the medial type B cell) of Hermissenda in animals (paired group, n = 8) that had been associatively trained with paired visual and vestibular stimuli. These changes were measured in comparison with neurons from animals receiving either no training stimuli (naive group, n = 8) or to neurons from animals receiving stimuli that were explicitly unpaired at randomly varying temporal intervals (unpaired group, n = 8). Because there were no differences between type B cells from naive animals compared with cells from unpaired animals, the anatomic changes described below for the paired group were specific to the temporal association of the training stimuli and, therefore, are not due to the...