Non-psychotic affective symptoms are important components of psychotic syndromes. They are frequent and are now thought to influence the emergence of paranoia and hallucinations. Evidence supporting this model of psychosis comes from recent cross-fertilising epidemiological and intervention studies. Epidemiological studies identify plausible targets for intervention but must be interpreted cautiously. Nevertheless, causal inference can be strengthened substantially using modern statistical methods.

Directed Acyclic Graphs were used in a dynamic Bayesian network approach to learn the overall dependence structure of chosen variables. DAG-based inference identifies the most likely directional links between multiple variables, thereby locating them in a putative causal cascade. We used initial and 18-month follow-up data from the 2000 British National Psychiatric Morbidity survey (N = 8580 and N = 2406).

We analysed persecutory ideation, hallucinations, a range of affective symptoms and the effects of cannabis and problematic alcohol use. Worry was central to the links between symptoms, with plausible direct effects on insomnia, depressed mood and generalised anxiety, and recent cannabis use. Worry linked the other affective phenomena with paranoia. Hallucinations were connected only to worry and persecutory ideation. General anxiety, worry, sleep problems, and persecutory ideation were strongly self-predicting. Worry and persecutory ideation were connected over the 18-month interval in an apparent feedback loop.

These results have implications for understanding dynamic processes in psychosis and for targeting psychological interventions. The reciprocal influence of worry and paranoia implies that treating either symptom is likely to ameliorate the other.