Working Memory and Short-Term

Memory Abilities in Accomplished

Multilinguals

ADRIANA BIEDRON
The Pomeranian Academy in Supsk
Arciszewskiego 22a
76-200 Supsk
Poland
Email: adrianabiedron@wp.pl

ANNA SZCZEPANIAK
The Pomeranian Academy in Supsk
Arciszewskiego 22a
76-200 Supsk
Poland
Email: ania.sl@wp.pl

The role of short-term memory and working memory in accomplished multilinguals was investigated. Twenty-eight accomplished multilinguals were compared to 36 mainstream philology
students. The following instruments were used in the study: three memory subtests of the
Wechsler Intelligence Scale (Digit Span, Digit-Symbol Coding, and Arithmetic, which constitute a memory and resistance to distraction index); two short-term memory tests of the Modern
Language Aptitude Test (Part I [Number Learning] and Part V [Paired Associates]); and the
verbal Intelligence Quotient (IQ), the nonverbal IQ, the general IQ, and a working memory test, the Polish Reading Span (PRSPAN). The results of the accomplished multilinguals
were compared to the results of 1st-year English philology students (mainstream). The analysis
revealed that short-term memory and working memory abilities in the accomplished multilinguals were higher than in the mainstream philology students. Results might contribute to the
understanding of the controversial role of working memory and short-term memory abilities
in accomplished multilinguals. A suggestion that the two components of working memory
(the phonological loop and the central executive) are significant factors in determining the
outcome of learning a foreign language is discussed.

IN
THE
LITERATURE
ON
SECOND
language acquisition (SLA), it is often stated
that working memory plays a role in determining the outcome of foreign language learning
(Dornyei, 2005; Ellis, 2001; Miyake & Friedman,
1998; Robinson, 2009; Sawyer & Ranta, 2001).
Although there is a large body of research on
what combinations of individual characteristics
contribute to success in the learning of foreign
languages (i.e., Dewaele, 2007; Ehrman & Oxford, 1995; Griffiths, 2008), there is a marked
lack of research on the memory abilities of
accomplished multilinguals.
The purpose of the study reported herein was
to identify and analyze the characteristics of the
The Modern Language Journal, 96, ii, (2012)
DOI: 10.1111/j.1540-4781.2012.01332.x
0026-7902/12/290306 $1.50/0
!
C 2012 The Modern Language Journal

memory of accomplished multilinguals. (We characterized learners as accomplished multilinguals

if they spoke at least one foreign language at level
C1/C21 and a number of other languages at different levels of proficiency, varying from A1 to B2.)
In particular, the extent to which working memory
capacity can serve as a strong predictor of foreign
language learning success was investigated.

LITERATURE REVIEW
Working Memory and Short-Term Memory
Memory is probably the most significant factor in the theory of foreign language aptitude,
often called memory ability in the contemporary literature (Dornyei, 2005; Ellis, 1996; Sawyer
& Ranta, 2001). The term short-term memory
(STM) is used to describe a sort of static memory
that is held for a short period of time (less than

Adriana Biedron and Anna Szczepaniak

20 seconds). In contrast, working memory (WM)
involves the temporary storage and manipulation
of information that is necessary for the performance of a wide range of cognitive tasks (Baddeley, 2003). Hence, WM is seen as fundamentally
a form of memory, but it is more than memory,
for it is memory at work (Conway, Jarrold, Kane,
Miyake, & Towse, 2008, p. 3). WM comprises a
number of components that perform several cognitive functions that encompass mechanisms for
the storage of information and mechanisms for
executive control of information. It is the mechanisms of executive control that differentiate WM
from STM (Conway et al., 2008).
The multicomponent WM model is now accepted universally. It was formulated by Baddeley
and Hitch (1974). They originally proposed dividing memory into three subsystems: (a) the phonological loop, which processes verbal and acoustic
information, (b) the visuospatial sketchpad, which
processes visual information, and (c) the central
executive, which is a supervisory attention-limited
control system. Later, they proposed a fourth factor, the episodic buffer , which stores information
(Baddeley, 2000). In subsequent research on WM,
the findings of correlation analyses have provided
evidence that WM plays an important role in a
number of complex cognitive abilities, such as language learning, reasoning, comprehension, and
cognitive control, and that WM measures are an
indicator of intellectual ability (Kane, Conway,
Hambrick, & Engle, 2008).
The concept of WM is crucial in research on individual differences because it plays a central role
in intellectual functioning in general. It has been
shown to be relevant to many everyday tasks, such
as reading, making sense of spoken discourse,
problem solving, and mental arithmetic. Given its
importance, it is now the focus of considerable
research efforts in cognitive psychology and cognitive neuroscience (Conway et al., 2008, p. vii).
WM has limited capacity, which constrains cognitive performance. Individuals differ with respect
to their WM capacity. People with greater WM
capacity perform better on a variety of cognitive tasks, such as complex learning, reading and
listening comprehension, and reasoning, than
people with smaller WM capacity. The capacity depends on a number of variables, such as age, brain
damage, and disease in general. Older children
outperform younger children, healthy adults outperform patients with frontal-lobe damage, and,
generally, WM deteriorates from around age 65.
This variation, measured by span tests, is believed
to affect human cognitive functioning (Conway
et al., 2008).

291
Many issues regarding WM have yet to be investigated fully. What causes variation in WM performance is one of them. Several candidates for the
cause of the variation have been proposed, such
as mental speed, inhibition of attention, executive
attention, goal maintenance, and conflict resolution (Conway et al., 2008). Kane et al. (2008),
for example, proposed a theory of WM capacity
as being determined by executive attention, using individual differences among healthy young
adults as a basis. They postulate that sources of
variation are multiple, including domain-specific
skills and strategies, and a domain-general attention capability. Attention capability accounts for
the predictive validity of WM span tests and underlies other cognitive abilities, including fluid
intelligence (Gf). The issue of whether WM is domain general or domain specific remains controversial, but many arguments support the idea that
the source of variation in WM is domain general
(Conway et al., 2008).
Working Memory, Short-Term Memory, and the
General Cognitive Factor
There is an ongoing discussion about the relationship between STM, WM, and Gf (Conway
et al., 2008; Engle, Laughlin, Tuholski, & Conway,
1999; Kane et al., 2008). The first to address the
question To what extent are the terms STM and
WM different terms for the same construct and to
what extent do they refer to different but more or
less related constructs? were Engle et al. (1999,
p. 309). Having tested 135 participants for their
WM capacity, STM, and Gf, they concluded that
WM and STM are separate, but substantially correlated constructs (correlation .68). The researchers
found that the correlation was based on the
shared feature representing storage, coding, and
rehearsal, although some shared variance was also
likely to be due to executive attention. It was the
increased demand on executive attention that was
found to cause the unique, residual variance in
WM.
The accuracy of measures of WM capacity and
STM span is debatable. Kane et al. (2008) argue
that although WM span tasks are reasonably good
measures of executive attention, due to their dual
nature, a measure of WM capacity does not need
to be dual to measure control of attention. A
dual task requires participants to maintain access
to information outside of conscious awareness in
the face of proactive interference. However, some
STM span tasks seem to measure executive control, for example, some spatial STM tasks or STM
tasks that include long lists of verbal items. Due

