Recent years have seen dramatic advancement in the measurement of biology at a systems level. Researchers routinely obtain thousands or millions of simultaneous measures of dynamic systems. In humans, this includes neuroimaging, which can be used to probe the brain bases of affect and emotion in increasingly sophisticated ways. Neuroimaging can provide measures of activity in 300,000 brain locations and 60 billion functional associations every second. However, the complexity of these measures presents new challenges in maintaining scientific transparency and reproducibility. In this talk, I describe several new models of the brain bases of affective processes, including models that predict the intensity of negative affect, autonomic responses, prosocial emotions, and pain. These models reduce complex neuroimaging data to measures that can be readily replicated and generalized across laboratories. They can be tested prospectively on new participants, providing unbiased estimates of effect size that are often dramatically larger than single regions from standard brain maps. By asking which stimuli and psychological states these measures respond to across studies, we can induce the nature of their associated psychological constructs, providing a foundation for understanding how affect and emotion are generated in the brain.