The effects of gaze direction on memory for faces were studied in children from three different age groups (6-7, 8-9, and 10-11 years old) using a computerized version of a task devised by Hood, Macrae, Cole-Davies and Dias (2003). Participants were presented with a sequence of faces in an encoding phase, and were then required to judge which faces they had previously encountered in a surprise two-alternative forced-choice recognition test. In one condition, stimulus eye gaze was either direct or deviated at the viewing phase, and eyes were closed at the test phase. In another condition, stimulus eyes were closed at the viewing phase, with either direct or deviated gaze at the test phase. Modulation of gaze direction affected hit rates, with participants demonstrating greater accuracy for direct gaze targets compared to deviated gaze targets in both conditions. Reaction times (RT) to correctly recognized stimuli were faster for direct gaze stimuli at the viewing phase, but not at the test phase. The age group of participants differentially affected these measures: there was a greater hit rate advantage for direct gaze stimuli in older children, although RTs were less affected by age. These findings suggest that while the facilitation of face recognition by gaze direction is robust across encoding and recognition stages, the efficiency of the process is affected by the stage at which gaze is modulated.