An experiment is reported which is designed to test the hypothesis that selection errors in the abstract form of the Wason four-card selection task derive from an attentional bias. Specifically, it is hypothesized that subjects' choices are based upon the perceived relevance of the exposed sides of the cards, as determined by linguistic cues, and that they do not reflect any analysis of the possible hidden sides of the card as required by the logical task. The methodology used derives from an earlier, unpublished experiment in which the task was presented by a microcomputer which recorded the order in which decisions were made about each of the four cards. It was predicted that subjects would consider first those cards deemed to be ?relevant?. The earlier study had found predicted associations between decision order and decision frequency based upon the well-established ?matching bias? effect. This finding was replicated and extended to another linguistic bias in the experiment reported here which improved upon the design of the previous experiment in several respects. In the earlier study, a strong left-to-right influence of spatial layout of the cards on decision order was also observed. In the present experiment a rectangular layout reduced but did not eliminate the effects of spatial location. Subjects' decisions to accept or reject cards were also biased by their spatial location, in a surprising manner. The results are discussed with reference to contemporary theories of reasoning.