Two experimental approaches--the cellular and the neurogenetic--to the study of molecular mechanisms of elementary memory implicate the cyclic AMP cascade in general, and adenylate cyclase in particular, in the processes of acquisition and short-term memory. Models of learning and memory should account for four basic phenomena: persistence of memory, stimulus convergence, temporal specificity and memory decay. These phenomena thus place constraints on the structure and operation of any postulated memory apparatus. The relevant experimental data derived from the study of memory mutants in Drosophila and of the gill and siphon withdrawal reflex in Aplysia are discussed and analyzed in light of the above mentioned constraints.