When humans perform actions that have a predictable effect in the environment, the intensity of these action-effects is attenuated. This phenomenon is thought to be related to motor based sensory prediction such that when the observed effect matches the prediction, the action-effect is attenuated. In the present paper we develop a new model to describe how this prediction might be implemented in the brain. This model supposes that voluntary action selection involves the preactivation of learnt action-effects. By modeling motor induced preactivation in sensory pathways we were able to generate a number of novel predictions regarding participants' performance in a contrast discrimination task. In order to test these predictions we trained participants to learn action-effect contingencies between left and right hand button presses and letter stimuli. We found a significant reduction in contrast discrimination sensitivity for stimuli that were congruent with these learnt action-effect associations. Furthermore, using participants' contrast ratings we were also able to show that this reduction in contrast sensitivity was driven by an increase in the internal response for lower contrast stimuli, consistent with the notion that sensory attenuation results from preactivation of learnt sensory action-effects. This provides a novel account of how motor prediction drives sensory attenuation of action-effects.