The effect of reward is thought to be pervasive. It affects our motor actions and our sensory judgements. We have investigated the influence of reward on perceptual decision-making at multiple levels in humans and macaque monkeys. We set up a behavioural paradigm that could be employed with both species. The perceptual decision-making task involved reporting the direction of rotation of a structure-from-motion cylinder made up of two transparent sheets of random dots. Cylinder rotation could be disambiguated by applying binocular disparity to separate the front and back surface of the cylinder. This allowed us to present a range of stimuli in each experiment, including the case of an ambiguous cylinder stimulus whose direction of rotation could either be perceived as clockwise or counterclockwise on a given trial. Two Rhesus macaques were each implanted with a headpost, eye coils and a recording chamber under general anaesthesia (7-8 mg/kg/hr alfaxalone/alfadolone acetate i.v. or 1-3% isoflurane gas, to effect) (Dodd et al 2001; Read and Cumming 2003). They were trained to report the direction of rotation of the cylinder by making a saccade to one of two choice targets. They received a fluid reward for a correct decision; for ambiguous stimuli they received a fluid reward on 50% of randomly selected trials. The amount of fluid they received increased in a predictable fashion after two and then four consecutive, correct choices. We used electrical microstimulation in extra-striate visual area V5/MT to distinguish the separate roles of sensory neural signals arriving in V5/MT (and therefore not influenced by microstimulation) from the combined neuronal signals downstream of V5/MT (which are influenced by microstimulation). Electrical stimulation of a cluster of neurons selective for a particular direction of cylinder rotation biased visual perception. Psychometric functions for discrimination of cylinder rotation were consistently shifted, with the animal making more choice in the preferred direction of cylinder rotation at the stimulation site. Importantly, the size of this electrical microstimulation effect on the psychometric function was different when the reward available at the end of a trial was large or small (Wilcoxon signed-rank test: p<0.001, n = 48 stimulation sites). This result suggests that reward affected sensory signals separately from the artificially introduced electrical signal rather than simply acting upon the combined neural signal downstream of V5/MT. When we fitted psychometric functions with a simplified drift diffusion model, the best fitting model required a significant change in drift rate as well as a change in bound, indicating that reward affects both the early sensory stages and the later decision-related stages of cortical processing. Further evidence for this conclusion comes from a study with six human participants carrying out the same perceptual task under different reward conditions, in which the reward is points converted into money at the end of the experiment. Preceding each trial, human participants were cued to indicate whether both choices were rewarded equally or one higher than the other (8 versus 1 point). Only correct responses were rewarded; for ambiguous stimuli, half the trials were rewarded at random. Feedback of points scored was given at the end of every trial and the cumulative score was given at regular intervals during the experiment. Under these conditions, participants showed a consistent behavioural response bias towards more highly rewarded choices for the same cylinder stimulus. Analysis with a drift diffusion model again implicated changes in drift rate to account for the effect of reward, the stimulus parameter that represents the strength of the sensory evidence being integrated to the decision (Model fit t-test comparison across individuals, p<0.05). Human magnetoencephalography (MEG) recordings with a beamformer analysis showed that after the cue onset but prior to stimulus onset, there was activation for biased reward conditions over balanced reward in both V5/MT at around 475 msec and dorso-lateral prefrontal cortex (dPFC) at 590 msec. When the cylinder stimulus on a given trial had a direction of rotation that was highly not lowly rewarded, we found cortical activation in the precuneus in visual cortex at 85 msec after stimulus onset followed by dPFC at 125 msec and other prefrontal areas during stimulus presentation. These results, obtained with humans and monkeys, show that signals in visual cortex are affected by the reward available on a given trial. This suggests that in perceptual decision tasks, one mechanism by which reward could influence behaviour might be through biasing of signals in the visual processing areas of the brain.