Logical inference tasks were presented to 112 police officers. Two logical forms were employed, converse and contrapositive. Three levels of activation were identified by a Q-Sort and incorporated into the design. A third factor was also considered, four styles of negative. Results of 24 items were analyzed by means of sign tests. Converse tasks were found to be easier than contrapositive tasks. Increased activation improved performance on the converse tasks but seemed to interfere with performance on the more difficult contrapositive tasks. Of special interest was the finding that contrapositive tasks with more negatives were easier than those with fewer negatives. (LS)