Traits, Genes, and Coding 391the cause could have been interpreted in many other ways by the flower. Thus the
cause in this arrangement counts as arbitrary, it’s environmentally located, and
it looks like the kind of factor that really shouldn’t count as a representation of
a developmental outcome; so, with respect to the arbitrariness condition alone,
excessive liberality emerges as a genuine danger. Of course, I have characterized
arbitrariness not in terms of a cause potentially having a range of different ef-
fects, but in terms of the equivalence class of different physical factors that could
have played the same causal role being fixed by informational rather than brutely
physical considerations. Nevertheless, the content ‘start to flower’ could clearly
have been carried by environmental factors other than increase in day length, so
it looks as if arbitrariness in my sense is present too, and in the same worrying
place. The solution (in the framework I am promoting) is to take seriously the
conceptual interlock between arbitrariness and homuncularity with respect to the
justification of coding-talk. For while, in the flower case, it might well be said that
there is a consumer system that digests the putative information (by interpreting
the increase in day length as an instruction to flower), it is hard to see how to make
sense of the claim that the overall arrangement contains a producer system that
has performed the role ofencoding that information inthe relevant causal factor,
namely in the increase in day length. So the environmental factor in question does
not emerge as being representational in character.
This response to Godfrey-Smith’s example does not establish that non-genetic
factors could not ever qualify as vehicles of representational content in develop-
ment, once arbitrariness and homuncularity are plugged in as necessary condi-
tions. Take animal signalling systems. If one could specify the appropriate causal
co-variations (that is, between the signals and the construction of developmentally
downstream structures), those systems will contain noises, marks, and so on, that
will count as environmentally located vehicles of representational content. (Of
course, the producer subsystem will be in one individual animal, while the con-
sumer subsystem will be in another, but nothing I’ve said rules out such a state of
affairs.) However, notice that the existence of such elements does not violate the
weakened uniqueness constraint. It is neither unreasonable, nor extravagant, nor
explanatorily inefficacious to claim that the developmental contribution of such
factors is representational in character. What needs to be ruled out is the system-
atic inclusion of illegitimate factors (such as an increase in day length). And that,
I think, is plausibly achieved by a proper recognition of the part played by the
producer subsystem. However, that recognition also brings us to what, I suspect,
is the most controversial claim that I shall make in this paper
5 A BULLET TO BITEStrictly speaking, according to the proposal currently on the table, it’s not the
DNA molecules that constitute the representational vehicles that play a coding
role in development, but rather the nucleotide triplets (thecodons!) that make