Cocaine Addiction Doesn't Blur the Odds. It Blunts the Bill.

The standard account of bad decisions in addiction is a story about arithmetic. People who use drugs, the reasoning goes, cannot hold probabilities in their heads, so they overestimate the upside and underestimate the risk. A study published November 5 in Translational Psychiatry tested that account directly, by handing people the arithmetic.

One hundred and five patients with moderate-to-severe cocaine use disorder and 80 healthy volunteers completed a decision-making task in which the size of the reward, the size of the loss, and the probability of each outcome were all stated explicitly. No hidden contingencies, no learning curve. Just numbers on a screen and a choice to make.

Patients used the probability information exactly as well as the control group did. Their difficulty was somewhere else entirely.

The deficit is in the consequences, not the odds

Where patients diverged was in how much they let the stated outcomes move their choices. Their integration of outcome values was reduced, and the reduction was lopsided: losses were discounted more heavily than rewards. A person with cocaine use disorder, in this task, could tell you precisely how likely a bad outcome was and still not let that knowledge change what they did.

A subset of participants also underwent diffusion-weighted MRI, and here the result cuts against the intuitive explanation. The orbitofronto-striatal white matter tracts that support decision-making did not differ between patients and controls. What varied — in everyone, regardless of diagnosis — was how individual structural differences within those tracts predicted the degree to which reward and loss information shaped choices. The study was led by Tsen Vei Lim with Rudolf Cardinal, Elijah Mak and Karen Ersche at the University of Cambridge, funded by the Medical Research Council, and published as an early-release version of record pending final typesetting.

So the finding is not that cocaine damages the wiring that weighs consequences. It is that the weighing happens differently in people with the disorder, and that the loss side is where it tilts.

What that means for treatment design

The clinical implication is narrower and more useful than "patients need to understand risk better." They already do. What they need is support translating consequences they can describe into choices they actually make — which is a different therapeutic task, and one that existing treatments address unevenly.

It also fits, uncomfortably well, with why contingency management outperforms almost everything else for stimulant use disorder. Contingency management does not ask anyone to reason about a distant probabilistic loss. It pays a specific, immediate, reliable reward for a verified behavior, week after week. The one channel this study suggests remains intact — reward-driven choice, when the reward is concrete — is the channel contingency management uses. Loss-framed approaches, increasingly popular in digital health, assume the opposite channel works.

For people in treatment for stimulant use disorder, the practical takeaway is less about a new therapy than about honesty in counseling: rehearsing the cost of relapse in the abstract is asking a patient to use the exact faculty the study found compromised. Building consequences into the next 48 hours is not.

The study is cross-sectional, so it cannot say whether the pattern preceded cocaine use or followed it, and a laboratory task with printed probabilities is not a Friday night. But it does retire one explanation, and the field has leaned on that explanation for a long time.