292
to the fact that STM can hold only four items at
a time and the phonological loop can hold items
for only 2 seconds, tests with more than four items
would require some degree of executive attention, which means that, in fact, they measure WM
capacity. Given the foregoing, it would behoove
researchers to exercise great caution in analyzing
data and drawing conclusions regarding measurements of WM capacity.
Kane et al. (2008) found that only WM capacity, but not STM, predicted variance in Gf,
which suggests that greater attentional demands
of WM span tasks resulted in WMGf correlation. Researchers generally agree that variation
in WM capacity causes significant variation in
general cognitive ability and that executive attention is the central factor in this variation. No
matter how highly correlated they are, WM and
Gf are not the same (Conway et al., 2008; Kane
Wilhelm, & Sander,
et al., 2008; Oberauer, Su,
2008).
Although WM is generally considered to be domain general, some studies report a low correlation between individual verbal and spatial WM
abilities, which suggests that WM is domain specific. In this regard, Kane et al. (2008) found that
WM and STM were much more domain specific in
people in high-IQ groups than in lower IQ groups,
in that verbal and spatial WM abilities were correlated much less in high-IQ than in lower IQ
groups. They concluded that such differences between verbal and spatial WM abilities might result
from testing participants from a group in which
general cognitive ability is high and in which the
range of ability is small, for example, the majority of university students. The researchers suggest
that when the range of general cognitive ability is
narrow, any variability in cognitive performance
results from other factors, such as domain-specific
abilities, skills, or strategies.
The Role of Memory Abilities in Foreign Language
Learning
Some researchers claim that WM, in particular the phonological loop, is a significant factor determining a foreign language learning outcome (Baddeley, 2003; Baddeley, Gathercole, &
Papagno, 1998; Service, 1992). There is evidence
to suggest that both children and adults who have
poor memories (as measured by digit span and
nonword repetition) have poor language skills
(Baddeley et al., 1998).
If deficiencies in WM result in delayed language development and generally poor language
abilities in both children and adults (Gather-

The Modern Language Journal 96 (2012)

cole & Baddeley, 1990), might a talent for learning languages be the result of an exceptionally
efficient phonological loop? Research findings
suggest that it is. Papagno and Vallar (1995) compared gifted and average foreign language learners with respect to their memory abilities. The
results showed that better learners significantly
outperform poorer learners in phonological
memory tests, that is, digit span and nonword repetition, which means that their abilities to learn
new words are greater. The general intellectual
abilities, as well as nonverbal abilities, of the two
groups of learners were comparable. It may be
concluded that having a good phonological working memory makes it easier to learn unfamiliar
phonological material. The most important conclusion drawn by Baddeley et al. (1998) as a result
of Papagno and Vallars work is that [t]he case
of gifted language learners suggests that a natural talent for language learning may arise directly
as a consequence of excellent phonological loop
function (p. 166).
In line with Baddeley and his colleagues, Skehan (1998) concluded that it is excellent memory
abilities that underlie linguistic talent. They based
their conclusions on the results of research on
talented foreign language learners to the effect
that such learners do not have exceptional intelligence but all possess exceptional verbal memory (cf. Ioup, Boustagi, El Tigi, & Moselle, 1994;
Morgan, Smith, Tsimpli, & Woll, 2007; Obler,
1989; Sawyer & Ranta, 2001; Schneiderman & Desmarais, 1988).
Contemporary models of foreign language aptitude include memory as a relevant factor. Skehan (1998, 2002) proposed a Processing Stage
Model of Aptitude, in which memory is used at
basically all stages of SLA. In the first processing
stage, noticing , learners pay attention to some aspect of a language and rehearse it in WM. Noticing is considered to be an ability (cf. Robinson,
2002) and the capacity to notice differs among
learners. Skehan associates noticing with Carrolls
phonemic coding ability (Carroll & Sapon, 2002),
but he relates it primarily to WM capacity. In
the second stage, patterning , learners detect and
manipulate patterns in the foreign language. At
this stage, the learners analyze and generalize input. Skehan finds correlates to these processes
in Carrolls grammatical sensitivity and inductive language learning. However, the processes of
analyzing patterns and extrapolating from input
undoubtedly require WM and long-term memory engagement. The learning of new items
seems to be mediated by both the phonological loop and long-term phonological knowledge

Adriana Biedron and Anna Szczepaniak

(Baddeley et al., 1998). In addition, Stage 3, controlling , during which learners build a fluent linguistic repertoire by automatizing and proceduralizing information, relies on the retrieval of information from memory. Finally, Stage 4, lexicalizing , is also connected with the automatization of
language use and gaining fast access to linguistic
knowledge. Memory storage and retrieval seem to
be required for these processes. Skehans model
suggests that learners may possess different memory abilities that affect performance at the different stages of learning, which might affect the
overall learning outcome.
WM is often treated as a key component of
foreign language aptitude. For example, Miyake
and Friedman (1998) emphasize its central role
in language processing in both language production and language comprehension. Language
processing is a linear process that requires the
simultaneous integration and storage of incoming sequences of symbols (cf. Ellis, 1996). According to Miyake and Friedman, Baddeleys WM
theory matches this SLA model perfectly in that
it contains both the storage and the processing
component.
Robinson (2002, 2003, 2007, 2009) includes
memory in his Aptitude Complex Hypothesis.
He proposes four aptitude complexes, each of
which includes one or more of a variety of memory
factors. For example, aptitude for learning from
recasting contains the aptitude for noticing the gap
and memory for contingent speech. When learning
spoken language from a teacher, the learner first
utters a word or expression. Then the teacher repeats it with correct pronunciation (he recasts the
utterance). The learner holds the teachers recast
in memory (called memory for contingent speech)
and compares the teachers pronunciation with
his own, simultaneously noticing the gap between
them.
These abilities depend on the capacity and
speed of the learners WM. Robinsons conclusions are similar to Skehans. People differ with
respect to their ability to remember, which affects
their linguistic outcomes. Moreover, aptitude profiles for older learners might be more varied than
for younger learners (Robinson, 2002). Robinsons important contribution to the understanding of the role of WM in SLA is a postulate, based
on empirical evidence, that it correlates significantly and positively with language proficiency in
all learning conditions, including implicit and incidental learning. Therefore, his findings contradict the view of Krashen (1981) and Reber (1993)
that implicit and incidental learning are insensitive to individual differences.

293
A cumulative body of evidence demonstrates
a strong correlation between first language (L1)
WM capacity and second language (L2) WM capacity, and between WM capacity and L2 proficiency (Berquist, 1998; Daneman & Carpenter,
1980; Harrington & Sawyer, 1992; Mackey, Philip,
Egi, Fujii, & Tatsumi, 2002; Miyake & Friedman,
1998; Osaka, Osaka, & Groner, 1993; Robinson,
2002; Sagarra, 1998). Miyake et al. (1998) found
a positive relationship between WM, L2 linguistic knowledge, and L2 listening comprehension.
Harrington and Sawyer (1992), using a version
of Daneman and Carpenters reading span test,
reported a positive correlation between L2 WM
capacity and L2 reading proficiency. Miyake and
Friedman (1998) found a positive correlation between L1 WM capacity, L2 WM capacity, and sentence comprehension. All the aforementioned
studies, contrary to Engle et al.s (1999) findings,
reported no correlation between STM, as measured by digit span, and language proficiency or
WM. Moreover, a positive relationship was found
between L1 WM and L2 WM, which suggests that
performance on WM measures may be language
independent (Mackey et al., 2002). This finding
upholds the general vision in cognitive science
that WM is not domain specific (Kane et al., 2008).
A suggestion that was in agreement with the aforementioned studies was offered by Skehan (1982),
who did not find any correlation between STM
and learning outcomes using a span test. Having analyzed the results of research on WM (Harrington & Sawyer, 1992; Robinson, 2002), Skehan
(2002) argues that the reading span test is an efficient tool for measuring WM capacity because
it requires both storage and processing of information, thus involving the phonological loop and
the central executive.
Dornyei (2005) considers WM one of the most
promising issues in research on language aptitude. He upholds Baddeleys concept of WM,
especially its verbal component, the phonological loop, which he considers to be an ideally
suited memory construct for SLA (Dornyei, 2005,
p. 55). In addition to its STM component, it
contains an attention component, which is a significant source of individual difference in SLA
(cf. DeKeyser, 2003; Robinson, 2003). In line
with Dornyei, Ellis (2001), Miyake and Friedman (1998), and Sawyer and Ranta (2001) all
agree that WM capacity may be the concept
that will revolutionize research on language aptitude. If attention is necessary for learning and
if it is limited by WM capacity, there must be
a close relationship between WM capacity and
learning outcomes (Sawyer & Ranta, 2001). They

