An error analysis of the oral production of Korean adults learning English was performed on informant speech samples, using Corder's Algorithm for providing data for description of idiosyncratic dialects as a guide for determing error. The procedures of error analysis and morpheme acquisition studies were combined to address the following questions: (1) which errors occur most often in oral Korean English, (2) whether there is an order of difficulty of these morphemes for the group, and (3) whether a natural order of accuracy could be found for these morphemes when the orders of accuracy for three proficiency levels were compared. Examination and analysis of the data revealed the existence of a significant correlation between the three composite orders of accuracy. Explanations for shifts in the rank order of certain morphemes between levels are offered, and implications of the study for English as a second language teachers are explored. (Author/AMH)