Noting that children evaluate their reading comprehension more effectively when they are informed about the standards of evaluation they should use, a study examined whether training in the use of certain standards would show generalization from one level of processing to the other, and also maintenance of standard use. The study also examined whether subjects would benefit more from training in "macroprocessing" standards (structural cohesiveness, internal consistency, informational completeness) which require more global text integration than "microprocessing" standards (lexical, propositional cohesiveness, external consistency) which involve more local attention to words and sentences. Subjects, 80 average and above average readers in fourth grade, were assigned to one of three instructional groups: macroprocessing, microprocessing, or no instruction. Subjects in the training groups were given practice applying the standards appropriate to their condition. All groups then read and evaluated passages containing problems requiring the use of all standards. They were tested again two and a half weeks later. Results indicated that the children were more likely to identify problems of the types in which they were instructed, but both trained groups identified more noninstructed problems than did the control group, indicating generalization. The training advantage was also maintained over time. In addition, the macroprocessing group performed better than the microprocessing group on the delayed test, indicating a longer term benefit of global evaluation training. Reading skill affected performance, but did not interact with other factors. (JC)