The purpose of the study was to find out what spatial frequency information human observers use in the recognition of face images. Signal-to-noise ratio thresholds for the recognition of facial images were measured as a function of the centre spatial frequency of narrow-band additive spatial noise. The relative sensitivity of recognition to different spatial frequencies was derived from these results. The maximum sensitivity was found at 8–13 c/face width and the bandwidth was just under two octaves. Qualitatively similar results were obtained with stimuli in which Fourier phase was randomised within a narrow band of different centre spatial frequencies. This resulted in a considerable increase of energy threshold around 8 c/face width and less elsewhere. Further, contrast energy thresholds were measured as a function of the centre spatial frequency of band-pass filtered face images. As a function of object spatial frequency (c/face width), energy threshold first decreased and then increased. The lowest energy thresholds found around 10 c/face width were lower than the energy threshold for unfiltered images. This is what one would expect if face recognition is narrow-band, since band-pass filtered images of optimal centre spatial frequency do not contain unused contrast energy at low and high spatial frequencies. In conclusion, the results suggest that the recognition of facial images is tuned to a relatively narrow band (<2 octaves) of mid object spatial frequencies.