Statistics Question

Read Larson-Hall, 2008, Shaaban & Ghaith, 2000, and Valeo, 2013:

For each study create a summary table of what kinds of statistical analyses were done and what was the purpose of doing each of these statistical analyses. Please make sure to include descriptive statistics and reliability as well. I don’t want you to include details about each of the analyses. Just name them and explain WHY they were used.     
082981_SLR_035-064.qxd
12/12/07
3:03 PM
Page 35
Second Language Research 24,1 (2008); pp. 35–63
Weighing the benefits of studying a
foreign language at a younger starting
age in a minimal input situation
Jenifer Larson-Hall
University of North Texas
Received July 2006; revised December 2006; accepted January 2007
This study examined whether a younger starting age is advantageous in a situation of minimal exposure to an instructed foreign
language (ⱕ4 hours classroom contact per week). Previous theoretical and empirical studies indicated there should be no advantage
for an earlier start. Japanese college students who started studying
English between ages three and twelve (n ⫽61) were examined on
a phonemic discrimination (ɹ/l/w) and grammaticality judgement
task (GJT). After controlling for language aptitude and amount of
input, statistical correlations were found between starting age and
scores on the GJT (r ⫽– .38) but not the phonemic task (r ⫽.03).
These earlier starters were also compared to peers who began study
in junior high at age twelve or thirteen (n ⫽139) on the same measures. The earlier starters were found to score statistically higher on
the phonemic but not morphosyntactic measure, and this remained
true in an ANCOVA analysis where total amount of hours of study
input were controlled for. A robust ANCOVA testing for differences
at different levels of input found interesting interactions between
group affiliation and amount of input. Language attitudes were also
tested. The evidence shows there can be perceivable age effects for
linguistic measures even in a situation of minimal exposure to a foreign language, but these may not emerge until a substantial amount
of input has been gained.
Keywords: second language learning, age effects in L2, input effects
on L2, L2 English phonology, L2 English morphosyntax
Address for correspondence: Jenifer Larson-Hall, PO Box 311307, Denton, TX 76203, USA; email:
jenifer@unt.edu
© 2008 SAGE
10.1177/0267658307082981
082981_SLR_035-064.qxd
36
12/12/07
3:03 PM
Page 36
Studying a foreign language at a younger starting age
I Introduction
The question of what age is the optimal age to start learning a foreign
language has vast practical and economic consequences on the resources
of individuals and schools. However, as second language acquisition
researchers we are far from the goal of clearly understanding the role
that age plays when the amount of language input is very small. The present study examines the question of whether there is any advantage to
starting study of a foreign language at an early age by looking at the performance of present-day Japanese college students on one phonological
and one morphosyntactic measure and considering their language learning histories. One group of students began studying English at a median
age of nine, while the other group began their study when they entered
junior high school at age 12 or 13. The study is limited to learners who
received what I call ‘minimal input’, defined as no more than four hours
of instruction per week. Certainly, in such a situation learners will not
acquire a nativelike proficiency in any area, but the issue is whether such
study can ultimately advantage the language learners because of a younger
starting age.
Lay people may assume that the robust finding of ‘younger is better’
applies to second language acquisition in all situations. However, all
the critical period studies which show that ‘younger is better’ have been
conducted with immigrant populations, in which child learners were
immersed in the target language environment and presumably spent a
large percentage of their waking hours being exposed to the second language in school (Oyama, 1976; Johnson and Newport, 1989; Flege et al.,
1999; DeKeyser, 2000). The ‘younger is better’ phenomenon has no
guarantee of applying in situations of only minimal input; thus, it is
imperative that we look at the long-run consequences of minimal input
before making any conclusions.
Some researchers have argued that age effects would not be expected
to appear in situations of minimal input because critical period effects
are only seen in naturalistic or immersion classroom settings (Patkowski,
1994; DeKeyser, 2000; Lightbown, 2000). Theoretically, it could be
argued that if children and adults learn through fundamentally different
mechanisms (Bley-Vroman, 1988), and that in particular children learn
primarily in an implicit way (DeKeyser, 2003), then small amounts of
082981_SLR_035-064.qxd
12/12/07
3:03 PM
Page 37
Jenifer Larson-Hall
37
input will not be enough to trigger the formation of a morphological,
syntactic or phonological system.
Notice that this theoretical position might imply that younger starters
would perform even worse than later starters after the same amount of
input. Muñoz (2001) posits that earlier starters cannot gain the positive
effects of an early start if there is insufficient input for the kind of
implicit learning that is done by children, while they are not cognitively
advanced enough yet to benefit from explicit methods of instruction.
In order to examine the long-term effects of an earlier start, but in a
minimal input situation, test results from 200 learners of English in Japan
were obtained. These learners experienced at least 6 years of English study
and were all at least 18 at the time of testing. The participants may not
have reached their point of ultimate attainment in English, especially as
many were continuing to study English at the college level, but this time
period should be long enough for advantages, if there are any, to emerge.
In Japan, many parents spend large amounts of money to send their
children to private language schools. Koike and Tanaka (1995) assert
that Japan is one of the largest markets of foreign language teaching in
the world, with an estimated 30 billion dollars spent on English language teaching every year. English language learning is viewed as important enough to future success that educators in Japan have recently
decided to introduce compulsory English language instruction in the
elementary schools (Butler, 2005). The answer to the question of
whether early age is advantageous for these Japanese learners in a minimal
input situation will be of interest to linguists as well as educators and
administrators throughout many countries.
II Previous research on age advantages with minimal input
Several recent articles – using research done on two bilingual communities of third language learners of English in Spain (in the Basque
Country and Catalonia) – find absolutely no linguistic benefits for
younger starters (ages 4 or 8) vs. later starters (age 11) in a foreign
language environment (García Mayo and García Lecumberri, 2003).
Studies in the Basque Country (Cenoz, 2003; García Lecumberri and
Gallardo, 2003; García Mayo, 2003; Lasagabaster and Doiz, 2003)
examined students who had begun receiving 2–3 hours of English
082981_SLR_035-064.qxd
38
12/12/07
3:03 PM
Page 38
Studying a foreign language at a younger starting age
instruction per week at ages 4, 8 and 11 (not all studies examined the
four-year-old starters). Students were tested after 600 hours of school
exposure, at testing ages of 9, 13 and 16. Similarly, the study on bilinguals studying English in Catalonia (Muñoz, 2003) looked at students
who began at either age 8 or 11, and were tested once after 200 hours (at
ages 11 and 13) and again after 416 hours (at ages 13 and 15). These participants thus seem to be quite comparable to those in the present study,
in the sense that English was studied as a foreign language and input was
minimal. The authors of the articles tested a variety of modalities and
skill areas, but found no advantage for earlier starters in grammaticality
judgements (García Mayo, 2003), degree of foreign accent or discrimination tasks (García Lecumberri and Gallardo, 2003), production and
perception in oral testing (Cenoz, 2003; Muñoz, 2003) or composition
writing (Cenoz, 2003). In fact, in almost every reported comparison it
was found that earlier starters lagged behind later starters at least numerically and often statistically. In no cases were earlier starters statistically
better than later starters. The only advantage found for earlier starters
was in attitudes and motivation: Cenoz (2003) found statistically higher
motivation levels among the youngest starters (4 years old) after 600 hours
of instruction. However, in a different article on learners in Catalonia,
Muñoz and Tragant (2001) found no difference in attitudes between 3rd
grade (8–9 years) and 6th grade (11–12 years) starters.
The main problem with accepting the conclusion that early age gives
little advantage for situations of minimal input using the studies in the
García Mayo and García Lecumberri (2003) collection is that learners
were tested after the same amount of exposure, meaning that the earlier
starters were younger in chronological age than the later starters when
they took the tests. There is evidence in psychology, specifically with IQ
scores, that older learners will be better at test-taking than those who are
cognitively more immature (Bernstein et al., 1988) and, indeed, Cenoz
(2003) notes that cognitive maturity may be a possible explanation for
the findings of higher results overall for older starters. Thus, if the effect
for an earlier start is small, it may be hidden by the cognitive advantages
that the older starters hold on the tests. One might surmise that this
explanation should then only apply on tests where explicit knowledge or
test-taking is an issue (such as a grammaticality judgement test), and
would not be expected to apply in situations of spontaneous production.
