Recently researchers have debated the nature and significance of the cognitive style construct as a basis for understanding individual differences in behaviour in organizations. Two rival theoretical traditions prevail, one group of scholars arguing that cognitive style is best conceived within complex, multidimensional frameworks, others contending that the various facets of style can be meaningfully subsumed under a single, overarching dimension. The Allinson-Hayes Cognitive Style Index (CSI) is a 38-item instrument, predicated on the unitarist conception of the construct. This paper presents theoretical and methodological arguments as to why the previously hypothesized unifactoral structure of the CSI is potentially found wanting. Two variants (oblique versus orthogonal factors) of an alternative two-factor model, comprising separate analytic and intuitive dimensions, are developed and the results are reported of a series of principal components and confirmatory factor analyses (N=939) designed to overcome the limitations of previous research into the factor structure of the CSI. The results strongly indicate that the two-factor model with correlated factors provides a better approximation of responses to the CSI than previously reported unifactoral solutions. In the light of these findings we propose a revised scoring procedure, in which the analysis and intuition items are treated as separate scales, and consider the implications for future theory-building, research and practice.