This paper focuses on the particular kinds of difficulties which arise in the study of an emergent tone-system, exemplified by Tamang in Nepal, where pitch, phonation and other laryngeal features combine in the definition of a tone. As a consequence, conducting a well-ordered analysis in stages first of phonetic transcription, then variation in context, then interpretation is not possible. Rather we have to discover the contrasting categories first, and study their phonetic realization next, or do both at the same time. This also leads to questioning the validity of the traditional distinction of features into "distinctive" and "redundant" and proposing instead an analysis of an abstract "tone" as a bundle of cues. We will only sketch the second characteristic of the Tamang tone system, the extension of tone over the phonological word. The contributions of instrumental studies and of a comparative-historical perspective are discussed.