We have used mouse molecular genetic approaches to study memory mechanisms. We found that the autophosphorylation of alphaCaMKII (alpha-isoform of calcium/calmodulin-dependent kinase II), a major kinase in the excitatory post-synapse, is essential for NMDA receptor-dependent LTP (long-term potentiation) in hippocampal area CA1. This LTP deficit correlates with impaired one-trial learning in aversive conditioning tasks, suggesting that LTP is required for rapid, one-trial learning. However, despite impaired CA1 LTP, memory can be formed in aversive conditioning tasks after multiple training trials. We found evidence that this memory is dependent on increased PSD-95 expression followed by the generation of multi-innervated spines, a type of synapse where a dendritic spine receives more than one presynaptic input. We suggest that generation of multi-innervated spines is a slow learning mechanism when LTP is impaired.