Abstract The five factor model, a dominant model of personality today, is based on the pre-structuring of test-components and factor analytic (FA) interpretations. These procedures have been criticised, especially FA’s limitation to accurately identify the underlying structure of constructs. This paper examined whether the NEO-PI-R revealed the same trait structures when analysed through alternative statistical procedures. Non-metric multidimensional scaling (NMDS) was employed alongside FA techniques ( N = 384). FA indicated the presence of five factors, but NMDS suggested a more parsimonious arrangement, comprising three (similar to Eysenck’s PEN model), or even two “super structures”. Furthermore, a circular trait arrangement was a reasonable alternative, suggesting that traits can be meaningfully (re)classified based on their kind and their integrative centrality. Over-reliance on FA models might not be the most appropriate approach for a better understanding of personality.