THE MULTILINGUAL MIND
https://www.frontiersin.org/files/Articles/105591/fpsyg-05-01401-HTML/image_m/fpsyg-05-01401-g002.jpg
A. Introduction
1.
A General Language
Processing Model: Levelt’s Speaking Blueprint
Levelt’s Speaking Blueprint has become
the most complete and accepted one for a monolingual speaker. When
a person wants to express an idea, he/she has to find words for that
idea, put those words in a wellformed sentence, pronounce the words in the
right order, and in doing so she has
to co-ordinate thousands of tiny little muscles. To explain this complexity
of interacting events in speech processing, Levelt’s Speaking
blueprint (Levelt, 1993; Levelt, 1989) According to
this model (see Figure A4.1), the production of speech takes place
in three relatively distinct stages: the Conceptualiser, the Formulator and the Articulator.
·
The
Conceptualizer, this is the first stage, where the speaker which generates a ‘preverbal’
message and contains meaning intentions that have to be put into words and
sentences in the next two stages. This preverbal message contains a number of
conceptual characteristics.
·
The
Formulator, in this stage, the preverbal message from the stage ‘conseptualizer’
lead to the selection of a set of lexical items called ‘lemmas’ in the
Formulator. A ‘lemma’ can be seen as the ‘word to represent a concept’. In
addition to representing a concept and containing semantic information each
lemma contains all kinds of other information – most importantly, how this word
combines with other ones. In other words, is it a noun or a verb, and if it is
a verb, what type of complement does it take? Or, is this word formal enough
(register), or is this word appropriate in this context (pragmatic
information)? Once the appropriate lemmas have been selected, they have to be
combined into a well-formed sentence. This process is called ‘grammatical
encoding’, which Levelt describes as ‘solving a set of simultaneous equations’
(1993: 4).Grammatical encoding results in a surface structure of a sentence in
which all the properties of all the lemmas selected are satisfied. However, the
surface structure has not yet been specified for its phonological
characteristics. This is taken care of in the next stage ‘phonological
encoding’, where the phonological information associated with the selected
lemmas is matched to phonologically encoded word frames.
·
The
Articulator, phonological encoding from the previous stage is actualized in the
form of sound (spoken language).
The processes is illustrated in a figure below.
In
addition, the lexicon in Levelt’s model consists of two separate elements: the
lemma which contains conceptual, semantic, syntactic and pragmatic information,
and the lexeme, which is the phonological form associated with the lemma. It is
important to realise that an entry in the lexicon can be:
§
a single word (school),
§
a compound word (high school),
§
a fixed expression (go to school, graduate from high
school),
§
an idiom (to be of the old school = ‘to have an
old-fashioned or traditional opinion’)
§
or any other group of words that are stored as a
conventional unit.
2.
Towards a Dynamic
Model of the Multilingual Mental Lexicon
Attempts have been made to adjust Levelt’s
model for the multilingual speaker. For example, De Bot (1992) argues that the
Conceptualiser is most likely to be language-independent, whereas the
Formulator is the most likely candidate to be language-dependent because it
contains information about grammar. However, selection of the words from the
right language requires the inclusion of language-related information in the
preverbal message (De Bot, 2002). As the lexicon plays a central role in
language processing, we will discuss these and other matters from the
perspective of the multilingual mental lexicon.
Related to this, a well-known theory, The central assumption in Weinreich’s approach said that
concepts and words are stored separately. With this assumption in mind,
Weinreich argued that there are three different ways in which the multilingual
lexicon could possibly be organised: as a compound, as a co-ordinate and
as a subordinate one. In a compound organisation, it is assumed that
there is one common concept with a different word in each language. In a
co-ordinate organisation, there is a complete separation between the different
languages: each word in each language has its own concept. In a subordinate
organisation, there is just one set of concepts, but the items in the second
language can only be reached via the items in the first language: there are no
direct connections between the concepts and the words in the second language.
Another interesting discussion is
about whether a language could be switched on and off or not. In this case, an
influential proposal was that of Green (1986), who proposed three states in
which languages can be at a certain moment in time: selected, active and
dormant. The language that is used at a certain moment is the selected language;
languages that, at that particular moment, play a role in the background are
labelled active; languages that do not play a role at that moment are dormant.
