Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

SECOND LANGUAGE LEARNING: AN

INFORMATION-PROCESSING PERSPECTIVE'

Barry McLaughlin
University of California, Santa Cruz

Tammi Rossman
Concordia University

Beverly McLeod
University of California, Santa Cruz

I t is a byword in contemporary cognitive psychology that humans are limited-


capacity processors. This paper discusses ways in which children and adult learners
manage to make the most of their limited processes in dealing with the complex input of
a second language. An information-processing approach to second language learning is
proposed and evidence in support of this approach is presented. We also discuss the
implications of a n information-processing perspective for second language pedagogy
and research.

This paper is a n attempt to explore the relationship between cognitive


processes and the development of a second language. Without denying the
validity of other points of view, we shall look a t second language learning
from the perspective of human information processing. This perspective
derives from contemporary cognitive psychology, with its concern for the
processes of learning, perception, memory, problem solving, and decision
making. It is our contention that such a n approach raises new questions
that are empirically testable, and sheds light on findings from a number of
areas of second language research.

HUMAN INFORMATION PROCESSING

Any communication task involves the assessment and coordination of


information from a multitude of perceptual, cognitive, and social domains.
The speaker must learn to obey a large number of conversational

' The authors wish to thank Mike Sharwood Smith. Ann Peters. Sacha Felix, and two
anonymous reviewers for their helpful comments and criticisms of earlier drafts of this
manuscript. The comments of Thomas Carr were especially helpful in writing the final version
of this paper.

I35
136 Language Learning Vol. 33. No. 2

conventions and must communicate the intended message unambiguously.


Since humans are limited-capacity processors, such a task requires the
integration of a number of different skills, each of which has been practiced
and routinized. Each of these notions-that humans are limited-capacity
processors, that complex skills involve the integration of elements, and that
skills become routinized-requires elaboration.

Processing limitations

The limitations that exist on how individuals process information can be


described along two dimensions. The first relates to the focus of attention,
and is largely a function of task demands. The second relates t o
information-processing ability, and is largely a function of how the
individual deals with the information on the basis of past experience.
That human beings are selective in focusingattention was pointed out by
William James (1890:403): “It is the taking possession by the mind, in clear
and vivid form, of the one out of the seemingly several simultaneously
possible objects or trains of thought.. . . ”This selectivity is required by the
limited capacity of the human mind to process information. T o attend t o
one thing, James continued, “implies the withdrawal from some things in
order to deal effectively with others”( 1890:404). Very few people can listen
attentively to two simultaneous conversations.
Since James, psychologists have wrestled with the phenomenon of
selective attention. Wundt ( 1896) divided mental processes into two
levels-a central focus level and a peripheral diffuse level. Recent cognitive
psychologists have made similar distinctions, speaking of focal attention
versus preattentive processes (Neisser 1967), the selective system versus the
perceptual system (Broadbent 1958), and active thought versus passive
thought (Rapaport 1957). At present, cognitive psychologists tend to use a
multilevel system in which, for example, attention may be divided into a
central scanning process and a brief buffer storage, with connections to
various memory systems such a s the sensory register, short-term memory,
and long-term memory (see Blumenthal 1977). Whether this multilevel
system is conceptualized in terms of specific components (e.g., Baddeley
1978) or in terms of types or levels of processing (e.g., Craik and Lockhart
1972), allowance is made for the fact that not everything reaching the
organism through various input channels becomes a n object of attention.
Some scanning system selects one line of input over others and gives it
direct access t o the central cognitive system; other input sources remain on
the periphery of attention. The central cognitive system can be thought of
McLaughlin. Rossman. and McLeod 137

as an executive control system (Baddeley 1981; Carr 1979) that has as its
functions goal setting and task organization (Shallice 1972).
This brings us to a second dimension that relates to how individuals
process information: information-processing ability. The success of the
system in dealing with a given input depends on the characteristics of the
input and the information-processing ability (including knowledge and
expectancies) of the perceiver. If, for example, one is exposed to a rapid
flow of speech in a language one does not know, the effect is that
information-handling capacity becomes overloaded and one eventually
“switches off.” Similarly, young children d o not attend long to speech well
beyond their capacities, which is one reason why parents and othersadjust
their speech in various ways when talking to young children.
In short, humans are limited-capacity information processors, both in
terms of what they can attend to a t a given point in time and in terms of
what they can handle on the basis of knowledge and expectations.
Attention focus-what the individual can attend to at a given point in
time-can be focal or peripheral. Information-processing ability-how the
individual deals with incoming information-is a function of past
experience and the characteristics of the input.

Information-handling procedures: integration of elements

How is it then that individuals process large amounts of information and


develop complex social, linguistic, and cognitive skills? Miller ( 1956), after
all, argued that human short-term memory is limited t o seven plus or minus
two items. Yet Miller went on to point out that items in short-term memory
are not equivalent to “bits” of information as defined by information
theory. That is, items in memory can contain more than minimal amounts
of information depending on the individual’s learning history and on
characteristics of the input. This is well illustrated by De Groot’s research
(1965) with expert chess players. These experts could almost perfectly
reconstruct a chess board illustrating the chess positions of two players,
even after only five seconds of exposure. This did not mean that they had
better short-term memory than chess novices who could not reconstruct
the board positions; they simply had more efficient ways of encoding the
information into meaningful chunks. De Groot demonstrated this by
showing the chess experts boards with randomized chess positions. In this
case their performance was no better than chess novices.
How people chunk information has been the object of extensive research
in contemporary cognitive psychology. What is a t issue here is how
138 Language Learning Vol. 33, N o . 2

information is organized so that it can be utilized in short-term (working)


memory and transferred into long-term storage. According t o one
approach (Craik and Lockhart 1972), there are various levels at which
information can be processed, and the likelihood of long-term retention
depends on the depth of processing involved. Learners use different
strategies t o handle input, some of which involve "deeper" levels of
processing and result in superior long-term retention. While this approach
is vulnerable t o criticism on empirical grounds (Baddeley 1978), the levels-
of-processing framework is helpful in that it suggests that poor learners
may not suffer from inherent storage or capacity deficits, but rather need to
apply different learning strategies.
The organization of long-term memory systems is another factor that
affects the manner in which information will be handled. Cognitive
psychologists differ as to whether they regard the long-term memory
system as being organized in terms of associative networks or hierarchical
systems (see A n d e r s o n a n d Bower 1974). In either case, the
predetermination of relations between elements reduces the amount of
cognitive effort required t o handle isolated bits of information. By dealing
with related units of information rather than isolated bits, more efficient
processing becomes possible.
In recent years researchers have investigated the effects of practice,
rehearsal, and familiarity on information processing. A particularly
important variable appears to be the degree of attention involved. The
more attention required, the more resources are consumed and the slower
the processing. Greater practice, rehearsal, or familiarity with the material
allows information to be handled more routinely without as much cost in
attentional monitoring as when there is less practice, rehearsal, or
familiarity with the material.

