Most attempts to localize physical correlates of memory in the central nervous system (CNS) rely on ablation techniques. This approach has the limitation of defining just one of an unknown number of structures necessary for memory formation. We have used the Drosophila rutabaga type I Ca(2+)/CaM-dependent adenylyl cyclase (AC) gene to determine in which CNS region AC expression is sufficient for memory formation. Using pan-neural and restricted CNS expression with the GAL4 binary transcription activation system, we have rescued the memory defect of the rutabaga mutant in a fast robust spatial learning paradigm. The ventral ganglion, antennal lobes, and median bundle are likely the CNS structures sufficient for rutabaga AC- dependent spatial learning.