Adult versus Child L2 Acquisition:

An Information-ProcessingApproach
Barbara L. Kennedy
The University of Alabama at Birmingham

This article assumes that adult L2 learners cannot achieve

the same degree of proficiency in their L2 that child L2 learners
can. An information-processingapproach to this phenomenon is
presented; the aspects of unitization, faulty or incomplete de-
clarative knowledge, and limited working memory space are
used to account for the deficiencies often found in adult learners’
L2 competence. Once L1 procedural-knowledge productions
have become unitized, it is difficult to alter them. If an L2
production responds to the same cue as does an L1 production,
interference from the L1 often occurs. Conscious effort is
required to overcome these problems, as well as correct and
complete declarative knowledge; however, in an actual commu-
nicative situation, conscious effort is focused most often on the
semantics of the communication,rather than on the mechanical
forms, and the strongest unitized (automatic) productions take
over. Moreover, if the learner is able to focus on forms, hdshe
may not have all of the declarative-knowledgeof the L2 he or she
needs or enough working memory space to manipulate all that
is required. Justification for the application of a cognitive
psychology model t o L2 language learning is given. In addition,
the information-processing model is compared to a model that
requires a totally separate language acquisition device to ac-
count for the same phenomenon.

Adults seem not to be able t o learn languages as well

as children. Penfield and Roberts (1959) popularized the

477
478 Language Learning Vol. 38, No. 4

notion of critical period, related to brain plasticity, but it

was Lenneberg (1967)who focused attention on the rela-
tionship between critical period and brain lateralization
and a cut-off point in the language-acquisition process.
Those concerned with second-language acquisition in
particular have been debating the differences between
children’s and adults’ acquisition of a second language
ever since.
Although the controversy that child L2 learners are
able t o outperform older L2 learners is still unresolved,for
the purpose of this paper, the writer has assumed that
there is a period of time in human development beyond
which it becomes difficult for language learners to acquire
native-like proficiency in a language. Although the writer
believes there is strong justification for accepting the ex-
istence of child L2 acquisition superiority, that discussion
lies outside the scope of this paper. Given the assumption
that child L2 learners have an advantage over adult L2
learners, a reasonable explanation for this assumption is
still necessary. Providing a reasonable explanation is
confounded by social, psychological, and physical vari-
ables. To separate these variables and to try to determine
which part each one plays in the language-acquisition
process of L1 and L2 has been the intent of many research-
ers (e.g., Krashen, 1975; Schumann 1975; Felix 1978,
1981, 1985). It may be that these variables each have a
role in the development of language, and it also may be
that the significance of the role each one plays differs from
individual to individual, in which case no explanation
should be overlooked. The purpose of this paper is to
present one alternative that may account for this phe-
nomenon, an information-processing approach.
 Kennedy 479

AN INFORMATION-
PROCESSING APPROACH
Cognitive psychologists (e.g., Anderson, 1983) have de-
velopedinformation-processing modelsto account forlearn-
ing both language and other types of knowledge. Aspects
of these models: unitization, faulty or incomplete declara-
tive knowledge, and lack of working memory space, can ac-
count for some of the problems adult learners have achiev-
ing native-speaker-like competence in another language.
The left and right hemispheres of the brain have tradi-
tionally been viewed as verbal and nonverbal areas, re-
spectively. In Language Functions and Brain Organiza-
tion, Segalowitz (1983) makes it clear that such an exclu-
sive dichotomy is much too simplistic. Moreover, it be-
comes more and more obvious that although designating
specific areas of the brain as responsible for language
functions may refine our description of the many compo-
nent parts, the sum of these parts may not be representa-
tive of the whole we are trying to analyze. Not only may
the tests we use to locate specific hnctions be creating
artifacts (Valsiner, 19831, but if we take as given that
language acquisition is a developmental process that is
caused by an interaction between the organism and his or
her environment, then it should not be surprising that,
with the varied and different experiences of individuals,
we would find varied and different neurological structural
development. Although Geschwind and Levitsky’s (cited
in Segalowitz & Bryden, 1983, p. 342) discovery that the
planum temporal area involves language hnctions may
make a “direct link between brain morphology and neu-
ropsychology”inviting, this link seems,in reality, not to be
as productive as it first appeared.
480 Language Learning Vol. 38,No. 4