082981_SLR_035-064.qxd
12/12/07
3:03 PM
Page 39
Jenifer Larson-Hall
39
However, language learning strategies of memorizing vocabulary and
conscious imitation of pronunciation models have been shown to be
used more often by older than younger children (Victori and Tragant,
2003), and thus cognitive skills could also have an influence on other
types of tests such as spontaneous production.
Older studies, conducted in the 1970s, also concluded that there was
little linguistic advantage to beginning study of a second language earlier in a minimal input situation. Oller and Nagato (1974) examined
Japanese students who began studying English in elementary school
(1–2 hours a week) and compared them to peers who began English
study in junior high only. Using a cloze test, the two groups were tested
in 7th, 9th and 11th grades. The authors found statistical differences
between the two groups in the lower two grades but not in 11th grade,
and they concluded that the younger starters’ advantages had disappeared by that point. The Oller and Nagato study thus avoids the problem of testing participants at different ages, but I would argue that a
possible reason that no differences were found in 11th grade was statistical. The tests with 7th and 9th graders had approximately 50 students
in each group, but the test in 11th grade had only 25 students who had
started early. If effect sizes are small, statistical results can disappear in
a smaller sample size (Tversky and Kahneman, 1971). For example, the
power to find a medium effect size on a t-test with 50 participants is .7
(meaning there is a 70% chance of finding a difference if one exists),
but the power to find the same effect size with only 25 participants is .4
(that is, chances are only 40% that you will find a difference even if one
exists). The Oller and Nagato study has also been criticized because
earlier starters were in the same classes with later starters, which means
that if the earlier starters were more advanced in some areas, they had
to ‘wait’ for the other learners to catch up. While this is a legitimate criticism of the study, it is not something that the present study avoids since,
to my knowledge, regular Japanese schools do not distinguish levels
of students when English classes begin in junior high. Thus, earlier
starters would be put into the same classes as beginning learners.
Another well-known older study of British elementary students who
learned French (by Burstall, 1977) had very large numbers of participants, but found that those who began learning French at age 8 showed
no advantages over those who began at age 11 when both groups were
082981_SLR_035-064.qxd
40
12/12/07
3:03 PM
Page 40
Studying a foreign language at a younger starting age
tested at age 16, except on a listening comprehension test. Even on the
comprehension test, effect sizes were small (Burstall says the difference
was 2–4 points on a 28-item test). However, students who started French
earlier did have more positive attitudes towards French. Tests given in
this study were quite holistic (at least for advanced students), and
included reading, writing, speaking and listening comprehension tests.
This study has the same problem as Oller and Nagato in that, for the most
part, the earlier starters were put into the same classrooms as older students. However, Buckby (1976) shows that when students were separated
into classes of earlier and later starters, at age 16 the earlier starters performed statistically better than older starters on listening and reading
tests, and performed equally well on speaking and writing. Thus there did
seem to be some advantages found for an earlier start but, overall,
Burstall (1977) was pessimistic about the utility of a younger start for foreign languages. Although this is an important study, it is hard to compare
its results with the present study, as the tests used were fairly different.
III A critical period for second language acquisition?
Extensive testing has established that ‘younger is better’ when it comes
to both first and second language acquisition (for recent reviews, see
Hyltenstam and Abrahamsson, 2003; DeKeyser and Larson-Hall, 2005).
Large-scale studies such as Flege et al. (1999) have shown that age
affects different parts of language differently, with phonology being
affected at a much younger age than basic morphosyntax. Some authors
have concluded there is a critical period for some aspects of lexical
development as well (Mayberry and Eichen, 1991; Lederberg and Spencer,
2005). Even though the age phenomenon is robust, there is still considerable debate over the ultimate causes of this age-related phenomenon, with some claiming it is maturationally conditioned (Hyltenstam
and Abrahamsson, 2003; DeKeyser and Larson-Hall, 2005), others arguing for more socio-psychological factors such as quantity and quality of
input, motivation and affective variables (Jia et al., 2002; Moyer, 2004),
and some claiming that it is due to general cognitive declines that continue over the lifespan (Bialystok and Miller, 1999; Birdsong, 2005).
There are no definitive ages that have been pinpointed for possible
critical period endpoints, but the Flege et al. (1999) study found that
082981_SLR_035-064.qxd
12/12/07
3:03 PM
Page 41
Jenifer Larson-Hall
41
even Koreans who had arrived in the USA as young as age 4 were, as a
group, judged to pronounce sentences with more of an accent than a
group of monolingual English speakers. From the scatterplot of foreign
accent ratings given in that article (1999: 85), it can be noted that no
Korean immigrant who arrived after the age of 12 scored in the range
of native speakers. On the other hand, for a grammaticality judgement
test (GJT) similar to the one used in the present study, Korean immigrants who arrived up to the age of 5 1⁄2 were not significantly different
from native speakers. For the scatterplot of scores for the GJT test
(1999: 85), it is not until after age 18 that participants are not able to
score within the range of native speakers.
There is really no reason to a priori privilege any particular age
when doing a test of age effects, and the best methodology would be to
use correlational analysis to look for trends instead of grouping participants and ‘throw[ing] away information’ (Maxwell and Delaney, 2004:
400). Such a methodology will be employed in this article for the group
of younger starters, but because there is essentially no variation in age
for the participants who began their study of English when they entered
junior high, correlational measures will not be very informative.
Therefore, I will also examine whether the younger starters differ from
the older starters as a group.
Similarly to critical period studies, the present study compares longterm outcomes of younger vs. older starters. However, it is not a critical
period study, as it does not compare the ultimate attainment of the learners and does not compare participants across a range of children to adults.
It also differs from critical period studies because the context is that of
English as a foreign language. However, critical period studies such as
Johnson and Newport (1989) and DeKeyser (2000) provide examples of a
methodology that could be used to compare the morphosyntactic abilities
of large numbers of participants: the grammaticality judgement test (GJT).1
Although such tests have been controversial (Sorace, 1996; Chaudron,
2003), researchers have experimentally found them to be reliable (Gass,
1994) and valid (Leow, 1996; Mandell, 1999), and many researchers continue to use them. The GJT test used in this study examines basic English
1 Johnson and Newport (1989) tested 46 Chinese and Korean users of English and found a correlation
of r = –.77 between age and GJT scores for all learners. DeKeyser (2000) tested 57 Hungarian users of
English and found a correlation of r = –.63 between age and GJT scores across his entire group.
082981_SLR_035-064.qxd
42
12/12/07
3:03 PM
Page 42
Studying a foreign language at a younger starting age
morphosyntax such as article usage, third-person singular marking and
gender in pronouns. In their diagnostic role uncovering problems with
basic morphosyntax that native speakers find essentially uncontroversial,
this GJT is a much more direct and economical measure of grammatical
intuition than spontaneous speaking or writing would be.
In order to obtain one specific measure of phonological ability, a
receptive phonemic discrimination task was used. This listening test
involved distinguishing whether words began with an initial ɹ/l/w. This
phonemic discrimination task was previously used by Yamada et al.
(1996) and found to be difficult for Japanese learners.2 Thus, although
such a task certainly does not take a measure of phonological ability as
a whole, it was thought that it would be effective in separating learners
on a difficult point of phonological ability.
Another measure that was used was a language learning aptitude test.
Individual aptitude for learning a language has been shown to correlate
with scores on linguistic tests for adults (Dörnyei and Skehan, 2003),
and aptitude is thus an important factor that should probably be taken
into account in all language acquisition studies that try to separate factors responsible for variation in scores. Oller and Nagato (1974) used
an intelligence test in their study of younger vs. older starters and found
a statistical correlation between IQ and cloze test scores (r ⫽.40).
IV Research questions
The main issue investigated is whether an earlier starting age for studying
English results in any benefits for Japanese learners when the learning
context is that of receiving minimal input in a foreign language environment. Previous research and theory would indicate that there would be no
advantage, and perhaps even a disadvantage, for a younger starting age in
the minimal input situation, except for attitude towards language learning.
My first research question asks whether there is any direct relationship
between the age that students began studying English and their scores on
one phonological task and one morphosyntactic task. The answer to this
question involves a correlational analysis among the earlier starters and
can directly address the issue of age. In order to minimize the effects of
2 Yamada et al. (1996) was designed to test the effect of word familiarity on discrimination abilities,
since some of the words in their corpus were real and others were nonsense words. The authors found
word familiarity correlated highly with perceptual ability (r ⫽.55).
