A CENTRAL assumption about long-term potentiation in the hip-pocampus is that the two classes of glutamate-receptor ion channel, the N-methyl-D-aspartate (NMDA) and the kainate/quisqualate (non-NMDA) subtypes, are co-localized at individual excitatory synapses^1,2. This assumption is important because of the perceived interplay between NMDA and non-NMDA receptors in the induc-tion and expression of long-term potentiation: the NMDA class, by virtue of its voltage-dependent channel block by magnesium^3,4 and calcium permeability^5,6, provides the trigger for the induction of long-term potentiation, whereas the actual enhancement of synaptic efficacy is thought to be provided by the non-NMDA class^7,9. If both receptor subtypes are present at the one synapse, such cross-modulation could occur rapidly and locally through diffusible factors. By measuring miniature synaptic currents in cultured hippocampal neurons we show that the majority (∼70%) of the excitatory synapses on a postsynaptic cell possess both kinds of receptor, although to different extents. Of the remaining excita-tory synapses, ∼20% contain only the non-NMDA subtype and the rest possess only NMDA receptors. This finding provides direct evidence for co-localization of glutamate-receptor subtypes at individual synapses, and also points to the possibility that long-term potentiation might be differentially expressed at each synapse according to the mix of receptor subtypes at that synapse.