Valsiner proposes that we may need a more holistic

approach and should view the brain as composed of proc-
essing systems (1983,p. 236). The brain may not be spe-
cialized so much for language, per se, but for the different
types of processing: serial versus parallel processing or
analytic versus holistic (Vaid, 1983). Therefore, con-
sidering an information-processing approach, Vaid says,
However the differences between the two hemispheres are
characterized, an implication of the information-processing
approach for language processing is that both hemispheres
contribute to it; they differ only in the nature and degree oftheir
contribution. Thus, for example, if serial processing of linguistic
input is required by a given task,one would expect greater leR
hemisphere (LH) participation than when the same input is to be
processed holistically. (p. 318)

GAGNE’S INFORMATION-
PROCESSING MODEL
Even though an information-processing model for
language may seem appealing, many of the models devel-
oped have not taken the time t o account for phenomena
such as adult nonnative L2 acquisition;however, it is pos-
sible to apply Gagnb’s(1983)information-processingmodel
to the development ofproceduralfunctions of language in
relation to the age range when learning a new language
with native-like proficiency seems to become elusive.
In Gagnb’s model, there are two types of knowledge,
declarative and procedural. Declarativeknowledge refers
to the facts, anecdotes, and so forth about the world that
one knows (1983,p. 481, whereas procedural knowledge is
knowledge of “how to do something” (p. 48). Procedural
knowledge is constructed from declarative knowledge
 Kennedy 481

originally. Procedural knowledge is of two types, pattern-

recognition and actiodsequence, and is stored in what are
termed productions. Pattern-recognition is exactly what
it sounds like, the ability to view something and be able to
classifyit. Actiodsequenceis being able to follow the steps
in a procedure until the end goal is reached. For proce-
dural knowledge, a conscious effort is made to construct a
production that will obtain the goal. For the construction
of a pattern-recognition production, generalization and
discrimination are used to collect the knowledge needed to
automatically recognize patterns.
An important difference between declarative and pro-
cedural knowledge is that “procedural knowledge is im-
portant for routine performance while declarative knowl-
edge is important during learning, transfer, and novel
problem solving” (Gagn6, 1983, pp. 49-50). Moreover,
procedural knowledge is more closely tied to the environ-
ment than is declarative knowledge (p. 51).
Automaticity is an important characteristic of pro-
ductions. Declarative knowledge is often very redundant.
However, when procedural knowledge is constructed from
declarative knowledge (i.e. when productions are formed)
the redundancies are deleted. As the learner becomes
more and more skilled at using any particular produc-
tions, the productions become more and more automatic,
and less and less conscious work is needed to reach the
goal. “The slow thinking process of deciding what is to be
done is short-circuitedbecause actions automaticallytake
place when certain conditions are noted” (p. 54).
Once small productions have been created, they can
be unitized into larger productions. Unitization is what
makes the productions automatic. The productions cre-
ated from the compilation of declarative knowledge are
not very large because they are consciously constructed in
482 Language Learning Vol. 38, No. 4

short-term or working memory, which is notably unable to

hold much information at one time. However, unitization
of these smaller productions allows larger productions t o
be made (p. 137). Gagn6 states, “the prolonged speed up
of cognitive procedures with practice provides evidence for
the unitization process” (p. 138). “The reason for the
prolonged effect of practice on the speed of performance is
that some productions are still being unitized even after
many, many practice trials” (p. 139). However, negative
side effects do exist.
It can make behavior more rigid. If individuals habitually solve
a given type of problem in one way, the solution used becomes
unitized. Once unitized, it is more difficult to learn alternate
solutions because the conditions that trigger productions in the
alternate solutions never enter working memory.
(Gagn6, 1983, pp.139-140)
In other words, the conditionsalways trigger the unit-
ized production because procedural knowledge that has
been unitized responds automatically to environmental
conditions. The unitization process, as was pointed out
earlier, takes time and practice; as a matter of fact, Gagn6
(personal communication, Fall 1983) says that becoming
a real expert in an area, a process that demands continual
unitization of many, many procedures in the area of
expertise, takes anywhere from ten to fifteen years.
Neurological evidence exists to support this condens-
ing of productions into smaller packages.
The literature on automatization of skill learning in other
cognitive areas...suggests that an extension of this notion to
languageprocessingwould be appropriate. And, indeed, Whitaker
hypothesizes that when a young child is learning language every
available neuron is recruited, but that as language skills are re-
fined, a more economical repackaging is effected in a core
language area. This hypothesis is based on the fact that early
brain damage tends to result in global, rather than specialized,
 Kennedy 483