082981_SLR_035-064.qxd
12/12/07
3:03 PM
Page 43
Jenifer Larson-Hall
43
language learning aptitude as a possible confound, aptitude scores are
factored out. As a further step in the correlational model, because there is
a possible confound between age and amount of input – in that those who
began learning at a younger age could have a larger amount of classroom
input than those who began later – a partial correlation controlling for
amount of input is also conducted.
My second research question is whether there are group differences
between participants who began studying English before formal study
in junior high school and those who did not. This result cannot speak
directly to the issue of whether age itself confers any advantages, but
does speak to the practical question of whether, as a group, any advantage can be found for Japanese students studying language before it
becomes mandatory in junior high.
The third question of interest is whether there is any difference in
attitude towards language learning in general or English in particular
between the two groups of Japanese students. Given previous research,
we would expect earlier starters to hold a more positive attitude towards
learning language.
V Method
1
Participants
Over 200 native Japanese speakers who were college students at a
national university in Japan took part in this study. The students tested
here represent an elite group and can be considered among Japan’s
brightest students. Japan has seven original national universities, and
getting into one of these universities is very difficult (Hughes, 1999).
The participants were students in 12 different classes and thus represent a sample of convenience, not a random sample. They were not paid
for their participation, and they performed the tests and filled out the
questionnaires during class time. No native speaker control group was
included in this research design, as the question was not how the
Japanese learners of English would perform relative to native speakers
of English, but rather, whether an early start would provide any advantage over those with a later start.
Table 1, which summarizes some characteristics of the participants,
can be found in the instruments section. Of the 200 participants, 102 were
male and 98 female. They ranged from ages 18 to 21, with the average
082981_SLR_035-064.qxd
44
12/12/07
3:03 PM
Page 44
Studying a foreign language at a younger starting age
age at testing overall of 19.4 years. Among the 61 students who reported
studying English before junior high school (‘earlier starters’), 20 were
male and 41 were female. The majority of earlier learners (59%) began
their study at age nine or later and studied English for 3 years or less.
For those participants who began English study in junior high (‘later
starters’), their age of first exposure to English was 12 or 13, and they
studied English for six years in school, as did the earlier learners. Fifty
percent of all learners additionally studied English outside of school
after they started junior high (66% of the earlier starters and 40% of the
later starters). No information was gathered about socioeconomic status,
although Burstall (1977) found this to correlate strongly with scores.
In general, Japanese society is less economically stratified than western
countries such as the USA, and Verba et al. (1987) reports there is only a
weak correlation between educational level and income in Japan.
Although it is not possible to say with certainty what method was used
to teach all of these Japanese learners of English, it is likely that English
teaching in the state-funded schools focused mainly on vocabulary, nonintegrated grammatical knowledge and translation exercises (Hughes,
1999; Morrow, 1987). The major reason for this is that, although the
Japanese Ministry of Education states that it supports the goal of communicative language abilities, university entrance exams are comprised
mainly of multiple choice items testing grammatical knowledge, translations and reading passages (Brown, 1995). Koike and Tanaka (1995)
assert that teaching will be tailored towards doing those things that help
students pass these difficult tests, and not towards achieving spoken fluency. Entrance exam test washback may not extend as far as conversation
or cram schools for eight-year-olds, so it is difficult to know what kinds
of teaching methodologies were employed for the earlier starters in this
article, although practical considerations of age probably mean that teaching contained less explicit grammatical explanation.
2
Instruments
Students took an oral grammaticality judgement test (GJT), a phonemic
discrimination test and an aptitude test.
a Grammaticality judgement test: The GJT used here was a version of
Johnson and Newport’s (1989) test of basic English morphosyntax,
082981_SLR_035-064.qxd
12/12/07
3:03 PM
Page 45
Jenifer Larson-Hall
45
adapted and used by DeKeyser (2000). The reliability – Kuder–Richardson
formula 20 (KR-20) internal consistency estimate – of DeKeyser’s test
administration was high, at over 90%. DeKeyser’s version was used
without further change because most of the changes made for Hungarian
learners are also challenging for Japanese learners. For example, the
absence of determiners in Japanese make these sections difficult, while
object pro-drop could also transfer from Japanese. This instrument was
designed to test judgements on only basic features of morphosyntax in
English that are unproblematic for native speakers. The 11 areas that
DeKeyser (2000) lists in the GJT are:
past tense (example: *‘Last night the old lady die in her sleep’)
● plural (*‘Teachers often give useful advices to their students’)
● third-person singular (*‘Our neighbor will goes to Europe next year’)
● present progressive (*‘Tom working in his office right now’)
● determiners (*‘Mrs. Johnson went to library yesterday’)
● pronominalization (pro-drop: *‘Mike wrote the letter but didn’t send’;
gender errors: *‘John knew but she did not tell’)
● particle movement (*‘Kevin called Nancy for a date up’)
● subcategorization (depends on these individual verbs: say, laugh,
say/tell, learn/study, want, hope, allow/let, enjoy; example: *‘I want
you will go to the store now’)
● yes–no questions (*‘Can ride the little girl a bicycle?’)
● wh-questions (*‘Who you meet at the park every day?’)
● word order (*‘The student eats quickly his meals’)
●
This test was therefore meant to ascertain whether early exposure
would prove an advantage for basic categories of morphosyntax.
Although there is no guarantee that earlier starters learned these morphosyntactic features in their early classes, since these areas are very
basic it seems likely that if they learned anything at all they would have
been using these features of the grammar.
b Phonemic discrimination task: For the receptive phonological measure, Yamada et al.’s (1996) phonemic discrimination test was used. The
test asked students to identify whether each of 96 items began with ɹ/l/w.
Participants heard a real word like ‘ring’ or ‘wing’, or a nonsense word
082981_SLR_035-064.qxd
46
12/12/07
3:03 PM
Page 46
Studying a foreign language at a younger starting age
like ‘ling’ (with the exception of the set royal/loyal/woyal, word groups
consisted of only one syllable). Listeners heard each word only once, and
there were 2 seconds between words. Between each group of 10 words
there was an 8 second pause. Participants saw the sequence ‘r l w’ on their
answer sheet and were told to circle the letter of the first sound of the
word they heard. Yamada et al. found 99.8% accuracy on the test among
10 native speakers of American English, but only 66.9% accuracy among
50 Japanese speakers. A sample of some of the words can be found in
Appendix 1. I do not have any information on how much attention was
paid to pronunciation in classes for earlier starters except to note that
among the early starters, 56% report that none of their teachers were
native speakers, while 44% had at least one native speaker teacher.
c Aptitude test: The aptitude test was a Japanese version of the
Modern Languages Aptitude Test (Carroll and Sapon, 1959) normed on
160 Japanese college students by Sasaki (1996). Only Parts 3 and 4,
Language Analysis and Sound–Symbol Association, were used. The
Language Analysis section consisted of 15 four-way multiple choice
items, while the Sound–Symbol Association test had 22 four-way multiple choice items (18 written in katakana and four in Roman script).
KR-20 estimates of reliability for Sasaki’s administration were .80 for
Language Analysis and .63 for Sound–Symbol Association.
d Background questionnaire: All students filled out a six-page questionnaire about their early experience with English, hours studying
English in junior high and high school both in school and doing homework, attendance at language schools after age 12, self-ratings in speaking and reading, use of English, and attitude towards language (the
questionnaire is available upon request). The self-ratings were measured on a 1–10 scale as shown below:
How well did you feel you could speak English on a conversational level before you
entered this university? Place a ✗ on the scale.
1…………2…………3…………4…………5…………6…………7…………8…………9…………10
Can’t
Can say
Can say
Can have
Don’t have Like a
speak
words
simple
a conversation to think
native
at all
only
sentences
about
speaking
082981_SLR_035-064.qxd
12/12/07
3:03 PM
Page 47
Jenifer Larson-Hall 47
How well did you feel you could read English (in general, not in any specific field)
before you entered this University? Put a ✗ on the scale.
1…………2…………3…………4…………5…………6…………7…………8…………9…………10
Can’t
Can only
Can understand
Can read
Can read Like a
read
read some
simple sentences
paragraphs
easily
native
at all
words
and naturally
A rough probe of attitude was measured through two questions on
the questionnaire. The questions were rated on a 1–10 scale where 1 ⫽
hate it, 5 ⫽ don’t care, and 10 ⫽ love it. The questions were phrased:
‘How much do you like studying languages?’ and ‘How much do you
like studying English?’ (these questions are similar to those in
Gardner’s AMTB for ‘Interest in foreign languages’ and ‘Attitudes
toward learning [a specific language]’; see appendix of Gardner,
1985). The use of English was based on the measure used in Flege et
al. (1999) and asked about average use of English at home, school,
part-time work and social situations on a 1–7 scale, where 1 ⫽ never
and 7 ⫽ frequently.
