The synaptic tagging and capture hypothesis of protein synthesis-dependent long-term potentiation asserts that the induction of synaptic potentiation creates only the potential for a lasting change in synaptic efficacy, but not the commitment to such a change. Other neural activity, before or after induction, can also determine whether persistent change occurs. Recent findings, leading us to revise the original hypothesis, indicate that the induction of a local, synapse-specific 'tagged' state and the expression of long-term potentiation are dissociable. Additional observations suggest that there are major differences in the mechanisms of functional and structural plasticity. These advances call for a revised theory that incorporates the specific molecular and structural processes involved. Addressing the physiological relevance of previous in vitro findings, new behavioural studies have experimentally translated the hypothesis to learning and the consolidation of newly formed memories.