The spatial distribution of visual attention is a yet unresolved question. One of the main topics is whether attention is distributed in a graded fashion around an attended location (e.g., Downing, 1988; Zimba & Hughes, 1987). The present experiments explore whether, and on which conditions, gradients of attention arise and contribute to perceptual facilitation. A masked or unmasked prime precedes one of two targets whose temporal order has to be judged. The prime captures attention, which shortens the perceptual latency of the primed target (perceptual latency priming; Scharlau & Neumann, 2003a; Shore, Spence, & Klein, 2001). No strong evidence for an attentional gradient was found. (1) Accuracy of temporal order judgements was independent of the distance between the two targets that were judged. That is, facilitation of the second target by the first target was spatially invariant. (2) With targets of short duration, facilitation was independent of prime-target distance. (3) With targets of long duration, gradients were found: Facilitation declined continuously with distance. Thus, long duration of stimuli may be a sufficient precondition for an attentional gradient. A control experiment showed that object-based attention contributes only marginally to perceptual latency priming.