associated with the use of business simulation materials, and an even bigger edge associated with video materials. In-house training of the instructors’ use of these materials serves to optimize the impact of the instruction. The second conclusion we would like to underscore is that unanalyzed archival data may still be valuable for program evaluation. The key to retrieval of such data isidentifying key characteristics of program organization that can be cast into survey protocols. These protocols can then be used to locate organizations with language training programs, with or without internal evaluation. Program characteristics can be reconstructed through the examination of curriculum documents and test records. Meta-analysis with ex post facto surveys can provide new interpretations of global program effects. Program designers can thus learn from retroactive analyses of previous programpolicies in evaluating options for future program improvement. Finally, even though we have shown that archival data can be exploited to provide broad generalizations about organizational effects on language training programs, we still encourage program managers to conduct rigorous internal evaluation on an on-going basis.