aphasia. On the other hand, random brain damage in an adult

is not nearly so likely to affect language,but a focal lesion, even
though small, can cause a great deal of damage to the language
system. This reduction in the size and concentration of the
primary language-processingareas of the brain seems to be due
to the automatizationofgrammaticalskills, which enablesmore
efficient language processing. (Peters, 1983, p. 86)

UNITIZATION AND
ADULT L2 ACQUISITION
GagnB’s information-processing model may be ap-
plied to the development of procedural functions of lan-
guage; moreover, this development may have a strong re-
lationship to the period in life when acquiring native-like
proficiency in a new language eludes the learner. First,
consider the second-language-learningsituation. If the
beginninglearner is beyond the early teen years, a foreign
accent will most likely be retained in the L2, as demon-
strated in phonological studies (Fathman 1975; Oyama,
1976);however, younger learners, especially as young as
four or five years of age, will oRen have native-like pro-
nunciation. This phenomenon may be explainedby saying
that the older learner has already unitized his or her com-
plete L1 phonological system. In the study done by Lewis
(cited in GagnB, 1983), it has been shown that once
procedural knowledge is unitized, it is extremely difficult
for the learner to alter it. If, in fact, the phonological
system is a system that can be unitized, and if the L1
phonological system has been unitized before the acquisi-
tion of L2 begins, then the unitized system of the first
language would resist any kind of alteration or ex-pansion
to include any new sounds from the L2. Perhaps, as is
observable in the real world, learners, instead, force the
484 Language Learning Vol. 38, No. 4

new sounds to conform to the closest sound in their first

language, thereby preserving the unitized system.
At first glance, this idea brings a behavioristicview of
language learning to mind. Behaviorists believe that
language learning is habit formation and practice is the
key to being a successhl language learner. They believe
that children mimic what they hear and those mimicked
sentences, af-ter a lot of practice, eventually become a
habit. Certainly, an information-processing approach
would agree that practice is necessary, but at some point
it is necessary for the learner to acquire some type of
abstract governor over the language. Behavioristic views
of language learning make no attempt to account for
abstraction. Behaviorists insist on remainingin the world
of concrete entities when accounting for how people learn
language, which implies that people can on-ly say sen-
tences they have heard before. However, advocates of an
information-processing approach look into the linguistic
functions ofthe mind, one ofwhichis the ability to abstract
linguistic rules from the language environment, an ability
that al-lows the individual to create sentences he or she
has not heard before.
An information-processingexplanation for the reten-
tion of foreign accents would not necessarily conflict with
many of the other theories that can contribute to an
explanation for this phenomenon (e.g., Hubel & Wiesel,
1970; Geschwind, 1974;Rosansky, 1975;Bruce, 1977; Fe-
lix, 1978,1981,1985;Seliger, 1978;Oyama, 1978;Bishop,
1981; Vaid, 1983; Scheibel, 1984);however, this informa-
tion-processing model may not account for all of the
evidence for all of the nonnative performance the older
language learner displays. Unitization would imply an
approach equivalent to a Contrastive Analysis Hypothe-
sis, an outgrowth of Behaviorism, for syntax, but analyses
 Kennedy 485

