Dual Coding Theory and Trilingual Implications on Recall

Dual Coding and Trilingual Implications
17
Dual Coding Theory and Trilingual Implications on Recall
Brittany Christiansen
Huron University College
In order to observe trilingual implications on
Paivio's dual-coding theory, 48 undergraduate participants,
half bilingual, and half trilingual, •were randomly assigned
to one of two conditions. These conditions were either
translating or copying a list of 16 words they had to
subsequently and unknowingly recall after an irrelevant
anagram task. It was found that translating the words
yielded superior recall, regardless of degree of
multilingualism. However, due to methodological issues
further discussed, the results were unable to significantly
outline concrete implieations trilingualism may have
regarding the dual-coding theory. Future suggestions are
also explored.
In the context of the unilingual memory and cognition, Paivio (1971) developed
what he called a dual-coding model. This theory is based on the premise that memory and
cognition are assisted by two distinct symbolic systems, with one addressing verbal
information and the other addressing nonverbal information (or imagery). Paivio also
assumed that the two systems were simultaneously interconnected and independent.
Interconnectedness or interacting referring to the ability of a representation in one to
activate a representation in the other, more concretely, the verbal system can arouse
images and the images can be described verbally. Independence implying that there
should be an additive effect on recall i f both systems are to be activated, since activation
in one system does not engage the other (Amedt and Gentile, 1986). Paivio and
Desrochers (1980) put forth an extended theoretical framework for the dual coding theory
that incorporates the independence approach and regards bilingual features of memory
and cognition. The independence aspeet would assume that bilingual individuals have
Dual Coding and Trilingual Implications
1g
two distinct memory stores and that the eonnection between the two occurs during
translation, thus having an additive effect on recall probability. Amedt and Gentile go on
to say that the verbal arrangement of multilingual individuals are further partitioned into
distinct linguistic systems, each independent of, however still interacting with, the other
system as well as with the non-verbal one. Paivio and Desrochers took this assumption
one step further when suggesting that the manner in which the systems interact is during
translation, maintaining an approximate one-to-one relation between languages (e.g. dogchien), as opposed to a one-to-many relation within languages (e.g. dog, canine, hound,
mongrel etc).
Paivio and Lambert (1981) performed a direct test of the independence and
interconnectedness aspects of the dual coding theory in French-English bilinguals as their
memory was assessed for words coded unilingually, bilingually, and pictorially. During
the experiment, participants copied 1/3 of a list of common English words, translated
another third into French, and drew pictures of the remaining third. A subsequently
unanticipated memory task then followed. They concluded that the results were
consistent with the predictions from the bilingual dual coding hypothesis meaning that
recall increased dramatically in the following order: copy condition, translate condition,
image condition. Paivio and Lambert however, went further in arguing that while the
difference between the copy and translation conditions could be explained by the Craik
and Lockhart (1971) depth-of-processing theory, such a theory could not account for the
distinction between tbe imagery and translation conditions. To understand the depth-ofprocessing theory, one must first understand what levels of processing are. As cited by
Craik and Lockhart, there is a general consensus that perception involves quick
Dual Coding and Trilingual Implications
19
interpretation of stimuli at a number of levels (stages) (Selfridge and Neisser, 1960;
Treisman, 1964; Sutherland, 1968). Preliminary levels concern themselves with the
evaluation of the physical realm such as lines, angles, brightness, pitch, and loudness,
while further stages compare the input with stored concepts from previous learning
(pattern recognition and derivation of meaning). This is what is known as depth of
processing since there is a sort of hierarchy of stages, where 'depth' refers to a more
substantial degree of semantic or cognitive analysis (Craik & Lockhart). As cited by Vaid
(1988), two subsequent experiments were able to replicate Paivio and Lambert's partem
of results using French-English bilingual participants (Vaid, 1982; Amedt & Gentile,
1986) which, taken together, have been interpreted by the psychological community as
evidence and support for a dual coding model of memory, and moreover, an
independence approach to bilingual linguistic memory.