The routinization of skills: controlled and automatic


processes
If we assume that some tasks consume more resources and involve
slower processing than others, on what basis is processing energy
allocated? Kahneman ( 1973) has suggested that the amount of processing
energy consumed depends on the nature of the task; certain mental
operations require greater processing capacity than others.
Several researchers (LaBerge and Samuels 1974; Schneider and Shiffrin
1977; Shiffrin and Schneider 1977) have conceived of the differences in the
McLaughlin, Rossman, and McLeod 139

processing capacity necessary f o r various mental operations in a


dichotomous way: Either a task requires a relatively large amount of
processing capacity, or it proceeds automatically and demands little
processing energy. Furthermore, a task that once taxed processing capacity
may become so automatic that it demands relatively little processing
energy.
In their discussion of human information processing, Shiffrin and
Schneider conceived of memory a s a large collection of nodes that become
“complexly interassociated” through learning. Each node is a grouping or
set of informational elements. Most of the nodes are inactive and passive
and, when in this state, the interconnected system of nodes is called long-
term store (LTS). When, because of some kind of external stimulus, a small
number of these nodes a r e activated, the activated nodes constitute short-
term store (STS).
There are two ways in which these nodes become activated: Shiffrin and
Schneider called these the automatic and the controlled modes of
information processing. Automatic processing involves the activation of
certain nodes in memory every time the appropriate inputs are present.
This activation is a learned response that has been built up through the
consistent mapping of the same input to the same pattern of activation over
many trials. Since a n automatic process utilizes a relatively permanent set
of associative connections in LTS, most automatic processes require a n
appreciable amount of training to develop fully. Once learned, an
automatic process occurs rapidly and is difficult to suppress or alter.
The second mode of information processing, controlled processing, is
not a learned response, but a temporary activation of nodes in a sequence.
This activation is under attentional control of the subject and, since
attention is required, only one such sequence can normally be controlled at
a time without interference. Controlled processes necessarily intrude on the
ability to perform simultaneously any other task that also requires a
capacity investment. Controlled processes are thus tightly capacity-
limited, and require more time for their activation. But controlled
processes have the advantage of being relatively easy to set up, alter, and
apply to novel situations.
Two aspects of this conceptualization of information processing are
especially important for the present discussion. First, learning involves the
transfer of information t o long-term memory a n d is regulated by
controlled processes. That is, complex skills are learned and routinized
(i.e., become automatic) only after the earlier use of controlled processes. It
140 Language Learning Val. 33, No. 2

is controlled processes that regulate the flow of information from working


to long-term memory. Learning involves time, but once automatic
processes are set up at one stage in the development of a complex
information-processing skill, controlled processes are free to be allocated
to higher levels of processing. Thus controlled processing can be said to lay
down the “stepping stones” for automatic processing as the learner moves
to more and more difficult levels (Shiffrin and Schneider 1977).
Second, the distinction between controlled and automatic processing is
not based on conscious experience. Both controlled and automatic
processes can in principle be either conscious or not. Since most automatic
processes occur with great speed, their constituent elements are usually, but
not necessarily, hidden from conscious perception. Some controlled
processes also occur with great speed, so that they may not be available to
conscious experience. Shiffrin and Schneider ( 1977) called these controlled
processes “veiled.” Other controlled processes, those they referred to as
“accessible,” are easily perceived by the learner.
This complexity is illustrated in Table I . Controlled processes may occur
with or without awareness, depending on the learner’s focus of attention.
While some attention is needed for all controlled processes, certain new
skills-e.g., learning to drive a car-may require that the learner focus
attention on the various component skills involved (Cell A). Other skills-
e.g., learning to speak grammatically in a first language-may involve only
peripheral attention (Cell C). Thus children learn the rules of grammar
“incidentally” as a byproduct of trying to communicate with others.
Automatic processes can also become the focus of attention, when one
deliberately attends to a skill one normally performs routinely-e.g.,
driving or typing (Cell B). Most automatic processes, however, are
performed without deliberate awareness-e.g., adults speaking their first
language-so that attention is freed for other tasks (Cell D). In Table 1 , the
“intentional”-“incidental” distinction refers to the extent to which task
demands on a particular occasion require focal or peripheral attention
(McLaughlin 1965), recognizing, of course, that there are shades of gray at
the point where focal and peripheral attention converge.
In short, the controlled process-automatic process distinction is not
based on conscious versus subconscious awareness, but instead relates to
the degree to which the skills in question have been routinized and
established in long-term memory. T o the extent that this has not occurred,
information-processing techniques can be said to be imperfectly mastered,
temporary, and subject to controlled processing; to the extent that the skills
McLaughlin. Rossman, and McLeod 141

Table I
Per/ormanc e UJ a /uric tion of information processing and fooc UJ of attention

Focus of Information processing


attention Controlled Auiomatic
(Cell A) (Cell B)
Focal “Intentional” performance “Intentional” performance
of a new skill of a well-trained skill

(Cell C) (Cell D)
Peripheral “Incidental” performance “Incidental” performance
of a new skill of a well-trained skill

are well-mastered and permanent, information processing can be said to be


automatic. Optimal performance requires a flexible blend of automatic
and controlled processing (Bock 1982; Shallice 1972). Routine activities
must be automatized so that resources are available for higher-order
processing, but a t the same time, failure to employ controlled processes can
lead to rigidity and sterile performance.

