favour of connecting nausea with ‘something I ate’ and either no bias at all, or a negative
inclination to learn correlations between nausea and flashes and bangs.
According to Deacon (1997: 72), the question whether some mark is iconic, indexical, or
symbolic, is not about the intrinsic properties of the mark itself, but is a question about
the system by which it is actively perceived. So a smile might be a part of some person’s
being happy (iconic) or it might be an indicator of happiness (indexical), or even
deployed, like Judas’ s kiss, as a conventionalised signal (symbolic).8 While agreeing
with Deacon’s general point, we note that the different types of reference each have their
own peculiar constraints which , to some extent, make a difference to what can count as a
mark. The word ‘hound’ cannot be iconic of dogs, because it cannot be relied upon to be
a part of doggy experiences in the same way as hairiness can. Further, wracking sobs are
iconic or indexical of misery in ways that conventional labels like “sad” can’t be (Frank
1998), because we don’t generally think anyone can just decide to burst into tears, even
though we do think that anyone can profess deep sadness.
Note also that on Deacon’s view the distinction between three types of reference implies
a distinction between (at least) three degrees of competence (Deacon 1997: 74). A being
which could make use of iconic reference to deal with its environment may not be able to
manage indexical relations, any more than one that has mastered some indexical relations
need be cable of dealing with symbolic ones. The transitions from iconic to indexical,
and from indexical to symbolic, are learning problems, with their own distinctive
demands. Our primary interest here is in these transitions, and the implied learning
problems.
In line with the ‘tale of two Clarks’ above, we note that Clark himself lacks an answer to
these questions. This is so even though parts of his work are clearly relevant to these
transitions, and highlight aspects of them considered from the perspective of concept
formation, and RR learning, that is learning involving ‘representational redescription’
(Clark and Karmiloff-Smith 1993, Clark 1993: especially Ch. 4). As we hope to show,
though, other parts of his work not specifically concerned with language, but with the
demands of robust real-time embodied responsiveness, help us make more headway with
approaching the how question.
II
How to do Things Without Words
Human infants are extraordinarily dependent. They are only able to support their own
heads at around three months, cannot reach until around four months, crawl until nine, or
walk until thirteen. Unlike other primates, they are unable to cling to their parents in
order to be moved around. Almost anything which takes place in accordance with their
needs, or, later, their goals, has to be done for them. For a being in such a situation there
are clearly advantages to be gained from being socially legible - that is from being
visibly hungry, distressed, uncomfortable, happy, and so forth, when nourishment,
comfort, concerned attention, play, etc., are appropriate. Infants need social relationships
in order to survive, and those who take care of infants, typically kin and paradigmatically
mothers, need social relationships in order to manage their own energy and resource
allocation when caring for the genetic and material investment represented by a child.
The relationships in question are, and have to be, more than simply affiliative. While
close mutual interest is undeniably crucial, caregivers have other demands on their
attention, especially when an infant has siblings, or is dealing with severe scarcity.9 And
even without siblings, there are times when no matter what a child seems to want, it is
more important to make it keep quiet, or wait for some other more urgent goal to be
8 / 23