Numerical taxonomy validity

We prefer to let historians of science pursue this argument. For although it was — and remains — important to trace the roots of the historical origins of an idea in science, the development of numerical taxonomy has so far outpaced the early primitive ideas on this subject that to have to rely on Adanson s views for a validation of modern numerical phenetics seems as irrelevant as to rely on Mendel s writings for a validation of the findings of the molecular geneticists. (Sneath and Sokal, 1973 23-24)... 

Their defensiveness is obvious, for surely no one who used the term Adansonian pretended that the views of that dead botanist certified the validity of numerical taxonomy. 

I believe the apparent inconsistency between issue 1 and 2 can be explained by a careful consideration of what history is and how the present relates to the past. I believe that both statements are deeply true and I intend to address this topic explicitly elsewhere. Here, I shall address only issue 3. Sokal and Sneath seem to assume the existence of an unbiased higher court populated by historians of science. They imply that the critics of their view of Adanson, who were biologists by profession rather than historians, had an axe to grind and were finding fault with Sneath s references to Adanson as a way of questioning the soundness of numerical taxonomy itself. That concern about objectivity is valid, though of course it also points to the possibility that their own interest in Adanson was likewise liable to be biased. 

NSF, see National Science Foundation Nuclear hormone receptor genes, 272 Null hypothesis, 58 Numerical taxonomy, 127 development of, 6 validity of, 7... 

In contrast to the numerical taxonomy quality function, the conceptual clustering quality function takes into account the utility of the concepts representing the clusters, as seen below, where the parameter CONCEPTS is the set domain of concepts which may be used to describe the clusters. In the context of conceptual clustering, a concept is a valid cluster description as specified within the cluster description language. Consequently, the set of possible concepts need not be explicitly enumerated. An important bias in this method is the use of the simplicity of a cluster description as a metric for the quality of clustering. 

---