of L2learners’errors indicate that the L1 is most often not

a contributing factor in their syntax errors. Therefore,
unitization might account only for some of the phonologi-
cal errors the older learner makes. Because, as is pointed
out by Seliger (1978),syntactic and semantic systems do
not have the finiteness of phonological systems and, as
yet, we have no real way to measure their completeness,
it may be that unitization only occurs in more restricted
systems like phonological systems.
However, it may be that the syntactic system is also
unitized. Because the learner is conscious that his or her
dependable L1 system may not be dependable for his or
her learning of the L2,what we may be seeing on the
surface in the learner’s attempt to use the L2 are syntactic
and semantic constructions from declarative knowledge.
These constructions would necessarily be taking place in
short-term memory, or working memory. Because, as was
pointed out earlier, working memory has limited space
and can only handle small amounts of information at one
time, the learner may, in addition, be using his or her
unitized L1 phonological system as it would not demand
work-ing memory space. Moreover, when constructing
speech from declarative knowledge, he or she may not be
able to account for all the grammar in his or her speech
production due to the confines of working memory and,
perhaps, also due to a lack of declarative knowledge about
the L2. Some errors may be a result of faulty declarative
knowledge due to overgeneralization or excessive dis-
crimination. This would explain why we see both errors
that do not appear to be a result of the L1 and, at other
times, when the learner backslides into a brief dependence
on his or her unitized L1 system, errors that seem to be
related to the L1.
The generalizations and discriminations the learner
486 Language Learning Vol. 38, No. 4

formulates about the L2 cannot be integrated into his or

her procedural knowledge language store if the proce-
dural knowledge is in conflict with the unitized knowledge
or demands any kind of expansion of concepts. Therefore,
the learner would be forced to store this knowledge in an
area that is, perhaps, less specialized for the function. Or
it might be that new areas within the range of that
particular function (which may involve several levels and
locations) may not be available due to neuron death in the
mature organism (Bruce, 1977) or a lack of synapse sites
(Bishop, 1981). This would account for the older L2
learner’s nonnative syntax. Individual differences in
neural structures would most likely play a role here.
The possibility that the new knowledge may have to
be stored in a different location, if the L1 system is
unitized, is supported by the Sussman, Franklin, and
Simon study (1982) of late bilinguals who showed signifi-
cantly less lateralization, that is, more bilateral symrne-
try, in their L2, compared to much more left-hemisphere
lateralization in their L1. Early bilinguals showed more
left-hemisphere lateralization of their L2 than did late
bilinguals, but less lateralization of their L1. Monolin-
guals showed strong left-hemisphere lateralization for
their L1. The implication of their findings for an informa-
tion-processing theory is that the early bilinguals may be
able to integrate the L2 system into their L1 system before
it has unitized. Certainly the laterality profiles for both
their L1 and their L2 are very similar, unlike late bilin-
guals who show very different lateralization profiles for
the two languages.
The information-processing hypothesis has value
where the Contrastive Analysis Hypothesis fails. The in-
formation-processinghypothesis does not claim to be able
to predict what the errors will be because they are o h n a
 Kennedy 487

result of not being able to work with all that is necessary

in the limited space of working memory; in addition, the
errors are, most likely, also a result of learning strategies
being used by the individual, rather than being exclu-
sively a result of the L1 system, incomplete declarative
knowledge, and/or restricted working memory space.
Because the information-processing model does not
discriminate between the way language is learned and the
way other knowledge is learned, the question, how can a
young child learn a system as abstract as language when
he or she appears to have problems learning other ab-
stract kinds of knowledge, remains. Bley-Vroman (1987)
rejects the possibility that a child learning an L l and an
adult learning an L2 could be using the same system. He
argues, “Language is a complicated abstract formal sys-
tem, and young children seem not to have the general
cognitive capacity to deal with such systems” (p. 22).
However, how abstract is the child’s system? As Peters
(1983)notes, working memory is where space is limited;
space is not limited in long-term memory. So, can we not
assume that the child stores nearly everything as a sepa-
rate entity in long-term mem-ory until the age when he or
she can work with abstractions and start condensing what
he or she knows? As previously mentioned, Peters points
out that there is a neurological correlate to the idea of
unitization. Geschwind (1974)gives the example of a ten-
year-old child who appeared able to recover his language
completely after left-hemisphere damage. Even if it is
argued that the tests given to the child were not sensitive
enough to reveal the remaininglanguage problems, adults
with the same damage would not have made an equally
strong recovery. Apparently, areas of the brain that are no
longer accessible to the adult are still accessible to the ten-
year-old child for language development.
488 Language Learning Vol. 38, No. 4