For calculation of hours of exposure to the second language (L2),
student reports of how many hours they spent per week studying
English and also doing homework in both junior high and high school,
and any time they studied English in formal classes outside school
(such as in ‘cram school’) were noted. These numbers were multiplied
by 44, for the number of weeks that students are in school. Because the
students were recalling events from the previous 6–8 years, there will
undoubtedly be some error in this measurement. However, student
reports generally lined up with the Japanese ministry guidelines for
how much English study is mandatory in junior high (three hours per
week) and high school (five hours per week). About half of all participants (96 out of 200) reported studying English outside school in junior
high and high school as well, and these hours were calculated into their
total study time. It is also important to acknowledge that hours of
instruction may not equal input. Recognizing the possible flaws of this
variable it nevertheless seemed preferable to ask for self-report data in
this case than to totally ignore possible differences in time spent
exposed to English input. Measures for all participants can be furnished
upon request.
082981_SLR_035-064.qxd
48
12/12/07
3:03 PM
Page 48
Studying a foreign language at a younger starting age
Table 1 Summary characteristics of participants with mean and standard deviation
in parentheses
Years of study
Amount of input (hours)
Age at testing
Age began formal instruction
Self-rating (speaking; 10 ⫽nativelike)
Self-rating (reading; 10 ⫽nativelike)
Use of English (4–28 points, higher ⫽more use)
Attitude toward studying language (10 ⫽ love it)
Attitude toward studying English (10 ⫽ love it)
Earlier (n ⫽61)
Later (n ⫽139)
3.0 (2.1) ⫹6*
1923 (719)
19.4 (0.9)
8.3 (2.5)
5.2 (1.1)
6.9 (1.1)
7.1 (2.7)
7.0 (1.9)
7.1 (1.8)
6*
1764 (727)
19.3 (0.8)
12.5**
4.8 (1.1)
6.6 (1.2)
7.0 (2.8)
6.4 (1.9)
6.5 (2.0)
* This number was not measured, but assumed as all students would have had to
have this amount of study of English in the six years of junior high and high school.
** This is an estimated number as I did not ask those who began study in junior high
what age they were when they entered junior high.
3
Procedure
The tests and questionnaires were given to students in their classrooms.
The order of the three tests (the GJT, phonemic discrimination and aptitude) were varied in each class. Testing took 90 minutes: the GJT took
66 minutes, the aptitude test took 17 minutes and the phonemic discrimination test took 6 minutes. There were three versions of the phonemic
discrimination test which were given in an ordered fashion (1–2–3) to
different classes to avoid order effects.
VI Results
The main issue investigated is whether an earlier starting age for studying English results in any benefits for Japanese learners when the learning context is that of receiving minimal input in a foreign language
environment. Previous research and theory would indicate that there
would be no advantage, and perhaps even a disadvantage, for a younger
starting age in the minimal input situation, except for attitude towards
language learning.
1
Correlational analysis among the earlier starters
The first research question asks whether there is any direct relationship
between the age that students began studying English and their scores
on a phonological task and morphosyntactic task. The answer to this
082981_SLR_035-064.qxd
12/12/07
3:03 PM
Page 49
Jenifer Larson-Hall
49
question involves a correlational analysis among the earlier starters
between age and test scores.
Before addressing the question of the relationship between age and
test scores, however, I examined whether there were any correlations
between language aptitude scores and test scores among the earlier
starters. If there were a statistical correlation, we would want to factor
aptitude out by doing a partial correlation, as Oller and Nagato (1974)
did when they found a correlation between IQ scores and cloze test
scores in their experiment. The correlation between aptitude scores and
phonemic discrimination scores among earlier starters is r ⫽ .37,
n ⫽58, p ⬍.05, power ⫽.84, and the correlation between aptitude
scores and GJT scores is r ⫽.32, n ⫽59, p ⬍.05, power ⫽.71.3 Because
there is a statistical correlation between aptitude and both test scores,
any further correlations will be partial correlations controlling for language learning aptitude.
Another factor to control for is amount of input. It may be the case
that earlier starters just have a larger cumulative amount of input, and
thus their advantage may have nothing to do with age per se. Figures 1
and 2 show scatterplots of the amount of input and scores on both tests.
The lines superimposed on the graphs are smoother (Loess) lines and
follow the trend of the data instead of fitting a straight line to the data
like a regression line does. The scatterplots show that the data do not follow a straight line, and thus there may be no statistical correlation
between input and test scores for the earlier learners: In fact, the correlation is not statistical for the phonemic discrimination task among the earlier starters (r ⫽– .07, n ⫽61, p ⫽.57, power ⫽.08), but it is statistical for
the GJT (r ⫽.27, n ⫽61, p ⫽.035, power ⫽.56). However, it may still be
worthwhile to examine how correlations change when amount of input is
partialled out.
The correlation between scores on the phonemic discrimination task
and starting age, controlling for aptitude, was not statistical (r ⫽.05,
df ⫽58, p ⫽.72, R2 ⫽.00, power ⫽.06). Controlling additionally for input,
the situation does not change substantially and the correlation is still not
statistical (r ⫽.03, df ⫽57, p ⫽.80, R2 ⫽.00). On the other hand, the
3 This is the value with outliers removed by using a robust correlation (Wilcox, 2001; Maronna
et al., 2006)
082981_SLR_035-064.qxd
50
12/12/07
3:04 PM
Page 50
Studying a foreign language at a younger starting age
Figure 1 Scatterplot showing the relationship between amount of input and scores
on the phonemic discrimination task for earlier and later starters
partial correlation (controlling for aptitude) between the grammaticality
judgement test and starting age is statistical (r ⫽– .42, df ⫽58, p ⬍.001,
R2 ⫽.18, power ⫽.93). The negative correlation means that scores
declined as the starting age increased. Controlling additionally for
amount of input, there is only a slight decline in the strength of r
(r ⫽–.38, df ⫽57, p ⫽.003, R2 ⫽.14).
2
Examining group differences
My second research question was whether a group of students who
began their English study at a younger starting age would hold any
advantage over those who only began English study in junior high. This
is rather a different question than the question of the relationship of age
to test scores, as was examined through the correlational analysis. Here,
because the cut-off age of the groups is arbitrary, the question is not so
082981_SLR_035-064.qxd
12/12/07
3:04 PM
Page 51
Jenifer Larson-Hall
51
Figure 2 Scatterplot showing the relationship between amount of input and scores
on the GJT for earlier and later starters
much whether age itself confers any benefits, but whether any benefits
accrue to the group of students who were motivated enough (or whose
parents were motivated enough!) to want to study English before it
became mandatory in school. This still speaks to the practical question
of whether any advantage can be found for Japanese students who
begin their study of English before junior high with minimal input.
In comparing all of the earlier starters to all of the later starters,
a series of three independent samples t-tests was conducted (a Bonferroni
adjustment was applied and an alpha level of .05/3 ⫽.017 was set).
Mean scores and standard deviations on the GJT, phoneme discrimination test and aptitude test are given in Table 2. Skewness and kurtosis
levels of the tests were within the normal range. A side-by-side graphic
comparison of the groups can be seen in Figure 3. Results showed no
statistical difference between groups for the aptitude test (t120.4 ⫽–.45,
p ⫽.65, ␩2 ⫽0; mean difference ⫽–.29 ⫾1.28, power ⫽.13). Although
082981_SLR_035-064.qxd
52
12/12/07
3:04 PM
Page 52
Studying a foreign language at a younger starting age
Table 2 Results on 3 tests by overall group and earlier learners vs. later learners
(standard deviation in parentheses)
GJT (200 points)
Mean
Range
R/L/W (96 points)
Aptitude (37 points)
Earlier
Later
Earlier
Later
Earlier
Later
114.7 (10.7)
91–142
112.3 (10.1)
87–135
54.7 (14.7)
19–90
49.3 (13.2)
21–78
31.6 (4.2)
19–37
31.2 (4.4)
19–37
Grammatical judgement
130
120
110
Number correct out of 200
70
60
50
20
90
30
100
40
Number correct out of 96
80
140
90
Phonemic discrimination
Later
Earlier
Later
Earlier
Figure 3 Scores of Japanese learners of English compared on two measures; actual
data points are overlaid on the boxplot and median lines are in white
we saw in the correlational analysis that there was a correlation
between aptitude and test scores, the t-test shows that the difference in
aptitude scores across groups is not statistical and thus should not affect
group comparisons.
