SUMMARY
Collins (1981) proposed evidence regarding the kinship of languages (ambelau, buru, sula and taliabo) and grouped them into middle-east Maluku group based on the qualitative evidence. However, the evidence is limited and not convincing. This article intends to explain Proto-Austronesian (PAN) reflexes to Sula language and their relation to Collins Hypothesis (1981) about the West-Central Maluku Subgroup. Field data collection has been carried out in the Sula Islands, especially Fatche isolates in the form of 200 basic vocabulary and 800 cultural vocabulary and Collins study (1981). The collected data was analyzed using a joint innovation top-down approach. The results showed that there was a regular and irregular change in PAN phonemes into Sula language. A PAN phoneme changes regularly and irregularly at once. PAN phonemes that undergo regular changes are * p, * t, * C, * k, * ?, * b, * d, * m, * n, * ?, * s, * S, * i, * a, * u, * ? * uy, and * ay, while those who do not experience regularly are * z, * j, * N, * h, * R, * r, and * w. PAN reflexes into Sula in some respects show a different pattern of change from the evidence put forward by Collins (1981). This does not mean negating Collins's (1981) hypothesis of the family tree of the group, but other relevant evidence needs to be identified. In addition, PAN's reflexes to Sula were found to be the same as Collins's (1981) proposed evidence of the historical relations of the West- Central Maluku Subgroup.