Kelleher et al's study is very interesting and raises some important questions, ^{Reference Kelleher, Keeley, Corcoran, Lynch, Fitzpatrick and Devlin1} but we think that it also has some confounding factors that need to be addressed before conclusions are made. In addition, there are some methodological issues which we would like to be clarified. The response rate in study 1 is 52%, which might not be enough to support the conclusion of this kind of study. Second, owing to the different inclusion criteria in studies 1 and 2, there is a strong case for non-response bias. The way in which the first interview sample (study 3) was assembled seems unclear. Also, the way in which the second interview sample (study 4) was composed raises questions as to whether it can truly be considered a sample that represents the general population as claimed in the article. As far as confounding factors go, there is no mention of psychoactive substance misuse. With the potential of drugs to produce hallucinogenic effects, and the known link between conduct disorder, depression and attention-deficit hyperactivity disorder with substance misuse comorbidity, ^{Reference Zeitlin2} there is a chance that this could lead to results that do not reflect the true nature of the link between psychotic symptoms and non-psychotic disorders.

Another thing that could possibly be of interest and could affect the overall conclusions of the study is whether the study made any kind of differentiation between hypnagogic, hypnopompic and daytime hallucinations. ^{Reference Ohayon, Priest, Caulet and Guilleminault3} Last, there is no mention on the effects of the hallucinations on the children and adolescents, whether they have perceived them as positive, negative or neutral, and whether they have sought any help or counselling because of them. There is also no mention of help-seeking or school and family problems among the children and adolescents who were classified as having a diagnosable non-psychotic disorder, which might have been a more precise way to link the severity of childhood and adolescent problems than the simple use of the number of comorbid diagnoses assessed in one interview in a non-clinical setting.