Many speakers with repaired cleft palate have reduced intelligibility, but there are limitations with current procedures for assessing intelligibility. The aim of this study was to construct a single-word intelligibility test for speakers with cleft palate. The test used a multiple-choice identification format, and was based on phonetic contrasts which are problematic for speakers with cleft palate. Fifteen Cantonese speakers with repaired cleft palate served as speakers. The intelligibility score of each speaker was calculated, based on the number of words correctly identified by eight naive listeners. The most problematic contrasts, based on mean error proportions, were place of articulation (stops and nasals), stop versus fricative, and stop versus affricate. The most robust contrasts were also determined, and physiological and language-specific interpretations offered. Single-word intelligibility could be predicted with 91% accuracy using three phonetic contrasts. The importance of an 'explanatory' intelligibility test, for focusing treatment and monitoring severity, is discussed.