In looking at group differences on the two linguistic tests, there was no
difference between groups on the GJT (t109.2 ⫽–1.51, p ⫽.13, ␩2 ⫽.01;
082981_SLR_035-064.qxd
12/12/07
3:04 PM
Page 53
Jenifer Larson-Hall
53
mean difference ⫽–2.43 ±3.19, power ⫽.82). There was, however, a statistical difference between the groups for the phonemic discrimination
task (t104.0 ⫽–2.47, p ⫽.015, ␩2 ⫽.03; mean difference ⫽–5.40 ⫾5.40,
power ⫽1.00). The magnitude of the difference in the means (the ␩2
value) for the phonemic discrimination test can be classified as small,
according to Cohen (1988), and explains about 3% of the variance in the
scores.
Again, as the earlier learners had spent more time involved in study
and might have received more input, it is informative to examine group
differences while controlling for total hours of input. In order to see
whether there would be a difference between the earlier and later starters
with equal amounts of input, a one-way ANCOVA with test scores (GJT
or phonemic discrimination) as the dependent variable, status as earlier
or later starter as independent variable, and total hours of input as the
control variable was used to statistically match learners in both groups
with equal amounts of input. An ANCOVA controls for the variability in
scores due to the total hours studied, and is preferable to a matched pairs
sample that may match learners unequally (Maxwell and Delaney, 2004).
ANCOVA analysis confirmed the results of the t-tests, in that there was
no effect for group (earlier vs. later starters) on the GJT (F1, 197 ⫽1.69,
p ⫽.20, partial ␩2 ⫽.01, power ⫽.25), while there was an effect for
group on the phonemic discrimination test (F1, 197 ⫽6.55, p ⫽.01, partial
␩2 ⫽.03, power ⫽.72). The covariate of total hours of study was statistical for the GJT (F1, 197 ⫽6.24, p ⫽.013, partial ␩2 ⫽.03, power ⫽.70),
but not for the phonemic discrimination test (F1, 197 ⫽.00, p ⫽.99, partial
␩2 ⫽.00, power ⫽.05).
A closer look at the factor of hours of input can be helpful to understand what is happening between the two groups. Looking back to the
scatterplots of total hours of study and test scores in Figures 1 and 2, it
can be seen that the trend of the data for earlier learners was usually
above the later starters in both tests when the total hours were somewhere between 1500 and 2000. In fact, a robust ANCOVA (20%
trimmed means bootstrap-t ANCOVA) that compares points along the
Loess line found statistical advantages for the earlier starters on the
GJT at 1833 and 2000 hours, but a statistical advantage for later starters
at 800 hours (other points did not show statistical differences). Given
that none of the studies reported on in the García Mayo and García
082981_SLR_035-064.qxd
54
12/12/07
3:04 PM
Page 54
Studying a foreign language at a younger starting age
Lecumberri (2003) volume examined learners who had had more than
600 hours of input, my results do not conflict with those of those
researchers, but show that more hours of input may be necessary to
begin to see differences between an early and later start. A robust
ANCOVA examination of the phonemic discrimination scores found
statistical advantages for the earlier starters at 1300, 1555, 1833 and
2000 hours (other points did not show statistical differences). Again, it
seems to take a substantial accumulation of input to see advantages for
an early start to begin to show.
3
Differences due to native speaking teachers
Given that there was a difference between earlier and later starter groups
on the phonemic discrimination test, we might wonder whether having
had a native-speaking English teacher (NEST) would have been helpful
for the earlier starters in their phonemic ability. Of the 61 earlier starters,
34 reported that at least one of their teachers was a native speaker. Those
with a NEST scored more highly on the phonemic discrimination test but
had more variance (mean ⫽56.1; sd ⫽16.1) compared to non-NEST
students (mean ⫽53.0; sd ⫽12.8). A t-test examining the difference
between these two groups was not statistical, however (t59.0 ⫽.84, p ⫽.40,
␩2 ⫽.01; mean difference ⫽3.11 ±7.41, power ⫽.23). A scatterplot in
Figure 4 shows the distribution of phonemic discrimination test scores
and the age that study began among the earlier starters. Looking at
the smooth line, there seems to be a tendency for the very early starters
(ages 4–6) to do much better on the test if they had NESTs, but the numbers are too small to test this intuition statistically. Further research with
larger numbers of participants who started very young could help elucidate this area.
4
Attitudes towards language learning
My third research question was whether there would be any differences
in attitudes towards language learning between the earlier and later
starters. On the question of learning languages in general, the earlier
starters had higher scores than the later starters on the 10-point scale (for
scores, see Table 1). A t-test revealed that this difference was nearly statistical (t197 ⫽–1.9, p ⫽.06, ␩2 ⫽.02, power ⫽.77) for the question of
082981_SLR_035-064.qxd
12/12/07
3:04 PM
Page 55
Jenifer Larson-Hall
55
Figure 4 Scatterplot of beginning age for earlier starters and score on the phonemic
discrimination test with division by whether or not teacher was a native English
speaking teacher (NEST) or not (non-NEST)
learning languages in general as well as for learning English in particular (t197 ⫽–1.9, p ⫽.06). A correlation just among the earlier starters did
not find any statistical relationship between age that study started and
attitude towards learning English specifically (r ⫽–.16, n ⫽60, p ⫽.22,
power ⫽.23) or languages in general (r ⫽.04, n ⫽60, p ⫽.74).
VII Discussion
Contrary to many previous studies, this study found some modest effects
for an early starting age in both grammatical and receptive phonological
abilities. Age was not entirely separate from amount of input, however,
and in fact interacted with it. Earlier starters gained an advantage on the
grammaticality judgement test the younger they started their study of
English, with differences in age accounting for 14% of the variation
when controlling for total hours of input. An ANCOVA with the GJT
082981_SLR_035-064.qxd
56
12/12/07
3:04 PM
Page 56
Studying a foreign language at a younger starting age
found total hours of input but not status as an earlier or later starter to be
a statistical factor in explaining scores, and a robust ANCOVA further
elucidated that earlier starters scored more highly than later starters on
the GJT when they had a range of about 1600–2200 hours of input. In
other words, morphosyntactic abilities could be enhanced by an early
start, but only after a substantial amount of hours of input (this means
from 6–8 hours a week if we calculate the 1600–2200 hours of input
over 6 years and 44 weeks per year). When input levels became very
high there were no statistical differences between the groups.
Earlier starters as a group scored statistically better than later starters
on a test of phonemic discrimination, but no clear advantages for a
younger starting age among the earlier starters was seen. The difference
between groups accounted for about 3% of the variation among the
earlier-starting and later-starting groups. There were no clear effects for
having a native-speaking English teacher. Although the data graphics
seemed to indicate that having a native speaker might be most helpful
at quite young ages (4–6), the number of participants was too small to
ascertain this statistically. A robust ANCOVA indicated an interaction
between amount of input and group affiliation. It found that earlier
starters held an advantage over later starters when they had a range of
about 1200–2200 hours of input.
The fact that this study found some modest advantages to an earlier
start in studying a classroom-instructed language, even with minimal
input, can probably largely be attributed to the fact that learners in this
study had larger amounts of total input than previous studies. Even
though by definition the study was limited to classroom instruction of
no more than 4 hours a week, participants reported significant amounts
of homework and study outside the classroom that added to their total
hours of study, and effects for being in the earlier starter group were not
seen until the amount of input was fairly large. In fact, at least for the
GJT, later starters scored statistically higher at 800 hours of input than
earlier starters. This would fit in with the results from the studies in the
García Mayo and García Lecumberri collection, since students in these
studies were tested after at most 600 hours of classroom exposure
(although it is not known whether these students would have had substantial amounts of homework and out-of-school classroom instruction
in English as well).
082981_SLR_035-064.qxd
12/12/07
3:04 PM
Page 57
Jenifer Larson-Hall
57
We may ask why age differences would not become apparently
advantageous until individuals had acquired a significant amount of
input. One answer may be statistical. Looking at the scatterplots in
Figures 1 and 2, it is clear that the majority of participants fall in the
middle range (1000–2500) for total hours of input. Therefore, there
may simply be more power to find differences in this range because
there is a larger n. Given this line of reasoning, we might expect to find
an advantage for earlier starters at every input level, if the group sizes
were large enough. However, previous studies such as Burstall et al.
