Abstract
This article explores the developmental history of TOPIK over the last 20 years and its recent trends. Over the last two decades, TOPIK underwent two major systematic reforms, achieving both quantitative and qualitative growth over the course of its 52 testing sessions. TOPIK has utilized a six-level evaluation system from its inception to the present. The evaluation system was amended from the earlier six-test set - whereby each level corresponded with a separate test (1997~2005: the $1^{st}{\sim}9^{th}$) - to the three-test set (2006~2014: the $10^{th}{\sim}34^{th}$), and finally to the two-test set (2014~present: the $35^{th}{\sim}42^{nd}$). In the earlier exams, abilities in Vocabulary Grammar, Writing, Listening, and Reading were assessed. However, beginning with the $35^{th}$ TOPIK, abilities in Listening, Reading, and Writing (only in TOPIK II) were assessed and the evaluation of the writing section was changed to a task-based process, improving TOPIK into a more qualified analysis of proficiency. Over the last 20 years, the number of countries TOPIK is administered in has greatly increased from 4 to 73 countries, and the annual number of test-takers has also significantly increased from 2,692 to 250,141. The distribution of proficiency levels of the test-takers has shifted over time - initially "Beginner>Intermediate>Advanced" in the earlier exams, to "Intermediate>Advanced>Beginner" after the mid 2000s - as the number of those studying the Korean language for specific purposes and continuing education increased. Test-takers have indicated a shift in their purpose for taking the exam, initially citing "to assess proficiency" and more recently selecting "to study abroad," and this could also correlate with changes in the proportion of test-takers' proficiency level. In general, 85~95% of beginner, 50~65% of intermediate, and 45~60% of advanced test applicants passed the respective proficiency level. To date, no practices have yet been implemented to standardize the difficulty level longitudinally across test sets.