Abstract
A great deal of research has shown that small changes in question wording, format, orordering can sometimes substantially alter people's reports of their attitudes. Althoughmany scholars have presumed that these so‐called response effects are likely to be morepronounced when the attitudes being measured are weak, a number of studies have disconfirmedthis notion. This paper presents several new tests of this hypothesis using a variety of measuresand analytic techniques. The findings largely replicated previously documented effects andnon‐effects but also uncovered new effects not previously tested. No single strength‐relatedattitude attribute emerged as a consistent moderator of all response effects. Rather, differentindividual attributes moderated different effects, and a conglomeration of strength‐relateddimensions did not emerge as a reliable moderator. Taken together, these results support theconclusions that different response effects occur as the result of different cognitive processes,and that various strength‐related attitude attributes reflect distinct latent constructs rather than asingle one.