The assumption of the middlemost level – that of the active language – is
required by the observation that when speaking a particular language, a speaker
may use words from another language, either because that word is more appropriate
or because the speaker cannot immediately find the word in the selected
language. This code-switching is very common, and models of lexical
processing must be able to account for this phenomenon.
Another important theory is ‘A
Dynamic Model of Multilingusl Mental Lexicon’. The model shows that the lexicon
can be seen as a dynamic system in its own right because it constantly changes,
influenced by external and internal forces, and self-organises. All the
possible information associated with a lemma and the degree of activation of a
lexeme depend on the input and output a speaker has experienced to a greater or
lesser degree not only in his entire life, but also in very recent times.
3.
A Dynamic Model of
The Multilingual Mental Lexicon and Its Implications for SLA
The dynamic
model of the multilingual mental lexicon is correct, it will have
certain implications for SLA. The key words are ‘association’ and ‘activation’. The
words that are heard, seen or used most often are the words that are most easily
accessed again and will have the most associations with other information such
as how it is used. Words that are heard, seen or used the least will be the more
difficult to retrieve.
certain implications for SLA. The key words are ‘association’ and ‘activation’. The
words that are heard, seen or used most often are the words that are most easily
accessed again and will have the most associations with other information such
as how it is used. Words that are heard, seen or used the least will be the more
difficult to retrieve.
Additionally,
the
literature on effective and efficient vocabulary acquisition generally points
to the necessity of ‘association’ and ‘activation’ with the terms elaboration and rehearsal (Hulstijn, 2000). Elaboration means that a maximum
number of associations is made in relation to a lexical item; the
word should be seen in several different contexts
and the learner should pay explicit attention
to all characteristics of a word, from
orthography and prosody to the word’s syntactic and
semantic features. Besides elaboration, rehearsal is
required to make access to the lexical item more automatic. The best results will be achieved when rehearsal
takes place with increasing time
intervals, from minutes up to about one month.
B.
Extension
Unit B4
The Multilingual
Mind Extended
Judith F. Kroll conducted
several researches to expose the issues regarding bilinguals and multilinguals.
In this part, the first two articles were conducted by Kroll in different year
with different partner. The first article, Kroll and Steward (1994), deals with
Revised Hierarchical Model (RHM) which becomes the reference for the current
resources. The model deals with the connection between the concepts, the L1
lexicon and L2 lexicon. In short, L1 words can directly be connected to the
speaker’s conceptual representation (graphic 1) while L2 words cannot be dirrectly
connected and it should be bridged by L1 (graphic 2).
The
second article conducted by Kroll and Dijkstra (2002) which focuses on the
correlated questions towards the multilingual lexicon by using Bilingual
Interactive Activation (BIA). The burning issue is whether selective access
must be assumed or all language we know is active in processing. Although these
articles were conducted by the same researcher, the second one appears as the
critics for the previous one especially on the argument that the lexicons of
different languages are not processed at the same time.
‘Category
interference in translation and picture naming – evidence for assymetric
connections between bilingual memory representations’ by Judith F. Kroll and
Erika Steward (1994) in Journal of Memory
and Language, 33, 149-174
Many
scholars still cite this article since it argue that the lexicons of L1 and L2
are detached. The aim is to obtain the answer of whther the lexical items of L2
is directly connected to the relevant concepts or they are still interceded by
the lexical items of L1.
Task
B4.1
In
another research, Kroll and Curley (1988) stated that if eloquent bilinguals
were conceptually interceded L2 words, they can possibly
Also, they argued that
only eloquent subjects of bilinguals in L2 should get adavantages from the
semantic organization. On the other hand, it is confusing because the
assumption is different from the result. The assumption is if the sematic
category is well-arranged, the subjects spend longer time to name the pictures.
There are two results of this study. The first is that the eloquent subjects
spend longer time to find the lexical words in L2 when the list is categorized
rather than randomly mixed. The second is all subjects spend longer time to
name the pictures in L1 when the list is categorized rather than mixed
randomly. To emphasized, the results were influenced by the semantic
organization. They finally assumed that there was category of interference
rather than facilitation in the process.
