Memory is essential for our day-to-day life. It has long been hypothesized that memories are stored in the brain as "engrams", yet the identity and nature of such engrams remains elusive. Our studies centered on three lines of investigation: isolating memory traces (engrams) at the level of neuronal ensembles to tag them for subsequent rapid manipulation; testing whether these neuronal ensembles are sufficient for memory recall; and finally, testing whether it is possible to optically activate a specific memory trace during the simultaneous presentation of an emotionally salient event to induce an artificial association between the two. To identify and control the neuronal ensembles underlying a particular memory, we combined optogenetics with an activity-dependent, doxycycline-regulatable system. Using channelrhodopsin-2 (ChR2), we labeled a population of neurons active during contextual fear conditioning in the dentate gyrus (DG) of the hippocampus, and later selectively activated these cells with light stimulation. By doing so, we artificially induced fear recall in a context different from the context used for fear conditioning. This suggests that DG neuron populations active during contextual fear memory acquisition is sufficient for fear memory recall upon reactivation, causally pinpointing these neurons as the cellular basis of a memory engram. Light-activation of these engram-bearing cells allowed us to alter the memory or even create memories that never existed in reality. This powerful system grants us the ability to dissect and control memory engrams at cellular and neuronal ensemble level, thus greatly enhances our understanding for the mechanisms of learning and memory.