With three data sets-one comprised of school children (n = 272) and two comprised of university students (n = 453, n = 650)-the present study examined the factor structure of Cacioppo and Petty's (1982) Need for Cognition Scale (NCS) and its relation to standardized measures or ability and academic performance. The results from testing five models (one-factor, two-factor, two-factor correlated, hierarchical, and trait-method solutions) were consistent across the three samples. In all cases a trait-method model best fit the data. In addition to, and orthogonal to, the trait factor common to all items, two uncorrelated method factors, based on the wording polarity of the item, were identified. Furthermore, it appears that the reliable correlation found between the NCS and academic performance and measures of verbal ability is, at least in part, the result of the presence of the negative polarity items on the NCS. (c) 2005 Elsevier Ltd. All rights reserved.