Representing the probability and uncertainty of outcomes facilitates adaptive behavior by allowing organisms to prepare in advance and devote attention to relevant events. Probability and uncertainty are often studied only for valenced (appetitive or aversive) outcomes, raising the question whether the identified neural machinery also processes the probability and uncertainty of motivationally neutral outcomes. Here, we aimed to dissociate valenced from valence-independent (i.e., generic) probability (p; maximum at p=1) and uncertainty (maximum at p=0.5) signals using human neuroimaging. In a Pavlovian task (n=41; 19 females), different cues predicted appetitive, aversive, or neutral liquids with different probabilities (p=0, p=0.5, p=1). Cue-elicited motor responses accelerated, and pupil sizes increased primarily for cues that predicted valenced liquids with higher probability. For neutral liquids, uncertainty rather than probability tended to accelerate cue-induced responding and decrease pupil size. At the neural level, generic uncertainty signals were limited to occipital cortex, while generic probability also activated anterior ventromedial prefrontal cortex. These generic probability and uncertainty signals contrasted with cue-induced responses that only encoded the probability and uncertainty of valenced liquids in medial prefrontal, insular and occipital cortices. Our findings show that the brain processes probability and uncertainty in a generic fashion. Moreover, the behavioral and neural dissociation of generic and valenced signals indicates that the brain keeps track of motivational charge and highlights the need and usefulness of characterizing the exact nature of learned representations.Significance StatementEncoding the probability and uncertainty of outcomes is important for adaptive behavior. Here we ask to what extent the brain represents probability and uncertainty regardless of whether the predicted outcomes are valenced (i.e. motivationally relevant) or generic (i.e., valence-independent). We dissociate generic from valenced variables by using not only cues that predict appetitive or aversive outcomes, but also cues that predict neutral outcomes. Our data reveal distinct behavioral effects and largely separate neural representations of valenced and generic variables. For example, valenced probability activated more proximal parts of medial prefrontal and occipital cortex whereas generic probability activated more distal parts. Thus, the representation of probability and uncertainty is multiplexed, allowing for tailored information processing according to computational needs.