Temple et al. [1] highlight correctly many problems with the existing literature on cannabis, including problems about how cannabis use and cannabis-related problems are defined, and problems of measuring and surveying these activities consistently. Relevant theories about cannabis use are impoverished to the point that, as Temple et al. [1] point out, basic definitions are lacking. For instance, critics of the diagnostic classification approach to mental health argue that mental health ‘diagnoses’ are inherently subjective, somewhat arbitrary and that different diagnoses lack predictive or discriminatory power [2]. There are reasons to be concerned that complex, varied human activities, such as cannabis use, are packed into containers more suitable for palpable disease states. Conducting the classification more consistently risks propagating systematic bias rather than advancing understanding.

This extends to meta-analysis and systematic review. These techniques work very well when there are clear, objective definitions of disease, intervention and outcome, so that it is possible, for example, to identify that one medicine reduces mortality more than another. It is less likely that the techniques apply well to mental health problems, where any clear, objective definitions are imposed on a complex reality and potentially create as many difficulties as they resolve. Temple et al. [1] usefully discuss the dilemma of the conditions under which daily cannabis use might not constitute dependence.

What is truly lacking in cannabis research is sufficient understanding of cannabis use to even begin to specify standards. Addiction research sometimes forgets that scientific inquiry requires a theory of how one thing causes another. In the absence of theory, consistency and meta-analysis will not do instead. One ends up confusing correlations with causes. For instance, we do not know whether and how it is possible to consume cannabis heavily and regularly without dependence, because we lack a theory of ‘dependence’ that fits cannabis-using behaviour and can explain why many users do not become dependent, yet some do.

Nor is there any theory of how cannabis might cause psychosis. There is psychopharmacological theory about how cannabis causes its acute effects, which have some features in common with psychotic symptoms. Not enough is known about subclinical ‘psychotic’ symptoms, or about the experience of cannabis use, to understand why most cannabis users do not become ‘psychotic’, but some do. The scope of these problems is illustrated by the fact that standard questionnaire assessments for psychosis explicitly exclude drug-related experiences. This is pragmatic and sensible, but not theoretically based. For long half-life drugs such as the cannabinoids, there is a challenging issue of deciding for how long a drug might have direct effects. Is a hallucination 24 hours after last use of cannabis a ‘drug’ experience or not? It is possible that the correlation between cannabis use and psychotic symptoms is inflated in some surveys by participants failing to recognize that cannabis can have psychoactive effects for hours, or perhaps days, after use.

Developing such theories would require research on the details of cannabis use, which was sufficiently controlled or experimental to actually pose and test hypotheses. Making the juggernaut of epidemiological psychiatry more road-worthy is useful, but it will not lead to conceptual breakthroughs. An analogy will illustrate this. The massive advances made in neuropsychology have come about in part by studying the brains and behaviours of people with acquired brain injury in considerable detail and developing working models of the psychological impact of brain damage. Surveying the general population or, for example, people who have had strokes, would have generated some fuzzy correlations between brain insult and behavioural changes, but there would be no causal model. Most cannabis research is correlational, and it is not enough. Rather than working towards premature consistency and consensus, what is needed are better-designed studies of cannabis use that look in detail at the drug, the psychological and developmental state of the user and the social settings of use. The world's favourite illegal drug deserves this much attention.

Declaration of interest