(1974) have had quite large group sizes (n ⫽5000⫹) and still did not
find many advantages for earlier starters.
Another answer may be that increasing amounts of input allow a
more coherent language system to form, and as this occurs then advantages are able to more easily be demonstrated. This may be easiest
to understand for the morphosyntax. For example, with very small
amounts of exposure grammatical knowledge may still be fragmentary
and tests of explicit judgement may favour those who learned a language in a more fragmentary, discrete-grammar-point type of way when
they were older. Thus, the accumulation of many hours of input for the
earlier starters may have finally allowed them to use their more integrated knowledge in an effective way. For phonemic discrimination,
one might expect that the ability to discriminate phonemes of the language would not depend so much on an integrated knowledge of the
entire language, but studies of Japanese learners of English have shown
that productive pronunciation can definitely improve with more exposure to the language (Riney and Flege, 1998), so receptive phonological
abilities may take some time to develop as well.
The effect sizes of the correlations and t-tests show that age plays
only a modest explanative role, explaining from 3–14% of the variance
in scores. Given the findings cited above that the ability to sound
nativelike in phonology ends at an earlier age than the ability to do well
on grammaticality judgement tests, we might have expected to find a
stronger role for age in phonology. One possible reason for the discrepancy may be that there was a very wide range of exposure to nativelike
phonology within this group, as opposed to what would be found for
immigrants in an immersion situation. In other words, although we can
be fairly sure that immigrants will be exposed to a variety of nativelike
082981_SLR_035-064.qxd
58
12/12/07
3:04 PM
Page 58
Studying a foreign language at a younger starting age
accents and an abundance of spoken language, this is not clearly the
case for children studying a foreign language. On the other hand, it is
more likely that any input in English, whether spoken or written, would
have contained abundant amounts of nativelike grammatical information, so the grammar is less likely to be affected by variations in quality and quantity of available spoken input than the phonology. Further
research in the effects of early exposure to a language should try to better detail the type and amount of nativelike phonological input that
learners received, and try to gather larger groups of learners who began
at a very young age, where nativelike input may provide the most
benefit.
As for attitudes towards learning languages in general or English
specifically, which was the only advantage that earlier starters in García
Mayo and García Lecumberri (2003) volume were found to have,
there was a strong trend towards a statistical difference when looking
at group differences, with earlier starters tending to have more positive attitudes towards studying languages in general and English in
particular.
VIII Conclusions
This study was conducted to ascertain whether there would be any
relationship between starting age and scores on a phonemic and
morphosyntactic measure in a situation of minimal input. Contrary to
predictions that age only plays a role in naturalistic or immersion environments, the present study found evidence that a younger starting age
makes a modest difference to both phonological and basic morphosyntactic abilities, even in a situation of minimal input. The additional
factor of input that was included in this study may be crucial in understanding the divergence of results from previous studies. These studies
did not test participants with large amounts of input, and this study
showed that advantages only emerged when total input increased.
Contrary to what a lay person may think, age does not confer a ‘magical’
ability to learn a second language quickly and natively in a situation of
minimal input. However, age does seem to play a non-negligible role in
improving second language acquisition, given that language learners
receive enough input. Starting to study a language at a younger age is
082981_SLR_035-064.qxd
12/12/07
3:04 PM
Page 59
Jenifer Larson-Hall
59
one way to ensure larger amounts of language input, so the present
experiment overall finds a beneficial effect for starting to study a language at a younger age, even when input is only minimal.
Recommendations for educators and policy-makers
Because this study is in the minority of studies in that it found a positive effect for a younger starting age in a situation of minimal input,
readers should be cautious in interpreting it too broadly. However, my
recommendation would be to begin foreign language study as young as
possible, with as many hours of input as are possible. I agree with Lee
(1977) that few other school subjects, such as mathematics or reading,
have been held up to the scrutiny that language learning has as to
whether an earlier start is more effective than a later start. It is likely
that older students are more effective at learning almost any kind of
cognitive endeavour than younger students, but this does not stop us
from putting young children in school and teaching them these subjects.
With evidence of even modest benefits to a younger start for foreign
language it then seems worthwhile to include this subject in the curriculum, especially as most studies concur that positive attitudes towards
language learning develop with an early start. In fact, better morphosyntactic and phonemic discrimination ability may not be the only
benefit that earlier learners receive. Other factors not tested for here –
such as an ability to produce nativelike phonemes and morphosyntax,
an increased vocabulary, more nativelike intonation, better speaking
fluency or so on – may also result from an earlier start.
Acknowledgements
I would like to thank the students of Kyushu University who participated in this study, and the English department at that university for
financial help provided to conduct this study. I benefited greatly from
the comments of Greg Mizera as well as several anonymous Second
Language Research reviewers, but of course any errors still found here
are my own. Richard Herrington deserves special recognition for being
an important influence on my statistics. An earlier version of this article was presented at the 2005 Second Language Research Forum
conference.
082981_SLR_035-064.qxd
60
12/12/07
3:04 PM
Page 60
Studying a foreign language at a younger starting age
IX References
Bernstein, D., Roy, E., Srull, T. and Wickens, C. 1988: Psychology. Boston,
MA: Houghton Mifflin.
Bialystok, E. and Miller, B. 1999: The problem of age in second-language
acquisition: influences from language, structure, and task. Bilingualism:
Language and Cognition 2, 127–45.
Birdsong, D. 2005: Interpreting age effects in second language acquisition. In
Kroll, J. and de Groot, A.M.B., editors, Handbook of bilingualism:
psycholinguistic approaches. New York: Oxford University Press,
109–27.
Bley-Vroman, R. 1988: The fundamental character of foreign language learning. In Rutherford, W. and Sharwood-Smith, M., editors, Grammar and
second language teaching: a book of readings. New York: Newbury
House, 19–30.
Brown, J.D. 1995: English language entrance examinations in Japan: problems and solutions. Unpublished paper presented at the Proceedings of
the JALT 1995 Conference.
Buckby, M. 1976: Is primary French really in the balance? Modern Language
Journal 60, 340–46.
Burstall, C. 1977: Primary French in the balance. Foreign Language Annals
10, 245–54.
Burstall, C., Jamieson, M., Cohen, S. and Hargreaves, M. 1974: Primary
French in the balance. Windsor: NFER Publishing.
Butler, Y.G. 2005: Comparative perspectives towards communicative activities among elementary school teachers in South Korea, Japan and
Taiwan. Language Teaching Research 9, 423–46.
Carroll, J.B. and Sapon, S. 1959: Modern language aptitude test: form A.
New York: The Psychological Corporation.
Cenoz, J. 2003: The influence of age on the acquisition of English: general
proficiency, attitudes and code-mixing. In Garcia Mayo, M. and Garcia
Lecumberri, M., editors, Age and the acquisition of English as a foreign
language. Clevedon: Multilingual Matters, 77–93.
Chaudron, C. 2003: Data collection in SLA research. In Doughty, C. and
Long, M.H., editors, The handbook of second language acquisition.
Malden, MA: Blackwell, 762–828.
Cohen, J. 1988: Statistical power analysis for the behavioral sciences.
Newbury Park, CA: Sage.
DeKeyser, R.M. 2000: The robustness of critical period effects in second
language acquisition. Studies in Second Language Acquisition 22,
499–533.
—— 2003: Implicit and explicit learning. In Doughty, C. and Long, M.H., editors, The handbook of second language acquisition. Malden, MA:
Blackwell, 313–48.
082981_SLR_035-064.qxd
12/12/07
3:04 PM
Page 61
Jenifer Larson-Hall
61
DeKeyser, R.M. and Larson-Hall, J. 2005: What does the Critical Period
really mean? In Kroll, J. and de Groot, A.M.B., editors, Handbook of
bilingualism: psycholinguistic approaches. New York: Oxford University
Press, 88–108.
Dörnyei, Z. and Skehan, P. 2003: Individual differences in second language
learning. In Doughty, C. and Long, M.H., editors, The handbook of second language acquisition. Malden, MA: Blackwell, 589–630.
Flege, J.E., Yeni-Komshian, G. and Liu, S. 1999: Age constraints on secondlanguage acquisition. Journal of Memory and Language 41, 78–104.
García Lecumberri, M. and Gallardo, F. 2003: English FL sounds in school
learners of different ages. In García Mayo, M. and García Lecumberri,
M., editors, Age and the acquisition of English as a foreign language.
