In Chapters 2 and 3 we saw that the segment inventories of languages can be divided into subgroups. Thus, we have separated the group of vowels from the group of conson ants, and when we discussed the way languages ‘build up’ their inventories, we dis tinguished groups of voiceless obstruents (e.g. [p,t,k]) from voiced groups that are otherwise the same (i.e. [b,d,] in our example). This would appear to suggest that the segment is not the smallest constituent of phonological structure. In this chapter, we will motivate this assumption. We will introduce the distinctive features that will in principle enable us to describe the segments in the world’s languages, and to refer to those groups of segments that play a role in their characteristic phonological processes and constraints. The latter consideration will be shown to provide an important motivation for the assumption of distinctive features. In this perspective, these features are the elements by which we can refer to natural segment classes, groups of segments that are treated as groups by languages.