534 
Telopea Vol. 6(4): 1996 
the patterns produced by one suite of characters are consistently different from the 
patterns discernible using all the other characters. Hence, they might look much harder 
at the discordant characters to see if they might have been individuated incorrectly. 
Embryological investigations might show that these suspect characters have been 
misidentified. Probably the most difficult aspect of cladistic analysis is to distinguish 
primitive from derived characters and to organize them into transformation series. 
The trouble with the views of scientific method set out by Popper (conjectures and 
refutations) and Hennig (reciprocal illumination) is that they portray scientific method 
as being messy. There is no one preferred way to begin scientific investigations. You 
start out wherever you happen to be and try one damned thing after another. If one 
hne of research doesn't work out, try some other avenue. It may turn out to be a 
dead end too, or you might succeed in organizing a large area of science into a 
single system. In short, there is no such thing as the scientific method. As Johnson 
(1972:12) characterized his own method: 
Since we accept, as indeed most pure pheneticists do, that the characters of 
organisms which are important to them, and to us, are determined largely by 
their evolutionary history, we inevitably become involved in partial circularity of 
argument if we base our classificabons themselves to some degree on phylogenetic 
considerations and interpretations. The building of such partially phylogenetic 
classifications involves some positive feedback from conclusions to argument, 
and hence incurs stern disapproval from those who seem to think that scientific 
investigation and interpretation should depend on simple elementary logic. 
Numerical taxonomists were not the first to see that messy lines of inference run the 
danger of becoming circular. However, the only alternative is to have a Descartian 
'straight line' methodology in which first one does A, totally and perfectly, and then 
one does B, just as totally and perfectly, and so on. No errors ever get introduced 
into science. Hence, none ever have to be eliminated. Although Johnson admits that 
sometimes systematists have lapsed into arguing in circles, he insists that traditional 
phylogenetic methods themselves are not inherently circular. 
Back in 1968 1 agreed with Johnson, and I still do. But as I mentioned previously, 
systematists such as the pheneticists and pattern cladists, who seemed to have been 
arguing for straight-line scientific methods that exclude all theoretical speculations 
in the early stages of classification, now claim that they have been seriously 
misinterpreted. Although Sokal and Sneath (1963:22) rejected Hennig's method of 
reciprocal illumination as one more example of the 'much-condemned vertical 
construction of hypothesis upon hypothesis,' later they set out a method that looks 
very much like reciprocal illumination: 
It should be clear that generalizations about the taxa cannot be made before 
one has recognized the taxa; that taxa cannot be recognized before the 
resemblances between organisms are known; and that these resemblances 
cannot be estimated before organisms and their characters have been examined 
(Sneath & Sokal 1973; 5). 
Thus, systematists must start with examining organisms and their characters, then 
proceed to estimating resemblances between organisms, then to the recognition of 
taxa, and finally to generalizations about taxa. This procedure looks very much like 
a straight line methodology, but Sneath and Sokal (1973: 5) go on to acknowledge 
that 'some of these steps may be in effect combined in certain computational methods, 
or the whole procedure may be repeated a second time for some special reason.' 
However, whenever such repetition occurs, the 'order of the steps within the 
procedure cannot be changed without destroying the rationale of the classificatory 
process' (Sneath & Sokal 1973: 5). 