INFORMATION PROCESSING A N D SECOND


LANGUAGE LEARNING

It is currently popular in discussions of second language learning to


distinguish between two processes: learning and acquisition. This
distinction has been made by a number of people, but is principally
associated with Krashen’s M o n i t o r Model of second language
performance (Krashen 1976, 1977). As Krashen and others use it, the
learning/ acquisition distinction is based on whether the processes involved
are “conscious” (as in learning) or “subconscious” (as in acquisition). As
Krashen noted, “special experimental conditions are necessary”( 1979: 152)
to validate such a distinction. Yet no such procedures exist. Krashen’s
attempt to have subjects distinguish whether they are operating by “rule”or
“feel” (Krashen, Butler, Birnbaum, and Robertson 1978) is woefully
inadequate (McLaughlin 1978). and arguments as to whether particular
constructions a r e acquired o r learned ( K r a s h e n 1979: 152-153;
McLaughlin 1978:317-3 18) merely underscore the difficulty of operating
on the basis of “conscious” experience. Introspection and subjective
reports are notoriously unreliable and easily mistaken. The windows to our
minds are unfortunately rather clouded (Shatz 1977).
142 Language Learning Vol. 33, No. 2

Controlled and automatic processes in second language


learning

Because the learningiacquisition distinction is so difficult to pin down,


we believe that the pie should be cut in a different way. Instead of making a
distinction between learning and acquisition that rests on whether the
processes are conscious or subconscious, we propose that the distinction
between controlled and automatic processes be used to differentiate
processes that are capacity-limited and temporary and those that are
relatively permanent and nearly always become active in response to a
particular input configuration. As we have just seen, controlled and
automatic processes can be either the focus of attention (which is usually,
but not always, true of controlled processes) or on the periphery of
attention (which is usually, but not always, true of automatic processes).
Since, however, controlled and automatic processes may or may not be
subject to conscious awareness, the distinction cannot be equated with
Krashen’s learning/ acquisition distinction (McLaughlin 1978).
In a n information-processing perspective, learning in the initial stages is
seen to involve the use of controlled processes with focal attention to task
demands. As the learner becomes more familiar with the situation,
attention demands are eased and automatic processes develop, allowing
other controlled operations to be carried out in parallel with automatic
processes as performance improves. In second language learning, for
example, the initial stages of learning involve the slow development of
skills as the learner attempts to automatize the various components of
performance.
Does this mean that second language learning inevitably involves the
initial assimilation of rules that are then applied with focalattention t o task
demands? This is one strategy, but there are others depending on how the
individual approaches the task of learning a second language and what the
demand characteristics of the situation are. Table 2 outlines some
possibilities. A learner may, for example, initially focus attention on
formal rules (Cell A), but this attention t o formal properties can also be
peripheral (Cell C ) . In the second case, learning may take the form of
“implicit learning,” which involves the relatively passive apprehension of
linguistic structure and a focus on other (e.g., communicative) task
demands, or “analogic learning,” whereby learners generalize around
instances and build up their knowledge of formal linguistic structure from
instances rather than from formal rules (see McLaughlin 1980a). Once
McLaughlin, Rossman. and McLeod 143

learning is routinized, information processing is automatic and may


involve attention t o formal properties-as in a test situation (Cell B)-or
may not involve such attention-as in normal communication in natural
settings (Cell D).

Table 2
Possible second language performan1.e as a,func,rion ofin/ormarion-processinl: procedures
and arrenrion rojormal properries v / language

Attention to
formal properties Information processing
of language Controlled Automatic
(Cell A ) (Cell B)
Focal Performance based on Performance in a
formal rule learning test situation

(Cell C) (Cell D)
Performance based Performance in
Peripheral on implicit learning communication
or analogic learning situations

It is important to recall that the attentional dimension is viewed as a


continuum, with sufficient allowance for the fact that in many instances it is
difficult to know exactly how attentive the learner is t o formal linguistic
properties. There are also many gradations along the controlled-automatic
axis, with native language processing at the automaticend and a beginning
second language learner at the controlled end of the continuum.
Another example might be helpful for understanding Table 2. One might
deliberately eavesdrop on a conversation in a language that one knows
imperfectly (Cell A). Understanding the content of the conversation in such
a case would differ from unintentionally overhearing a conversation in that
language, where the person might not be fully aware of overhearing and
understanding something until afterwards (Cell C ) . Similarly, one might
eavesdrop on a conversation in one’s native language (Cell B), and this
would be different from unintentionally overhearing a conversation in that
language (Cell D).

Workload requirements in second language learning

At this point it is useful to summarize the basic assumptions of the


approach we are taking t o second language learning. First, it is assumed
that humans are limitedcapacity information processors. As we saw
144 Language Learning Vol. 33, No. 2

earlier, this limitation reflects restrictions on what can be attended to at a


given point in time, and on how information is processed in terms of a n
individual’s knowledge and expectations. Second, it is assumed that in
order t o function effectively, humans develop ways of organizing
information. Some tasks require more attention; others that have been
well-practiced require less. The development of any complex skill involves
building up a set of well-learned, automatic procedures so that controlled
processes will be freed up for new tasks. In this way limited resources can be
spread thinly t o cover a wide range of task demands.
In this conceptualization, complex tasks are characterized by a
hierarchical structure. That is, such tasks consist of subtasks and their
components. The execution of one part of the task requires the completion
of various smaller components. The various subroutines may, however, be
learned independently. People usually try to learn the various subtasks
more o r less concurrently. That is, the attention given to various
subroutines is typically on a time-shared basis, with the learner attending
now t o one feature and then to another. But eventually the subroutines
need to be integrated to attain the higher-order goal.
Levelt (1977) maintained that speaking is a n excellent example of a
hierarchical task structure. The first-order goal is to express a particular
intention. T o d o this, the speaker must decide on a topic and select a certain
syntactic schema. In turn, the realization of this schema requires
subactivities, such as formulating a series of phrases to express different
aspects of the intention. But to utter the phrases there is need for lexical
choice, activation of articulatory patterns, utilization of appropriate
syntactic rules, etc. Whether the process is strictly hierarchical is open to
debate, since there may be a n interaction between the various subactivities,
with, for example, lexical decisions influencing syntactic choice (Bock
1982). Nonetheless, each of the component skills needs to be executed
before the higher-order goal can be realized.
To attain the higher-order goal the learner needs a plan, whereby the
selection of subactivities is regulated according to overriding goals. The
plan controls the order in which a sequence of operations is to be
performed (Lichtenstein and Brewer 1980). Plans include goal states,
procedures for reaching the goal state from any initial state, and tests for
evaluating if the goal state has been or will be reached (Miller, Galanter,
and Pribram 1960). An automated plan can consist of lower-level
components that may or may not be automated and whose selection is
governed by the master plan. Levelt (1977) pointed out that the choice o f a
particular lexical entry during the execution of a n automatic phrase-
McLaughlin, Rossman, and McLeod 145

