cognitive development. Nativism is primarily supported, not by thepov-
ertyof the stimulus, but by thedegree of convergencein the outcome of the
developmental process given varying stimuli.
This general point in favour of nativism is what we shall have in mind in
speaking ofdevelopmental rigidity. Consider the case of language acquisi-
tion. In the industrialised West, parents tend to be assiduous in providing
their oVspring with helpings of ‘Motherese’, but there are many other
communities in which adults take the view that there is little point in
talking to pre-linguistic children – with no apparent ill eVects (Pinker,
1994; especially chs. 1 and 9). In general the linguistic input to which a
child is exposed makes all the diVerence to whichlanguage the child
acquires. It also seems that a certain minimum level of linguistic input is
necessary for the child’s language to develop at all, as evidenced by rare
cases of complete deprivation – such as wolf-boys (Malson, 1972), and the
self-suYcient twins studied by Luria and Yudovich (1956). But the propen-
sity to acquiresomelanguage is so strong that it can survive remarkably
severe levels of degradation in input. For instance, children of pidgin-
speakers spontaneously develop a Creole with genuine grammatical struc-
ture (Bickerton 1981, 1984; Holm, 1988); deaf children born to hearing
parents and not taught any form of Sign manage to develop their own
gestural languages (‘home-sign’ – Goldin-Meadow and Mylander, 1990;
Goldin-Meadowet al., 1994); home-sign-using deaf children brought to-
gether into communities spontaneously elaborate their gestural systems
into fully grammatical sign-languages (Pinker, 1994); and some deaf-blind
subjects can learn a language through the input they receive by placing
theirWngers on the throat and lower lip of a speaker (the Tadoma method:
C. Chomsky, 1986).
There is much more evidence in favour of the Chomskian position on
language acquisition (see Cook, 1988; Chomsky, 1988; Carruthers, 1992,
ch.6). For example, the characteristic mistakes children make are not at all
what would be expected if they were deploying a domain-general learning
strategy without any pre-speciWed constraints on possible grammatical
structures. Rather, the patterning in their mistakes shows that a rule-
hungry, powerfully constrained, device for acquiring language has been at
work. Moreover, a recent large twin-study by Plomin and colleagues found
that the factors underlying severe language-delay are largely genetic, with
three-quarters of the variance in delay amongst twins being attributable to
genes, and only one-quarter to the environment (Daleet al., 1998).
It is also instructive to contrast learning to speak (and to understand
speech) with learning to read. Considering only what has to be learnt,
learning to read should be assessed as a relatively trivial task for those with
normal sight and hearing – for it is, essentially, just amapping problem,
54 Modularity and nativism