Midbrain dopamine seems to play an outsized role in motivated behavior and learning. Widely associated withmediating reward-related behavior, decision-making and learning, dopamine continues to generatecontroversies in the field. While many studies and theories focus on what dopamine cells encode, the questionof how the midbrain derives the information it encodes is poorly understood and comparatively less addressed.Recent anatomical studies suggest greater diversity and complexity of afferent inputs than previouslyappreciated, requiring rethinking of prior models. Here we elaborate a hypothesis that construes midbraindopamine as implementing a Bayesian model selector in which individual dopamine cells sample afferentactivity across distributed brain substrates, comprising evidence to be evaluated on whether current distributedneural activity is 'important.' To effectively generate a temporally resolved phasic signal, a population ofdopamine cells must exhibit synchronous activity. We argue that synchronous activity across a population ofdopamine cells signals consensus, invigorating responding to recognized opportunities and facilitating furtherlearning. In framing our hypothesis, we shift from the question of how value is computed to the broaderquestion of how the brain achieves coherence across distributed, parallel processing. We posit the midbrain ispart of an 'axis of agency' in which the prefrontal cortex, basal ganglia and midbrain form an axis mediatingcontrol, coherence and consensus, respectively.