The low reliability due to raters' subjective scoring is a matter of concern in a construct response or write-answer type test. For detecting items in which raters' scoring bias might exist, the applicability of the methods of differential item functioning (DIF) and the Mantel-Haenszel (MH) procedure were examined in the case of a language test for elementary school children. The test consisted of a) write-answer type items requiring a phrase or a short sentence, b) short-snswer items calling simple words, and c) multiple-choice items. Each of the five raters scored more than three -hundred answer sheets guided by the scoring manual. Although the write-answer type test items here had no more significant indications on raters' differences than the other two types of test items, this study suggested that the DIF, and the MH procedure in particular, could be useful tools of item analysis for the improvement of testing containing subjective scoring items.