building plan may depend on the choice of particular words in an earlier


phrase, on the topic of discussion, and so on. Consequently, even for native
speakers, lexical decision may require substantial attention (see Bock
1982).
In the case of a beginning second language learner a considerable
amount of cognitive effort may be needed simply to realize a correct, or at
least an adequate, phonetic expression of individual words. At the same
time, the learner needs to employ appropriate syntactic rules and must
draw on a limited lexical system. Thus each component requires more or
less work depending on how well-learned it is. The more well-learned a
component skill is, the less effort (and processing time) required for its
execution. The execution of new skills is costly in terms of workload
involved and will occur only when other tasks and cognitive demands are
minimized.
Tarone (1979) has argued that data gathered in informal settings reveal
the most systematic interlanguage style and the style least subject to
interference. The informal, or “vernacular,” speech style is a product of
language skills that are relatively more automatic (Tarone 1982). When the
learner directs attention to speech, controlled 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 theseautomatic
skills are given too much attention. When skills are not completely
automatized, however, performance can be improved by giving the learner
more time to apply controlled processes. Thus learners in test situations
sometimes d o better when given more time because the test items are not
automatized and require the application of controlled processes.
Another factor affecting workload requirements is the complexity of the
task. Initially, of course, the second language learner is confronted with a
“blooming, buzzing confusion.” The tendency in this situation is to fall
back on old skills, to use one’s first language to crack the code of the
second. For example, the learner at this stage may execute those routinized
operations that are normally designed to produce utterances in the first
language, although attempting to employ the sound system and lexicon of
the second language. At some intermediate stage, target language routines
may become more automatic, thereby lessening cognitive demands. As
workload requirements lessen, there should be less reliance on old skills,
which is in accord with the finding that interference errors predominate
early in learning (Taylor 1975) or when learners confront intractable
problems that impose significant new cognitive demands (Wode 1978).
Note that in thisaccount of the second language learning process, what is
I46 Language Learning Vol. 33, No. 2

learned does not consistently display itself in performance. Situational


demands a n d the cognitive demands of the task itself affect the likelihood
that new skills will be revealed. Interference errors and deviations from
systematic developmental patterns are seen as evidence that new skills are
inhibited because of excessive workload demands. Only when the cognitive
workload is reduced does one get a veridical picture of what has been
learned.
Most adult second language learners are familiar with the experience
described by Brown (1973) of knowing words, phrases, a n d sentences in a
second language perfectly for classroom use, but being unable to utilize
them outside of the classroom when in contact with native speakers of the
target language. O n e explanation of this phenomenon is t h a t the
individual’s full range of linguistic abilities is not displayed because the
workload involved in maintaining conversational interaction has
overloaded information handling capacities. Thus individuals such as
Krashen’s “P” (see Krashen 1979), who display accurate knowledge of
formal rules but d o not use them in conversational speech, cannot handle a t
one time the two competing demands of maintaining the flow of
conversation a n d speaking accurately in a formal sense.
In addition, person variables interact with task and situation demands,
as can be illustrated through one of psychology’s classic principles, the
Yerkes-Dodson law (1908). According t o this law, performance on any
task is a n inverted U function of motivation level. As motivation increases,
performance increases-up to a point. Beyond this point, performance
decrements are incurred a s increasing levels of motivation, accompanied
by increasing arousal and anxiety, disrupt performance. For example,
some motivation is beneficial t o performance on a n examination, but t o o
much disrupts performance. Figure 1 illustrates how individuals differ in
their tolerance for higher motivation levels. Some people (e.g., B) perform
best under motivation conditions (e.g., point X) that seriously impair the
performance of other individuals (e.g., A).
This picture is complicated by the fact that different tasks have different
optimal levels of motivation. This can be seen in Figure 1 if curve A is
thought t o represent a task with high information-processing demands and
curve B a task with low information-processing demands. Lower levels of
motivation disrupt performance on the complex task: At point X ,
performance on task A is disrupted, whereas this level of motivation
produces optimal performance on the less complicated task B.
Although the Yerkes-Dodson law was expressed in terms of levels of
motivation, it could be expressed equally well in terms of information-
McLaughiin, Rossman. and McLeod 147

High

\
\
\
B

Low

1.ow Point High


X
Motivational level

Figure 1. The Yerkes-Dodson law (1908). There is an optimal degree of


motivation for performance on a task, after which point further increases in
motivation disrupt performance. Individuals (andtasks)diffPrin the levelof
motivation neededfor optimal performance.

processing workload requirements (Kahneman 1973; Levelt 1977).


Individuals differ in the amount of information they can process at any one
time; what constitutes a n overload (i.e., degrades performance) for
individual A may provide the ideal cognitive demands for individual B.
Similarly, different tasks and situations involve different information-
processing workload requirements; an equal cognitive effort can produce
very different results in two distinct situations. Note, however, that these
variables never operate in isolation in the real world. Individual differences
interact with situational factors in a complex manner (McLaughlin 1980b).

