The purpose of this research is to investigate whether negative worded items influence the test criterion-related validity for Likert scale. The "Learning Adjustment Scale" was used for this research. Three classes of elementary school students and three classes of junior high school students were randomly selected for this research. The findings are: (1)The item response consistency (Kappa consistency) between the positively and negatively worded test versions were higher for junior high school students than elementary school students. (2)The test mean scores between the positively and negatively worded test versions were significantly different. The mean scores of the positively worded test version were higher than the mean scores of the negatively worded test version. The differences of the test mean scores and standard deviations were greater for junior high school students than elementary school students. (3)The α coefficients were higher for the positive worded test version than for the negative worded test version. The differences of α coefficients between the positively and negatively worded test versions were greater for junior high school students than elementary school students. (4)The positively and negatively worded test versions resulted in different criterion-related validity for elementary students. The negatively worded test version was more precisely to predict the criterion (academic achievement/grades) than the positively worded test version. However, both the positively and negatively worded test versions had no prediction of the criterion for junior high school students. Therefore, this research suggests that negative worded items tend to result in meaning confusing for younger students and reduce the test reliability. However, positive worded items may not attract examines to pay attention to the test and reduce the test validity.