There is currently considerable controversy within philosophy over how a natural kind term's extension is determined. Adherents to traditional theories of meaning argue that extension is determined by intensional properties which usually consist of identifying descriptions. Recently a new, essentialist alternative to these traditional accounts has been advocated; proponents of this view maintain that a term's extension includes all objects which are essentially the same as a given paradigm instance of the term's use. In the present thesis it is argued that both description theories and essentalism describe not how classification must proceed but rather two alternative strategies for how a classification of natural kinds might be attempted. A term's extension is not determined in advance by either identifying descriptions or hidden essential properties since stress on either of these is itself a choice. This claim is exemplified by using empirical material drawn from the recent (post 1900) history of a classifictory science, botanical systematics. By means of this evidence it is shown that both "descriptionist" and "essentialist" strategies of concept application have been (and still are) pursued by different groups of taxonomists. One consequence of the position outlined above is that classifications are conventions and that they are evaluated instrumentally. The force of this argument is best illustrated by conceiving of classifications as part of a wider network of beliefs which are socially transmitted and sustained. Changes in networks are designed to further the interests to which a network is being put. It is argued here that the main kinds of interests which have been important in twentieth century systematics are 1) interests in technical prediction and control and 2) professional vested interests. An important aspect of both these kinds of interests is that they are normally considered to be "internal" to science. It follows that to understand scientific knowledge from a sociological perspective does not, of necessity, entail commitment to "external" explanations of scientific change.