PROCESSING A SECOND LANGUAGE: SOME


RESEARCH FINDINGS

For some time now, experimental psychologists have approached


language as linguistic information that must be processed in order to be
understood. There has been relatively little research on either first or
second language learning from this conceptual point of view. There are,
however, some relevant studies. In this section we refer to some research
results that bear upon the approach to second language learning we
advocate in this paper.
i 48 Language Learning Vol. 33, N o . 2

Automaticity in lexical retrieval

Much recent psycholinguistic research has focused on the relevance of


the automatic-controlled dichotomy to the manner in which the strength of
association between a word and its meaning is established (e.g., Posner and
Snyder 1975). One experimental paradigm is the lexical decision task, in
which subjects are required to make a decision about a target stimulus (e.g..
o a k ) for which they have been given a prime signal that is either
semantically related to the target stimulus (e.g., e h ) or not (e.g., truck).
Favreau (198 1 ) used this paradigm to investigate the extent to which there
is a semantic facilitation effect (i.e., where a prime that is semantically
related to a target word will increase the speed with which the target word is
processed) for both balanced bilinguals and dominant bilinguals highly
skilled in their second language. Subjects were shown a prime that was
either semantically related or unrelated t o the target. The target sometimes
was a word and sometimes a nonsense syllable. The time between the
showing of the prime and the target was manipulated and the subjects'task
was t o decide as rapidly as possible whether the target formed a word.
Favreau found that both balanced and dominant bilinguals showed
evidence of semantic facilitation in their first language and in the long time
interval condition in their second language, but only balanced bilinguals
showed the same semantic facilitation in the short time interval condition
in their second language. This was interpreted as evidence that the
dominant bilinguals, who were judged to be fluent in their second
language, nonetheless d o not process lexical items as efficiently as balanced
bilinguals. This lesser efficiency in automatized lexical processing was
thought to be the reason why the dominant bilinguals were found to read
more slowly in their second than in their first language, while this was not
the case for the balanced bilinguals.
More proficient second language learners also differ from less proficient
learners in that they use different techniques for encoding lexical items in
memory. Henning (1974) found that adult second language learners made
more or less use of acoustic encoding in a vocabulary recognition task as a
function of their degree of proficiency in the second language. More
advanced learners and native speakers tended t o make errors that indicated
semantic clustering for lexical items, whereas less advanced learners
showed evidence of a predominance of acoustic clustering rather than
semantic clustering. In our terms, the less advanced learners have not yet
automatized formal-in this case acoustic and orthographic-aspects of
McLaughlin, Rossman, and McLeod 149

the language, and so have less cognitive energy available for semantic
avpects. More advanced learners have automatized the acoustic and
orthographic aspects of the task, having already achieved automaticity
with respect to these subcomponents.
Dornic (1979) reviewed a series of studies dealing with automaticity of
processing in bilingual subjects. Using speed of processing as a measure of
automaticity, Dornic found that speed increases as a function of experience
with language. This was true with respect to both decoding and encoding
efficiency. Typically, the semantic content of words tended to be decoded
more slowly in the second language, and the subjects’ ability to encode
information in the second language (as measured by naming latencies)
tended t o be inferior to performance in the first language. Presumably the
second language had not attained the degree of automaticity that
characterizes the first, even in the case of subjects who are overtly balanced
bilinguals.
In a cross-sectional study of 13- t o 18-year-old German-Swedish
bilinguals, Magiste (1979) found that it took a somewhat shorter time to
develop proficiency in the decoding than in the encoding process of a
second language. In her research, reaction time measures to verbal stimuli
were used t o assess both decoding and encoding ability. Monolingual
subjects had faster reaction times, especially in encoding tasks, than did
bilingual subjects in their second language, leading Magiste to hypothesize
that the poorer performance of bilinguals is due to their failure to achieve
the same levels of automatic familiarity with the words of their second
language as monolingual subjects achieve.
Dornic (1979) reported that high information load enhances the
dominance of the bilingual’s stronger language system. Noise and other
stressors interfere significantly with the weaker language, speech in that
language becoming slowed down, rendered less precise, or even entirely
blocked. In the terminology of our earlier discussion, the workload
requirements degrade performance in the second language when cognitive
and situational demands are increased beyond some optimal point.

Automaticity in syntactic processing

Hatch, Polin, and Part (1970) asked university-level native and


nonnative speakers to cross out all of the instances of certain letters (e.g., all
e’s) occurring in a text. They reported that nonnative speakers found the
instances of occurrence of the letters with equal frequency in both content
I so Language Learning Vol. 33, No. 2

and function words, but that native speakers ignored the letters more often
in the function words than in the content words. This phenomenon can be
explained in terms of the amount of cognitive work required by the
nonnative speaker to process the language. Nonnative speakers expended
equal cognitive effort in processing both syntactic (as represented by the
function words) and semantic (as represented by the content words) aspects
of language. Native speakers, on the other hand, more frequently missed
the target letters found in function words, suggesting that they generally
did not focus their attention on such words. Instead, native speakers focus
their attention on meaning (on content words), processing syntactic
elements automatically when scanning a text.
Psycholinguistic evidence for differential processing of semantics and
syntax by native speakers comes from Sachs’ classic finding (1967) that
recognition memory for semantic features of a n utterance was superior to
recognition memory for syntax. Rossman (198 1) hypothesized that this
would not be the case for nonnative speakers, who have not yet achieved
the degree of automaticity in processing syntax that characterizes native
speakers. Rossman compared the performance of native speakers with two
groups of nonnative speakers of English on a reading recognition test.
Native speakers showed better recognition for semantic than for syntactic
changes. Nonnative speakers, however, showed a greater ability to
recognize whether the form of the sentence was changed than t o recognize
that its meaning was altered.
In a similar study, Wolfe (198 I ) tested 55 English-speaking children
learning French as a second language in California schools. The children
read a paragraph and were then shown a target sentence that they were to
identify as “the same” or “different” from a sentence in the paragraph. The
paragraphs were either entirely in English or entirely in French or mixed
with sentences in both languages. The target sentences were either in the
same or in a different language and had either the same or a different
meaning. Children who were more proficient in the second language
recognized more changes in meaning correctly, but made more errors in
recognizing changes in the language of the sentence than did less proficient
children. This suggests that the more proficient children had achieved a
degree of automaticity with respect t o processing form (in this case the
specific language of the sentence) that the less proficient children had not
yet attained.
Mc,Laughiin, Rossman, and Mc.Leod 151

