A mixed group containing university students, mostly studying mathematics, and some senior school pupils took two paper-and-pencil conditional reasoning tests in varying order with a one-week interval between the tests. One test consisted of five problems presented entirely in written form. The second test took one of two forms, both identical to the first test except that in the first form inessential drawings were added to the presentation of four of the problems and in the second form drawings, which acted as concrete referents and were essential, were added to the same four problems. Irrespective of the order of testing, subjects' performance on the problems with drawings used as concrete referents was significantly worse than on the corresponding problems presented in written form. Furthermore, the results of the initial testing session showed the school group performed significantly worse on both the form of test with inessential drawings and the form with concrete referents than on the written form. In addition, a group of students with recent training in reasoning significantly outperformed all others. Finally, negative conclusion bias was only observed for the Modus Tollens inference.