Aims  To model consumption patterns and problems associated with alcohol, cannabis, ecstasy, amphetamine and cocaine hydrochloride use in a non‐treatment sample of young polysubstance users.

Design  A cross‐sectional survey of 364 16–22‐year‐old (56.3% male) polysubstance users recruited and interviewed by peer interviewers.

Measurements  Structured questionnaires were used to gather identical datasets on the five target psychoactive substances, recording patterns of substance use; adverse consequences from use; negative effects; functions for substance use; and perceived peer use.

Findings  Functions for substance use strongly predicted intensity of use in all five substances when peer use, age of first use and demographics were controlled, explaining an additional 11–19% of the variance in scores. Functions also explained an average of 22% of the variance in problem scores over and above the effects of background variables and current intensity of use. In particular, functions concerned with relief from negative mood states were strong predictors of problem scores in alcohol, cannabis and cocaine.

Conclusions  The potential implications of using a functional approach to explaining and responding to substance use are considerable. This could help to enhance our understanding of how experimental substance use becomes regular and how regular use becomes problematic, and could thus inform prevention, education and intervention efforts.