The purpose of this study is to analyze reliability of holistic and analytical assessments teachers as raters implemented on sixth-graders' persuasive speaking. Raters tended to give higher marks in holistic assessment than in analytical assessment. All the students but D got higher marks in holistic assessment, which was statistically significant for B alone (p=.015). Only B showed significant correlation in reliability between holistic assessment and analytical assessment of speaking. To improve reliability of speaking assessment methods, it is essential to train raters. Also necessary is a research in authentic speaking assessment elements as the foundation for speaking assessment and in setting assessment standards which is easy to apply.