The induction of long-term potentiation (LTP) is generally assumed to be triggered by Ca2+ entry into dendritic spines via NMDA receptor-gated channels. A previous computational model proposed that spines serve several functions in this process. First, they compartmentalize and amplify increase in [Ca2+]i. Second, they augment the nonlinear relationship between synaptic strength and the probability or magnitude of LTP induction. Third, they isolate the metabolic machinery responsible for LTP induction from increases in [Ca2+]i produced by voltage-gated Ca2+ channels in the dendritic shaft. Here we examine this last prediction of the model using methods that combine confocal microscopy with simultaneous neurophysiological recordings in hippocampal brain slices. Either of two Ca(2+)-sensitive dyes were injected into CA1 pyramidal neurons. Direct depolarization of the neurons via the somatic electrode produced clear increases in Ca2+ signals within the dendritic spines, a result that was not predicted by the previous spine model. Our new spine model suggests that some of this signal could theoretically result from Ca(2+)-bound dye diffusing from the dendritic shaft into the spine. Dye diffusion alone cannot, however, explain the numerous cases in which the Ca2+ signal in the spine was considerably larger than that in the adjacent dendritic shaft. The latter observations raise the possibility of voltage-gated Ca2+ entry directly into the spine or else perhaps via Ca(2+)-dependent Ca2+ release. The new spine model accommodates these observations as well as several other recent experimental results.