Most studies on CV co-occurrence patterns in early phonetic development have been based on Indo-European languages. Data from infants learning Mandarin, which has a substantially different phonological system from Indo-European languages, can confirm or refute the findings of previous studies, thus shedding further light on the theoretical bases of CV association. Spontaneous vocalizations were recorded in 45-minute sessions for each of 24 Mandarin-learning infants aged 0;7 to 1;6. In addition, the speech production of 24 caregivers was audio-recorded during their natural daily interactions with the infants at home. Both infants' vocalizations and adults' speech were transcribed and analysed for consonant–vowel co-occurrence patterns. These patterns were similar to those in other language groups, but language-specific patterns were evident by 1;0. Combinations of alveolars+front vowels and velars+back vowels confirm Davis & MacNeilage's (1990, 1995) frame-then-content theory and Clements's (1991) model of unified features for consonants and vowels. However, our finding of a language-specific pattern (labials+back vowels) suggests the need to reexamine the ‘pure frame’ of Davis & MacNeilage and Clements's grouping of labials and rounded vowels.