Many philosophers and medical scientists assume that disease categories or entities used to classify concrete cases of disease, are often defined by disease mechanisms or causal processes. Others suggest that diseases should always be defined in this manner. This paper discusses these standpoints critically and concludes that they are untenable, not only when 'disease mechanism' refers to an objective mechanism, but also when 'mechanism' refers to a pragmatically demarcated part of the total "objective" causal structure of diseases. As an alternative to principles that use the concept of disease mechanism or analogous concepts, a pragmatic approach is suggested and described. This approach has been suggested before, but in problematic or inadequate versions. This paper proposes a version compiled of two "pragmatic principles" and shows that they are much more adequate than the principle of disease mechanism. With reference to a case study of a still ongoing international discussion of various candidates for a classification system for malignant lymphomas, including REAL (Revised European-American Classification of Lymphoid Neoplasms) in which the concept of disease mechanism or analogous concepts plays a very small part, it is shown just how pivotal these two pragmatic principles can be in actual discussions of definitions of diseases. Finally, it is pointed out that with regard to modern philosophy of language it may, at least in some cases, be problematic to distinguish between the two pragmatic principles as they stand.