Studies of neurons from human epilepsy tissue and comparable animal models of focal epilepsy have consistently reported a marked decrease in dendritic spine density on hippocampal and neocortical pyramidal cells. Spine loss is often accompanied by focal varicose swellings or beading of dendritic segments. An ongoing excitotoxic injury of dendrites (dendrotoxicity), produced by excessive release of glutamate during seizures, is often assumed to produce these abnormalities. Indeed, application of glutamate receptor agonists to dendrites can produce both spine loss and beading. However, the cellular mechanisms underlying the two processes appear to be different. One recent study suggests NMDA-induced spine loss is produced by Ca2+-mediated alterations of the spine cytoskeleton. In contrast, dendritic beading is not dependent on extracellular Ca2+; instead, it appears to be produced by the movement of Na+ and Cl- intracellularly and an obligate movement of water to maintain osmolarity. A decrease in dendritic spine density was recently reported in a model of recurrent focal seizures in early life. Unlike results from other models, dendritic beading was not observed, and other signs of neuronal injury and death were absent. Thus, additional mechanisms to those of excitotoxicity may produce dendritic spine loss in epileptic tissue. A hypothesis is presented that spine loss can be a product of a partial deafferentation of pyramidal cells, resulting from an activity-dependent pruning of neuronal connectivity induced by recurring seizures. The dendritic abnormalities observed in epilepsy are commonly suggested to be a product and not a cause of epilepsy. However, anatomical remodeling may be accompanied by alterations in molecular expression and targeting of both voltage- and ligand-gated channels in dendrites. It is conceivable that such changes could contribute to the neuronal hyperexcitability of epilepsy.