Automaticity in reading

One aspect of second language performance where the automatic/con-


trolled processing distinction is especially relevant is reading. Shiffrin and
Schneider ( 1977) argued that in learning to read, children utilize controlled
processing to lay down “stepping stones” of automatic processing as they
move on to more and more difficult levels of learning. The transition from
controlled to automatic processing at each stage results in reduced
discrimination time, more attention to higher-order features, and ignoring
of irrelevant information.
If the reading process can be viewed as a sequence of transitions from
controlled to automatic processing, one would expect second language
learners to differ from native speakers in processing text. Cziko (1980)
compared French oral reading errors of two groups of seventh grade
English-speaking students with those committed by native French-
speaking students. Students of intermediate proficiency in French made a
significantly higher proportion of substitution errors that graphically
resembled the text than did advanced second language speakers or native
speakers. Moreover, the intermediate group also made a significantly
lower proportion of deletion and insertion errors than did the other two
groups. That is, the least proficient group, when reading a sentence such as
She shook the piggy bank and out came some money, would be likely to
substitute many for money (relying on graphic information), whereas more
advanced second language learners and native speakers were more likely to
insert dimes or dollars for money (relying on contextual and semantic
information). This lends support to the notion that a sequence of stages is
involved in the reading process and that less proficient second language
learners have not yet automatized skills as more advanced learners have.
Cummins (198 1) has suggested that the development of linguistic skills
be conceptualized as a continuum, with some skills requiring more and
some less active cognitive involvement. ‘Thosetasks for which the linguistic
tools have become largely automatized (mastered) require little cognitive
involvement for appropriate performance. Those tasks where this is not the
case require a greater cognitive involvement. In reading, word decoding
skills are the first that require cognitive involvement. As the learner
progresses, a degree of automaticity is achieved with respect to word
decoding, and this process can be short-circuited as the learner engages in
the process of sampling from the text to confirm predictions.
152 Language Learning Vol. 33, No. 2

Much of the difficulty that children in bilingual programs experience in


learning to read can be viewed as a failure to automatize lower-level skills
that are needed in the reading process. Children who learn t o read in a
second language must use a new vocabulary and syntactic system t o make
predictions about the text. The disharmony created by failure to predict
slows the reader t o a painful process of word-by-word decoding (Kaminsky
1976). with attendant discouragement.

IMPLICATIONS FOR SECOND LANGUAGE


PEDAGOGY AND RESEARCH

An information-processing approach stresses the limited cognitive


capacities of human learners, the use of various information-handling
techniques to overcome these limitations, and the integration of subskills in
mastering complex skills. The acquisition of a complex skill, such as
learning a second language, is thought to involve the gradual integration of
lower-level skills and their accumulation as automatic processes in long-
term storage. In the learning process component skills that require more
mental work become routinized and thereby free up controlled processes
for other functions. As automaticity develops, controlled processing is
bypassed and attentional limitations are overcome. This transition from
controlled to automatic processing is central to learning.
But how is this transition achieved? There are a number of ways of
describing the language learning process from a n information-processing
perspective. One way is through what is fashionably known in thejargon of
cognitive psychology as fop-down information processing; another is
through boffom-up information processing. The terms derive from the
field of artificial intelligence and are used in a number of different senses. In
one meaning, a t o p d o w n , or knowledge-driven, system is thought to use
higher-level information t o facilitate processing of incoming data. A
bottom-up, or inductive, system relies principally on the information
carried by the input. A t o p d o w n , or deductive, system uses information
from wider contexts (higher-level knowledge). From our point of view,
both bottom-up and t o p d o w n processing are subtypes of controlled
processing directed at achieving automaticity.
Speech t o young children tends t o be concrete and limited to events in the
immediate perceptual field (Hatch 1978). Data-driven, bottom-up
processing is adequate to deal with such relatively clear input and children
usually d o not seek out abstract, higher-level information about language.
M c Laughlin. Rossman. and M c Leod 1.53

The best pedagogical results are likely t o come from instruction that
prmides input within the child’s range of acceptance (i.e., input that is
neither so easy nor so difficult that the child “switches off‘-see Krashen
1980). One way to d o this is by focusing on tasks that require meaningful
communication.
Ordinary speech to adults, on the other hand, tends to be more abstract
and therefore more difficult than that which children hear. In this case,
interactive bottom-up and top-down processing is likely t o produce the
best pedagogical results. That is, some higher-level abstract knowledge of
linguistic structure is likely to help adult learners process a second
language. Indeed, Seliger (1975) found that adult learners given prior
knowledge of a rule system for the English article displayed greater long-
term gains in mastering the article than subjects lacking this knowledge.
Furthermore, abstract knowledge of rule systems may serve second
language learners as a shortcut in the learning process, saving them from
the trouble of generating false hypotheses about underlying rules.
By this we d o not mean to advocate any one particular teaching method.
As we argued earlier, there are a number of strategies learners can use in
tackling a second language, depending on the degree of focal attention
given to formal rules. The success of any particular strategy depends on the
characteristics of the situation and on individual learning styles. When the
input is relatively clear, an “implicit 1earning”strategy (McLaughlin 1980a)
might be most successful. That is, the learner might d o best to ignore
explicit considerations of form, and focus instead on communication. By
keeping the input clear (i.e., not too much in advance of learnercapacities)
and by concentrating on the kinds of “meaningful” and “communicative”
drills advocated by Krashen (1980), the teacher might produce the best
results, even with adult learners. Individual learning styles enter in here,
however. Some adult learners have a greater need than others to work from
knowledge of the rules-Hatch (1974) called such individuals “rule
learners.” Others prefer to work from the input-Hatch’s “data gatherers.”
Optimizing outcomes involves fitting the instruction treatment to the
individual’s learning style-which is the ultimate practical goal of research
on second language development (McLaughlin 1980b).
It is often thought that a n approach that focuses on the integration of
skills invariably stresses routine drills. It is true that in our conception of
the language learning process, repeated performance of the components of
the task through controlled processing leads to the availability of
automatized routines. As was mentioned earlier, however, these
I54 Language Learning Vol. 33, No. 2

automatized routines should be conceptualized as higher-order plans.


