The essays in this volume address foundational questions in phonology: What sorts of phenomena comprise the explananda of the field? How should phonological objects be represented? What is the optimal architecture for phonological theory? These questions cut across different schools of thought within the discipline, and they remain largely open after a half century of research. A main theme is that to study phonology productively, one must ask what modules are necessary, how these modules interact with each other and with other components of linguistic theory, and what the representational and computational resources of each module are. Computation and representation are inherently linked; as John McCarthy sagely remarked,‘‘Simply put, if the representations are right, then the rules will follow’’(1988: 84).

The modular approach seems natural simply because phonology is a component of human cognition, ultimately a biological object; all biological entities more complex than viruses are arguably best understood in a modular framework. The modular approach also enables us to break the larger questions of phonology into smaller ones. We can ask of apparently bewildering arrays of complex surface phenomena: What components are responsible for the facts at hand? Which aspects of the behavior are directly due to operations within components and which emerge from interactions between components? Rough answers to these questions help develop more precise questions about the individual components and the architecture that houses them. The following sections introduce three examples showing how the modular approach advances our understanding of phonology. The first example (section 1.2) illustrates the benefits of a modular perspective by showing that advances in the understanding of phonetic modules allow us to remove from phonology’s purview a classic and formerly vexing problem, the North American English vowel length alternations before voiced and voiceless obstruents. The efficacy of modularity within the phonology proper is shown in the next example (section 1.3), where we argue that