on the upper triangle, circular triad score on the lower triangle, and inconsistency score from a comparison of upper and lower triangles, all scores from a pairedcomparison matrix. Results confirmed a previous finding that type of score was a much more significant source of variation than other instrument characteristics (type of content, number of items or item format). Other findings cast doubt on the utility of the types of scores used as measures of a response inconsistency trait. RESPONSE inconsistency in comparison measures of preference may appear in at least two forms: as logical inconsistency or as temporal inconsistency. Logical response inconsistency (or more properly, intransitivity, as it is known in psychometric theory; see, e.g., Tversky, 1969) appears as circular triads (the choice of A over B, B over C, and C over A). Temporal response inconsistency appears as inversions of preference or choice on retest (the choice of A over B on test, B over A on retest). Recent findings have generated interest in the hypothesis that response inconsistency is interpretable as a personality trait. Pemberton (1966), in a study of the work preferences of 100 profes-