Constant amplitude and narrow-band Gaussian loadings are applied to extruded 6063 aluminum crack-growth specimens in an effort to characterize the effective stress intensity levels during random loading. Crack-growth rates are determined for constant amplitude loadings at stress ratios (R) of 0.09, 0.3 and 0.5, and for a variable amplitude loading simulated to match a narrow-band Gaussian spectrum. Crack-opening stress levels measured by the compliance method during the constant amplitude loading are found to differ substantially for -T5 and -T6 heat treatments due to a change from intergranular to transgranular crack growth. Crack-opening load ratios correlate well with the maximum applied stress intensity factor, Kmax, for the -T5 material. The Kmax dependence leads to an effective halving of the crack-growth exponent. Calculated variable amplitude lives are much shorter when this correlation is taken into account (an acceleration effect) and show a greater difference between loading blocks condensed by racetrack filtering at threshold levels of two and four standard deviations, similar to what was observed in the tests. Crack-opening-load measurements in one specimen with the narrow-band Gaussian (variable amplitude) loading failed to detect any closure. A substantial difference in the closure behavior of nominally identical R = 0.3 tests indicates that closure may occur irregularly in the extruded aluminum. Calculated crack-growth lives, assuming no closure in the variable amplitude tests, are much shorter than the test results. Including closure in the variable amplitude loadings greatly improves the predictions.