If the child’s repackaging, that is, unitizing, which re-

duces the neurological space needed for language, is still
in progress andnot completed, or, perhaps, not yet started,
the child’s ability to work with language abstractly may be
waiting for the cognitive development that would allow
him or her to build abstract systems. Piaget (Gallagher &
Reid, 1981) gave the young child credit for possessing
generalizative and discriminative abilities, and these may
be all that are necessary for the child to establish his or her
first language systems. Moreover, recently, cognitive
psychologists have argued that a child may be able to work
with abstractions; the idea that children‘do not possess
the ability to abstract, shown by earlier studies, may be an
artifact of those studies (D. L. Schallert, personal commu-
nication, June 16, 1987). If this is so, it would certainly
give rise to a completely new perspective on the way
childrenlearn. One of Chomsky’s(1959)arguments against
B.F. Skinner was that child language is more than habit
formation because children produce sentences they have
never heard before.Certainly, allowing children the abil-
ity to abstract rehtes a behavioristic view of language
learning, but does not impede an information-processing
approach. If young children cannot abstract, it nonethe-
less remains a question whether generalizative and dis-
criminative abilities might allow for creative play with
language. For example, the child who overgeneralizes and
calls a butterfly a bird, has most likely never heard a
butterfly referred to as a bird before, and appears to be
creatively searching for linguistic boundaries.
Bley-Vroman (1987) argues against the possibility of
a child using the same linguistic system for L1 as an adult
uses for L2, saying that “theorieswhich assume an essen-
tial similarity between adult foreign language learning
and child language acquisition have been driven to posit
 Kennedy 489

additional apparatus to account for the obvious differ-

ences” (p. 32). However, with an information-processing
model, no additional apparatusis necessary or useful. Use
of declarative knowledge, construction of procedures, and
unitization are said to take place with all types of knowl-
edge, not only language. Working memory space is lim-
ited, no matter what type of knowledge is being processed.
These are all natural features of our cognitive sys-tem.
The model Bley-Vroman suggests, however, requires chil-
dren to possess a complete system specifically for lan-
guage and nothing else, a system that becomes inoperable
once the L1 (or childhood L2) has been acquired. This
appears to be equivalent to additional apparatus.
Until now, this discussion of the ceiling effect posed by
an information-processing model has focused on produc-
tion, rather than perception. There is little difficulty in
imagining the way perceptual processes might be incorpo-
rated into a model such as this, however. As Diehl and
Kluender (1986) point out, it is easy to see how phoneme
recognition, a perceptual process, might be automatized
under a model such as this. From the perspective of the
present information-processing model, a task such as
phoneme recognition would be subsumed under the pat-
tern recognition type of procedural knowledge. The actual
analysis of the phonemes would not require working mem-
ory space; it would be automatic. Intuitively, it makes
more sense to view working-memory space as occupied by
the semantics of the communication, rather than by the
analysis of the structures conveying the semantics.
Working memory is the bottleneck of the whole processing
operation; there may not be room in working memory for
the analysis of all the segmental and suprasegmental
structures, and the semantics of the message as well.
Consequently, due to the automatic response of unit-
490 Language Learning Vol. 38, No. 4

ized L1 systems to environmental stimuli, due to lack of

declarative knowledge about the L2, and due to limited
space in working memory in which to employ all of the de-
clarative knowledge about the L2 that might be necessary,
late language learners may find errors unavoidable when
communicating in the L2.
McLaughlin, Rossman, and McLeod (1983) account
for differencesbetween L1 and L2 language learners from
an information-processing perspec-tive. They say that al-
though the L2 learners build automatized processes in the
L2, the learners slip back into L1 processes when overload
occurs. These researchers characterize this backsliding
into L1 automatic responses as an initial L2 learn-ing
phenomenon. However, with reference to lexi-cal re-
trieval, they say that although advanced L2 learners seem
to create some automatized procedures, the second lan-
guage does not attain “the degree of automaticity that
characterizes the first, even in the case of subjects who are
overtly balanced bilinguals” (p. 149).
If an L2 learner is aware that his or her first L1 sys-
tem may not be relied upon in the second-language situ-
ation and suppresses the L1 system, the productions from
procedural knowledge in the L2 would be allowed to sur-
face. Unfortunately, because conscious effort, or working
memory, would be required to allow this to happen, other
problems may arise. McLaughlin et al. point out that
when the learner directs attention to speech, controlled Cwork-
ing memory] processes come into play and performance is likely
to be interfered with-just as driving or typing are likely to be
interfered with (or at least slowed down) when these automatic
skills are given too much attention. (p. 145)
However, because McLaughlin et al. make no distinc-
tions between child and adult second-language learners,
whether child or adult L2 learners make the greatest
 Kennedy 491

