This paper offers a fuzzy-set model for interpreting vague frequency expressions, such as "rarely" and "sometimes," in the context of self-reported marijuana use. This model distinguishes between "intersubject fuzziness," reflecting differences between persons in the numeric interpretation of frequency expressions; "intrasubject fuzziness," indicating intervals of possible frequencies that a person refers to with a frequency expression; "conjoint fuzziness," reflecting differences between persons in the size of these intervals; and "random error." These types of fuzziness were examined for seven frequency expressions in two reporting conditions (confidential research, job interview). Intrasubject fuzziness consistently accounted for the largest proportion of response variance (approximately 50%); intersubject and conjoint fuzziness accounted for approximately 20% and 15%, respectively; random error accounted for about 15% of the variance. Intersubject, intrasubject, and conjoint fuzziness make the numeric interpretation of frequency expressions problematic. They reflect that different persons apply different frequency expressions to describe the same levels of actual marijuana use and that one person may use different expressions to describe marijuana use in different situations. The validity of self-reports may be enhanced by interpreting and analyzing frequency expression as fuzzy sets. © 1994.
Matt, G. E., & Wilson, S. J. (1994). Describing the frequency of marijuana use: Fuzziness and context-dependent interpretation of frequency expressions. Evaluation and Program Planning, 17(4), 357–369. https://doi.org/10.1016/0149-7189(94)90036-1