The article presents problems that arise when we try to define cluster concepts by listing their necessary and sufficient features. This definition does not work because for cluster concepts none of the features is necessary, but all together the features are sufficient. The article considers a very original way of defining cluster concepts developed by David Cooper, called sufficiency definition. Although this approach is disputable, it is interesting in the sense that it allows a better understanding of the nature of cluster concepts. Cooper proposed to apply to them the definition of the form: "A =dfh Something having .sufficient of the properties P1, P2, Pn". The main difference between such definitions and traditional ones is that they contain an indication of a sufficient number of properties. The reference to a sufficient number of properties is justified by the fact that in relation to cluster concepts it is impossible to list necessary conditions for their application. When we need to define one or another cluster concept, we list only those properties, the presence or absence of which allows us to apply it. A sufficient definition is the most appropriate way in this case to define cluster concepts. Interesting objections and sophisticated arguments against the theory of sufficient definitions are raised by Steven Boёr. In particular, he suggested to differentiate "strongly backed" and "weakly backed" cluster concepts and showed what difficulties the theory of sufficient definitions faces. His conclusion is that sufficient definitions can be applied only to a limited number of "strongly backed" cluster concepts. But this conclusion is not fatal for the theory of sufficient definition, because its main conclusions do not rest on the semantic aspect of the definition, but on the pragmatic one. The reason is that all depends on how the definition is used in various linguistic practices. Cooper called this reason the "pragmatics" of definition.