The threshold hypothesis regarding creativity and intelligence suggests that these two constructs are positively correlated except at the higher end of the IQ distribution, where they are unrelated. Much of the support for this hypothesis comes from comparisons of correlations within average and high-IQ groups. However, a common methodological flaw in these comparisons is that the high IQ group is smaller in variance than the average IQ group. In the present study we equated variances for a better test of the threshold hypothesis. College students in two IQ groups completed tests of creativity and intelligence. IQ and creativity correlated at least as strongly in the high IQ group as in the average IQ group, generally failing to support the threshold hypothesis. However, breakdown of composite IQ into fluid and crystallized IQ indicated minimal support for the threshold hypothesis with crystallized IQ, but support for an "inverse" threshold effect with fluid IQ. Fluid IQ correlated significantly with creativity in the high IQ group but not in the average IQ group, and this pattern was corroborated by regression analyses. Possible explanations for the pattern of results are offered. (Contains 2 tables.)