Green, Goldman, and Salovey (1993) challenged the view that "positive affect" and "negative affect" are largely uncorrelated dimensions. On the basis of factor analytic studies of happiness and sadness, and of positive and negative emotional activation (PA and NA), they claimed that a "largely bipolar structure of affect" (p. 1029) emerges when random and nonrandom error are taken into account. A reappraisal of their own findings and confirmatory analysis of additional data do not support this claim. Happiness and sadness Sonn a largely unidimensional bipolar structure, but PA and NA are relatively independent. However exploratory analyses yield a three-level hierarchy incorporating in one structure a general bipolar Happiness-Versus-unhappiness dimension, the relatively independent PA and NA dimensions at the level below it, and discrete emotions at the base. We emphasize the heuristic value of a hierarchical perspective.