Prior expectations can alter one's decisions about a stimulus. Expectation signals may modify sensory encoding of the stimulus, and/or may alter how the decision emerges from sensory evidence. Since area LIP has been proposed as a site for evidence accumulation in a direction discrimination task (Roitman and Shadlen 2002), we examined how expectations might modify activity in LIP. In our task, an arrow-shaped cue, pointing in one of two opposite directions, is presented for 200ms at the fovea. Following a 150 to 800ms delay, a foveal random-dot motion stimulus is displayed for 1000ms. The stimulus contained variable-coherence global motion either along or opposite the cued direction (Britten et al. 1992). The monkey is then free to make a saccade to one of two targets and is rewarded for choosing the target in the direction of the stimulus motion. The cue's direction matched that of the motion on 67% of the trials. The monkey learned to significantly but incompletely bias his choices towards the cued direction, especially on trials with weak, low-coherence motion. This bias suggests that the monkey integrates its prior expectation with the motion information to improve its performance. LIP neurons were selected based on spatial tuning and maintained activity in a memory saccade task. In our discrimination task, the motion axis and target locations were chosen so that one of the saccade targets (T1) lay within the neuron's response field (RF). Consistent with prior work, LIP responses ramped up prior to T1 choices, and the rate of ramping activity increased with motion coherence. The main finding was that LIP responses were elevated when the monkey expected motion towards the RF. Specifically, from roughly the onset of the motion, the neural response was increased on trials with cues towards the RF, relative to trials with cues pointing away from the RF. In addition, we tested some cells with a motion stimulus lasting only 250ms (instead of 1000ms). We observed a much larger increase in LIP activity starting around the motion onset on trials with cues towards the RF. Because there is less motion signal to evaluate on these short-duration trials, the choice is influenced more by the expectation, and this is reflected in the LIP activity. It is notable that the slope of the ramping activity prior to T1 choices was not altered by the direction of the prior cue. This is consistent with our previous finding (SFN 2006) that the cue had essentially no effect on responses in area MT. Together, these findings suggest that expectation in our task acts on the computation of decision variables rather than on the sensory representation of motion.