Such plans are thought t o be flexible entities that allow for integrated
execution of various complex tasks. As Levelt (1977) noted, while a n
essential object of training is automatization of lower-level components, it
is incorrect to conclude that this should be done exclusively by frequent
repetition of one and the same activity. Training should involve the
frequent use of a particular sentence structure in varied lexical settings, not
in the frequent use of particular sentences (Levelt 1977; McLaughlin 1979).
Once automaticity is developed and controlled processing bypassed, the
learner should be able to call up different lexical items and different
syntactic subcomponents, depending on the task. Thus the same plan can
generate different realizations depending on the lexical setting.
To summarize, the information-processing perspective does not offer
prescriptive guidelines t o the hard-pressed second language instructor. At
this stage, we d o not see our approach as providing much definitive advice
to second language pedagogy. Instead, we believe the information-
processing perspective provides a valuable research point of view. By using
objective measures developed by cognitive psychologists to measure
information processing, researchers interested in the second language
learning process can determine whether individuals or groups are
processing language in different ways,
T h e hallmarks of automatic processing a r e increased speed and
reallocated attention. These are empirically testable phenomena. Dornic’s
research (1979) is a good example of how reaction time measures can be
used to determine efficiency in processing linguistic material in first and
second languages. Dornic’s research also indicated that under conditions
where workload requirements were increased, attention was shifted from
the less dominant language so that performance in that language, but not in
the more dominant language, deteriorated.
In addition t o reaction time measures, there are other objective means of
assessing information processing in second language learners. In Favreau’s
study ( 198 I), the presence of semantic facilitation with short intervals
between prime and target words indicated highly efficient (automatic)
processing, whereas its absence indicated less efficient processing. Another
procedure is the Stroop task, in which subjects are shown color words
written in incongruent ink colors and are asked t o name the color in which
the word has been written. Presenting this task to bilingual subjects yields
information about the degree of automaticity of lexical retrieval in the two
languages, since the inability to inhibit an interfering process strongly
McLAaughlin, Rossman, and McLeod 1.5.5

indicates that the lexical retrieval process is a n automatic one (Posner and
Snyder 1975). Another paradigm involves ambiguous words. Monolingual
subjects have been shown t o be able t o process the multiple meanings of
ambiguous words even when the context is compatible with only one
meaning (e.g., Conrad 1974), which suggests that meaning activation is
automatic in such subjects. The same paradigm could be used to investigate
the degree to which meaning activation is (or is not) automatic in second
language learners with differing degrees of proficiency in a particular target
language. Finally, analysis of the sources of errors in sentence production
can provide information about the conditions under which the automatic
operations that support sentence formulation fail (Bock 1982).
A long-term goal for research on second language learning from an
information-processing perspective is a “components skills analysis”(Carr
1981). The first step in such a n analysis is to specify the component
information-processing skills that make up the task of learning a second
language, the second is t o determine which specified skills potentially are
involved in determining individual variation in overall success on the task,
and the third step is to determine the relative contribution t o variation in
overall success that is in fact made by each skill or skill group. Carr’s
analysis (198 1 ) of the reading process is a helpful model, but research on
second language learning has not achieved the sort of experimental
sophistication that has been attained in research on reading.
In this paper we have focused primarily on one aspect of information
processing-the role of automaticity in second language learning. We have
not elaborated other aspects of this approach, such as the focus of attention
dimension or the role of the executive control system. An adequate
information-processing theory of second language learning would include
not only specification of how automatic and controlled processes are
coordinated, but would also require a n understanding of selective attention
and the role and function of consciousness. For example, such a theory
would have to be able t o account for how second language learning occurs
with “peripheral” as opposed to “focal” attention t o the formal properties
of language (e.g., “implicit learning”-see McLaughlin 1980a).
We d o not mean t o imply that a n information-processing approach to
second language learning is a complete one. Such a perspective is only one
way of looking at language learning. In this respect we agree with Lakoff
and Johnson (1980) that “truth” is always relative t o a conceptual system
defined in large part by metaphQr. There is no single absolute truth about
second language learning; we are all like the blind Indians describing a n
156 Language Learning Vol. 33, No. 2

elephant. Our approach has certain advantages for research purposes, but
it does not say much, for example, about why certain formal aspects of
language are more difficult for certain learners to acquire than are other
structures. For this we turn to the linguist. On other issues one must consult
developmental psychologists or educators. Second language learning is a
complex phenomenon and there are many legitimate points of view. The
trouble begins when one starts to claim that a particular point of view is the
total one.

REFERENCES
Anderson, J.R.. and G.H. Bower. 1974. Human Associative Memory. New York: Wiley.
Baddeley. A.D. 1978. The trouble with levels: a reexamination of Crdik and Lockhart's
framework for memory research. Psychological Review 85: 139-152.
Baddeley, A.D. 1981. The concept of working memory: a view of its current state and
probable future development. Cognition 10:17-23.
Blumenthal. A.L. 1977. The Process of Cognition. Englewood Cliffs. N.J.: Prentice-Hall.
Bock. J . K . 1982. Toward a cognitive psychology of syntax: information processing
contributions to sentence formation. Ps.vchologii,al Review 89: 1-47,
Broadbent. D. 1958. Percepprion and Communication. New York: Pergamon.
Brown, R. 1973. Development of the first language of the human species. Amerkan
Psychologist 28:97-106.
Carr. T.H. 1979. Consciousness in models of human information processing: primary
memory. executive control and input regulation. In Aspwts q / Consciousness. Vol. I.
Ps.vc,hological Issues. ed. G . Underwood and R. Stevens. London: Academic Press.
Carr. T.H. 1981. Building theories of reading ability: on the relation between individual
differences in cognitive skills and reading comprehension. Cognition 9:73-1 14.
Conrad. C. 1974. Context effects in sentence comprehension: a study of the subjective lexicon.
Mernor.v and Cognition 2: 130-138.
Craik. F.I.M.. and R.S. Lockhart. 1972. Levels of processing: a framework for memory
research. Journal of Verbal Language and Verbal Behavior I 1:671-684.
Cummins, J. 1981. The role of primary language development in promoting educational
success for language minority students. In Schooling and Language Minoripry Srudenrs: A
Theoretical Framewmork. Los Angeles: Evaluation. Dissemination and Assessment Center.
California State University.
Cziko. G.A. 1980. Language competence and reading strategies: a comparison of first- and
second-language oral reading errors. Language burning 30: 101-1 16.
De Groot. A.D. 1965. Thought and Choice in Chess. The Hague: Mouton.
Dornic. S. 1979. Information processing in bilinguals: some selected issues. PsycholoRical
Research 40: 329-348.
Favreau. M. 1981. Automatic and Conscious Attentional Processes in First and Second
Languages of Fluent Bilinguals: Implications for Reading. Doctoral dissertation,
Concordia University.
Hasher. L.. and R.T. Zacks. 1979. Automatic and effortful processes in memory. Journal(/
Experimental Ps.vcholog.r.. General 108:356-388.
Hatch, E. 1974. Second language acquisition universals'? Working Papers on Bilingualism
~