Task
B4.2
Then,
another experiment was conducted to imitate Kroll and Curley’s study with a
bigger subjucts population and a better design. The subjects are tested in all
conditions (a within subject design). This uses a translation task which
resulted the same with the previous one the eloquent subjects took longer
duration to translate into L2 when the lexical word are categrized than
randomly mixed. Category interfrence
effect is the reason that is caused by the linking (apping) between the
concepts and the words. In short, the eloquent bilinguals need more time to
find the right and appropriate words rather than answer it at random with
shallow knowledge. For example, there are three rows of pink color. Less fluent
bilinguals will say that all colors are pink. However, the fluent ones are able
to differentiate the color such as pink, fuschia, and magenta.
A
|
B
|
C
|
To
obtain more satisfying ansswers, two additional experiment were conducted using
different research technique.
Experiment 1: Picture Naming Task
The concern of this
experiment is the link between concepts and lexical items for native speakers.
The participants were showed picture and they have to say the picture’s
representation. In this experiment, the duration between the moment the picture
appears and their reactions is measured. The result replicated the category
interference effect in picture naning by Kroll and Curley (1988). Pictures
produced category interference when they were named in a semantically
categorized list.
Experiment 2: Naming Pictures and
Words in Alternation
In this experiment, the
hypothesis was tested using spacing picture naming trials to reduce the
condition from repeated conceptual access. The major argument of the experiment
is that the word naming can be finished without conceptual access.
Task
B4.3
Method
The stimulus materials were identical to
Experiment 1. The stimulus materials for experiment 1 were 120 line drawings of
objects and the frequency ranged from 0-413x/million with a mean of 31.7. 16 UG
students with normal or corrected-to-normal visual acquity. The sequence of
alternation was equilibrated through stimulus list.
Task
B4.4
Reaction
times
Generally, the
reliability was slower in picture naming than the word naming. Here are the
detail measurement of the reaction times.
Ø Analysis by Subjects
F(1,15)=150.87, p<.001 (reliability)
F(1,15)=1.02, p<.10 (stimulus modality
and type of list)
F(1,15)<1 (effect of list type on
naming)
Ø Analysis by Items
F(1.238)=94=32.95, p<.001(reliability)
F(1,238)<1(stimulus modality and type
of list)
F(1,238)<1 (effect of list type on
naming)
Percentage of Errors
The total error was fewer in Experiment 1
than in Experiment 2.
Ø Experiment 1 (3.4%)
Picture
naming (13,1%)
Word
Naming (2,2%)
Ø Experiment 2 (7.6%)
Category
inference effect in picture naming
Alternation
of word and picture naming
Task
B4.5
In
conducting the third experiment, the researchers were trying to figure out the
reasons of the different rate of translation from L2 into L1. The arguments
strongly stated that bilinguals traslate
from L2 into L1 in shorter duration. In short, bilinguals translate faster from
L2 to L1 than from L1 to L2. It is
claled assymetric translation which needs modification on the concept mediation
and word association. However, the magnitude of the difference between L1 and
L2 naming was small relative to the difference between the 2 forms of
translation. Translation from L2 to L1 completed on a lexical basis. On the
other hand, translation from L1 to L2 needs concept mediation. In the process
of concept mediation, additional time should be considered since responding to
the picures needs longer duration than to the words.
Kroll
and Steward (1990) through RHM served explanation that both lexical and
conceptual links are active in bilingual memory. However, regarding the
purpose, one of them can be more dominant. Since all people including fluent
bilingual and mutilingual master L1 earlier and deeper than L2, the words or
the vocabularies are richer in L1. Thus, the lexical association from L2 to L1
are stronger. It is also counted for the links between words and concepts in
which L1 is stronger than L2.
In a
case of childhood, the younger they learn a second language, the better it is.
The mapping of L2 can be said similar to the L1 mapping to conceptual memory.
In younger age, L2 words are embeded to the system by lexical links with the
L1. The more fluent they are in L2, L2 can produce a direct mapping to the
concepts while the lexical connection into L1 is still exist. The conceptual
links and the lexical links are two-way direction. However, the lexical link from
L2 into L1 is stronger than the conceptual links because basically L2 words are
connected and mediated by L1.
