Three experiments investigated a dissociation originally described by Neumann, Esselmann, and Klotz. Stimuli were geometric shapes, preceded by similar shapes that were masked by metacontrast. Each experiment consisted of three parts. In the reaction time (RT) part, participants saw an array of geometric shapes, one of which was marked by bars, and had to respond to the marked shape's position by pressing an appropriate button. A prime (a similar, but smaller stimulus) preceded either the marked or an unmarked stimulus. In the temporal order judgment (TOJ) part, the task was to judge the temporal order of the marked and the unmarked stimulus. In the detection part, detectability of the prime was tested. Although its detectability was zero or close to zero, the prime affected both RT and the apparent onset as measured by TOJ. The effect on RT was significantly larger than the effect on TOJ (Exp. 1). Increasing the spatial context (number of non-target stimuli in the display) did not affect this pattern (Exp. 2). By contrast, reducing the temporal context (range of stimulus onset asynchronies) abolished the prime's effect in the TOJ task, although the prime affected RT under identical conditions. It is concluded that partially different mechanisms mediate the prime's effect in the two tasks and that the effect of stimulus context on TOJ found in the Neumann et al. study was due to temporal, not spatial context.