3:l-17.
Mc Laughlin, Rossman, and Mc,Leod 15 7

Hatch, E. 1978. Discourse analysis and second language acquisition. In Swond Language
Arquisition. ed. E. Hatch. Rowley, Mass : Newbury House.
Hatch. E.. P. Polin. and S . Part. 1970. Acoustic Scanning or Syntactic Processing. Paper
presented at meeting of the Western Psychological Association, San Francisco.
Henning. G. H . 1974. Remembering foreign language vocabulary: acoustic and semantic
parameters. Language Learning 23: 185-196.
James. W. 1890. The Princ,iples ?/ f.vycholog.r.. New York: Holt.
Kahneman, D. 1973. Arrention and EJort. Englewood Cliffs. N.J.: Prentice-Hall.
Kaminsky. S. 1976. Bilingualism and learning to read. In The Bilingual Child. ed. A. Simoes.
New York: Academic Press.
Krashen. S . 1976. Second Lmguage Acquisiiion. Survqv of Linguistic Science, ed. W .
Dingwall. Stamford. Conn. Greylock.
Krashen. S. 1977. The Monitor Model for adult second language performance. In Vienpoints
on English as a Second Language, ed. M. Burt. H. Dulay, and M. Finocchiaro. New York:
Regents.
Krashen. S. 1979. A response to McLaughlin. “The Monitor Model: Some methodological
considerations.” Language Learning 29: 15 1-167.
Krashen. S. 1980. Relating theory and practice in adult second language acquisition. I n
S e w n d Language Development. ed. S . Felix. Tuebingen: Narr.
Krashen. S.. J . Butler, R. Birnbaum and J . Robertson. 1978. Two studies in language
acquisition and language learning. ITL: Review, of Applied Linguistics 3940.
LaBerge, D.. and S.J. Samuels. 1974. Towards a theory of automatic information processing
in reading. Cognitive fs,rchology 6:293-323.
Lakoff. G . . and M. Johnson. 1980. Metaphors we live by. Chicago: University of Chicago
Press.
Levelt. W.J.M. 1977. Skill theory and language teaching. Studies in S e w n d Language
Aquisirion I :53-70.
Lichtenstein, E.H., and W.F. Brewer. 1980. Memory for goaldirected events. Cognriive
Psychology 12:412445.
Magiste. E. 1979. The competing language systems of the multilingual: a developmental study
of decoding and encoding processes. Journal of Verbal b u r n i n g and Verbal Behavior
I8 :79-90.
McLaughlin. B. 1965. “1ntentional”and “incidental”learning under conditions of intention to
learn and motivation. Psr,vhologicd Bulletin 63:359-376.
McLaughlin, B. 1978. The Monitor Model: some methodological considerations. Language
b u r n i n g 28:309-332.
McLaughlin. B. 1979. Linguistic input and conversational strategies in L I and L2. Studies in
S e w n d Language Acyuisrrion 2: I - I 6
McLaughlin, B. 198Oa. On the use of miniature artificial languages i n second-language
research. Applied fsyc~holinguisrics1:357-369.
McLaughlin, B. 1980b. Theory and research in second-language learning: an emerging
paradigm. Language b u r n i n g 30:331-350.
Miller. G.A. 1956. The magic number seven. plus or minus two: some limits on our capacity
for processing information. fs.r.c.hologic,al Review 63:8 1-97.
Miller. G.A., E. Galanter, and K . Pribram. 1960. Plans and rhe Structure vf6ehuvior. New
York: Holt. Rinehart. & Winston.
Neisser. U. 1967. Cognitive Psjschology. New York: Appleton-Century-Crofts.
Posner. M.I.. and C.R.R. Snyder. 1975. Attention and cognitive control. In lnjormation
Processing and Cognition: The Lqyola Symposium, ed. R.L. Solso. Hillsdale. N.J.:
Erlbaum.
Rapaport. D.. ed. 1957. Cognirive Srrui,rures: Collected Papers. New York: Basic Books.
158 Language Learning Vol. 33* N o . 2

Rossman. T. 1981. 7 he Nature of Linguistic Processing in Reading a Second Language.


Masters Thesis. Concordia University.
Sachs. J . 1967. Recognition memory for syntactic and semantic aspects of connected
discourse. Perception and Ps.vc.huphy.5ic.s2:437442.
Schneider. W.. and R.M. Shiffrin. 1977.Controlled and automatic processing. I . detection,
search. and attention. Psjrhu/ugic.u/ Review 84:1-66.
Seliger. H.W. 1975. Inductive method and deductive method in language teaching: a re-
examination. I R A L 13: I- 18.
Shallice. T. 1972. Dual functions of consciousness. Psj~cho/ogicdReview. 79:383-393.
Shati. M. 1977. The relationship between cognitive processes and the development of
communicative skills. In Nebraska Symposium on Morivarion. ed. H.E. Howe and C.B.
Keasey. Lincoln. Nebr: University of Nebraska Press.
Shiffrin. R.M.. and W. Schneider. 1977. Controlled and automatic human information
processing. I I . perceptual learning, automatic attending. and a general theory.
P.yyc.ho/ugica/Review, 84:127-190.
Tarone. E. 1979. Interlanguage as chameleon. Language Lanrning 29:181-191
Tarone. E. 1982.Systematicity and attention in interlanguage. Language Learning 32.
Taylor, B.P. 1975. 7 h e use of overgeneralization and transfer learning strategies by
elementary and intermediate students in ESL. Language Learning 25:73-108.
Wode, H. 1978.Developmental principles in naturalistic L2 acquisition. I n Sewnd Longuage
Ac.quisitiun: A Buuk v / Readings. ed. E. Hatch. Rowley. Mass.: Newbury House Press.
Wolfe. S. 1981. Bilingualism: One or Two Conceptual Systems. Masters Thesis. San
Francisco State University.
Yerkes. R.M..and S.D. Dodson. 1908.Therelationofstrengthofstimulustorapidityofhabit
formation. Journal of Cumpararive Neuro/ug.t, 18:459482.

You might also like