This study examines the effects of three different violation instructions in the abstract selection task. Margolis (1987) hypothesized different selection patterns for each of these instructions. The results of Expt 1 supported all of his predictions, including one for a large number of not-p and not-q selections, an extremely infrequent choice for the abstract selection task. In Expt 2, the violation instruction leading to the not-p and not-q selections was used in an abstract permission schema selection task problem. In this case, it enhanced the facilitation normally observed for such problems instead of leading to not-p and not-q selections. The present results are discussed as support for attentional cueing models of reasoning, specifically Margolis's scenario/linguistic ambiguity theory and Evans's heuristic/analytic processing theory.