Calcium ions enter through discrete ion channels at presynaptic nerve terminals before binding to and activating transmitter release sites. Opposing models hold that release sites are gated either by calcium domains of single, closely associated channels or by extensive, overlapping domains from many remote channels. At the chick calyx synapse we find a linear relation between transmitter release and the number of open calcium channels, favouring single domain activation. This finding is consistent with results from the squid giant synapse but contrasts with steep power dependences reported in rodent synapses, suggestive of activation by extensive overlapping domains. These different reports were reconciled by plotting 'per cent domain overlap' against the external calcium concentration used for each species. This relationship predicts the involvement of local channels in the activation of release sites in all species. Further, it suggests that each release site is activated by calcium ions from its immediately associated channels and not by ions that enter through channels associated with a neighbouring release site.