294
hypothesize that WM is a system that integrates all
other components of foreign language aptitude
(Sawyer & Ranta, 2001).
Summing up, there is cumulative evidence that
individual differences in WM affect SLA. Generally, there is consensus among researchers on the
adequacy of Baddeleys model of WM, with its
phonological loop and central executive components as subsystems that are decisive in learning a
foreign language.
STUDY
The purpose of the study reported herein was
to examine the role of STM and WM in accomplished multilinguals. The study consisted of two
parts. First, the accomplished multilinguals cognitive factors (STM, WM, foreign language aptitude, and IQ) were presented in order to provide
evidence of their exceptionality. Second, 28 accomplished multilinguals were compared to 36
mainstream English philology students in order
to identify differences in memory factors between
the two groups.
Hypotheses
Using the findings described in the previous
section as a basis, we formulated the following
hypotheses.
1. Accomplished multilinguals will score
higher on STM and WM tests than mainstream
philology students.
2. Accomplished multilinguals will score
higher on items with linguistic material on
STM and WM tests than mainstream philology
students.
3. A positive pattern of intercorrelations
among memory tests scores will be observed.
4. The WM score will correlate to general IQ
score.
METHODOLOGY
Operationalization of WM and STM Constructs
WM was operationalized as the ability to mentally maintain information in an active and readily accessible state, while concurrently and selectively processing new information (Conway et
al., 2008). STM was operationalized as a kind
of static memory that is held for a short period
of time (less than 20 seconds). The mechanisms
of executive control differentiate WM from STM
(Baddeley, 2003; Baddeley et al., 1998; Conway
et al., 2008; Engle et al., 1999; Kane et al., 2008).

The Modern Language Journal 96 (2012)

Participants
There were two groups of learners: accomplished multilinguals and mainstream philology
students. The accomplished multilinguals group
consisted of 28 students who were identified based
on proficiency scores, the number of languages
they had learned, language learning history, recommendation of their teachers, and the Modern Language Aptitude Test (MLAT) score. The
mainstream philology students were 1st-year English philology students at a Polish university
the Pomeranian Academy in Supsk. They were
intermediate learners of English as a foreign language.
Accomplished Multilinguals
The sample consisted of 28 accomplished multilingual foreign language learners. Twenty-one
were female and seven were male. All of them
were native speakers of Polish, but two were dominant in other languages: one in German and
one in French. They were mainly philological
students from Polish universities (n = 17), but
there were also teachers in foreign language departments at university (n = 6), school teachers
of English (n = 2), one student of mathematics,
and two translators. The participants were mainly
undergraduates and postgraduates who had been
nominated by their teachers; 6 were doctoral students. Their age varied from 21 to 28 years; the
mean age was 24. All the participants were experienced language learners. The age of onset
varied for different learners and for different languages. In the case of English, the average age
of onset was 11 years. For Chinese and Japanese,
it was generally 19 years. For other languages, it
varied from 19 to 27 years. The period of learning also varied, from 15 years to a few months.
Many of the participants had a break in their
learning. Some learned for some time and then
stopped. Some declared that they had learned a
language well, but had forgotten most of it because of lack of practice. Some learned and used
languages regularly, some others practiced them
only once a week or occasionally. The participants
learned in formal settings, such as school and university, and most of them had a chance to study
abroad. The length of their stay abroad varied
from a few weeks to 5 years. One participant was
born in Austria (German dominant, but fluent
in Polish because her mother was Polish) and
one attended primary school in Switzerland (balanced in French and Polish). Each learners language learning history was different, so it would

Adriana Biedron and Anna Szczepaniak

be difficult to categorize them into subgroups.
The level of proficiency of the sample in at least
one foreign language was advanced (C1/ C2). All
the participants were highly advanced in English.
Twenty-three were highly advanced in two foreign languages (including English), three were
fluent in three languages, one in one language,
and one in four. If they spoke more than two
foreign languages, their level of proficiency in
the additional languages was usually communicative (A2/B1+). The number of languages they
were learning varied from one to ten (four average) and included European and non-European
languages. European languages included English
(n = 28), German (n = 20), French (n = 11),
Italian, Swedish, Danish, Norwegian, Spanish,
Portuguese, Irish, Welsh, Russian, Hungarian, Romanian, Croatian, and Latin. Non-European languages encompassed Chinese (n = 10), Japanese
(n = 5), Tibetan, Hindi, Turkish, Arabic, Mongolian, Korean, and Hebrew. Altogether, the group
consisted of 1 bilingual, 13 trilinguals, 6 quadrilinguals, and 3 pentalinguals; 5 participants spoke
more than five languages, the highest number being ten languages. The terms bilingual, trilingual, and quadrilingal were adopted from Dewaele (2007) and do not refer to proficiency
levels. All the achievements were formally confirmed by official documents: certificates acknowledged in Poland and diplomas from universities2 in the case of advanced level of a language.
If elementary/communicative level was declared,
end-of-course grades were accepted as a proof
of the level of advancement. In the case of Chinese (n = 8) and Japanese (n = 3) philology
students, both their native and nonnative teachers evaluated their rate of progress in comparison to their classmates. One of the Chinese language teachers evaluated the level of Chinese
of her students according to the Council of Europe norms. The best students were nominated
to participate in the research. These measures
were undertaken to ensure the advanced level of
the participants. Only participants whose general
MLAT score placed them within at least the 95th
percentile were accepted for the research.
Mainstream Philology Students
There were 26 females and 10 males in the sample. The mean age of the mainstream philology
students was 20 years. They were monolingual Polish learners of English as a foreign language. At
the time the study was conducted, they had been
learning English for 710 years. All of them had
private lessons in addition to their regular course