Task B4.6
Since
thetranslation from L2 to L1 is faster than L1 to L2, the connection to the
lexical is required. On the other hand, the last experiment needs mediation
concept. Also, interference does not alsways appear in all situation. It only
occurs in the conceptually based translation from L1 to L2. The researchers
also found the the additional time is needed. If the translation between L2
into L1 is stronger and L2 into L1 needs conceptual access, so the cognate
words in L1 does not include the L2 lexical representation. If a L2 learner
already recahed the connection of lexical between L2 and L1 before being able
to facilitate the L2, they are supposed to be able to translate from L2 to L1
first.
Conclusions
-
In categorized pictures, there were found category interference effect.
-
For bilinguals who still cannot balance the L1 and L1, an assymetry occurs betwee two directionals of translation
which influence describes differential reliance.
-
Translation from L1 into L2 is conceptually mediated and from L2 to L1 is
lexically mediated.
-
Crosslanguage mappings between lexical representation , and between lexical
representation and concepts, are asymmetric.
B4.7
The
second article in this chapter, Kroll and Dijkstra (2002), criticised the first
article which argument is that the lexicons of different languages are accessed
separately and the language should be decided first before the words. The main
question from the second article was the way bilinguals realize and say the
words in each two languages.
Model
of the Bilingual Lexicon
The
Revised Hierarchical Model
In order to understand
how people acquire or learn a second language, several studies have been
conducted. One of these studies was carried out by Kroll & Stewart (1994)
who proposed Revised Hierarchical Model (RHM). This model focused on whether
the words in the bilingual’s two languages are connected via the lexical representations
or by direct access to the conceptual representations. In proving this
assumption, initial evidence suggested that the connections between lexical
forms in L1 and L2 might be active early in L2 acquisition and then as the
second language learners improved their skill in L2, the words in each language
(first and second language) could access concepts directly. In addition, this
model also assumed that lexical links were stronger from L2 to L1 than the
reverse, as this was the initial direction of transfer during acquisition, and
L1 was assumed to have stronger connections to concepts than L2.
The RHM assumed
independent lexical representations for words in each language. Nevertheless,
more recent studies comprehension and production of words in two languages
suggested that RHM assumption was incorrect. However, as these recent studies
rejected the assumption proposed by Kroll & Stewart (1994), these studies
then supported an integrated lexicon which produced asymmetry occurrence
between L1 and L2.
The
Bilingual Interactive Activation Model (BIA)
The bilingual interactive
activation (BIA) model is a bilingual extension of the well-known interactive
activation (IA) model for monolingual visual recognition. This model differs
from the original IA model in two main respects: 1) it incorporates an
integrated lexicon with words of two languages rather than one, and 2) it
includes an extra layer of two language nodes that can be considered as
language label (tags) that indicate the language membership of each word.
The BIA model differs
from the RHM since BIA model accounts for asymmetries observed in unbalanced
bilinguals (stronger effect from L1 on L2 than vice versa). The BIA is
implemented in terms of the model’s resting level activations, which are generally
lower for words in L2 than L1. As a result, L2 words on the whole become
activated more slowly and to a lesser extent than L1 words.
Comprehension
What
codes are activated?
The presentation of a
word to monolingual induces activation not only of orthographic codes but of
phonological and semantic codes as well. Additionally, several studies have
been done in investigating what kind of codes that are activated as we speak in
the first and second language. First, in monolingual studies, it was found that
ambiguous words (e.g. ‘bug’ which can meant as an insect, a spy, or programming
error) could initially activated during recognition. Second, in as study of
word naming, Gottlob, Goldinger, Stone & Van Orden (as cited in De Bot,
Lowie, & Verspoor, 2005) found that English homographs (words with separate
pronunciations and meanings) were read slower than homonyms (with a single
pronunciation but separate meanings) and control words. These studies have
proved that during monolingual word recognition there is intralingual co-activation of lexical candidates with overlap in
meaning or form.
Some assumptions also
have been made in relation with the studies above. A nonselective access view
predicts that there will be interlingual
activation of similar words during bilingual word recognition as well. On the
other hand, selective access view assumes that a presented word will activate
the form and the meaning representations only from the language that is
currently selected. However, the study carried out by Dijkstra et al. (as cited
in De Bot, Lowie, & Verspoor, 2005) found that a presented word form leads
to the activation of all representations that it is associated with,
irrespective of the target language.
When
are these codes activated?
