Abstract Methodological arguments are usually invoked to explain variations in the structure of affect. Using self-rated affect from Italian samples (N=600), we show that individual difference variables related to affective differentiation can moderate the observed structure. Indices of circumplexity (Browne, 1992) and congruence coefficients to the hypothesized target were used to quantify the observed structures. Results did not support the circumplex model as a universal structure. A circular structure with axes of activation and valence was approximated only among more affectively differentiated groups: students and respondents with high scores on Openness to Feelings and measures of negative emotionality. A different structure, with unipolar Positive Affect and Negative Affect factors, was observed among adults and respondents with low Openness to Feelings and negative emotionality. The observed structure of affect will depend in part on the nature of the sample studied.