295
at school. Exactly 90% of them did not practice
English in a natural setting. Their proficiency level
was generally assessed as intermediate (B1/B2).
However, individual learners varied from intermediate to advanced with respect to the levels of
proficiency at particular skills. Their speaking and
listening abilities were higher than their reading
and writing skills, while grammar was the weakest point of the majority of the learners. Their
mastery of English was sufficient to complete only
Parts 1 and 5 of the MLAT, which do not require
advanced English. The information about their
level was based on end-of-semester grades.
DATA COLLECTION AND ANALYSIS
In the first part of the study, the accomplished
multilinguals were tested. Data were collected using the Wechsler Adult Intelligence Scale (WAISR [PL]), the MLAT(Carroll & Sapon, 2002), and
a working memory test, the Polish Reading Span
(PRSPAN), which was designed by the authors of
the study. The study lasted 16 months, from February 2008 until May 2009, and was conducted at
Polish universities. The tasks were administered
over 2 days. On Day 1, the participants completed
the MLAT and the PRSPAN, and on Day 2 the
Wechsler scale. The instruments are described
below under the heading Instruments. The intelligence test was conducted by a professional
psychologist in order to provide credibility and
validity, as well as to comply with formal requirements. The data were analyzed using Pearsons
coefficient of correlation and descriptive statistics. The descriptive statistics include calculation
of means, maximal and minimal results, and standard deviations.
In the second part of the study, the mainstream
philology students were tested and compared to
the accomplished multilinguals. The mainstream
philology students completed MLAT 1, MLAT 5,
the PRSPAN, Digit Span, and Digit-Symbol Coding tests. The scores of the accomplished multilinguals on MLAT 1, MLAT 5, the PRSPAN, Digit
Span, and Digit-Symbol Coding tests were compared to the scores of the mainstream philology
students. The results were submitted to a univariate analysis of variance (ANOVA) followed by a
posthoc Tukeys Honestly Significant Difference
test. The level of significance for all analyses was
p < .05.
INSTRUMENTS
The following instruments were used in this
study.

296
Wechsler Adult Intelligence Scale (WAIS-R [PL])
This scale is an adaptation of the Weschler scale
for use with the Polish population by Brzezinski, Gaul, Hornowska, Machowski, & Zakrzewska
(1996). The Wechsler Intelligence Scales are a
series of standardized tests used to evaluate intellectual abilities in adults. The test is used to
determine vocational ability, to assess adult intellectual ability in the classroom, and to determine
neurological deficiencies. Intelligence testing requires a clinically trained examiner. The scales
should be administered, scored, and interpreted
by a trained professional, preferably a psychologist. The complete test takes 6090 minutes to
administer. The test is standardized. The scales
have a mean score of 100 and a standard deviation
of 15. The standard deviation indicates how far
above or below the norm the participants score
is. A person taking the test receives a full-scale
(general) IQ score, a verbal IQ score, and a nonverbal (performance) IQ score, as well as scaled
scores on each of the subtests.
The Polish version of the Wechsler scales is composed of eleven subtests (six verbal and five nonverbal) comprising the full test.
The verbal scale involves six subtests:
1. Similarities: abstract verbal reasoning (e.g.,
In what way are a dog and a lion alike?)
2. Vocabulary: The degree to which one has
learned to comprehend and verbally express vocabulary (e.g., What is a tomato?)
3. Information: degree of general information
acquired from culture (e.g., What city is the capital of Italy?)
4. Comprehension: ability to deal with abstract
social conventions (e.g., Why should we not beat
children?)
5. Arithmetic: ability to solve mental mathematical problems. It tests working memory, attention, and numerical reasoning (e.g., How many
months are in three-quarters of a year?)
6. Digit Span: attention, concentration, and
mental control. In this subtest, participants are
given sets of digits to repeat initially forward then
backward (e.g., Repeat the numbers 2, 4, 9 in
reverse order).
The nonverbal scale involves five subtests:
1. Digit-Symbol Coding : visual motor speed and
short-term visual memory. The subtest involves
copying a coding pattern. Symbols are matched
with numbers according to a key.
2. Block Design: spatial perception, abstract visual processing, and problem solving.

The Modern Language Journal 96 (2012)

3. Picture Completion: ability to perceive visual
details quickly.
4. Object Assembly: the ability to create a whole
by discovering relations between elements. Involves jigsaw-type puzzles.
5. Picture Arrangement: pattern recognition. Involves arranging pictures into a logical sequence.
At the second stage of the analysis, the scores for
the three indices of verbal comprehension, perceptual organization, and memory and resistance
to distraction are determined. For the purpose
of this study, the subtests Digit Span, Digit-Symbol
Coding, and Arithmetic were analyzed. These subtests constitute a memory and resistance to distraction index.
Split-half reliabilities for the WAIS-R (PL) were
.88.93 for the full scale, .86.91 for the verbal
scale, and .79.88 for the nonverbal scale, depending on age. The validity coefficients (correlations
with other intelligence tests, e.g., Ravens Matrices) were .39.60, depending on a test for the full

scale (Brzezinski
et al., 1996).
Modern Language Aptitude Test (MLAT)
The MLAT (Carroll & Sapon, 2002) is a language aptitude test that is useful for predicting
success in learning a foreign language (Skehan,
1998). The MLAT is entirely in English and is
suitable for native and near-native speakers of English. It measures aptitude traits by five scores:
1. Number Learning : measures verbal STM, in
particular, auditory alertness, which might play
a role in auditory comprehension of a foreign
language.
2. Phonetic Script: measures the ability to associate sounds with symbols, that is, the ability to
learn the correspondence between speech sounds
and orthographic symbols. It also measures memory for speech sounds and the ability to mimic
speech sounds.
3. Spelling Clues: this partly measures the examinees native vocabulary knowledge and partly the
ability to associate sounds with symbols, but to a
lesser extent than subtests 1 and 2.
4. Words in Sentences: measures sensitivity to
grammar structure and the students ability to
learn the grammar of a foreign language.
5. Paired Associates: measures the rote memory
aspect of foreign language learning.
Split-half reliabilities for the MLAT were
.92.97, depending on the grade or age.
For college students, the validity coefficients
(correlations with course grades) provided in the

Adriana Biedron and Anna Szczepaniak

MLAT Manual (Carroll & Sapon, 2002) were
.18.69.
The MLAT was used for two reasons. First, all
the participants were highly advanced in English.
Second, the MLAT is considered the best available predictor of language learning success and
of extremely good and bad language learners
(Ehrman, 1998). The U.S. version was used because there is no Polish version of the MLAT,
or its equivalent. For the purpose of this study,
subtests 1 and 5 and the general score were analyzed.
Polish Reading Span (PRSPAN)
The PRSPAN is a Polish adaptation of the American Reading Span (RSPAN) (Engle et al., 1999),
designed by the authors of the study. The RSPAN
is referred to as a prototypical WM test, which is a
modified version of the reading span task (Daneman & Carpenter, 1980). The reading span test is
considered to be a valid and reliable instrument
for measuring WM capacity (Conway et al., 2008).
In accordance with Daneman and Carpenters
(1980) classical reading span test, the PRSPAN is a
dual task that requires the participant to read a series of sentences and, simultaneously, keep track
of the last word displayed, so that the words can
be recollected later.
The test comprises eight sets of sentences,
which contain three, four, five, six, seven, eight,
nine, and ten sentences in Polish, respectively.
Some of the sentences are sensible in the context of everyday life and others are not, but all are
grammatically correct. The length of each sentence is approximately ten words. There is an unrelated word at the end of each sentence, which is
a two-syllable noun. Participants read aloud sentences that are shown on the monitor, while trying to remember the unrelated words at the end
of the sentence. The sentences are displayed one
sentence at a time. A participant reads the sentence aloud and marks whether the sentence is
acceptable (A) or unacceptable (U) in an answer
sheet. Beforehand, the participants are told that
the criterion that they should use for determining
whether or not a sentence is acceptable is whether
they think it fits easily into everyday communication. For example, the sentence It is a nice day
today is acceptable, whereas the sentence A frog
said that it is a nice day today is not. The participants are asked to judge the acceptability or otherwise of the sentences to ensure that they attend
to the sentences. After each sentence, the experimenter presses a key that causes the next sentence
to be presented. The sentences are presented at