It is important to know at
which moment in time is the necessary lexical candidate selected. Then, based
on a study carried out by Dijkstra, Timmermans, & Schriefers (as cited in
De Bot, Lowie, & Verspoor, 2005), it is revealed that co-activation of
lexical candidates from different languages occurs until relatively late in the
word of recognition process. In addition, this study also reveals that the time
needed by the participants in the word recognition process depends on the
relative word frequency of the two readings of the interlingual homograph. For
bilingual, this subjective frequency is lower for items that belong to their L2
than to their L1, especially when their L2 proficiency level is low. Then, if
the subjective frequency of the L2-reading is negligible relative to its L1
frequency, the L2 reading will not be able to affect the lexical processing to
any considerable extent; meaning that the effect of the L2 on the L1 lexicon is
weak.
Critical
Factors that Affect Lexical Selection
It is also important to
consider the critical factors that may affect the selection of lexical
candidates during the bilingual word recognition process. The factors that may
have been identified in earlier including L2-proficiency, language intermixing,
task demands, and instruction.
Language
Intermixing and Task Demands
According to De Bot,
Lowie, & Verspoor (2005), “language intermixing refers to whether an
experiment contains exclusively items that belong to one language (blocked
presentation) or items from two languages (mixed presentation)”. A series of
three lexical decision experiments is conducted by Dijkstra, Van Jaarsveld,
& Ten Brinke (as cited in De Bot, Lowie, & Verspoor, 2005) in
investigating more about language intermixing and task demands. In Experiment
1, mixed presentation was carried out to Dutch bilingual participants; they
were asked to perform an English lexical decision task on a list that included
English/Dutch homographs, cognates, and purely English control words. In this
experiment, it was found that the response times to interlingual homographs
were unaffected by the frequency of the Dutch reading and did not differ from
monolingual controls. However, cognates were recognized faster than controls.
Then, in Experiment 2,
Dutch participants were asked to respond ‘no’ to Dutch words. This experiment
revealed that strong inhibition effects were obtained depended on the relative
frequency difference of the two readings of the homograph. The result of this
experiment evidently rejected the selective access view of bilingual word
recognition. Next, in Experiment 3, the participants performed general lexical
decision task by responding ‘yes’ if a word of either language is presented. In
this experiment, frequency-dependent facilitation effects were found for the
interlingual homographs. Thus, it could be concluded that the representations
for homographs in different languages could lead to both inhibition and
facilitation effect.
Effect
of Instruction
There are two types of
factors that we need to know. First, the bottom-up factors, such as: language
intermixing, stimulus characteristics, etc. Second, the top-down factors, such
as: participant expectancies based on the instructions of the experimenter. A
study carried out by Dijkstra, De Bruijn, Schriefers, & Ten Brinke (as
cited in De Bot, Lowie, & Verspoor, 2005) was tried to compare the effect
of instruction-induced expectancies and language intermixing in an English
lexical decision task performed by Dutch–English bilinguals. In the first
experiment, participants were explicitly instructed to respond ‘yes’ to
interlingual homographs and exclusively English words and ‘no’ to English
nonwords and to exclusively Dutch words. Then, in the second experiment, Dutch
items were introduced.
The result of the study
revealed that there was no significant differences were found between
interlingual homographs and controls in the first part of the experiment, while
strong inhibition was obtained for interlingual homographs in the second part.
This effect is demonstrated for words with a low-frequency reading in Dutch and
a high-frequency reading in English. In conclusion, Dijkstra, De Bruijn,
Schriefers, & Ten Brinke (as cited in De Bot, Lowie, & Verspoor, 2005) suggested
that language intermixing, rather than instruction-based expectancies, drives
the bilingual participants’ performance.
Discussion:
Similarities and Differences between Comprehension and Production
De Bot, Lowie, &
Verspoor (2005) stated that “the most striking similarity between comprehension
and production in bilinguals is the overwhelming evidence for nonselective
access to words in both languages”. They further explain comprehension and
production share the consequences of the lower L2 than L1 proficiency in
unbalanced bilinguals.
However, even though the
two domains share the aspect of language nonselective access, this does not
imply that orthographic, phonological, and semantic codes are used in the same
way or at the same moment in processing. For instance, the bottom-up nature of
comprehension requires that orthographic codes play a larger and earlier role
in word recognition than they do in word production. Likewise, the role of
phonology is likely to be more critical in production than in comprehension.
