The dynamic bipolarity of the positive and negative affective activation, measured with the PANAS scales, was studied using a pre-post design with an intervening experiment. The correlations between (a) the initial positive and negative constructs and (b) the respective change scores were estimated, random and systematic error being removed owing to a convenient structural equation modeling technique. Results demonstrated that a moderate perturbation may induce a medium correlation between latent change scores. Both strict dynamic independence and bipolarity were rejected. This result highlights the importance of individual differences in the way people perceive their affective changes. It is concluded that the PANAS two-factor model of affect provides only an approximate view of the structure and dynamics of mood.