Stimulus repetition can produce neural response attenuation in stimulus-category selective networks within the occipito-temporal lobe. It is hypothesized that this neural suppression reflects the functional sharpening of local neuronal assemblies which boosts information processing efficiency. This neural suppression phenomenon has been mainly reported during conditions of conscious stimulus perception. The question remains whether frequent stimuli processed in the absence of conscious perception also induce repetition suppression in those specialized networks. Using rare intracranial EEG recordings in the ventral occipito-temporal cortex (VOTC) of human epileptic patients we investigated neural repetition suppression in conditions of conscious and unconscious visual processing of words. To this end, we used an orthogonal design manipulating respectively stimulus repetition (frequent vs. unique stimuli) and conscious perception (masked vs. unmasked stimuli). By measuring the temporal dynamics of high-frequency broadband gamma activity in VOTC and testing for main and interaction effects, we report that early processing of words in word-form selective networks exhibits a temporal cascade of modulations by stimulus repetition and masking: neuronal attenuation initially is observed in response to repeated words (irrespective of consciousness), that is followed by a second modulation contingent upon word reportability (irrespective of stimulus repetition). Later on (> 300 ms post-stimulus), a significant effect of conscious perception on the extent of repetition suppression was observed. The temporal dynamics of consciousness, the recognition memory processes and their interaction revealed in this study advance our understanding of their contributions to the neural mechanisms of word processing in VOTC.