The globalization of industrialized economies has increasingly required corporations to train employees for
overseas posting. It has also involved the need to conduct in-house foreign language training. Such
programs are frequently evaluated for cost effectiveness, leading program designers to search for methods of
providing evidence of language gain accruing from instruction. Accountability in such programs is often
contingent on a comparison of standardized test scores before and after the program of instruction. A
parallel need in language program evaluation is in describing not only that there has been language gain, but
also in providing information about features of the program that are related to gains. An inferential problem
with pre-and-post comparisons comes when programs are evaluated without reference to comparable controls.
One methodological alternative to conventional language program evaluation is a meta-analysis of cumulative
archives of records gathered from a variety of programs. The present meta-analysis of language programs is
one such example. Tests and surveys from 36 language programs in Japan and Korea provided descriptors of
course objectives, language curriculum features, materials, duration of instruction, information about
instructors, and pre-and-post instruction TOEIC scores. In all, more than 3200 records of pre-and-post test
scores were coded for salient program characteristics. These characteristics were used in 'dummy codes' for
input to multiple regression analyses devised to estimate effect sizes for different combinations of program
characteristics. The results of the meta-analyses of aggregated data describe salient patterns of teacher training,
effective material use, and program duration that are linked to benchmark TOEIC outcomes in corporate
programs. The outcomes of this study extend to implications for optimal language program organization.