Amedt and Gentile point out, however, that due to the general nature of the dual
coding model, there could be a lot of evidence supporting parts of the model while
simultaneously refuting or not proving other features. An example of this is the
realization Vaid had prior to conducting their study, since the shortcomings of the Paivio
and Lambert experiment were taken into account and instead, a synonym generation task
was compared to translation, copying, and imaging. There were 28 undergraduate
psychology students bilingual in Spanish and English who participated, as well as 24
English monolingual students as the control group. They found a similar partem of recall
noted in previous studies (such as the Paivio and Lambert, and Amedt and Gentile) with
bilinguals, where imaged words were recalled at least twiee as well as words coded
verbally, which is consistent with the dual coding model of representation. However, they
Dual Coding and Trilingual Implications
20
highlight that in emphasizing qualitative distinctions between representational coding, the
dual coding theory needs further elaboration of kind of processing, whether it be verbal,
nonverbal, or both in terms of its visual representation. Without this specification, the
dual coding approach escapes the perseverant confound of levels-of-processing.
No previous research has been done to test implications that the trilingual brain
has on the dual coding theory and the present study will introduce that novel factor.
Previous experimental designs testing the dual coding model ran into confounds since
they contrasted unilinguals with bilinguals, despite finding a seemingly analogous
condition to translation involving unilingual synonym generation. The advantage of using
trilingual individuals and comparing them with bilinguals is that they are both able to
translate the words and a true test of coding will be possible when reviewing whether
translating twice rather than once will enhance recall.
The present study had a list of 16 words with two lines underneath each where the
words had to be either copied in English or translated into the languages known to the
participant. A n irrelevant anagram task took up the second page to ensure that the
participants did not anticipate having to recall the list. On the final page of the
questionnaire, the participants were asked to recall as many words as they could in an
unlimited amount of time. The two conditions of each group were to write the equivalent
words in their respective languages or to simply copy them twice in English. In order to
assure that the bi and trilingual groups were spending an equal amount of time on the first
task, the bilingual individuals wrote the English word once as well as the translated
equivalent in the second language, while the trilingual individuals wrote the two other
translated equivalents and disregarded writing it in English.
Dual Coding and Trilingual Implications
21
The consequence of the independence in memory coding systems in bilingual and
trilingual individuals is that the translation condition should yield an additive effect on
recall probability, thus recall should be superior in tbe condition where the words were
translated than the condition where repeating the same word does not have an additive
effect (Clark, Lambert and Paivio, 1988). With that said, the trilingual individuals have
further translations than do bilingual partieipants so they theoretically should be able to
recall more words as well. The rationale behind this assumption is that translating is
presumably engaging two verbal systems, while simply copying activates only one. A
further rationale would be the view the dual coding theory holds about memory for
translated words being better than that for copying because of the hypothesized weaker
connection of within-language verbal representations relative to between-language
connections (Paivio, 1981).
The present hypothesis is that in both bi and trilingual individuals, recall will be
superior in the end i f on the first page of the questionnaire, the words were translated
rather than copied. A secondary hypothesis will be that trilingual participants will
outperform bilingual participants since they performed an additional translation. For the
purpose of the current study, the independent variables were defined as the conditions
under which the participants either translated or copied the words, and whether the
participants were bi or trilingual. The dependent variables were the amount of words
recalled on the last page and were measured as the amount of correctly recalled words.
Thus, we expect to see the translation condition yielding a higher probability of words
recalled, and the trilingual individuals recalling more words than the bilinguals.