297
3-second intervals. After the last sentence in each
set, a blank slide is displayed as a cue for the participant to write down the words he remembered
in the answer sheet. The sentences are presented
from the shortest set (three) to the longest (ten)
one. The PRSPAN score is the cumulative number
of words recalled perfectly in all the trials.
For example:
.
Od wielu lat jego zona
i rodzina pracowali na farmie.
RYBA
(His wife and family have worked at a farm for many years.
FISH)
Poniewaz. byo duszno wyszam na dwor zaczerpna
!c
.
troche! s wiezego
powietrza. WATA
(Because it was stuffy I went out to get some air.
COTTON)
Jedzenie warzyw i czasopism bogatych w witaminy
sprzyja odpornosci organizmu. DRZEWO
(Eating vegetables and magazines rich in vitamins improves natural immunity. TREE)

The validity of the test was assessed by seven

competent judges. Highly convergent results were
interpreted as evidence that the judges agreed on
whether the participants answers met the criteria
of evaluation. The competent judges came from
two groups. In one group were four cognitive psychologists who conduct research in the field of
visual processing, memory, problem solving, and
language. They are practicing professionals who
are familiar with the methodology of psychological testing. In the other group were three members of faculty in the Philological Department of
Gdansk University who do research on applied
linguistics. The data were analyzed using Friedmans ANOVA and Kendalls coefficient of concordance. The Kendalls coefficient of concordance for the whole test was 0.898, p < .05. All
four of the psychologists decided that the test is
a valid measure of WM (the mean result on a
scale of 1 to 7 was 6.75). The concordance of the
judges was very high, which strongly indicates that
the test is valid.
The concurrent validity of PRSPAN was tested
with four standard STM tests. The PRSPAN correlated significantly with all the memory tests. The
correlations with MLAT 1, MLAT 5, Digit-Symbol
Coding, and Digit Span were .45, .47, .30, and .49,
respectively, p < .05.
A factor analysis was conducted to verify the
hypothesis that more than one factor is needed
to explain the variance among the test results.
The factor analysis (varimax) indicated that one
factor, memory, was present in all the tests: WM
(PRSPAN), MLAT 1, MLAT 5, Digit-Symbol Coding, and Digit Span. This factor accounted for

298

The Modern Language Journal 96 (2012)

TABLE 1
Descriptive Statistics for the Accomplished Multilinguals (n = 28)
MLAT 1
MLAT 5
MLAT G
WM
D_SPAN
ARITHM
DS_COD
MEM
IQ VERB
IQ NONV
IQ G

Mean

Minimum

Maximum

40.892
22.571
161.428
40.357
15.392
14.428
14.428
130.142
129.892
118.321
125.357

28.000
16.000
149.000
23.000
10.000
8.000
11.000
102.000
113.000
96.000
108.000

43.000
24.000
178.000
52.000
19.000
18.000
19.000
147.000
145.000
137.000
139.000

SD
3.247
1.989
9.378
7.171
2.499
2.588
2.044
9.800
8.891
9.951
8.215

Note. MLAT G = general score, WM = PRSPAN, D_SPAN = Digit Span, ARITHM = Arithmetic, DS_ COD =
Digit-Symbol Coding, MEM = Memory and resistance to distraction, IQ VERB = Verbal IQ, IQ NONV =
nonverbal IQ, IQ G = general IQ, SD = standard deviation.

52% of the variance altogether, which means that

all the tests measure a similar factormemory.
To establish whether the PRSPAN is reliable,
the testretest method was applied. The correlation between the test and the retest, which took
place 2 weeks apart, was 0.89 (n = 36), p <
.05. The Cronbach alpha for the reliability of
internal consistency for the PRSPAN was .69 and
the standardized item alpha was .76.
RESULTS AND DISCUSSION
The descriptive statistics for the accomplished
multilinguals are presented in Table 1.
The results presented in Table 1 distinctly show
that the accomplished multilinguals have strong
ability. Their mean MLAT score is 161.4 (99th
percentile). The minimal result is 149 (95th percentile), the maximal 178 (99th percentile, 92%
of correct answers). In comparison to the results obtained by Ehrman (1998) in research on
the language aptitude of participants in intensive
language courses, these results are very high. In
Ehrmans research, the mean general score for
the participants who were the best at speaking
(14 persons selected out of 295) was 151.2.
The results for all the memory tests indicate
high, although not homogenous, abilities in this
field. Very high results were obtained in MLAT
1 (95.3%), MLAT 5 (94.6%), and WM (76.5%).
The results for the Wechsler memory subscales
are high, hence, the memory and resistance to
distraction index is also high, but not as high as in
the MLAT results. The verbal intelligence of the
participants is high, and higher than nonverbal intelligence. The results for the PRSPAN that mea-

sured WM capacity are varied (Mean = 40.357,

SD = 7.171). However, some of the participants
managed to recollect about 90% of the words,
the highest result being 100%. This score was
recorded by 2 participants. One of these was a 28year-old university lecturer, a polyglot working in a
Sinology department. This person knows one language at level C2 (near-native) (Chinese), three
languages at level C1 (advanced) (English, Turkish, and German), and is able to communicate in
six other languages: Japanese, Hebrew, Swedish,
Russian, Croatian, and French. It is worth mentioning that she started to learn Chinese at the
age of 19. The other participant was a 21-yearold Japanese philology student, learning six languages. Her MLAT score was 175 (91% of correct
answers).
The PRSPAN was correlated (using a Pearson
correlation) with standard STM tests: the first
part of the MLAT (Number Learning), the fifth
part of the MLAT (Paired Associates), and the
memory subtests of the Wechsler Adult Intelligence Scale, that is, Digit-Symbol Coding, Digit
Span, and Arithmetic, which compose the memory and resistance to distraction index. Moreover, the PRSPAN results were correlated with the
verbal and nonverbal (performance) subscales,
as well as with the general intelligence quotient
and the general MLAT score. The results for
the accomplished multilinguals are presented in
Table 2.
As expected, there is a general pattern of
positive intercorrelations among measures. The
PRSPAN correlates positively with MLAT 1 and
MLAT 5, Digit Span, memory and resistance
to distraction index, and verbal and nonverbal

Adriana Biedron and Anna Szczepaniak

299

TABLE 2
Correlation Matrix for the Accomplished Multilinguals (n = 28)
Test
MLAT 1
MLAT 5
MLAT G
WM
D_SPAN.
ARITHM
DS_ COD
MEM
IQ VERB
IQNONV
IQ

MLAT 1 MLAT 5 MLAT G

1.00
.02
.48
.40
.45
.60
.04
.59
.33
.51
.49

.48
.35
1.00
.23
.18
.52
.09
.43
.28
.26
.33

.02
1.00
.35
.40
.21
.37
.16
.40
.29
.25
.32

DS_
D_SPAN ARITHM COD

WM
.40
.40
.23
1.00
.49
.25
.11
.38
.40
.49
.53

.45
.21
.18
.49
1.00
.48
.15
.77
.69
.20
.55

.60
.37
.52
.25
.48
1.00
.11
.79
.68
.40
.65

.04
.16
.09
.11
.15
.11
1.00
.26
.08
.27
.09

MEM

IQ
VERB

IQ
NONV

IQ

.59
.40
.43
.38
.77
.79
.26
1.00
.74
.46
.72

.33
.29
.28
.40
.69
.68
.08
.74
1.00
.45
.87

.51
.25
.26
.49
.20
.40
.27
.46
.45
1.00
.83

.49
.32
.33
.53
.55
.65
.09
.72
.87
.83
1.00

Note. MLAT G = general score, WM = PRSPAN, D_SPAN = Digit Span, ARITHM = Arithmetic, DS_ COD =
Digit-Symbol Coding, MEM = memory and resistance to distraction, IQVERB = verbal IQ, IQNONV =
nonverbal IQ, IQ = general intelligence quotient.

