language faculty as a module

language faculty as a module. There are debates over whether language
is a separate faculty or a part of general cognition. Traditional learning
theories are firm in the belief that language is a learned verbal behavior
shaped by the environment. In other words, language is not unique in its
own right. By contrast, nativist theorists insist on language being an independent,
innate faculty. Chomsky even advocates that, being one of the
clearest and most important separate modules in the individual brain, language
should be viewed internally from the individual and therefore be
called internal language or “i-language,” distinct from “e-language” or the
external and social use of language. Nativists also insist on language being
unique to humans, because even higher-order apes, though they have intelligence
(such as tool using, problem solving, insights) and live a social life,
do not possess a true language.


The view of language as an independent faculty has received support
from works in cognitive neuroscience, speech-processing studies, data associated
with aphasia (language impairment due to brain damage), and
unique case studies. Specific word and grammatical categories seem to be
registered in localized regions of the brain. Some empirical studies have
suggested that lexical access and word-meaning activation appear to be autonomic
(modular). As noted, Broca’s aphasia and Wernicke’s aphasia display
different language deficit symptoms. In 1991, Jeni Yamada reported the
case of Laura, a retarded woman with an IQ score of just 41 when she was in
her twenties. Her level of cognitive problem-solving skill was comparable to
that of a preschooler, yet she was able to produce a variety of grammatically
sophisticated sentences, such as “He was saying that I lost my battery powered
watch that I loved; I just loved that watch.” Interestingly, Laura’s normal
development in phonology, vocabulary, and grammar did not protect
her from impairment in pragmatics. In responding to the question, “How
do you earn your money?,” Laura answered, “Well, we were taking a walk, my
mom, and there was this giant, like, my mother threw a stick.” It seems that
some components of language, such as vocabulary and grammar, may function in a somewhat autonomic manner, whereas other parts, such as pragmatics,
require some general cognitive capabilities and social learning experiences.
Cognitive psychologists hold that language is not a separate module but a
facet of general cognition. They caution people against hasty acceptance of
brain localization as evidence for a language faculty. Arshavir Blackwell and
Elizabeth Bates (1995) have suggested an alternative explanation for the
agrammaticality in Broca’s aphasia: Grammatical deficits might be the result
of a global cognitive resource diminution, rather than just the damaged
Broca’s area. In 1994, Michael Maratsos and Laura Matheny criticized the
inadequate explanatory power of the language-as-a-faculty theory pertaining
to the following phenomena: comprehension difficulties in Broca’s
aphasia in addition to grammatical impairment; semantically related word
substitutions in Wernicke’s aphasia; the brain’s plasticity or elasticity (the
flexibility of other parts of the brain adapting to pick up some of the functions
of the damaged parts); and the practical inseparability of phonology,
semantics, syntax, and pragmatics from one another.
Some information-processing models, such as connectionist models,
have provided another way to discuss language, not in the traditional terms
of symbols, rules, or cognitive capacity but in terms of the strengths of the
connections in the neural network. Using computer modeling, J. L. McClelland
explains that knowledge is stored in the weights of the parameter connections,
which connect the hidden layers of units to the input units that
process task-related information and the output units that generate responses
(performance). Just like neurons at work, parallel-distributed processing,
or many simultaneous operations by the computer processor, will
result in self-regulated strength adjustments of the connections. Over extensive
trials, the “learner” will go through an initial error period (the selfadjusting,
learning period), but the incremental, continual change in the
connection weights will give rise to stagelike progressions. Eventually, the
machine gives rulelike performance, even if the initial input was random,
without the rules having ever been programmed into the system. These artificial
neural networks have successfully demonstrated developmental
changes or stages in language acquisition (similar to children’s), such as
learning the past tense of English verbs. As a product of the neural network’s
experience-driven adjustment of its connection weights, language
does not need cognitive prerequisites, or a specific language faculty in the
architecture (the brain). Although emphasizing learning, these models
are not to prove the tabula rasa (blank slate) assumption of traditional behaviorism,
either, because even small variations in the initial artificial brain
structure can make qualitative differences in language acquisition. The interaction
between the neural structure and environment (input cues and
feedback patterns) is further elaborated in dynamic systems models. For example,
Paul van Geert’s dynamic system, proposed in 1991, is an ecosystem
with heuristic principles modeled after the biological system in general and