Previous research indicates that Likert items do not result in unidime
nsional ordinal. rating scales nor do they consistently permit the sep
aration of Content and Intensity dimensions. An alternative strategy i
s to present a separate Content (agree or disagree) item immediately f
ollowed by an Intensity (strongly or not strongly) item. Panel data fr
om the 1992 and 1994 American National Election Studies are used to st
udy this strategy. The results indicate that Content and Intensity are
sometimes confounded with the two-part questions, and also that the j
oint responses to the Content and Intensity questions do not generally
constitute a unidimensional scale. (C) 1997 Academic Press.