Clevedon: Multilingual Matters, 115–35.
García Mayo, M. 2003: Age, length of exposure and grammaticality judgements in the acquisition of English as a foreign language. In García
Mayo, M. and García Lecumberri, M., editors, Age and the acquisition of
English as a foreign language. Clevedon: Multilingual Matters, 94–114.
García Mayo, M. and García Lecumberri, M., editors, 2003: Age and the
acquisition of English as a foreign language. Clevedon: Multilingual
Matters.
Gardner, R.C. 1985: Social psychology and second language learning: the
role of attitudes and motivation. London: Edward Arnold.
Gass, S. 1994: The reliability of second-language grammaticality judgments.
In Tarone, E., Gass, S. and Cohen, A., editors, Research methods in second language acquisition. Hillsdale, NJ: Lawrence Erlbaum, 303–22.
Hughes, H.J. 1999: Cultivating the walled garden: English in Japan. English
Studies 6, 556–68.
Hyltenstam, K. and Abrahamsson, N. 2003: Maturational constraints in
SLA. In Doughty, C. and Long, M.H., editors, The handbook of second
language acquisition. Malden, MA: Blackwell, 539–88.
Jia, G.X., Aaronson, D. and Wu, Y. 2002: Long-term language attainment of
bilingual immigrants: predictive variables and language group differences. Applied Psycholinguistics 23, 599–621.
Johnson, J.S. and Newport, E.L. 1989: Critical period effects in second language learning: the influence of maturational state on the acquisition of
English as a second language. Cognitive Psychology 21, 60–99.
Koike, I. and Tanaka, H. 1995: English in foreign language education policy
in Japan: toward the twenty-first century. World Englishes 14, 13–25.
Lasagabaster, D. and Doiz, A. 2003: Maturational constraints on foreignlanguage written production. In Garcia Mayo, M. and Garcia
Lecumberri, M., editors, Age and the acquisition of English as a foreign
language. Clevedon: Multilingual Matters, 136–60.
Lederberg, A.R. and Spencer, P.E. 2005: Critical periods in the acquisition
of lexical skills: evidence from deaf individuals. In Fletcher, P. and
082981_SLR_035-064.qxd
62
12/12/07
3:04 PM
Page 62
Studying a foreign language at a younger starting age
Miller, J.F., editors, Developmental theory and language disorders.
Philadelphia, PA: John Benjamins, 121–45.
Lee, W.R. 1977: For and against an early start. Foreign Language Annals 10,
263–70.
Leow, R.P. 1996: Grammaticality judgment tasks and second-language development. Georgetown University Round Table on Languages and
Linguistics, 126–39.
Lightbown, P.M. 2000: Classroom second language acquisition research and
second language teaching. Applied Linguistics 21, 431–82.
Mandell, P.B. 1999: On the reliability of grammatical judgement tests in second
language acquisition research. Second Language Research 15, 73–99.
Maronna, R.A., Martin, R.D. and Yohai, V.J. 2006: Robust statistics: theory
and methods. Hoboken, NJ: Wiley.
Maxwell, S.E. and Delaney, H.D. 2004: Designing experiments and analyzing
data: a model comparison perspective. 2nd edition. Mahwah, NJ: LEA.
Mayberry, R.I. and Eichen, E.B. 1991: The long-lasting advantage of learning
sign language in childhood: another look at the critical period for language
acquisition. Journal of Memory and Language 30, 486–512.
Morrow, P.R. 1987: The users and uses of English in Japan. World Englishes
6, 49–62.
Moyer, A. 2004: Age, accent and experience in second language acquisition.
Clevedon: Multilingual Matters.
Muñoz, C. 2001: Factores escolares e individuales en el aprendizaje formal de
un idioma extranjero [Scholarly and individual factors in the formal
comprehension of a foreign language]. Estudios de Linguistica, supplement 1, 249–70.
—— 2003: Variation in oral skills development and age of onset. In García
Mayo, M. and García Lecumberri, M., editors, Age and the acquisition
of English as a foreign language. Clevedon: Multilingual Matters,
161–81.
Muñoz, C. and Tragant, E. 2001: Motivation and attitudes towards L2: some
effects of age and instruction. Eurosla Yearbook 1, 211–24.
Oller, J. and Nagato, N. 1974: The long-term effect of FLES: an experiment.
Modern Language Journal 58, 15–19.
Oyama, S. 1976: A sensitive period for the acquisition of a nonnative phonological system. Journal of Psycholinguistic Research 5, 261–83.
Patkowski, M.S. 1994: The critical period hypothesis and interlanguage
phonology. In Yavas, M., editor, First and second language phonology.
San Diego, CA: Singular, 205–21.
Riney, T.J. and Flege, J.E. 1998: Changes over time in global foreign accent
and liquid identifiability and accuracy. Studies in Second Language
Acquisition 20, 213–43.
Sasaki, M. 1996: Second language proficiency, foreign language aptitude, and
intelligence: quantitative and qualitative analyses. New York: Peter Lang.
082981_SLR_035-064.qxd
12/12/07
3:04 PM
Page 63
Jenifer Larson-Hall
63
Sorace, A. 1996: The use of acceptability judgments in second language
acquisition research. In Ritchie and W.C., Bhatia, T.K., editors,
Handbook of second language acquisition. San Diego, CA: Academic
Press, 375–409.
Tversky, A. and Kahneman, D. 1971: Belief in the law of small numbers.
Psychological Bulletin 76, 105–10.
Verba, S. Verba, S., Kelman, S., Orren, G.R., Miyaki, I., Watanuki, J.,
Kabashima, I. and Ferree, G.D. 1987: Elites and the idea of equality: a
comparison of Japan, Sweden, and the United States. Cambridge, MA:
Harvard University Press.
Victori, M. and Tragant, E. 2003: Learner strategies: a cross-sectional and
longitudinal study of primary and high-school EFL teachers. In García
Mayo, M. and García Lecumberri, M., editors, Age and the acquisition
of English as a foreign language. Clevedon: Multilingual Matters,
182–209.
Wilcox, R. 2001: Fundamentals of modern statistical methods: substantially
improving power and accuracy. New York: Springer.
Yamada, R.A., Tohkura, Y. and Kobayashi, N. 1996: Effect of word familiarity on non-native phoneme perception: identification of English /r/, /l/,
and /w/ by native speakers of Japanese. In James, A. and Leather, J.,
editors, Second language speech. New York: Mouton de Gruyter, 103–18.
Appendix 1
Sample of words used in the phonological test
read
lead
weed
red
lead
wed
rook
look
wook
right
light
wight
royal
loyal
woyal
NOVE MR E IUD EC E MR E R 2000
632
Student Motivation to Learn
English as a Foreign Language
Kassim A. Shaaban
American University of Beirut
Ghazi Ghaith
American Univevsity of Beirut
Abstract: This study examined the motivation of 180 university-bound Lebanese students to
learn English as aforeign language (EFL). Data were gathered through administering a modified
version of the motivation scale developed by Wen (1997). Thefindings revealed that integrative
motivation, effort, valence, expectancy, and self-estimation of ability were internally related determinants of motivation for learning EFL. Instrumental motivation was found to be related to integrative motivation and valence only. The findings also revealed that female students were more
motivated than their male counterparts. Similarly, level 11 proficiency students were more motivated than were level III students. Howevel; thefindings did not show a significant effect on motivation related to either students’first foreign language or university majol:
Introduction
The role of learners’ motivation in the acquisition of languages other than their own has been a
subject of controversy in applied linguistics. Typically, second language acquisition (SLA) theorists tend to group motivation together with various aspects of personality and emotion as “affective” factors that play a role in language acquisition (Dulay et al. 1982; Stern 1983; Ellis 1985).
One corollary of this orientation is a focus on order of acquisition, developmental sequences,
and the role of biologically specified determinants of acquisition (universal grammar) over
which learners have no control.
However, the SLA theorists’ understanding of the role of motivation is limited, and our
understanding of motivation is likely to change, given widespread calls for helping students
become autonomous learners, that is, students who are involved in and responsible for their own
learning. Furthermore, many aspects of language learning are subject to learners’ active choices,
such as taking a course or not, communicating with native speakers, allocating attention, and so
forth. These considerations, coupled with the fact that learners’ cognitive strategies constitute a
strong determinant of acquisition (O’Malley et al. 1985), suggest that it is important to re-examine the role of motivation in the light of contemporary conceptualizations of language learning.
