The domains of natural information 61
and hence will pose too high a requirement. Given that organisms are always bound
to locally and temporally variable environmental conditions, this argument has
some prima facie plausibility. As de Wit et al. (2015) admit, locating ambiguities in
perception in the informational relations themselves will be a concession to “inferential” approaches to perception, for the sake of ecological credibility. However,
achieving such credibility does not require one to accept the inferentialist “doctrine
of intractable nonspecificity” of informational relations (as Turvey and Shaw 1979
call it). If informational relations may hold between all sorts of world affairs, and if
they are regular in the way outlined by Dretske, there will be relations in the environment that remain fully specific even under changeable ecological conditions.
Although highlighting the importance of the degrees of a perceiving organism’s epistemic contact with his environment is well-taken, acknowledging situations of underspecification does not require one to admit for a nonspecificity of
information. Just as the presence of invariants in the environment does not imply
invariance in ecological conditions for perception, variance in ecological conditions for perception does not amount to a weakening of invariant relations in the
environment. Indeed, camouflaged predators indiscernible from harmless twigs
may evolve, or humans may invent windows that happen to block a bird’s flight
path. In such cases, the optical invariants that are amenable to perception for the
animal and that had hitherto been sufficient for specifying to him an ecologically relevant condition in general or an affordance in particular remain in place,
whereas some of the distal conditions typically related to these optical invariants
have changed for a relevant subset of cases. However, if an optical invariant had
been sufficient to specify an open flight path to the bird before the invention of
glass windows but now is insufficient to do so, that invariant itself will not have
become compromised after that invention. What has happened is that this optical
invariant alone cannot be used as a reliable proximal signal of the distal ecological
condition anymore, and hence is now associated with a probability p < 1 within
the reference class of the type of proximal signals available to the bird.
Ecological changes affecting epistemic contact do not per se constitute a problem, yet there will be a penalty for being mistaken too often if and when the
receiver depends on discriminating between different conditions with a certain
degree of reliability. It is the requisite degree of reliability that determines the reference class for the signals and their probabilities. If the first and slightest mistake
is sufficient to get the organism into trouble, the reference class will be restricted
in such a way as to approximate Dretskean rigidity. If an individual or population
can live with 95% false alarms, conditions will be much more relaxed. Still, all
rounds of picking up the putative signal will be counted into the respective reference class. An organism is not typically in a position to gather information about
changes in ecological conditions or the degree of reliability of a signal. Depending on where the acceptable margin of error lies, he will either have to live with
increased uncertainty or undergo some adaptive reorganisation. He must then
either be or become able to detect where and when the reliability of the signals
attains specificity, which, under natural conditions for perception, seldom occurs,
or he must detect where it remains high enough to meet his or her requirements.
and hence will pose too high a requirement. Given that organisms are always bound
to locally and temporally variable environmental conditions, this argument has
some prima facie plausibility. As de Wit et al. (2015) admit, locating ambiguities in
perception in the informational relations themselves will be a concession to “inferential” approaches to perception, for the sake of ecological credibility. However,
achieving such credibility does not require one to accept the inferentialist “doctrine
of intractable nonspecificity” of informational relations (as Turvey and Shaw 1979
call it). If informational relations may hold between all sorts of world affairs, and if
they are regular in the way outlined by Dretske, there will be relations in the environment that remain fully specific even under changeable ecological conditions.
Although highlighting the importance of the degrees of a perceiving organism’s epistemic contact with his environment is well-taken, acknowledging situations of underspecification does not require one to admit for a nonspecificity of
information. Just as the presence of invariants in the environment does not imply
invariance in ecological conditions for perception, variance in ecological conditions for perception does not amount to a weakening of invariant relations in the
environment. Indeed, camouflaged predators indiscernible from harmless twigs
may evolve, or humans may invent windows that happen to block a bird’s flight
path. In such cases, the optical invariants that are amenable to perception for the
animal and that had hitherto been sufficient for specifying to him an ecologically relevant condition in general or an affordance in particular remain in place,
whereas some of the distal conditions typically related to these optical invariants
have changed for a relevant subset of cases. However, if an optical invariant had
been sufficient to specify an open flight path to the bird before the invention of
glass windows but now is insufficient to do so, that invariant itself will not have
become compromised after that invention. What has happened is that this optical
invariant alone cannot be used as a reliable proximal signal of the distal ecological
condition anymore, and hence is now associated with a probability p < 1 within
the reference class of the type of proximal signals available to the bird.
Ecological changes affecting epistemic contact do not per se constitute a problem, yet there will be a penalty for being mistaken too often if and when the
receiver depends on discriminating between different conditions with a certain
degree of reliability. It is the requisite degree of reliability that determines the reference class for the signals and their probabilities. If the first and slightest mistake
is sufficient to get the organism into trouble, the reference class will be restricted
in such a way as to approximate Dretskean rigidity. If an individual or population
can live with 95% false alarms, conditions will be much more relaxed. Still, all
rounds of picking up the putative signal will be counted into the respective reference class. An organism is not typically in a position to gather information about
changes in ecological conditions or the degree of reliability of a signal. Depending on where the acceptable margin of error lies, he will either have to live with
increased uncertainty or undergo some adaptive reorganisation. He must then
either be or become able to detect where and when the reliability of the signals
attains specificity, which, under natural conditions for perception, seldom occurs,
or he must detect where it remains high enough to meet his or her requirements.