Issues in the dimensional structure of affect - Effects of descriptors, measurement error, and response formats: Comment on Russell and Carroll (1999)

In their article "On the Bipolarity of Positive and Negative Affect," Russe ll and Carroll (1999) correctly noted that several factors-including the us e of a strictly unipolar item response format-can obscure the underlying bi polarity of certain positive and negative affect pairs. However, same of th eir other conclusions require clarification. First, bipolarity is evident e ven in uncorrected data obtained with unipolar formats. Second, polychoric correlations and appropriate item-analytic models can circumvent problems t hat may be associated with unipolar scales; consequently, it is unnecessary (and undesirable) to use bipolar formats in affect assessment. Third, no e vidence supports their claim that general mood ratings are less valid than other measures of long-term affect. Finally, aspects of their circumplex mo del are not well supported by the data; in its place, the authors suggest a more inclusive 3-level hierarchical structure.