Page 114 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

102 CHAPTER 5 From Synapses to Ephapsis

independently controllable components (the lips, the blade, the body and root of
tongue, the vellum, and the larynx), unique to our species, opened the way for the
selection of gestures that could be overlapped, interleaved, and mergeddthat is,
coarticulated. Assumption is that the coarticulation of gestures is controlled by
species-specific specialization for languageda phonetic module. The listener effort-
lessly perceives gestures the speaker produced, not because the process is simple,
but because the phonetic module is so well adapted to its complex task.
Traditional theory of speech cannot explain how production and perception devel-
oped together in evolution. George H. Mead’s [15] requirement for parity means that
language communication can succeed only if two parties have common understanding
of what countsdwhat counts for the speaker, must count for the listener. From motor
theory point of view, speaker produces gestures that are managed by machinery that
evolved in connection with a specifically communicative (linguistic) function, from
which it follows that gestures are phonetic by their very nature. They are recovered
from the acoustic signal by the phonetic component of the larger specialization for
language. There is no need to appeal to cognitive connections between an initial audi-
tory representation and some more abstract phonetic unit to which it is somehow
linked (bound), because the initial representation is already phonetic.
In a two-way communication, where speaker and listener exchange roles, motor
theory stipulates that parties conduct their business in common currency of phonetic
gestures; the gesture that is the unit of production for the speaker is the unit of percep-
tion for the listener. In contrast, traditional theory needs to explain how representa-
tions of the speaker (motor) and representations of the listener (auditory) which are
not alike, marched through evolutionary development in perfect step. Motor theory
holds that the primary representations of speaker and listener are exactly the same;
accordingly, a change in one is inevitably mirrored by the same changes in the other.
Motor theory, unlike traditional theory, explains clearly the gulf which separates
speech from reading and writing. The preliterate child is a prodigy of phonologic
development, commanding thousands of words; he readily produces their phonolog-
ical structures when speaking, and just as readily, parses them when listening. He
exploits the particulate principle quite naturally when speaking; it is enough to be
a member of the human race and to have been exposed to mother tongue. By
contrast, applying the particular principle to the task of reading and writing is an
achievement of distinctly intellectual kind. Relationship between the alphabet and
speech is entirely arbitrary and reading is always a translation. On the traditional
view, one might expect that reading and writing would be easier than speech. After
all, alphabetic characters are clearer signals than sounds of speech.
According to Liberman, evolution came up with a distinct phonetic module not
in the higher reaches of cognitive machinery, but down among the “nuts and bolts”
of action and perception, that is, motor structures. By creating distinctly phonetic
motor structures to serve as the ultimate constituents of language, the phonetic
specialization enables speech to meet requirements for parity, as well as those for
particulate communication, while also giving it biological advantage over the
reading and writing of its alphabetic transcription.