contribution to the findings they discuss is clouded. Be-

cause many researchers (Rosansky, 1975;O y m a 1978;
Patkowski, 1980)have shown that adult learners exhibit
less proficiency in their ultimate L2 attainment, it would
not be surprising to have the adult L2 learner weighing
the most heavily in the differences found between L2
learners and native speakers in the de-gree of auto-
maticity found in the use of the language. Thus, even
though adult L2 learners may be able to form procedural
productions, which underlie automatic processes, they
would be required to actively suppress their L1 system
and bring the L2 procedural productions into working
memory more often than would the speaker who learned
the L2 as a child. Pulling the “automatic”response into
working memory interferes with performance.
The L2 learner may often have to consciously choose
the response to be used if he or she wants to avoid an L1
procedural production. Many second-language students
are sometimes able to produce correct forms in the class-
room, even when errors while producing those forms have
previously been persistent, but as soon as these same stu-
dents are on the street and actually communicating, the
old inaccurate forms reappear. This problem, from an in-
formation-processingviewpoint, may be that the old form
is coming from a unitized procedure ofthe L1.If this is the
case, not only is the response an automatic L1 response,
but, also, the pathway to the response may be strong. The
older the learner, the stronger the L1 pathways to often-
used procedures. Even when a procedure may have been
constructed to automatize the new knowledge, if both the
old L1 procedure and the new L2 procedure are respond-
ing to the same environmental cues, the old procedure
surfaces unless the choice between responses involves
working memory, that is, conscious response. Procedures
492 Language Learning Vol. 38, No. 4

that are used most often are selected first when a cue has
two possible procedural choices and when working mem-
ory is not controlling the choice of response.
For teaching to be most effective and to facilitate
learning, determining the actual needs of the students is
important. The process of identifying needs must include
not only recognizing students’abilities, but also recogniz-
ing students’limitations. Overcoming some of the limita-
tions may involve more than what can be done in an in-
dividual classroom. Overcoming some limitations may
mean changing the system; that is, the ceiling effect in
adult second-language learning might stimulate ques-
tions about when foreign languages should be introduced
in school systems.
Although implications, such as when foreign lan-
guages should be introduced, may appear sim-plistic, an
information-processing viewpoint could account for sev-
eral different language learning perspectives. McLaugh-
lin et al. (1983) and Kasper (1984) discuss the differences
between bottom-up and top-down processing as strategies
for approaching various types of L2 material, but con-
sideration of these strategies is beyond the scope of this
paper. As McLaughlin et al. point out, much depends on
individual learning strategies; an information-processing
approach to language learn-ing does not imply anything
about method or about the success ofindividual strategies
(1983, p. 153). McLaughlin et al. indicate that aninforma-
tion-processing model is important however in providing
a productive perspective for research.
One of the basic goals for research is to gain an under-
standing of how the human mind works, es-peciallywhen
a second language is being acquired. The information-
processing model allows for empirical research. Even
though the results of such studies may have no direct
 Kennedy 493

application in the classroom, the knowledge gained from

them would most likely influence what takes place in the
classroom. Teaching without understanding the work-
ings of the human mind is certainly possible; however,
when communication breaks down and the information
channel is blocked, it would certainly be comforting, if not
helpful, for the teacher to have a greater understanding of
the problem.

