For the test phase, associated distractors were developed for each symbol and each knowledge type. Distractors for the verbal label were selected from incorrect answers provided in prior open-ended comprehension tests. Distractor actions were constructed to be related to the distractor verbal labels. However, since similar consequences can result from many different actions, it was difficult to find plausible candidates that could be considered unequivocally incorrect. Therefore, distractors for consequences were produced by randomly pairing symbols and consequences and then checking to ensure that the members of each pair were unrelated (see Fig. 2). Since distractor consequences lacked the plausibility of the other distractor types, it was expected that they would be more easily rejected as incorrect. However, this design issue is not relevant to the assessment of training effects since these effects are assessed by comparing pre- and post-test performance on the same items