Dimensionality of the Revised Life Orientation Test and the status of filler items

Dispositional optimism was originally construed as unidimensional (Scheier & Carver, 1992). However, LOT-R data (Scheier, Carver, & Bridges, 1994) generally appeared bidimensional as a number of studies suggest a two-correlated-factor model representing optimism and pessimism. Attempts at corroborating one-factor models suggest that correlated errors between positively worded items are required for an adequate account of the data. This article explains bidimensionality by the influence of social desirability (i.e., being positive is desirable). Namely, in the present study, correlated errors are interpreted as the presence of individual differences related to the tendency to present oneself in a positive manner. Moreover, response styles can be corroborated by appropriately modelling the entire covariance matrix (i.e., including fillers), by checking that fillers with positive meaning correlate with the faking-good group factor. Students (N=442) responded to a French adaptation of the LOT-R. The data were submitted to SEM analyses. The traditional two-correlated factor model (optimism-pessimism) was outperformed by a model including a common factor ("optimism") plus a factor grouping positive items only ("faking positive"). In addition, reliability analyses showed that the choice of the model clearly impacts the reliability estimates based on the model. The entire dataset was modelled for exploring the relationships between the fillers and the measurement model (i.e., the set of all relationships between factors and their indicators). The specific correlations of fillers whose meaning is positive with the faking-good group factor corroborated its substantial interpretation. It is concluded that there is no empirical necessity for hypothesizing that the dispositional optimism construct must be split into optimism plus pessimism.