An object that is relevant to the current task is more readily encoded than a similar, less relevant object (e.g., Wolfe, Cave, & Franzel, 1989). However, it is highly debated whether relevance-based selection is feasible at short exposure durations (e.g., Theeuwes, 2010). It has been argued that high local feature contrast attracts attention independently of task-relevance. Yet, recent studies have provided evidence that effects of task-irrelevant feature contrast interact with the task-relevance of the object in question (Nordfang, 2011). In a new experiment, display size was kept constant while target-distractor configuration was varied (2 targets and 6 distractors vs. 8 targets and no distractors). The targets were letters and the distractors were digits. Displays were postmasked and briefly presented at durations varying from 10 to 100 ms. On some trials, one of the characters (target or distractor) was a color singleton, but the probability that the singleton was a target was just the same as the probability that a nonsingleton was a target. Participants showed significant effects of both feature contrast and task relevance. The probability of correctly reporting a singleton target was significantly higher than the probability of reporting a nonsingleton target. The probability of correctly reporting a given target also was significantly higher for the displays with 2 targets and 6 distractors than for the 8-target displays, revealing selectivity based on task-relevance. This effect was significant even at an exposure duration of 30 ms. Furthermore, as expected from the results of Nordfang (2011), the presence of a singleton distractor decreased the probability of correctly reporting a given nonsingleton target significantly less than did the presence of a singleton target. In summary, the study provided new evidence that both contrast and relevance influence attentional selection at very short exposure durations and the effects of the two factors are interdependent.