Abstract Activation of the cAMP pathway was found to be implicated in the memory process. In the context-signal learning paradigm of the crab Chasmagnathus, the protein kinase (PKA) activator Sp-5,6-DCl-cBIMPS facilitated long-term memory (LTM) induced by spaced training while the PKA inhibitor 8-chloroadenosine-3′, 5′-monophosphorothioate, Rp-isomer (Rp-8-Cl-cAMPS) produced amnesia. In the present report the effect of the PKA inhibitor on long-term retention was assessed when administered (systemic injection of 2 μg/animal) at various times after training. According to previous results obtained with a lower dose, retention is impaired when the drug is administered immediately pretraining. An effect on acquisition was ruled out considering that the drug did not affect the performance during training. On the contrary, no effect of the PKA inhibitor was found with an immediately posttraining injection and amnesia was observed only when training was shortened from 15 to 12 trials (training duration from 45 to 36 min). At 2 and 12 h posttraining Rp-8-Cl-cAMPS injection failed to impair retention, but amnesia was found when the drug was injected at 4 and 8 h after training. In order to assess a possible effect of the drug in retrieval, the PKA inhibitor was administered 15 min before testing, and no amnestic effect was observed. These results suggest that two phases of PKA activity are required during consolidation of LTM, one during training and the other between 4 and 8 h after training. The link between these two periods of PKA activation and the two phases of the transcription factor NF-κB activation previously found in this model, as well as the similar time course found in rodents, is discussed. An amnestic effect of the drug was not found when administered immediately before a massed training protocol that yielded an intermediate-term memory, suggesting that in this type of memory PKA activation is not required.