When scientific studies are observational, based purely on measurement – however sophisticated – the possibility will remain that correlations between two things of interest do not reflect a direct causal connection. For example: suppose you find that a subject’s self-report of auditory hallucinations is associated with increased BOLD signal in brain region X. This doesn’t prove that there’s an abnormality of X (the abnormality might be in region Y, which influences X) or even that activity in X reflects the hallucination itself (perhaps it relates to an emotional response to the hallucination, or to the motor activity required to report it). A true experiment involves manipulating one or more independent variables, with appropriate control conditions, and examining the effects.