Punishment is not a good learning tool

In Altered Neural Responses to Punishment Learning in Conduct Disorder, researchers examined how young people learn from punishment and reward using fMRI and computational modelling, and the findings show that punishment-based approaches produce weak and unreliable learning signals in a significant subset of youth.

The study followed 174 young people (ages 9–18) who completed a reinforcement-learning task while their brains were scanned, and the researchers compared typically developing youth with youth diagnosed with conduct disorder — a population routinely subjected to the most punitive forms of school discipline.

Core finding

Across the entire brain, responses to reward appeared intact, while the anterior insula — the region that supports the registration of negative outcomes and the behavioural shifts that usually follow — expressed a distinct pattern during punishment learning for youth with conduct disorder, and the study shows that:

• Punishment learning proceeds more slowly and less efficiently for many young people, even in the presence of a strong insula signal.
• Reward learning remains fully available.

This means that punitive inputs generate limited behavioural learning while reward-based signals continue to guide adjustment and growth.

Why this matters

The insula supports the capacity to sense discomfort, anticipate harm, and shift away from unsafe choices, and when this system expresses atypical patterns, punitive responses create confusion rather than guidance, and fear circulates as noise rather than usable information. Behaviour reorganises around reactivity and risk rather than safety, and young people become more volatile, more distressed, and more exposed to harm.

The authors describe this as a pattern of punishment hyposensitivity, and this pattern appears across girls and boys and persists across variations in cognitive profile, socioeconomic context, and clinical complexity.

Broader implications

Punishment-centred behaviour strategies — detentions, loss of recess, collective punishment, sticker removals, exclusion — rely on the belief that children integrate negative consequences and shift accordingly, and this study deepens the evidence that:

• Punitive environments cultivate fear rather than learning.
• Reward-based or scaffolded approaches align with the neural pathways that actually support behavioural growth.
• Persistent reliance on punishment functions as a scientifically incoherent choice in systems serving children with atypical punishment-response patterns.

For school systems, this means punitive discipline expresses a design failure rather than a behavioural solution, and genuine safety arises through environments that support learning, regulation, and meaningful relational repair.