The vast majority of drugs fail when trying to go from animals to humans. Some fields fail more than others, with psychiatry being at the pinnacle of a competition no field wants to win. Understanding behavior has been difficult enough, delineating what went awry at the genetic and environmental level throughout development, to give rise to a myriad of disorders characterized by their behavior – is there a haystack big enough for that one needle? Then developing drugs targeting those mechanisms? First and foremost it is our job to delineate neural mechanisms underlying normal behavior. To do that, we must utilize tasks with domain-task specificity, wherein we are confident that the cognitive domain measured in animals matches that tested in humans. We can validate these tasks, examining their face, predictive (including pharmacological), and neurobiological validities. For treatment development, we also need to make sure that it is a domain worth targeting – do patients exhibit deficient performance, does the task have clinical sensitivity? I will present data we have generated over the past two decades demonstrating this process, how translational task-development can work in forward and reverse, how working with clinicians enabled testing the clinical sensitivity of even some of the simplest tasks in rodents such as exploration, and effortful motivation. This work culminates in testing potential treatments that patients guide us toward, such as cannabis for attenuating risk-taking in people with bipolar disorder. Thus, from potential pharmacotherapies, more targeted treatments can be through this use of domain specificity in testing.