A considerable body of sensory research has addressed the rules governing simultaneity judgments (SJs) and temporal order judgments (TOJs). In principle, neural events that register stimulus-arrival-time differences at an early sensory stage could set the limit on SJs and TOJs alike. Alternatively, distinct limits on SJs and TOJs could arise from task-specific neural events occurring after the stimulus-driven stage. To distinguish between these possibilities, we developed a novel reaction-time (RT) measure and tested it in a perceptual-learning procedure. The stimuli comprised dual-stream Rapid Serial Visual Presentation (RSVP) displays. Participants judged either the simultaneity or temporal order of red-letter and black-number targets presented in opposite lateral hemifield streams of black-letter distractors. Despite identical visual stimulation across-tasks, the SJ and TOJ tasks generated distinct RT patterns. SJs exhibited significantly faster RTs to synchronized targets than to subtly asynchronized targets; TOJs exhibited the opposite RT pattern. These task-specific RT patterns cannot be attributed to the early, stimulus-driven stage and instead match what one would predict if the limits on SJs and TOJs arose from task-specific decision spaces. That is, synchronized targets generate strong evidence for simultaneity, which hastens SJ RTs. By contrast, synchronized targets provide no information about temporal order, which slows TOJ RTs. Subtly asynchronizing the targets reverses this information pattern, and the corresponding RT patterns. In addition to investigating RT patterns, we also investigated training-transfer between the tasks. Training to improve SJ precision failed to improve TOJ precision, and vice versa, despite identical visual stimulation across tasks. This, too, argues against early, stimulus-driven limits on SJs and TOJs. Taken together, the present study offers novel evidence that distinct rules set the limits on SJs and TOJs.