p <. 05.
TABLE 3
Means and Standard Deviations for the Accomplished Multilinguals and the Mainstream Philology Students
(n = 64)
MLAT 1

Accomplished
Mainstream
All

MLAT 5

WM

D_SPAN

DS_COD

Mean

SD

Mean

SD

Mean

SD

Mean

SD

Mean

SD

40.892
31.805
35.781

3.247
8.481
8.070

22.571
16.750
19.296

1.989
5.748
5.341

40.357
26.944
32.812

7.171
8.860
10.518

15.392
12.750
13.906

2.499
2.442
2.781

14.428
12.361
13.265

2.044
2.002
2.255

Note. D_SPAN = Digit Span, DS_COD = Digit-Symbol Coding, WM = working memory.

IQ. The highest correlation observed is with

general IQ, at .53. In addition, the memory and
resistance to distraction index shows high and
very high correlations with all the measures.
These results are in agreement with previous
research results (Engle et al., 1999) and the
theory of the working memory capacity, which
states that WM shares a common field with STM
and is related strongly to intelligence through the
attentional component (central executive)
(Conway et al., 2008; Engle et al., 1999, Kane
et al., 2008).
The PRSPAN correlates with memory subscales
of the aptitude test, the MLAT. Part 1 (Number Learning) requires of the testee to memorize twelve numbers in an artificial language and
then recollect them in three-digit combinations,
such as 123, or 401 in a listening task. This activity seems to require a reasonable amount of
attention; more, for example, than a digit span
task. Listening recall is regarded as the central executive measure (Pickering & Gathercole, 2001).
In contrast, Part 5 (Paired Associates) is a classi-

cal STM task that requires the testee to memorize twenty-four words of an unknown language.
Digit Span, which is considered to be a standard
STM task, also correlates with the WM capacity.
This task requires repeating strings of digits after the researcher, forward and backward. Both
digit recall and word list recall are considered to
be measures of the phonological loop (Pickering
& Gathercole, 2001). It is likely that backward
repetition engages the central executive more
than forward repetition. Arithmetic is considered
to be a measure of WM in the U.S. version of
the Wechsler scale. Thus, it is correlated strongly
with memory and resistance to distraction and
with intelligence. WM correlates moderately positively with Arithmetic. However, the correlation
is not statistically significant. The lack of significance might be due to the specificity of the tasks:
The PRSPAN includes lexical material, whereas
Arithmetic draws on the ability to do mathematical calculations. Generally, Arithmetic is the
weakest component for the accomplished multilinguals (Table 1), which significantly lowered

300

The Modern Language Journal 96 (2012)

TABLE 4
One-Way ANOVA of Differences Between the Accomplished Multilinguals and the Mainstream Philology
Students (p < .05)
Test

Sum of Squares

MLAT 1
MLAT 5
WM
D_SPAN
DS_COD

Test of Significance

Effect

Error

F(1;58)

1300.62
533.75
2833.43
110.01
67.32

2802.32
1263.61
4136.32
377.43
253.16

28.776
26.189
42.471
18.071
16.487

0.001
0.001
0.001
0.001
0.001

Note. D_SPAN = Digit Span, DS_COD = Digit-Symbol Coding, WM = working memory.

FIGURE 1
Accomplished MultilingualsMainstream Philology Students Group Differences for Working Memory
WORKING MEMORY
12

10

Number of obs.

0
10

15

20

25

30

35

40

45

50

55

LEARNER: AM

their memory and resistance to distraction index. In addition, WM does not correlate significantly with the general MLAT score. It seems
that Part 3 of the MLAT, which requires knowledge of English vocabulary, and Part 4, which
measures sensitivity to grammar, and thus relies
on analytical ability, contributed to the lack of
correlation.
The only subscale that does not correlate with
other measures is Digit-Symbol Coding. This task
measures visual memory and visual motor rate
of learning, which are abilities that are probably not connected directly to foreign language
abilities.

60 10

15

20

25

30

35

40

45

50

55

60

LEARNER: MPS

Group Differences Analysis

The results of the accomplished multilinguals
were compared to the results of 36 mainstream
philology students. It was decided to compare the
results of three tests that are based on linguistic
material, that is, the PRSPAN, MLAT 1, and MLAT
5, and the results of two tests that are based on numerical material, that is, Digit-Symbol Coding and
Digit Span. The purpose of this comparison was
to demonstrate whether there are any differences
in the memory abilities between the two groups of
learners and whether the differences depend on
the type of material in a test.

Adriana Biedron and Anna Szczepaniak

301

FIGURE 2
Accomplished MultilingualsMainstream Philology Students Group Differences for MLAT 1

MLAT 1
25

Number of obs.

20

15

10

-5

10

15

20

25

30

35

40

45

50 -5

10

15

20

25

30

35

40

45

50

22

24

26

LEARNER: MPS

LEARNER: AM

FIGURE 3
Accomplished MultilingualsMainstream Philology Students Group Differences for MLAT 5
MLAT 5
20

Number of obs.

15

10

10

12

14

16

18

LEARNER: AM

20

22

24

26 4

10

12

14

16

18

LEARNER: MPS

20

302

The Modern Language Journal 96 (2012)

FIGURE 4
Accomplished MultilingualsMainstream Philology Students Group Differences for Digit Span
DIGIT SPAN
8

Number of obs.

9 10 11 12 13 14 15 16 17 18 19 20 7

LEARNER: AM

9 10 11 12 13 14 15 16 17 18 19 20
LEARNER: MPS

FIGURE 5
Accomplished MultilingualsMainstream Philology Students Group Differences for Digit-Symbol Coding
DIGIT SYMBOL
16
14
12

Number of obs.

10
8
6
4
2
0

10 11 12 13 14 15 16 17 18 19 20 8
LEARNER: AM

10 11 12 13 14 15 16 17 18 19 20
LEARNER: MPS

Adriana Biedron and Anna Szczepaniak

303

FIGURE 6
Mean Scores for Accomplished Multilinguals and Mainstream Philology Students for MLAT 1, MLAT 5, WM,
Digit Span, and Digit-Symbol Coding
Means
45
40
35

Values

30
25
20
15
10
5

AM

MPS

MLAT_1
MLAT_5
WM
D_SPAN
DS_COD

LEARNER

Note. D_SPAN = Digit Span, DS_COD = Digit-Symbol Coding, WM = working memory.

