The Coloured Progressive Matrices Test (CPM), which is intended to measure inductive reasoning ability, has been reported to correlate .91 with WISC Full Scale IQ. Each item of the CPM Test displays a pattern with a section missing; below it are six response alternatives, one of which would correctly complete the pattern. In a previous study, remedial self-instructional training led to item-specific and temporary increases on the CPM Test scores of first-graders. In the present study, new and presumably better self-instructional sequences were used with both first- and third-grade Ss. Trained Ss did no better than control Ss on a CPM posttest. The test, however, appeared to confound "pure" inductive ability with ability to imagine how a response alternative would look if inserted into the pattern. A new test format was developed to eliminate the role of the latter ability by having Ss judge the correctness of already completed patterns. Dramatic increases in test scores were found. Apparently the advantage of two years' mental growth that third-graders have over first-graders on the conventionally administered CPM derives in large part from improved power of imagining how a distractor would look in a new location, rather than of inductive reasoning ability. The new test format also showed that Ss were significantly better able to accept correctly completed than to reject incorrectly completed matrices. Among first-graders, this tendency was significantly stronger for trained than for control Ss.