Since its beginnings, phonological research has sought to discover constants below the phonetic flow associated with the linguistic expressions produced by speakers. Saussurean distinctiveness as the core property of phonemic units in European structuralism on a par with the functional criteria at the base of American phonemics have led to an interpretation of phonological phenomena in terms of distributional constraints on the phonetic segments, placing them in relation to some abstract level. We argue for the inadequacy of this approach in order to interpret the relation between lexical properties and the mapping onto the sensory-motor level. Specifically, the main formal requirements such as linearity, invariance, and biuniqueness of structural phonemics are called into question by the fact that harmonies, metaphonies, and other phonetic mechanisms provided by Romance systems escape such procedures and require a more powerful theory involving a cognitive level of organization. In fact, the speaker is able to master very complex phonological knowledge, including processes neutralizing underlying differences and types of overlapping, for which the notion of phoneme is of no use. On the contrary, these processes support an analysis in which lexical representations are connected to surface occurrences by rules and derivations as a part of the language capacity.

Chapter 3 first summarizes the differences between phonetics (Chapter 2), which looks at human speech sounds in general, and phonology, which examines how a subset of possible sounds is used and distributed in specific languages. It introduces the concept of the phonological unit, the phoneme, and how phonemes can be identified by minimal pairs, that is, comparing words that differ by only one sound yet have distinct meanings. It also explains the difference between phonemes and allophones that are typically the result of sound processes, as seen in Chapter 2, but may also be in free variation. We then explain how phonemes can be reduced to a limited set of distinctive features that allow us to organize them into natural classes to which phonological rules may be applied. We see how these rules can be represented easily to capture important generalizations about phonology. The chapter also examines the structure of syllables and explains how different languages may syllabify words and phrases in different ways, while at the same time obeying universal principles. The chapter ends with a section on how to discover the sound system of a language.