Different groups of college students wrote one solution or many solutions to problems, like the plot-title problem, of verbal, numerical, and pictorial material. Instructions to write many solutions yielded solutions of lower mean quality, but more superior solutions. Information about criteria for good solutions raised quality. Large quantity was associated with low quality, both for variations in conditions and for individual differences within conditions. Differences between problems in the quality-quantity relation were dependent on the number of superior solutions. These three types of judgment training: (1) individual, (2) dyadic, and (3) tutorial, interpolated between production of solutions and selection of best solution, were generally successful and, under certain favorable conditions, improved over-all performance. The response hierarchy model can be modified to apply to the present data, and transfer of training can be demonstrated between tasks. (BP)