Research on spatial cueing has shown that uninformative cues often facilitate mean response time (RT) performance in valid- compared to invalid-cueing conditions at short cue-target stimulus-onset-asynchronies (SOAs), and robustly generate a reversed or inhibitory cueing effect at longer SOAs that is widely known as inhibition-of-return (IOR). To study the within-trial time course of IOR we employ discrete-time hazard and conditional accuracy analyses to describe and model the shapes of the RT and accuracy distributions measured in two experimental tasks. In contrast to the mean performance measures, our distributional analyses show that (a) the uninformative cue generates response channel activation, (b) which continues during the cue-target interval so that the cue location must be stored in spatial working memory, (c) the premature cue-triggered response is selectively inhibited before target onset, (d) the IOR effect (valid versus invalid cueing) emerges around 160 ms after target onset in the hazard functions when cue-target SOA exceeds ~200 ms, quickly increases and decreases in size, and is gone within 120 ms, (e) the inhibitory component does not diminish over the course of the experiment, and (f) the location of an additional central cue relative to the current focus of spatial attention can generate response channel activation as well. These distributional data show that mean performance patterns conceal crucial information about behavioral dynamics, and suggest that sensory IOR is the direct result of encoding the cue location in spatial working memory to promote change detection, instead of attention leaving an inhibitory tag to promote visual search.