The results of the analysis showed that despite

the high level of differentiation of the group
(ages, languages, and professions), the standard
deviations of the accomplished multilinguals are
lower than those of the mainstream philology
students for memory tests that are based on
linguistic material, which indicates a balanced
high level among members of the accomplished
multilinguals group. As expected, there are big
differences in the memory abilities of the two
groups, with those of the accomplished multilinguals being higher (Hypothesis 1). The difference
in the PRSPAN (WM) is significant: 24.7%. The
results for comparison are presented in Table 3.
The mean score for the accomplished multilinguals for MLAT 1 is 41 points, which is close to
the maximum score of 43. The result of MLAT 1
for the mainstream philology students is significantly lower: 32 points. The same pattern occurs
for MLAT 5. The results are higher for the accomplished multilinguals: 23 points; and lower for the
mainstream philology students: 17 points (max
24). The biggest difference between the accomplished multilinguals and the mainstream philology students is for WM, as measured by PRSPAN:

40 and 27, respectively. Differences between the

mean scores for Digit Span and Digit-Symbol Coding are less striking.
A statistical analysis one-way ANOVA was conducted in order to estimate differences in
memory factors between the two groups. The
one-way ANOVA revealed statistically significant
differences between the two groups of learners
with respect to all the memory measures. The
biggest differences were for lexical tests: WM,
MLAT 1, and MLAT 5. The differences were
smaller for tests based on numerical material:
Digit Span and Digit-Symbol Coding. The results
are presented in Table 4.
Differences between WM scores for the two
groups are presented in the histogram in Figure 1,
differences for MLAT 1 in Figure 2, differences for
MLAT 5 in Figure 3, differences for Digit Span in
Figure 4, and differences for Digit-Symbol Coding
in Figure 5. Figure 6 illustrates the mean scores
for the accomplished multilinguals and the mainstream philology students for MLAT 1, MLAT 5,
WM, Digit Span, and Digit-Symbol Coding.
The posthoc Tukeys Honestly Significant Differences tests showed significant differences

304
between the accomplished multilinguals and
mainstream philology students as follows: for WM:
mean multilinguals = 40.357, mean mainstream
= 26.944, p = 0.001; for MLAT 1: mean multilinguals = 40.893, mean mainstream = 31.806,
p = 0.001; for MLAT 5: mean multilinguals =
22.571, mean mainstream = 16.750, p = 0.001; for
Digit Span: mean multilinguals = 15.393, mean
mainstream = 12.750, p = 0.001; and for DigitSymbol Coding: mean multilinguals = 14.429,
mean mainstream = 12.361, p = 0.001.
It is evident that both WM and STM abilities
are much higher in the accomplished multilinguals than in the mainstream philology students.
The differences are especially high for memory
tests based on linguistic material, in particular for
the WM test, which is in Polish, and for which,
therefore, the results were not influenced by a
knowledge of English.

CONCLUSION
The analysis shows that the WM and STM abilities of the sample of accomplished multilinguals
are greater than those of the sample of mainstream philology students. The mean memory
and resistance to distraction index of the accomplished multilinguals is high (129.83). Thus,
Hypothesis 1 is confirmed. The accomplished
multilinguals obtained higher scores than the
mainstream philology students on memory tests
that are based on linguistic material (MLAT 1,
MLAT 5, PRSPAN) than on tests based on numerical material (Digit Span, Digit-Symbol Coding, Arithmetic). Thus, Hypothesis 2 is confirmed.
The results of the present study are in agreement
with Kane et al.s (2008) hypothesis that WM and
STM are much more domain specific in people
with high IQ, for example, university students,
than in those from lower IQ groups. The group
of accomplished multilinguals comprised highly
selected foreign language learners, characterized
by high general and high verbal IQ.
There was a positive pattern of intercorrelations
among the memory tests, which confirms Hypothesis 3. These results are in line with Engle et al.s
(1999) conclusion that neither STM nor WM tasks
are pure reflections of these constructs. To the extent that STM tasks demand controlled attention,
they also reflect the WM construct. Moreover, the
test results reflect individual differences between
the participants with regard to their intelligence
or level of cognitive development. As a result, what
is clearly a STM task for one participant could be
a WM task for another.

The Modern Language Journal 96 (2012)

As far as Hypothesis 4 is concerned, the WM
score correlated significantly with verbal IQ, nonverbal IQ, and general IQ scores. This result supports the view that WM is a significant source of
variation in general cognitive ability and that the
factor of executive attention is the central aspect
of this variation (Conway et al., 2008; Kane et al.,
2008; Oberauer et al., 2008).
According to Conway et al. (2008), WM is a
structure comprising a number of components
and that performs several cognitive functions.
Those cognitive functions encompass mechanisms for storage of information (the phonological loop) and mechanisms for cognitive control
(the central executive). The mechanisms of executive control differentiate WM from STM. Given
that the accomplished multilinguals gained generally high scores in all memory tests, we may
conclude that they have superior memory abilities with respect to the phonological loop and the
central executive component. At this point, it cannot be determined whether the superior memory
abilities of the accomplished multilinguals are the
cause of their achievement or if it is, rather, their
experience in foreign language learning that predisposed them toward a certain kind of memory
tasks, that is, those that involve linguistic material. Given that, in general, memory improves with
practice, it is likely that the two factors are intertwined.
The results confirm Robinsons (2003) point
of view that WM capacity is a powerful factor in
determining SLA outcome. Further research on a
bigger sample of learners and analyzing other variables, such as the number of languages learned,
the proficiency level in particular languages, and
the age of onset, might extend our understanding
of the effect of WM capacity on success in SLA.

NOTES
1 Levels

of advancement are described in the Common European Framework of Reference for Languages: Learning, Teaching, Assessment (CEFR). issued by the Council
of Europe (Retrieved from http://www.coe.int/t/dg4/
linguistic/Source/Framework_EN.pdf). Levels A1 and
A2 denote elementary level, which means the ability to
communicate in simple, everyday situations. Level B1
denotes intermediate level, which is limited operational
proficiency; B2 means that a person can fluently and
spontaneously communicate with a native speaker in a
natural conversation. Levels C1 and C2 indicate proficiency level. C1 refers to the ability to communicate
complex, difficult, and detailed messages. A person at
level C1 can study in the target language. C2 mastery
level denotes near-native proficiency.

Adriana Biedron and Anna Szczepaniak

2 At Polish universities, modern European languages
are taught and evaluated according to the Council of
Europe norms. It might, therefore, be assumed that a
person who has graduated from the best Polish universities, such as Warsaw University, Gdansk University, the
Jagiellonian University in Krakow, or Adam Mickiewicz
in a European language and with
University in Poznan,
the highest grade, will be at level C1 or C2. In the case
of Oriental languages, the norms can be less transparent, due to the specificity of the process of learning. In
addition, students usually start learning at the 1st year
of studies, from level 0. In the case of Chinese (to use
an example), the HSK exam result can be taken into
account, as well as the rate of progress, as evaluated by
teachers.

