auditory–phonetic interface is only one of several specialized interfaces that take auditory signals as input.
Each of these interfaces“sees”different parts of the auditory signal. The auditory–phonetic module pays attention to
timing and distribution of formant transitions and to short-term variations in intensity and pitch—the distinctions it
needs to derive phonetic structure. But it is oblivious to overall amplitude, pitch, and timbre, which serve instead as
cues to voice and affect recognition. Conversely, the modules for identifying voice and affect are oblivious to the
distinctions thatidentifyphoneticsegments, but theydo detectthecues relevanttotheirowndomains. That is, each of
these interface modules makes use of only a subset of the distinctions available in the auditory input. Conversely, an
interface module need not provide all the distinctions possible in its output domain: think again of how the auditory-
phonetic module does not provide word boundaries.
This characteristic is replicated in other interface modules. The phonology–syntax module correlates linear order in its
input and output structures. But, as observed in section5.4, it knows nothing about aspects of phonological structure
such as syllabification, whichare relevanttopronunciation but notto syntax, and itknows few details of embedding in
syntactic structure, which are more relevant to semantics than to phonology. Similarly, the interface that relates
conceptual structure to visuo-spatial understanding knows nothing about such important aspects of conceptual
structure as scope of quantification, conditionals, illocutionary force, or value. It recognizes only those aspects of
conceptual structure that deal with objects, their parts, their location, their motion, and perhaps the forces among
them.Italso does notknoweverythingthereis toknowabout spatial structure: inparticular, itprobablydoes nothave
access to detailed analogue shape information about objects (Landau and Jackendoff 1993; Jackendoff 1996c; see also
section 11.5).
Thus, in general an interfacemodulecan be more domain-specific thantheintegrativemodulesthat constructits input
and output structures. Its“bi-domain specificity”is limited precisely to those aspects of the two formats that can be
directly correlated, while other aspects of each format are treated as irrelevant,“invisible”to the interface. Hence, to
reinforcea pointmade insection 1.6, weought tothinkoftheconnectioncreatedbyan interface notas a“translation”
fro mone for mat to another, but rather as a sort of partial ho mology.
In this light, consider some visual phenomena that interact with language perception. In the McGurk effect (McGurk
and MacDonald 1976; Massaro 1997) experimental subjects are presented simultaneously with visualbaand acoustic
da; they actually experience a heardba, with no awareness of the