REFERENCES
Anderson, J. R. (1983).Architecture of cognition. Cambridge, Massa-
chusetts: Harvard Press.
Bishop, D. V. M. (1981). Plasticity and specificity of language
localization in the developing brain. Developmental Medicine and
Child Neurology, 23,251-255.
Bley-Vroman, R. (1987).Logical problem. Linguistic Inquiry. Manu-
script submitted for publication.
Bruce, R. L. (1977).Physiological Psychology. New York, New York:
Holt, Rinehart, and Winston.
Chomsky, N. (1959).Review of Skinner's verbal behavior. Language,
35,26-58.
Diehl, R., & Kluender, K. R. (1986).On the categorization of speech
sounds. Unpublished manuscript.
Fathman, A. (1975).The relationship between age and second lan-
guage productive ability. Language Learning, 25,245-254.
Felix, S. (1978).Some differences between first and second language
acquisition. In N. Waterson & C. Snow (Eds.), The development of
communication (pp. 469-479).New York, New York: John Wiley &
Sons..
Felix, S. (1981).Competingcognitive systems in language acquisition.
Paper presented at the First European North American Workshop
on Cross-linguistic Second Language Acquisition Research, Los
Angeles, California.
Felix, S. (1985).More evidence on competing cognitive structures.
Second Language Research, 1,47-72.
494 Language Learning Vol. 38, No. 4

GagnB, E. (1983).The cognitive psychology of school learning. Vol. 1.

Unpublished manuscript.
Gallagher, J. M., & Reid, D. K. (1981).The learning theory of Piaget
and Inhelder. Belmont, California: Wadsworth, Inc.
Geschwind, N. (1974). Late changes in the nervous system: An
overview. In D. G. Stein, J. J. Rosen, & N. Butters (Eds.), Plasticity
and recovery of function in the central nervous system (pp. 467-508).
New York, New York: Academic Press.
Hubel, D. H., & Wiesel, T. N. (1970).The period of susceptibility to the
physiological effects of unilateral eye closure in kittens. Journal of
Physiology, 206,419-436.
Kasper, G. (1984).Pragmatic comprehension in learner-native speaker
discourses. Language Learning, 34,1-20.
Krashen, S. D. (1975,December). A model of adult second language
performance. Paper presented at the Winter meeting of the Linguis-
tic Society of America, San Francisco, California.
Lenneberg, E. H. (1967).Biological foundations of language. New
York, New York: John Wiley and Sons.
Levitsky, ?. ?. (19??).
McLaughlin, B.,Rossman, T., & McLeod, B. (1983).Second language
learning: An information-processing perspective. Language
Learning, 33,135-158.
Oyama, S. (1976).A sensitive period for the acquisition of a nonnative
phonological system. Journal of Psycholinguistic Research, 5,261-
285.
Oyama, S.(1978).The concept ofthe sensitive period on developmental
studies. Working Papers on Bilingualism, 16,1-17.
Patkowski, M. S. (1980).The sensitive period for the acquisition of
syntax in a second language. Language Learning, 30,449-472.
Penfield, W., & Roberts, L. (1959).Speech and brain mechanisms.
New York, New York: Atheneum Press.
Peters, A. (1983).The units of language acquisition. New York, New
York: Cambridge University Press.
Rosansky, E. (1975).The critical period for the acquisition of lan-
guage: Some cognitive development considerations. Working Pa-
pers on Bilingualism, 6,93-100.
Scheibel, A. (1984,January). A histological correlationfor lateraliza-
tion of speech function. Paper presented at The Dual Brain Confer-
ence, Los Angeles, California.
 Kennedy 495

Schumann, J. H. (1975).AfYective factors and the problem of age in

second language acquisition. Language Learning, 25,209-235.
Segalowitz, S.J. (Ed.). (1983).Language functions and brain organi-
zation. New York, New York: Academic Press.
Segalowitz, S. J., & Bryden, M. P. (1983).Individual differences in
hemispheric representation of language. In S.J. Segalowitz (Ed.),
Language functions and brain organization (pp. 341-372).New
York, New York: Academic Press.
Seliger, H. W. (1978).Implications of a multiple critical periods
hypothesis for second language learning. In W. C. Richie (Ed.),
Second language acquisition research: Issues and implications (pp.
11-19).New York, New York: Academic Press.
Sussman, H., Franklin, P., & Simon, T. (1982).Bilingual speech:
Bilateral control? Brain and Language, 15,125-142.
Vaid, J. (1983).Bilingualism and brain lateralization. In S. J. Sega-
lowitz (Ed.), Language functions and brain organization (pp. 315-
339). New York, New York: Academic Press.
Valsiner, J. (1983).Hemispheric specialization and integration in
child development. In S. J. Segalowitz (Ed.), Language functions
and brain organization (pp. 231-244).New York, New York: Aca-
demic Press.