In order to identify dropout-at-risk students early in their elementary years, the researchers examined the literature of empirical dropout studies and interviewed specialist of Taitung in this matter and developed a 60-item Dropout Tendency Scale through different stages of pretests. Twelve thousand thirty one 6th graders and seven hundred 7th graders participated in this study. Results of this study showed that this scale has good reliability and validity. The test-retest reliability coefficients ranged from .65 to .82 for the subscales and .85 for the whole scale. The internal consistency as measured by Chronbach’s α’s also ranged from .79 to .89, and the α’s for the whole scale are .95 and .96 for the 6th graders and 7th graders repectively. This scale did not fair as good in factor analysis, which were run intented to examine the construct validity. However, the results that boys scored higher than girls, and that the aboriginal students scored higher than non-aboriginal students in this scale still indicate sound construct validity. Lastly, students’ scores in this Dropout Tendency Scale correlate substantially high and statistically significant with students’ math and Chinese achievement scores, truancy counts, and teacher’s prediction scores, which indicate good concurrent validity of criterion-referenced validity. Still in want is the predictive validity, which could only be available by follow-up studies in a couple of years to come.