A new model of intonational phonology is here developed, based primarily on the ‘tone sequence’ approaches of Bruce & Gårding 1978 and Pierrehumbert 1980. As in those works, contours are seen as sequences of pitch accents composed of high and low tones. Unlike those works, this model uses cross-classifying features to represent the scaling (height) of tones, and the alignment of tones in time relative to their associated syllables. This form of representation permits both the specification of phonetic detail and the formulation of phonological generalizations. Three features—[delayed peak], [down-step], and [raised peak]—are introduced; to illustrate the type of description they make possible, largely familiar data from English, Dutch, Hungarian, and Rumanian are re-analysed in these terms.