Abstract
The purposes of this paper are a) to investigate whether the Lunic Language Marathon sorts the test-takers into a normal distribution; b) to detect inappropriate items, which are misfitting items, the items with low discrimination power values, or the items with low or high item difficulty values; and c) to investigate whether the results of the LLM show the selection effects of college entrance examinations. A total of 420 college students from four groups participated in this study. The first group consists of 78 students majoring in information technology at a national university. The second group has 105 students majoring in English language at a co-ed private university. Other 122 students, which fall into the third group, are majoring in social science or humanities at the same university as the second group. The last group contains 115 students majoring in English language at a women's private university. The results of this study show that a) Part 1 and 3 of the LLM sort the college students into a normal distribution, but as for Part 2 and 4, the ceiling effect has been observed; b) three items in the LLM are judged as inappropriate items; and c) there is some effect of group difference on the performance of the participants (Wilks' Lamda=.82, F(15, 1137.75)=5.79, p .00, η^2=.07, Power=1.00), but apparent effects of college entrance examinations have not been observed. This paper further discusses the use of the LLM as a diagnostic test at a college, and points out that specifically the scores on Part 1 will tell the instructors much about the learners' variety.
