Objective This paper is to validate a speech intelligibility test in Chinese mandarin for the evaluation of speakers with cleft palate, and to determine the intelligibility deficit patterns based on minimal pairs. Methods Thirty cleft palate patients with dysarthria were included as speakers for the test. Then the speech intelligibility of each speaker was rated and the error proportions of each minimal pair were analyzed by SPSS 16.0. The stepwise multiple regression analysis was employed to determine which minimal pairs were the best predictors of reduced speech intelligibility. Results The intelligibility scores were obtained form the thirty speakers ranged from 37% to 92.4%. The error proportion of the minimal pairs was ranged from 0. 13 to 0. 80. The three most problematic pairs, from the most to least severe were dental versus retroflex pair (0.80), retroflex versus palatal pair(0.68), and aspirated versus un-aspirated affricative pair(0.66). The speech intelligibility from cleft palate patients could be predicted at 71.5% accuracy using two pairs, aspirated versus un-aspirated affricative pair, dentals versus palatal pair. Conclusion The speech intelligibility test for Chinese Mandarin speakers with cleft palate can be used as an effective evaluation method, both qualitatively and quantificationally.