The purposes of this study were to investigate the effects of rater training in an L2 performance assessment and to examine the eligibility of L2 users of English as raters in L2 performance assessment. Rater training was conducted in order for raters to clearly understand the criteria, the evaluation items, and the evaluation procedure. In this evaluation, the rater training was conducted based on Common European Framework of Reference (CEFR). In the training, the raters watched the videos (North and Hughes, 2003), and discussed the learners' characteristics at each level. The analyses of the evaluations were done before and after the rater training based on Generalizability (G-Theory) and Multifaceted Rasch Analysis (MFRA). In the analyses based on G-Theory, the variance related to the items was reduced to about one sixth after the training, though no difference was found in the rater characteristics before and after the training in the analysis based on MFRA. Comparing the results of Kim (2009) with those of the present study, the raters are equally self-consistent with the raters of native speakers of English in Kim (ibid). Furthermore, it is legitimate to adopt L2 users as the raters, because, in countries where English is a foreign or second language, the non-native users teach and learn English. In this situation, teachers of L2 users are the most appropriate in L2 performance assessment if they are self-consistent in their ratings. (Contains 3 figures, 7 tables, and 2 footnotes.)