On Wed, 2 Aug 1995, Meredith Lane wrote:

I don't believe that Sokal, Sneath, and others ever said let's not try. What they objected to was the misguided attempt to interpret phenetic analyses (especially phenograms) as indicators of evolutionary relationships. In their books they devote considerable space to numerical cladistic methods. There was, and still is, a fundamental philosophical difference of opininon on hwat a classification should represent, and how we should construct classifications. Gilmour naturalness is a not unreasonable criterion and phenetic classifications can be very natural. In fact, phenetic classifications can be more stable and can have higher predictivity than cladistic classifications. It depends on the data used to construct the classification. A lot of hay was made by those who used special data sets to demonstrate the fallibility of both methods - it is easy to come up with a data set that will demonstrate superiority by either approach. One concern about cladistic approaches is that they have an extra layer of hypotheses between the data and the classification - generally, polarity is hypothesized for each character, a step not necessary for a phenetic analysis.

I will repeat a point I have made before. One reason for the swing was the apparent mathematical simplicity of numerical cladistics. In fact, no math is really necessary. However, eigen analysis cannot be intuited very easily from the data and certainly is not something that is easily done by hand (excpet in trivial cases).