Dual Coding and Trilingual Implications
22
Method
Participants
There were 48 participants in total, and the linguistic breakdown of the 24
bilingual individuals who spoke both English and the following languages were: 7
French, 6 Arabic, 2 Hebrew, 2 Persian, and one of each Punjabi, Russian, Korean,
Chinese, Japanese, Hindu and Spanish. The 24 trilingual participants were fluent in
English and in the following pairs of languages: 7 Hebrew-Arabic, 7 Urdu-Punjabi, 7
Spanish-French, and one of each Spanish-Romanian, French-Chinese, and RussianBelarusian. A brief history of the participants' language acquisition and fluency were
asked in order to assure accuracy when labeling the participants as tri or bilingual. There
were 36 men and 12 women who participated, and the average ages ranged from 19-29
years, typical for Undergraduate students at the University of Western Ontario. The
participants were selected at random or through networking, mostly in the Taylor library
in the Natural Sciences building of the University of Western Ontario. They were initially
asked i f they were either bi or trilingual, and i f they responded 'yes' they were randomly
assigned to either condition (whether they had to write the words in English only or in
their other languages). For the most part, the surveys were filled out in an individual
setting, with the exception of a few friends who filled them out together. A l l participants
who started the surveys completed them, however a few individuals were unable to
translate all the words or forgot to fill out a few lines, and no participant was
compensated for their time.
Dual Coding and Trilingual Implications
23
Materials
A letter of information, a consent form, a computerized three-page survey and a
debriefing form were handed out to the participants. The first page of the questionnaire
differed in order to create the independent variable, where one condition had "Write the
translated equivalents in your language(s) on the lines below" and the other condition had
"Write the following words in English twice on the lines below" (Appendix A). The
conditions for the independent variable also had an (s) because the bilinguals wrote the
word in English once and in their other language, and the trilingual individuals filled both
lines with the two other languages and not in English. Following this question were the
following 16 words written in capital letters with two lines to fill under each: APPLE,
FLOWER, PARENTS, FOOD, DRIVE, WINDOW, DOG, TODAY, ANIMAL, CITY,
SCHOOL, TEACHER, BANANA, NORTH, COUNTRY and RESTAURANT. The
second page had 6 anagrams (ancient, western, Israeli, Palestine, Palestine, Palestine) and
this page was the same for all conditions (Appendix B). The participants were given three
minutes to complete the anagrams and then given hints for the words they had left over.
The last page of the questionnaire had 16 lines to fill in the English words the participants
were able to recall (Appendix C). There was no additional testing done for reliability or
validity.
Procedure
Individuals studying at tables in the library were approached and asked whether
they were either bi or trilingual and i f so, were randomly segregated into one of the two
experimental groups. They read over the information page, signed the consent form and
proceeded to fill out the first page of the questionnaire without knowing there would be a
Dual Coding and Trilingual Implications
24
subsequent memory task. In order to account for extraneous variables, the few
participants who filled questionnaires out simultaneously were closely monitored to
ensure that they could not see one another's sheets or were not communicating. There
were limited instructions given to the participants due to the self-explanatory nature of
the questionnaire, with the exception of being told they could not go back to a previous
page once they had turned it over.
Results
Numbers of words recalled were subjected to a two-way analysis of variance
having two subgroups (bilingual, trilingual) and two levels of encoding (English only,
translated) (Appendix D). An alpha level of .05 was used for all statistical tests. The main
effect of translating the words vs. only writing them in English yielded an F ratio of F
(1,44) = 15.88,p < .05, indicating that the mean number recalled was significantly greater
for the translating condition ( M = 10.71, SD = 2.76) than for the English only condition
(M = 6.96, SD = 3.63) (Eigure 1). The main effect of bilingualism vs. trilingualism
yielded a statistically insignificant F ratio of F (1,44) = 0.95, p> .05, indicating that the
mean number of words recalled by bilinguals ( M = 8.37, SD = 3.86) were not
significantly different than the amount recalled by trilingual individuals ( M = 9.29, SD =
3.57). The interaction effect was non-significant, F (1,44) = 0.95,p > .05. Thus, the
hypothesis was supported that translating the words aided in higher recall, however the
secondary hypothesis was not supported where there was no significant difference in
recall between bi and trilingual individuals. A Pearson correlation analysis addressed the
relationship between words recalled and translated in trilingual participants. The outcome
demonstrated a statistically insignificant positive correlation between amount of words
12
English copying
Conditions for encoding
Translation
Figure 1. Mean values for words recalled in English and translation conditions in both bilingual
(dotted line) and trilingual (solid line) participants
Dual Coding and Trilingual Implications
translated and later recalled, r (10) = M,p>
26
.05. Thus indicating that the amount of
words translated did not relate to how many would be recalled.
