This research takes advantage of combined cognitive and anatomical studies to ask whether attention is necessary for high-level word processing to occur. In Experiment 1 we used a lexical decision task in which two prime words, one in the fovea and the other in the parafovea, appeared simultaneously for 150 msec, followed by a foveal target (word/nonword). Target words were semantically related either to the foveal or to the parafoveal word, or unrelated to them. In one block of trials subjects were also required to perform an auditory shadowing task. From PET studies we know that shadowing activates the anterior cingulate cortex, involved in selective attention. If the anterior attention system is always involved in semantic processing, shadowing should reduce semantic priming obtained from both foveal and parafoveal words. In contrast, if semantic priming by parafoveal words is independent of activation in that attention area, priming will not be affected by shadowing. Our results supported the latter hypothesis. A large priming effect arose from foveal primes, which was reduced by shadowing. For parafoveal primes a smaller priming effect arose, which was not affected by shadowing. In Experiment 2 prime words were masked. Semantic priming was reliable for both foveal and parafoveal words but there were then no differences between them. Most important, the size of priming was similar to that obtained from parafoveal words in Experiment 1. We conclude that the anterior attention system increases the potency of processing of consciously perceived stimuli, but there is a component of semantic priming that occurs without both focusing of attention and awareness, involving different cerebral areas to those involved in attention to language.