A fundamental scientific question concerns the neuronal basis of perceptual consciousness, which encompasses the perceptual experience and reflexive monitoring associated with a sensory event. Although recent human studies identified individual neurons reflecting stimulus visibility, their functional role for perceptual consciousness remains unknown. Here, we provide neuronal and computational evidence indicating that perceptual and reflexive consciousness are governed by an all-or-none process involving accumulation of perceptual evidence. We recorded single-neuron activity in a participant with a microelectrode implant in the posterior parietal cortex, considered a substrate for evidence accumulation, while he detected vibrotactile stimuli around detection threshold and provided confidence estimates. We found that detected stimuli elicited firing rate patterns resembling evidence accumulation during decision-making, irrespective of response effectors. Similar neurons encoded the intensity of task-irrelevant stimuli, suggesting their role for consciousness per se, irrespective of report. We generalized these findings in healthy volunteers using electroencephalography and reproduced their behavioral and neural responses with a computational model. This model considered stimulus detection if accumulated evidence reached a bound, and confidence as the distance between maximal evidence and that bound. Applying this mechanism to our neuronal data, we were able to decode single-trial confidence ratings both for detected and undetected stimuli. Our results show that the specific gradual changes in neuronal dynamics during evidence accumulation govern perceptual consciousness and reflexive monitoring in humans.