Page 113 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

3. Embodied Cognition 101

and placing. The majority of neurons are active during the observation of one
action only.
Rizzolatti and Matteli [13] conclude that both space perception and action recog-
nition derive from preceding motor knowledge concerning external world and
actions. Support for this notion comes from the demonstration of the existence of
peripersonal space. Peripersonal space is an extension in space of body surface. It
is anchored to the body and moves when body and body parts move. From a physical
point of view there is no reason whatsoever why eyes should select light stimuli
coming exclusively from space located around the body of the perceiver.
The motor hypothesis implies that F4 neurons code potential motor actions
directed towards particular spatial location. When a visual stimuli is presented in
the peripersonal space, it evokes automatically a “potential motor action” which
maps the spatial stimulus position in motor terms. The results showed that in large
majority of the neurons the receptive field expanded in depth with the increase of
stimulus velocity. These results are not compatible with a geometrical space repre-
sentation. In contrast, if the space is coded in motor terms, since time is inherent to
movement, the spatial map must have dynamic properties and vary, according to the
change in time of the object spatial location. Supporting evidence is in the finding
that movement plays an essential role in establishing ocular dominance. Signals
from retina are not enough; if kittens are paralyzed when exposed to visual patterns
or when visual signals are not usable to control behavior (gondola-driven kittens)
they do not lead to changes in cortical function.
Actions that do not belong to the motor repertoire of the observer (e.g., barking)
are recognized on the basis of their visual properties. In contrast, actions that do
belong to the motor repertoire of the observer activate the parietal-frontal circuits
that are involved in action programming. These actions are understood through
the mirror neuron system. The visual representation of the observed actions evokes
the corresponding action representation whose outcome is known to the acting indi-
vidual (Liberman). Personal knowledge is based not on a visual description of
observed events, but on the resonance of the observer motor system. Thus perception
is not a process categorically distinct from action, but, on the contrary, our funda-
mental knowledge, our personal knowledge of the other, is based on motor system.
These findings strongly support Merleau-Ponty’s concept of intentional arc.
A.M. Liberman in the 2000 paper coauthored with Whalen [14] conceived a
motor theory of speech, which is compatible with Rizzolatti’s and Gibson’s views.
Traditional theory of speech identifies sounds (phonemes) as the primitives that
are exchanged when linguistic communication occurs. Liberman motor theory of
speech claims on the other hand that the phonetic elements of speech, the true prim-
itives that underlie linguistic communications, are not sounds but rather, the articu-
latory gestures that generate those sounds.
The gestures are changes in the cavities of the vocal tractdopening or closing,
widening or narrowing, lengthening or shortening. Liberman states that these ges-
tures are phonetic ab initio, requiring no cognitive translation to make them so. Par-
ticulation of the vocal tractdthat is, the division of the vocal tract action into