Abstract
In this paper I discuss some considerations which enter into the definition of a linguistic period, i.e. a particular phase in the history of a language. I propose that historical varieties of a language be treated more on a par with synchronic variation. Our definition of a period must therefore avoid employment of arbitrary, extra-linguistic boundary-markers and should recognize and allow for graded types of categorization to supplement the usual metaphor of periods as discrete entities. What I am proposing is to bring our studies of historical variation closer to what has long been recognized as necessary in synchronic variation work. In what follows I will show how synchronic and diachronic variation have been treated differently by linguists. Then I put forward some programmatic suggestions on how to revise our definition of linguistic periods in the light of this comparison. Finally I provide a brief illustration of how this all works by considering a problem of periodization in the history of Welsh.