This paper outlines a novel approach to studying children's singing that enables a more accurate and global picture to be drawn of children's actual level of singing achievement across a range of tasks. It enables predictions to be made about their potential singing development in a formal school setting. It draws on a two-year longitudinal field study of children aged 4 to 8 years. The results highlight considerable diversity in actual singing achievement in different tasks and different test contexts. Children's singing development is both varied and multifaceted, and this has considerable implications for teaching and assessing singing as well as other forms of music-making that depend on singing in the first few years of school. (Contains 8 figures and 2 tables.)