Studies in the sources of variation in cloze scores: II. The verbal passages.

100-word speech samples of alcoholic and schizophrenic Ss were examined in order to relate characteristics of these samples to the differential predictability of (a) alcoholic speech under drug versus nondrug conditions (LSD-25), and (b) schizophrenic and alcoholic speech, using the cloze procedure. The deleted words in the samples were identified as to grammatical class form and frequency of usage, and these characteristics were related to predictability. Additionally, changes in predictability as the speaker continued to speak were studied. The ad hoc analysis in terms of grammatical class forms was suggestive, but the analysis in terms of frequency of usage was not. It was also of interest that predictability tended to rise as the alcoholics continued to speak, both in drug and nondrug conditions, but dropped in schizophrenic speech. 424 The predictability of verbal information has been studied in a variety of contexts by the