The use of positively and negatively phrased items and the fit of a factor solution

To avoid the response set phenomenon, in handbooks it is suggested to compose item batteries of both positively and negatively formulated items. However, while doing a factor analysis, researchers usually neglect the effect this may have on the fit of the factor solution. In that case, the fit between a plausibly interpretable factor solution and the observed correlations may prove to be very unsatisfactory. In our opinion, however when both positively and negatively formulated items are used, it seems unrealistic to base the factor analysis on the content of the items only; also the difference in phrasing may influence the intercorrelations of the items. In this paper we discuss some possible additional assumptions in consequence of this consideration and we examine to what extent the deriving analysis decisions may improve the fit.