Recent years have seen a rapid increase in opportunities for very young children in the UK to participate in musical activity. These opportunities are provided by a range of music professionals who are expanding their work into the early years sector. Support for this increase in early childhood music draws on recent research into aspects of musicality, fuelling a growing conviction that music is developmentally beneficial. This article, while not implying any diminution in the potential value of music in the upbringing of very young children, calls for a more rigorous and critical use of research evidence. The main argument presented is that research evidence is being used discriminately to divert attention onto the transferable benefits of music into domains such as basic skills--music as means--rather than direct attention to musicality as it is manifest in earliest childhood. The article draws on ideas evolved in practice during an action-research project which aimed to develop arts practice in early childhood settings. The project team of artists--recruited from a range of disciplinary backgrounds, including music--gradually developed a generic model of improvisatory arts activity which emphasized dynamic temporal-spatial activity. This model of practice connected with the corpuses of recent research into infant musicality at points which, I will argue, are more congruent with their findings than the discriminatory readings. The revised model is further discussed and expanded with reference to some priorities and theoretical positions which are currently influential in early childhood arts practice.