Previous deception detection training studies have compared people receiving training in nonverbal behaviors associated with deception to control groups receiving no training and found that people who are trained are slightly to moderately more accurate than people who have not been trained. Recent research on the relationships between source veracity and specific nonverbal behaviors, however, suggests that those relationships are weak, inconsistent, and limited to high stakes lies. If specific nonverbal behaviors are not reliable indicators of deception, then one might wonder why training improves accuracy. This paper tests the speculation that the simple act of training, independent of the training content, may improve accuracy simply because those in training conditions process messages more critically. This speculation was tested in three experiments that included both no training and bogus training control groups. The bogus training group was most accurate in Study 1, but this finding failed to replicate in Study 2. A coding study (Study 3) examined behavioral differences in the stimulus tapes. The predicted differences were observed in a final experiment (Study 4) were training was based on the coded stimulus tapes. The results suggest that the effects of training are generally small and highly variable from message to message, that valid training does not produce large improvements over a bogus training control, and that bogus training can produce statistically significant improvements over a no-training control. © 2005 Western States Communication Association.