In 5 experiments, the authors tested whether the processing of nonconscious spatial stimulus information depends on a prior intention. This test was conducted with the metacontrast dissociation paradigm. Experiment 1 demonstrated that masked primes that could not be discriminated above chance level affected responses to the visible stimuli that masked them. Experiments 2 and 3 showed that this effect was abolished when the task instruction was changed in such a way that the primes ceased to be task relevant. Experiments 4 and 5 demonstrated that a prime's effect depended on whether it was associated with the same response as the target or with an opposite response.