Automaticity is usually discussed in terms of its benefits. Automaticity has, however, a cost that manifests itself in procedures that are highly routinized but require close attention, such as verbal checklist procedures. In such procedures, errors occur because the routine leads to automaticity. In three paper-and-pen experiments, we tested this manifestation and investigated ways to decrease automaticity in verbal checklist procedures. In the experiments, subjects proofread sets of multiplication problems to detect erroneous operations, simulating the checklist procedure. In Experiments 1 and 2, two conditions were compared: a fixed-order condition (in which each set contained operations in the same order) and a varied-order condition (in which the operations were in a different order in each set). In Experiment 1, proofreading times were measured to establish the role of fixed sequential order as a consistent environment promoting the emergence of automaticity. In Experiment 2, we introduced errors into the material, and in Experiment 3 we introduced "alerting" conditions to interfere with the development of automaticity. The results indicated that the subjects in the varied-order and alert conditions detected significantly more errors than did those in the fixed-order condition. The implications of the findings for current theories of automaticity are discussed as well as those for the design of checklist procedures.