The domains of natural information 67
condition, namely the presence of some individual, population or set of entities
to be tracked by R (see NI-5b). The domain of a type of signals is a natural way
of carving out a reference class for the informational relations involved, and its
spatio-temporal boundaries.
In order to illustrate the problem to be solved by introducing the concept of
domains of natural information, a brief discussion of Millikan’s own example in
(2004, Chapter 3) which she, in turn, borrowed from Dretske (1988), will be helpful: if tracks of a certain shape in the woods of her (or mine, or someone else’s)
home state have hitherto always and exclusively been created by quail (or, to
abstract from concrete species, “q-birds”), they carry information about q-birds
with a probability of 1 and, one might believe, they do so with lawful causal
determinacy, as no other animal in those woods would be able to create tracks of
the same shape, size and relative placement. The domain of the signal will be the
entire set of q-tracks in those specific woods, and the reference class will comprise
those tracks, the q-birds and the regularities by which q-birds give rise to q-tracks,
which will be quite strict under this perspective. Nothing, however, rules out the
possibility that pheasants (or “p-birds”) may happen to cause tracks of the very
same appearance, save for the fact that there are no p-bird populations in that area.
What information, then, would tracks of that shape carry if a population of p-birds
happened to migrate, say, from Tyrol to Carniola? And what would warrant the
lawfulness of the correlation between tracks of that specific shape, size and relative placement in the woods of Tyrol, save for the fact that q-birds have not happened to migrate the other way? Finally, even if there were no such species as the
p-bird to exist at the time of observation in the first place, what would rule out the
possibility for them to evolve at a later time, or to have roamed the same and other
woods in bygone days but now having become extinct? Under this perspective,
the domain of the signals involved might remain the same in terms of the set of
q-tracks in a given wood under investigation, whereas the regularities and possible source conditions included will have changed. The reference class will now
encompass q-birds and the patterns of p- and q-bird population dynamics.
If there are q-birds as populations with a limited extension in space and time,
and if there is a possibility of p-birds existing at other times and in different places
while leaving behind identically shaped tracks, the information carried by those
tracks would end up being equivocal and fall prey to the problem of disjunctive
content that, according to Jerry Fodor (1990), bodes ill for an evolutionary naturalism about content. Thus, on Dretske’s definition, it would not be information
at all (although he allows for a softer notion of information in other places).
3
Millikan’s alternative is to characterise natural information C as probabilistic and
locally bound – which, as I have begun to argue in the previous section, are not
precisely the same thing.
On this background, domains of natural information or “natural signs” are to be
understood in accordance with the mathematical meaning of the term “domain” in
the first place. The mathematical meaning is that of the domain of a function, that
is the set of input values for which the function is defined (e.g. all real numbers
or all positive integers), and from which the output values, as the “image” of the
condition, namely the presence of some individual, population or set of entities
to be tracked by R (see NI-5b). The domain of a type of signals is a natural way
of carving out a reference class for the informational relations involved, and its
spatio-temporal boundaries.
In order to illustrate the problem to be solved by introducing the concept of
domains of natural information, a brief discussion of Millikan’s own example in
(2004, Chapter 3) which she, in turn, borrowed from Dretske (1988), will be helpful: if tracks of a certain shape in the woods of her (or mine, or someone else’s)
home state have hitherto always and exclusively been created by quail (or, to
abstract from concrete species, “q-birds”), they carry information about q-birds
with a probability of 1 and, one might believe, they do so with lawful causal
determinacy, as no other animal in those woods would be able to create tracks of
the same shape, size and relative placement. The domain of the signal will be the
entire set of q-tracks in those specific woods, and the reference class will comprise
those tracks, the q-birds and the regularities by which q-birds give rise to q-tracks,
which will be quite strict under this perspective. Nothing, however, rules out the
possibility that pheasants (or “p-birds”) may happen to cause tracks of the very
same appearance, save for the fact that there are no p-bird populations in that area.
What information, then, would tracks of that shape carry if a population of p-birds
happened to migrate, say, from Tyrol to Carniola? And what would warrant the
lawfulness of the correlation between tracks of that specific shape, size and relative placement in the woods of Tyrol, save for the fact that q-birds have not happened to migrate the other way? Finally, even if there were no such species as the
p-bird to exist at the time of observation in the first place, what would rule out the
possibility for them to evolve at a later time, or to have roamed the same and other
woods in bygone days but now having become extinct? Under this perspective,
the domain of the signals involved might remain the same in terms of the set of
q-tracks in a given wood under investigation, whereas the regularities and possible source conditions included will have changed. The reference class will now
encompass q-birds and the patterns of p- and q-bird population dynamics.
If there are q-birds as populations with a limited extension in space and time,
and if there is a possibility of p-birds existing at other times and in different places
while leaving behind identically shaped tracks, the information carried by those
tracks would end up being equivocal and fall prey to the problem of disjunctive
content that, according to Jerry Fodor (1990), bodes ill for an evolutionary naturalism about content. Thus, on Dretske’s definition, it would not be information
at all (although he allows for a softer notion of information in other places).
3
Millikan’s alternative is to characterise natural information C as probabilistic and
locally bound – which, as I have begun to argue in the previous section, are not
precisely the same thing.
On this background, domains of natural information or “natural signs” are to be
understood in accordance with the mathematical meaning of the term “domain” in
the first place. The mathematical meaning is that of the domain of a function, that
is the set of input values for which the function is defined (e.g. all real numbers
or all positive integers), and from which the output values, as the “image” of the