ANCHOR POINT EFFECTS ON THE EQUIVALENCE OF QUESTIONNAIRE ITEMS

Frisbie and Brandenburg (1979) present an interesting observation that while we often spend considerable time and effort in the construction of questionnaire item stems to ensure their content validity, we pay little attention to the validity of the response alternatives used. Those authors then investigated two characteristics of response alternatives: (a) whether the options were letters or numbers, and (b) whether all intermediate response alternatives or only the endpoints were labeled. The number-letter distinction had no significant effect on the responses while the labeled versus unlabeled intermediate alternatives showed small but systematic effects on six of the eight items. Of these six significant differences, the three which involved ratings on an evaluative continuum produced separations in the means which ranged from .25 to .38 standard deviations. The remaining three significant differences were found on scales which were not rated on the evaluative continuum. These items produced differences of .11 to .16 standard deviation units. Thus, while the differences in general may be small and of questionable practical significance, the findings on the evaluative continuum are the strongest and warrant further investigation.