Although the correlation between general cognitive ability (*g*) and performance on speeded cognitive tasks is well-established, there is need for a better understanding of how successive stages of processing contribute to this relationship. Previous research suggests that *g* is primarily associated with the rapidity of decision-making rather than perceptual processing of stimuli; the implication is that *g* should statistically interact with a manipulation affecting the difficulty of the decision process, while simultaneously failing to interact with a perceptual manipulation. We applied Sternberg's method of additive factors to test this hypothesis in two reaction time tasks, each of which systematically manipulated the demands on perceptual acuity and decision-making. With a total of 773 participants, we found evidence of an interaction between a short-form measure of *g* and the decisional—but not perceptual—manipulations. This pattern was found in both number-comparison (Experiment 1) and tone-comparison (Experiment 2) tasks. Additionally, diffusion modeling of the Experiment 1 results revealed that the diffusion rate ($v$) is associated with *g* and affected by an informational attribute of the stimulus (numerical magnitude) but not a perceptual attribute (contrast), the nondecision time ($T_{er}$) is not associated with *g* and shows the opposite pattern of selective influence. Taken together, these findings add to the evidence for a theoretical framework partitioning reaction time into several processing stages, of which only the decision-making stage is associated with *g*.