314 Species
inclusion in a species classification is in effect the only alternative to a typological or
essentialist approach, with the treatment of variance as degradation that it involves.
The use of a term of morphological similarity due to ancestry—“family
resemblance”—by Wittgenstein indicates that the sort of classification principles
appropriate to the natural biological world (and by extension to the sociolinguis-
tic world) are very different from those appropriate in the mathematical sciences
that served as the source for so much of modern western thought. There are some
deep issues involved: lineages of information or structure appear to generate group-
ings that are complex, fractal, and stochastic, and the “linear” treatment of natural
kinds that may (or then again may not) be useful and effective in the physical sci-
ences, where history and individuality are not relevant to the state description of
a system or process, are less useful and less effective in biology. The older views
that consider species to be natural kinds of the physical/mathematical variety are
severely Procrustean, and eventually artificial and therefore subjective. I think that
Wittgenstein’s FRP is an abiological attempt to generalize a class of “nonlinear”
predicates, ones that are appropriate in biological work.
DO FAMILY RESEMBLANCE PREDICATES WORK FOR BIOLOGICAL SPE CIES?
The taxonomic movement known as “numerical taxonomy” at the time, of its formula-
tion in the 1960s and later as “phenetics,” attempted to use something like Wittgenstein’s
FRP. Indeed, the originators of phenetics, Sokal and Sneath, drew for inspiration on the
work of Morton Beckner, who had explicitly used the Wittgesteinian FRP to discuss
species.^91 However, phenetics ran up against a problem—their clustering through the
use of computer analysis of characters in a morphological metric space led to instability
and sensitivity of the results to the choice of characters chosen. Effectively, phenetics
was an attempt to find the natural groupings of the populations themselves, in the hope
that species would appear as clusters in a Cartesian metric that were isolated from other
populations, and that this would be possible no matter which characters were used.
The problem appears to rely on two facts about actual species: first, there is no
absolute measure of difference in, say, morphological metrics or genetic distance that
is either common to all separate species or absolutely enough to prevent interbreeding
in sexual organisms. Second, species do not all vary commensurately in all charac-
ters. Some characters are evolutionarily conserved for various reasons: they may be
strongly buffered against change due to developmental linkages (pleiotropy), or they
may be held constant by stabilizing selection because they are functionally important.
So, the choice of characters to use is significant. Typically, taxonomists use “useless”
(“nonfunctional”) characters, which can change more or less at a constant rate over
evolutionary time, because homoplasies (convergently evolving characters) lose phy-
logenetic information the way that a pathway that joins another pathway is no longer
as informative about the prior journey of a walker as a single pathway would be—the
walker may have traveled along two pathways now, and knowing the current location
of the traveler is no help in determining which way they went.
(^91) However, Chaney 1978 argues that the FRP and polythetic groups are different beasts, and the use of
polythesis by numerical taxonomists is independent of the FRP.