The transformation of sensory inputs to appropriate goal-directed actions requires estimation of sensory-cue values based on outcome history. To clarify how cortical neurons represent an outcome-predicting cue and actual outcome, we conducted wide-field and two-photon calcium imaging of the mouse neocortex during performance of a classical conditioning task with two cues with different water-reward probabilities. Although licking behavior dominated the area-averaged activity over the whole dorsal neocortex, dorsomedial frontal cortex (dmFrC) affected other dorsal frontal cortical activities, and its inhibition extinguished differences in anticipatory licking between the cues. In individual frontal cortical neurons, the reward-predicting cue was not simultaneously represented with the current or past reward, but licking behavior was frequently multiplexed with the reward-predicting cue and current or past reward. Deep-layer neurons in dmFrC most strongly represented the reward-predicting cue and recent reward history. Our results suggest that these neurons ignite the cortical processes required to select appropriate actions.individual neurons within subdivisions of M2 during simple tasks and to analyze how individual neurons encode the multiple types of information. We assumed that the neuronal representations of the cue value and outcome information in classical conditioning, a condition that does not require the learning of an association between a specific action and its consequence, is the basis for the processes of the transformation of sensory cues to appropriate motor actions. In classical conditioning tasks, striatal neurons represent information on reward-predicting cues (cue values), present reward (including reward prediction error), and recent reward history (reward in the preceding trial), as well as behaviors such as licking (Bloem et al., 2017;Yoshizawa et al., 2018).However, it is still unclear how the dorsal neocortex, including M2, represents such information, which is generally required for decision-making.To extract these information types, we trained head-fixed mice to perform a classical conditioning task with two sound cues assigned to different probabilities of water delivery (Bloem et al., 2017;Oyama et al., 2015;Shin et al., 2018). We conducted widefield calcium imaging of the entire dorsal cortex (Allen et al., 2017;Makino et al., 2017;Musall et al., 2019) during the training sessions. In addition, in the late stage of learning, we conducted two-photon calcium imaging of three dorsal frontal areas up to a depth of 800 µm from the cortical surface (dorsomedial frontal cortex [dmFrC] corresponding to antero-medial M2, dorsolateral frontal cortex [dlFrC] corresponding to antero-lateral M2 [or occasionally referred to as the anterolateral motor area; ALM], and the primary motor cortex [M1] corresponding to the caudal forelimb motor area (Franklin and Paxinos, 2007;Hira et al., 2013aHira et al., , 2013bKomiyama et al., 2010;Svoboda and Li, 2018;Tennant et al., 2011), and the medial prefrontal...