Programmers employ a multitude of languages to build systems. Some are general-purpose languages. Others are specific to individual domains. These assist programmers with at least three different tasks: domain modeling, system validation and representing the structure of their general purpose program. As a result, programming languages have become key factors in the software engineering process. They are, however, rarely codified into the process and treated systematically. My dissertation develops a framework to treat programming languages as software engineering artifacts. In this framework, languages are identifiable, reusable entities that programmers can compose and link to produce larger languages; furthermore, languages themselves meet the properties of software components. Programmers can augment this lateral growth of languages with vertical growth, by producing languages that synthesize languages. Thus, software construction becomes a multi-phase process. In later phases, programmers use languages to build programs; in earlier phases, they employ languages to construct languages. This treatment of languages as artifacts addresses several open questions.