Long-term potentiation (LTP) of synaptic transmission in the hippocampus is a much-studied example of synaptic plasticity. Although the role of N-methyl-D-aspartate (NMDA) receptors in the induction of LTP is well established, the nature of the persistent signal underlying this synaptic enhancement is unclear. Involvement of protein phosphorylation in LTP has been widely proposed, with protein kinase C (PKC) and calcium-calmodulin kinase type II (CaMKII) as leading candidates. Here we test whether the persistent signal in LTP is an enduring phosphoester bond, a long-lived kinase activator, or a constitutively active protein kinase by using H-7, which inhibits activated protein kinases and sphingosine, which competes with activators of PKC (ref. 17) and CaMKII (ref. 18). H-7 suppressed established LTP, indicating that the synaptic potentiation is sustained by persistent protein kinase activity rather than a stably phosphorylated substrate. In contrast, sphingosine did not inhibit established LTP, although it was effective when applied before tetanic stimulation. This suggests that persistent kinase activity is not maintained by a long-lived activator, but is effectively constitutive. Surprisingly, the H-7 block of LTP was reversible; evidently, the kinase directly underlying LTP remains activated even though its catalytic activity is interrupted indicating that such kinase activity does not sustain itself simply through continual autophosphorylation (see refs 9, 13, 15).