REFERENCES
Baddeley, A. D. (2000). The episodic buffer: A new component of working memory? Trends in Cognitive
Sciences, 4, 417123.
Baddeley, A. D. (2003). Working memory and language:
An overview. Journal of Communication Disorders,
36 , 189208.
Baddeley, A. D., Gathercole, S., & Papagno, C. (1998).
The phonological loop as a language acquisition
device. Psychological Review, 105, 158173.
Baddeley, A. D., & Hitch, G. J. (1974). Working memory.
In G. Bower (Ed.), The psychology of learning and
motivation (vol. 8, pp. 4790). New York: Academic
Press.
Berquist, B. (1998, March). Working memory models applied to second language acquisition. Paper presented at the annual conference of the American
Association of Applied Linguistics, Seattle, WA.
Brzezinski, J., Gaul, M., Hornowska, E., Machowski, A.,
& Zakrzewska, M. (1996). Skala inteligencji D. Wechslera dla dorosych. Wersja zrewidowana. WAISR
(Pl). Podrcznik. [Wechsler adult intelligence scale. A
revised version. WAIS-R (Pl). Manual ]. Warszawa:
Psychologicznych PTP.
Pracownia Testow
Carroll, J. B., & Sapon S. (2002). Modern Language Aptitude Test. MLAT. Manual 2002 edition. N. Bethesda,
MD: Second Language Testing.
Conway, A. R. A., Jarrold, C., Kane, M. J., Miyake, A., &
Towse, J. N. (2008). Variation in working memory:
An introduction. In A. R. A. Conway, C. Jarrold,
M. J. Kane, A. Miyake, & J. N. Towse (Eds.), Variation in working memory (pp. 317). Oxford: Oxford
University Press.
Daneman, M., & Carpenter, P. (1980). Individual differences in working memory and reading. Journal of
Verbal Learning and Verbal Behavior , 19 , 450466.
DeKeyser, R. (2003). Implicit and explicit learning. In
C. J. Doughty & M. H. Long (Eds.), The handbook of
second language acquisition (pp. 313349). Oxford:
Blackwell.
Dewaele, J. M. (2007). The effect of multilingualism,
sociobiographical, and situational factors on communicative anxiety and foreign language anxiety

305
on mature language learners. International Journal
of Bilingualism, 11, 391409.
Dornyei, Z. (2005). The psychology of the language learner .
Mahwah, NJ: Erlbaum.
Ehrman, M. (1998). The Modern Language Aptitude
Test for predicting learning success and advising
students. Applied Language Learning , 9 (12), 31
70.
Ehrman, M., & Oxford, R. L. (1995). Cognition plus:
Correlates of language learning success. Modern
Language Journal , 7 , 6789.
Ellis, N. (1996). Sequencing and SLA: Phonological
memory, chunking and points of order. Studies in
Second Language Acquisition, 18, 91129.
Ellis, N. (2001). Memory for language. In P. Robinson (Ed.), Cognition and second language instruction (pp. 3368). New York: Cambridge University
Press.
Engle, R. W., Laughlin, J. E., Tuholski, S. W., & Conway, A. R. A. (1999). Working memory, short-term
memory, and general fluid intelligence: A latentvariable approach. Journal of Experimental Psychology: General , 128, 309331.
Gathercole, S. E., & Baddeley, A. D. (1990). Phonological memory deficits in language disordered children: Is there a causal connection? Journal of Memory and Language, 29 , 336360.
Griffiths, C. (Ed.). (2008). Lessons from good language learners. Cambridge: Cambridge University
Press.
Harrington, M., & Sawyer, M. (1992). L2 working memory capacity and L2 reading skill. Studies in Second
Language Acquisition, 14, 2538.
Ioup, G., Boustagi, E., El Tigi, M., & Moselle, M. (1994).
Re-examining the CPH: A case study of successful
adult SLA in a naturalistic environment. Studies in
Second Language Acquisition, 16 , 7398.
Kane, M. J., Conway, A. R. A., Hambrick, D. Z., & Engle,
R. W. (2008). Variation in working memory capacity as variation in executive attention and control. In A. R. A. Conway, C. Jarrold, M. J. Kane, A.
Miyake, & J. N. Towse (Eds.), Variation in working
memory (pp. 2149). Oxford: Oxford University
Press.
Krashen, S. D. (1981). Second language acquisition and
second language learning . Oxford: Pergamon Press.
Mackey, A., Philp, J., Egi, T., Fujii, A., & Tatsumi, T.
(2002). Individual differences in working memory, noticing interactional feedback and L2 development. In P. Robinson (Ed.), Individual differences and instructed language learning (pp. 181
209). Philadelphia: Benjamins.
Miyake, A., & Friedman, N. P. (1998). Individual differences in second language proficiency: Working
memory as language aptitude. In A. Healy & L.
Bourne (Eds.), Foreign language learning (pp. 339
364). Mahwah, NJ: Erlbaum.
Morgan, G., Smith, N., Tsimpli, I., & Woll, B. (2007).
Classifier learning and modality in a polyglot savant. Lingua, 117 , 13391353.
H., Wilhelm, O., & Sander, N.
Oberauer, K., Su,
(2008). Individual differences in working memory

306
capacity and reasoning ability. In A. R. A. Conway,
C. Jarrold, M. J. Kane, A. Miyake, & J. N. Towse
(Eds.), Variation in working memory (pp. 4976).
Oxford: Oxford University Press.
Obler, L. (1989). Exceptional second language learners.
In S. Gass, C. Madden, D. Preston, & L. Selinker
(Eds.), Variation in second language acquisition: Psycholinguistic issues (pp. 141159). Clevedon, England: Multilingual Matters.
Osaka, M., Osaka, N., & Groner, R. (1993).
Language-independent working memory: Evidence from German and French span tests.
Bulletin of the Psychonomic Society, 31, 117
118.
Papagno, C., & Vallar, G. (1995). Verbal short-term memory and vocabulary learning in polyglots. Quarterly Journal of Experimental Psychology, 38A, 98
107.
Pickering, S., & Gathercole, S. (2001). The working memory test battery for children. London: Psychological
Corporation.
Reber, A. S. (1993). Implicit learning and tacit knowledge:
An essay on the cognitive unconscious. Oxford, England: Clarendon Press.
Robinson, P. (2002). Learning conditions, aptitude complexes and SLA: A framework for research and
pedagogy. In P. Robinson (Ed.), Individual differences and instructed language learning (pp. 113
133). Philadelphia: Benjamins.
Robinson, P. (2003). Attention and memory during SLA.
In C. J. Doughty & M. H. Long (Eds.), The handbook of second language acquisition (pp. 631679).
Oxford: Blackwell.
Robinson, P. (2007). Aptitudes, abilities, contexts, and
practice. In R. M. DeKeyser (Ed.), Practice in second

The Modern Language Journal 96 (2012)

language (pp. 256286). Cambridge: Cambridge
University Press.
Robinson, P. (2009). Individual differences, aptitude
complexes, SLA processes and aptitude test development. In Pawlak, M. (Ed.), Studies in pedagogy
and fine arts. New perspectives on individual differences in language learning and teaching (vol. 8, pp.

89109). Poznan-Kalisz:
Faculty of Pedagogy and
Fine Arts in Kalisz Adam Mickiewicz University in

Poznan.
Sagarra, N. (1998, March). The role of working memory in
adult L2 development: A longitudinal study. Paper
presented at the annual conference of the American Association for Applied Linguistics, Seattle,
WA.
Sawyer, M., & Ranta, L.(2001). Aptitude, individual differences, and instructional design. In P. Robinson (Ed.), Cognition and second language instruction
(pp. 319354). Cambridge: Cambridge University
Press.
Schneiderman, E. I., & Desmarais, C. (1988). The talented language learner: Some preliminary findings. Second Language Research, 4, 91109.
Service, E. (1992). Phonology, working memory, and
foreign language learning. Quarterly Journal of Experimental Psychology, 45A, 2150.
Skehan, P. (1982). Memory and motivation in language aptitude testing (Unpublished doctoral dissertation).
University of London.
Skehan, P. (1998). A cognitive approach to language learning . Oxford: Oxford University Press.
Skehan, P. (2002). Theorising and updating aptitude.
In P. Robinson (Ed.), Individual differences and instructed language learning (pp. 6995). Philadelphia: Benjamins.