arnold on Sep 16, 2008
This could be answered best by the Horan et al (2003) article.
Their Research Question: Do wording-effects represent a distinctive response style rather than a purely artificial result? How do we interpret the two-factor exploratory structure produced by the use of both positively and negatively worded items? Should these factors be interpreted as two separate substantive components or should they be interpreted as a substantive component plus a systematic method effect associated with pos/neg wording of individual items?
Purpose/Contribution: The distinction between artificial and substantive interpretations of such wording effects has important implications for Structural Equations Modeling (SEM).
Findings: Wording effects associated with negatively worded items can be estimated as a distinct latent variable or factor and that such a negative-wording factor was seen for more scales than esteem (Questions 1 and 2). The negative-wording factors were related across different content areas (Question 3). Also these negative-wording effects were stable over two consecutive 2-year time periods (Question 4).
Therefore the main hypothesis is supported and these wording effects in survey instruments satisfy the criteria to be considered a type of response style rather than a substantively irrelevant artifact.
Response Style: a behavioral consistency operating across measures of several conceptually distinct content traits (identified by the existence of a latent variable)
Horan, DiStefano, and Motl. (2003) “Wording Effects in Self-Esteem Scales: Methodological Artifact or Response Style?” Structural Equation Modeling. 10.3. 435-455.