Background‘Cloze procedure assesses the predictability of text by deleting words at set intervals and having a panel of raters fill the blanks. Two refinements to the procedure are described: ‘modified Cloze and ’ ‘reverse Cloze’, the latter examining the subject's ability to interpret a partial transcript.MethodSamples of speech were obtained from patients with schizophrenia, manic-depression and orthopaedic disorders using a standard prompt, and the output analysed using modified Cloze. In addition, subjects completed a reverse Cloze passage.ResultsBoth modified Cloze and reverse Cloze discriminated the performance of patient groups, with patients with schizophrenia performing worst.ConclusionsThese techniques extend earlier findings suggesting reduced predictability in psychotic speech, particularly for patients with schizophrenia. The reverse Cloze procedure suggests a ‘mirror-image’ deficit and is a potential objective index of psychopathology that is much simpler to apply than traditional Cloze.