hand, where do they come from? Do all these pathways of informationflow, complete with all their constraints, have
to be innate?
I see a couple of possibilities. My inclinationis to thinkthat all thelevelsof structure—the integrative processors—are
innate, and that some of the interface processors are innate but others are not. The process of reading, for instance,
acts like a module in an accomplished reader, but it requires intensive training for most people in a way that the
phonology–syntax module does not. At this point I don't fully understand the logical and neurological issues involved
in making a claim that an interface module is learnable, so I will have to leave it at that.^116
7.5.4 Multiple inputs and outputs on the same“blackboard”
Let us return briefly to the function of workingmemory as a“blackboard.”Because of the independence of modules,
any single levelof structure can be fed simultaneously by multiple interface processors, all of whose outputs are taken
intoaccount bytheintegrativeprocessor in constructinga maximallycoherentstructure.We have seenthisabundantly
within thelanguage processor already: for instance, in perception, the phonologicalintegrativeprocessor is workingon
material being provided by the acoustic–phonetic interface, by the lexicon, and (through feedback) by the
syntax–phonology interface.
We can now add to this mix the McGurk effect, where an interface processor with visual inputs (specifically mouth
configuration) provides information to the phonological“blackboard”about the composition of phonetic segments.
The phonology integrative processor doesn't know which interface processor has provided the input; it just puts
together the fragments of structure on the“blackboard”as best it can. (And in the case of conflict between auditory
and“McGurk”input, it sometimes chooses the latter!) In other words, the phonology integrative processor is still
domain-specific and informationally encapsulated.
A parallel case in an altogether different domain is the sense of body orientation (Lackner 1981; 1988; Lackner and
DiZio2000). Here inputs converge fro ma widevarietyofsensory syste ms: stretchreceptorsinthe muscles, touchand
pressure sensors in the skin, the semicircular canals and otolithic organs in
IMPLICATIONS FOR PROCESSING 227
(^116) A possibly wild suggestion: Supposethere are cortical areas for whose use more than one integrative module competes, and which shift allegiances depending on attention.
Such an area would be in a position to detect and record fortuitous correlations between two modules that are capable of using it. It might therefore be possible for it to
develop into an interface processor without having any special use for that purpose wired into it fro mthe start. This speculation is encouraged by the fact that in the
congenitally deaf, visual function can invade areas that are normally auditory cortex, and vice versa in the blind (Neville and Lawson 1987; Kujala etal. 2000).