Furthermore, in both domains, there is evidence for semantic processing, but
again the contribution of meaning is generally more reliable in production than
in comprehension. In comprehension, semantics appear to play a role when there
is a consistent correspondence between lexical form and meaning.
Moreover, the time course
over which these lexical codes are activated must also be different for
comprehension and production, especially the different nature of orthographic
and conceptual representations makes such a difference all the more likely. For
example, it may be that more lexical alternatives are initially activated in
comprehension than in production because there are simply more orthographic
neighbors of the input word than semantic alternatives for the output concept.
The last, in production,
the language of speaking can and must be determined by the language user; in
comprehension the requirement to determine the language in which the task is
performed depends in a more complex way on the nature of the task itself.
World
Retrieval in a Second Language
One of the crucial
aspects of language production is retrieval of words. Many second/foreign
language learners still find it difficult to find words in the target language
than in the first language. In proving this statement, we can conduct a simple
test called ‘verbal fluency’ which is widely used in research on aphasia. In
this test, the participants should list as many words from a specific category
as possible within one minute. What is meant by specific category here are
phonological (e.g. words beginning with letter ‘d’) or semantic (e.g.
‘vegetables’, ‘fruits’, etc.).
Experiencing
L2 Processing
One of the most complex
linguistic activities humans can do is to simultaneously interpret in the
target language. As we know that simultaneous interpreter must at the same time
listen and understand what is being said and translate that and produce
coherent speech as fast as possible. Then, in getting used to interpreting
simultaneously, you can do the following task: 1) choose a text in the first
language, 2) read it slowly and record your reading, 3) play the record and
start to interpret it by using the target language with another recorder, 4)
transcribe your rendering and compare it with the original text, and 5) analyze
it; you may want to know what went well and what went wrong, etc.
Reading
Test
As we speak in the target
language, we may experience tip-of-the-tongue phenomena; a common occurrence
where the speakers do know the number of syllables and the stress pattern of a
word but fail to fill in the empty skeleton with segmental information. Then,
reading test task here is a small-scale experiment in which you test lexical
access in the multilingual mental lexicon. Study the following text (De Bot,
Lowie, & Verspoor, 2005, p. 260):
Aoccdrnig to rscheearch at Cmabrigde
Uinervtisy, it deosn’t mttaer in waht oredr the ltteers in a wrod are, the olny
iprmoetnat tihng is that the frist and lsat ltteer be at the rghit pclae. The
rset can be a total mses and you can sitll raed it wouthit porbelm. Tihs is
bcuseae the human mnid deos not raed ervey lteter by istlef, but the wrod as a
wlohe. Amzanig huh?
In conducting this test,
you can do the following steps: 1) create two texts as the one above (leaving
the first and the last in its place and randomizing the in-between letters);
one in the L1 of your subjects and one in the L2 of your subjects, 2) find the
subjects with different level of proficiency in the L2, and 3) ask your
subjects to read both the L1 and L2 texts aloud and time the reading time.
References
Bot, K.D., Lowie, W., & Verspoor, M. (2000). Second Language
Acquisition. An Advance Source Book. New York: Routledge
Kroll, J.F and Curley, J. (1986). Picture Naming and Bilingual Translation.
Unpublished Manuscript. South Hadley, MA: Mount Holyoke College.
Kroll, J.F and Curley, J. (1988). Lexical Memory in novice bilinguals: The
role of concepts in retrieving second language words. In M. Gruneberg, P.
Morris, and R. Sykes (Eds), Practical
Aspects of Memory (pp. 389-395). London: John Wiley and Sons.
Kroll, J.F and Dijkstra, T (2002). The bilingual Lexicon. In R. Kaplan
(Ed), The Oxford Handbook of Applied
Linguistics (pp. 301-321_. Oxford: Oxford University Press.
Kroll, J.F and Steward, E. (1989). Translating from one language to
another: The role of owrds and concepts in making the connection. Paper
persented at the Meeting of teh Dutch Psychonomic Society, Noordwikehout, The
Netherlands.
Kroll, J.F Steward, E. (1994). Concept Mediation in Bilingual Translation
and Picture Naming – evidence for asymmetric connections between bilingual
memory representations. Journal of Memory
and Language, 33, 149-174.
Comments
Post a Comment