Put another way, we don’t want to start by buying into a conception of symbols which is
too congenial to approaches viewing language largely or completely in terms of its
abstraction amenable aspects. The more one focuses on those aspects, we maintain, the
more difficult it is to see how language could possibly get started, or, perhaps, how
symbols could be ‘grounded’ (Harnad 1990).
Recall that utterance-activity embraces both analog (or non text-like) and non-arbitrary
elements. To balance this permissiveness, it is useful to adopt some way of
conceptualising how aspects of utterance-activity relate to the how question. For the
present occasion we’ll use an ‘off the shelf’ solution - the distinctions between iconic,
indexical and symbolic reference due to Pierce (1955), especially as appropriated by
Deacon (1997). Rather than directly defend the distinctions, we’ll simply take them on
board as a taxonomy, leaving aside the empirical question about the extent to which the
specified categories are occupied, or the taxonomic analysis is a useful or powerful one.
Iconic reference involves some kind of perceived resemblance, perhaps even to the extent
of failure to distinguish, between two features of the world. Deacon (1997: 75) uses a
camouflaged moth as an example, which is only successfully iconic of tree bark to the
extent that it is not perceptually distinguished from the bark on which it stands. The
iconic relationship is, given the range of ways in which two things might be said to
resemble one another, a relatively weak one.
Indexical reference on the other hand requires some degree of correlation between two
re-identifiable types. Again there is a wide range of possible types of correlation,
including spatial adjacency and temporal succession. In order for there to be an indexical
relationship, a perceiver must be able to identify phenomena as instances of the two
types (smoke and fire, say), and note a relationship between them so that, for example,
identification of the first can lead to anticipation (or production) of the second.
With symbolic reference, the idea is that (to a significant extent conventional) symbols
stand in a distributed network of relationships with one another, where the ‘positive’
reference of any symbol is, at least potentially and partly, cashed out in terms of
indexically determined equivalence classes. Symbolic reference is, because of the
importance of ‘horizontal’ relationships to other symbols, much less hostile to vagaries
of correlation than indexical reference, so the boy who cried ‘wolf!’ undermined the
indexical value of his utterances, while not changing the symbolic reference of ‘wolf’
(Deacon 1997: 82). Symbolic representation also permits the construction of higher order
types not directly grounded in experience (‘unicorn’) but which do nonetheless partly fix
experiential criteria (‘looking like a unicorn’), and others (‘prime number’) which would
be impossible, or nearly so, to fix in indexical terms.
Deacon’ s view is that symbolic referential relationships are constructed out of indexical
ones, which in turn are constructed out of iconic ones, so he envisages a pair of
‘thresholds’ with characteristic cognitive demands and developmental problems in
crossing them. For our part we are less confident that the icon, index, symbol taxonomy
need be related to cognition and development in such a way, partly because we’re
convinced that dispositions to track at least some iconic and indexical relations are
ontogenetically innate (see, Cowley et al., in press). That seems to fit with, for example,
the work of Garcia and Koelling (1966) who studied aversion responses to different
stimuli in rats. They showed that rats very easily learned to associate (a) a noise and light
signal with an electric shock, and (b) a distinctive flavour with (radiation induced)
nausea. In both cases the test populations fairly quickly acquired an avoidance response
to the initial signal. Garcia and Koelling also showed that the reversed combinations
(light and sound followed by nausea, and distinctive taste followed by a shock) were
more difficult for the rats to learn. The innate mechanism suggested here is a bias in
7 / 23