Discussion
The hypothesis was supported by the results of both bi and trilingual groups
where recall was far superior when the words were translated rather than copied. The
secondary hypothesis that trilingual participants would outperform bilingual participants
was not supported in either condition, and no correlation was significantly found between
amount of words translated and later recalled.
The group means did differ in the predicted fashion however not significantly,
more specifically, the trilingual participants did have a higher average of words recalled
in both the English and translated conditions, and those who translated a greater number
of words did end up recalling more however this relationship was not significant
statistically. The trilinguals outperforming bilinguals was expected since the participants
were initially asked whether they were bilingual or trilingual so many of them reported
having thought of the words in their other languages even i f they were not asked to
translate them (as was the case in the English condition). So the trilingual participants
were expected to outperform the bilingual ones simply due to the fact that they had an
additional language in which to translate and think about the word. The scores were more
widely distributed in the English condition and with the bilingual participants.
The significant results did align with previous findings such as Paivios, explaining
that two verbal systems were engaged when translating as opposed to the one being
activated during the copying task. However, we were unable to significantly prove that
Dual Coding and Trilingual Implications
27
trilingual individuals could outperform bilinguals, which would have furthered the theory
that stronger connections exist between-languages than within-languages to a theory that
encompasses trilingual individuals make more cormections than bilinguals. There were
several methodological issues that could have easily contributed to the lack of
significance in the findings, since they did vary in the predicted fashion. Such
methodological matters include the wide variation among and within the participant
groups in terms of history of second and third language acquisition, more importantly, the
selection of who participated was loosely based on their acquisition histories, specifically
the age at which bilingualism began was not generalized. This element has received
substantial attention in previous literature on bilingualism (Lambert, 1969) and emerged
as a significant factor in the second experiment in Paivio and Lamberts study where in
late bilingual individuals, the superiority of the image over the translate condition was not
significant.
It was difficult enough to encounter participants who were trilingual;
consequently, finding trilingual individuals who fit a tight criterion would have been even
harder. Linguistic background was a definite interest to Paivio and his colleagues, despite
not having offered any concrete predictions regarding this variable. Gentile and Amedt
state however, that it is possible to obtain predictions from the theory to further conduct
experiments concerning the dual coding theory and bilinguals. However, regarding the
task in the current study, and the similar one in the Paivio-Lambert experiment, Gentile
and Amedt find there should be no expectation that the translation and copying
conditions should interact with acquisition history since translating engages both
language and memory systems providing superior memory as opposed to unilingual
Dual Coding and Trilingual Implications
28
copying which engages the same linguistic coding system. They further this argument by
stating that it should be true regardless of skill, age and general use of the secondary
language so long as the precise vocabulary to be manipulated in the experiment is not
foreign to the person. Nevertheless, this notion of 'balanced bilinguals' should be taken
into account for future studies in order to rule out any possible confounds. An example
that could be used is the one employed in the Paivio and Lambert study where they
manipulated a Stroop test in order to assess the balance between both languages known to
the individual. A stroop test is one where words describing colours (such as red, blue,
green etc) are written in a different colour than the word describes. The challenge lies in
inhibiting oneself from reading the word since the goal of the task is to say to colour the
word is written in. Being able to do this task in both languages in the same amount of
time should indicate balance in the languages. Language proficiency could also be
assessed with something to the effect of self-ratings in the form of a detailed
questionnaire of language background and usage for speaking, reading, writing and
comprehension. For inclusion in the experiment, participants should be required to have a
minimal score such as 4 out of 7 (where 7 indicates a native-like competence in three of
the four modalities) like the one used in the Vaid study. Fishman and Cooper (1969) have
shown that self-ratings of proficiency have provided dependable indices of language
proficiency comparable to measures obtained with more objective tests of language
aptitude. There were also methodological issues that lay in the temporal aspect of the
questionnaire; rather than having an unlimited amount of time, there should have been
more structure due to the wide range of time spent on the reeall task among participants.
