Considerations for analysis of source monitoring data when investigating hallucinations in schizophrenia research

Abstract

Self/other (i.e., internal/external) source monitoring is one of the leading paradigms for the study of hallucinations in schizophrenia. The cognitive processes that underlie hallucinations are theorized to transform self-generated (internal) cognitive events into other-generated (external) cognitive events. These proposed cognitive operations also appear to play a role in producing analogous types of errors in self/other source monitoring, namely a memory bias whereby recalled material that was self-generated is misremembered as other-generated, referred to as an externalization bias. Externalization biases are more frequent in groups of hallucinating schizophrenia patients than in other groups. One source of measurement error that is inherent in the study of the externalization bias is that, even for never-previously viewed items, there is a tendency to guess an external source under conditions of uncertainty. If such guessing takes place in response to self-generated but forgotten items, these guesses will be summed along with true externalization biases in the frequency count of externalizations, producing measurement error. Multinomial modeling is a statistical technique that has been used to estimate the influence of external-source guessing in order to separate it from true externalization bias estimates. However, a number of challenges related to model choice and model validation are involved, and these challenges may render multinomial modeling impractical. We instead recommend analysis of covariance (ANCOVA), or difference score methodology, as an appropriate method for partialling external-source guessing rates (external-source false positives) out of externalization bias rates.