The task was also filled out under many different settings due to the simple nature of the
Dual Coding and Trilingual Implications
29
participants involved, since it was hard to find trilingual individuals, many exceptions
were made in order to have them complete the questionnaire such as online completion
and many of them were also in a hurry so they did not spend much time recalling. The
anagram task should have also been a bit easier or should be changed to a different type
of task for only one participant was able to get half of the words and the rest were given
clues. Even though the same clues were given to all the participants, this task was not
performed uniformly in a cognitive sense because many of them simply did not want to
try. There should also be a final page of the questionnaire asking whether the participants
knew they had to subsequently recall words before getting to that page because some had
an idea that it may be possible which could have confounded the results even further.
That final page should also ask them whether they thought in their other languages or not
since many reported having done so, even in the English condition. There should also
have been more words on the first page to translate or copy in order to observe a more
significant correlation between the amount of words recalled and translated. In the
trilingual condition where the words were translated, there was almost a ceiling effect
where the majority of the words were recalled. Further clarification is definitely needed
when performing experiments that are supposedly testing the dual coding theory simply
due to the vague nature of the theory itself.
Results of this study and further studies like this could yield practical tools which
multilinguals could benefit from and use to enhance recall in day-to-day tasks or while
studying. It may even create an additional reason as to why one would want to learn a
new language. These results could also aid in a further understanding of how the
multilingual brain funetions and where new languages are stored.
Dual Coding and Trilingual Implications
30
References
Amedt, C.S., & Gentile, J.R. (1986). A test of dual coding theory for bilingual memory.
Canadian Journal of Psychology, 40, 290-299.
Clark, J.M., Lambert, W.E., & Paivio, A. (1988). Bilingual dual-coding theory and
semantic repetition effects on recall. Journal of Experimental Psychology, 14,
163-172.
Cooper, R.L., & Fishman, J. A. (1969). Alternative measures of bilingualism. Journal of
Verbal Learning and Verbal Behavior, 8, 276-282.
Crial, F.I.M., & Lockhart, R.S. (1972). Levels of processing: A framework for memory
Research. Journal of Verbal Learning and Verbal Behavior, 11, 671-684.
Desrochers, A., & Paivio, A. (1980). A dual-coding approach to bilingual memory.
Canadian Journal of Psychology. 34, 388-399.
Lambert, W.E. (1969). Psychological studies of the interdependencies of the bilingual's
two languages. In J. Puhvel (Ed.), Substance and structure of language. Los
Angeles: university of California Press.
Lambert, W., & Paivio, A. (1981). Dual coding and bilingual memory. Journal of Verbal
Learning and Verbal Behavior, 20, 532-539.
Paivio, A. (1971). Imagery and verbal processes. New York: Holt, Rinehart &Winston.
Vaid, J. (1988). Bilingual memory representation: A further test of dual coding theory.
Canadian Journal of Psychology, 42, 84-90.
Dual Coding and Trilingual Implications
Appendix A
Write the words twice I N ENGLISH on the lines below.
APPLE
FLOWER
PARENTS
FOOD
DRIVE
WINDOW
DOG
TODAY
ANIMAL
CITY
SCHOOL
TEACHER
BANANA
NORTH
COUNTRY
RESTAURANT
Dual Coding and Trilingual Implications
Appendix B
Solve the following anagrams.
ANT NICE
(solution: ancient)
NEW REST
(solution: western)
SERIAL I
(solution: Israeli)
TALE PENIS
(solution: Palestine)
APE SILENT
(solution: Palestine)
ANTI SLEEP
(solution: Palestine)
Dual Coding and Trilingual Implications
Appendix C
How many English words do you remember?
Dual Coding and Trilingual Implications
34
Appendix D
Analysis of Variance Summary Table
MS
F
I
168.75
15.88
10.09
1
10.09
0.95
Interaction
0.11
1
O.II
0.01
Error
467.50
44
10.63
Total
646.45
47
Source
SS
Rows
168.75
Columns
df