In their classic study of the role of motivation in second language acquisition, Gardner and
Lambert (1959) identified two kinds of motivation: integrative and instrumental. These
researchers maintained that integrative motivation signifies the learners’ desire to identify with
the target culture, whereas instrumental motivation refers to the need to fulfill a practical objective – such as obtaining employment. Later on, Gardner and Lambert (1972) and Gardner
(1980, 1985, 1988) proposed a socioeducational model of motivation that emphasized the influ-
Kassim A. Shaaban (Ph.D., University of Texas) is Associate Professor ofEnglish at the American
University of Beirut, Beirut, Lebanon.
Ghazi Ghaith (Ph.D., Indiana University) is Associate Professor of Language Education at the
American University of Beirut, Beirut, Lebanon.
FOKElGN LANGUAGE ANNALS * VOI,. 33, NO.6
ence of cultural beliefs on the development of the integrative motive and suggested the presence of a positive causal
link between integrative motivation and second language
achievement, as opposed to learning a second language for
instrumental purposes only where no such link exists. A
basic premise in Gardner’s model is that successful learners
are active and integratively motivated and that integrative
motivation is independent of aptitude. In addition,
Gardner et al. (1979) and Gardner (1985) developed the
Attitudehlotivation Test Battery (AMTB), thus “setting
high research standards and bringing L2 motivation
research to maturity” (Dornyei 1994, p. 277).
Research evidence synthesized by Oller (1981),
Gardner (1985), and Au (1988) yielded mixed results
regarding some key assumptions in Gardner’s model. This
is especially so with respect to the integrative motive and
causality hypothesis. These studies did not clearly support the superiority of integrative motivation in different
contexts; Kruidenier and Clement (1986) and Belmechri
and Hummel (1998) even maintain that instrumental
motivation is generally more prominent in EFL contexts
than integrative motivation. These findings have led to
the conclusion that the relationship between motivation,
as conceptualized by Gardner, and language learning is
an “unstable non-linear function that varies greatly across
individuals, contexts, and learning tasks” (Oller 1981, p.
15). Consequently, researchers began to deal with motivation in the second language classroom from the perspective of motivation structures and a broad yet contextspecific approach (Crookes and Schmidt 1991; Dornyei
1990; Wen 1997).
Dornyei (1990), in a study of adult learners of English
in Hungary, postulated a motivational construct “consisting of (1) an Instrumental Motivational Subsystem, (2) an
Integrative Motivational Subsystem, ( 3 ) Need for
Achievement, and (4) Attribution about Past Failures” (p.
45). The second and third components were seen as essential for an intermediate level of proficiency in the target
language, whereas the desire for going further was associated with integrative motivation. Furthermore, Dornyei
identified three loosely related aspects of the integrative
motivational subsystem, namely, (1) interest in foreign languages, cultures, and people; (2) desire to broaden one’s
view and avoid provincialism; and ( 3 ) desire for new stimuli and challenges. Along similar lines, researchers such as
Crookes and Schmidt (1991) and Oxford and Shearin
(1994) questioned the validity of existing motivational
models and suggested that the notion of second language
learning motivation should be broadened to include the
general psychology theories of integrative motivation.
In 1994, Dornyei elaborated on his earlier model of
motivation and suggested that it functions at three levels:
(1) language level, which comprises integrative and
633
instrumental motivation subsystems; (2) learner level,
which comprises need for achievement and self-confidence (as determined by language use anxiety, perceived
second language competence, causal attributions, and
self-efficacy); and (c) learning situation level, which
includes course-specific, teacher-specific, and group-specific components. Similarly, Wen (1997) incorporated
expectancy-value theories in his investigation of the
motivation of Asian and Asian-American students learning Chinese at American universities. Wen identified four
motivational factors that had a bearing on the enrollment
and continuation of students in Chinese language classes;
these were motivation of instrumentality, intrinsic motivation, expected learning strategies and efforts, and passivity towards requirements.
In the present study, we used Gardner and Lambert’s
conceptualization of integrative and instrumental motivations as well as expectancy-value theories (suggested by
Wen, 1997) to investigate the motivation of universitybound Lebanese students to learn EFL. The expectancyvalue theory was originally proposed by Lewin (1951).
This theory was further explicated by Vroom (1964),who
postulated that the effort exerted toward any action is
determined by the valence and expectancy that the action
would lead to the desired outcomes. Vroom (1964)
defined valence as “an affective orientation toward particular outcomes” (p. 14), and Lewin referred to it as “the
psychological value of particular outcomes” (cited in
Wen 1997, p. 236). Expectancy is defined as the “the
probability of attaining successful performance.” (Oxford
and Shearin 1994, p. 21). Thus, according to the
expectancy value theories, learners’ motivation to acquire
a second language is determined by their effort, perception of the degree of attractiveness of the goals (valence),
perception of the probability of attaining the goals
(expectancy), and appraisal of their ability to achieve the
goals.
The present study addressed the following questions.
1. To what extent are the determinants of learners’ motivation to acquire a foreign language (integrative motivation, instrumental motivation, effort, valence,
expectancy, ability) internally related?
2. Is there a difference between male and female learners
in their motivation to learn EFL?
3. Is there a difference among students with different levels of proficiency in their motivation to learn EFL?
4. Is there a difference in the motivation to learn English
between students whose first foreign language is
French and those whose first foreign language is
English?
5. Is there a difference among students in various majors
at the university (Arts, Sciences, Engineering, and
other fields) in their motivation to learn English?
634
The first question was prompted by our interest in
investigating the degree to which the motivational constructs identified in the literature (integrative and instrumental motivations, effort, valence, expectancy, ability) are
internally related. The second purpose was to determine
the specific motivational constructs on which students
may differ across the variables of gender, proficiency levels,
first foreign language, and university major.
Our review of the literature yielded mixed results
regarding the effect of gender on learners’ motivation to
learn a language other than their own. Several studies have
reported that females are more motivated and determined
to do academic work than males (e.g., Karsenti and
Thibert 1994; Zammit 1993; Zughoul and Taminian
1984). Conversely, Coleman (1995) maintained that the
gender-based differences in motivation are rather marginal, and Suleiman (1993) reported negative motivation attitudes to studymg English as a foreign language among
female Arab students. Furthermore, Sung and Padilla
(1998) concluded that there is need for further research to
determine gender differences in motivation.
Similarly, the literature includes conflicting evidence
with regard to the effect of proficiency level on motivation
to learn a foreign language. For example, Coleman (1995)
reported a slight but measurable relationship between integrative motivation and higher levels of foreign language
proficiency These findings are corroborated by those of
Boykin and Tmngamphai (1987), who maintained that
motivation and language proficiency are positively correlated. Moreover, Sung and Padilla (1998) reported that
“advanced level students [studying Asian languages]
scored significantly higher in instrumentauintegrative
motivation to learn a foreign language than did beginninglevel students” (p. 215). But can these findings be generalized to university-bound students studymg EFL? This
question seems pertinent, especially in light of Tweles’
(1995) report that students’ level of motivation was not
shown to correlate highly with proficiency
The study also attempted to provide empirical evidence regarding the intuitive assumptions in the Lebanese
multilingual context that students who study French as a
first foreign language and those who major in the “fields of
sciences” tend to be better motivated than those who study
English as a first foreign language and those who major in
the arts.
In light of the preceding discussion of the literature,
the present study proposed the following null hypotheses:
H01. There is no significant difference in motivation
to learn EFL between male and female students.
HO2. There is no significant difference in motivation
to learn EFL among students with different proficiency.
H03. There is no significant difference in motivation
to learn EFL between students who studied English as a
NOVEMRER/DECEMRER 2000
first foreign language and those who studied French as a
first foreign language.
H04. There is no significant difference in motivation
to learn EFL among students with various university
majors.
Method
Participants
One-hundred-eighty students (n = 180) enrolled in the
University Orientation Program (UOP), which offers
intensive English classes at the American University of
Beirut (AUB), participated in the study. This program is
intended to help upgrade students’ English language proficiency so that they may pursue their studies in the majors
that they have been accepted into at the University, where
English is the medium of instruction. The acceptance of
students into various majors at the University is normally
determined on the basis of their scholastic record and their
scores on SAT I and SAT 11. However, those students who
do not get the required score on English language admission tests (TOEFL, SAT Verbal, or AUB-EN) are asked to
join UOP
All the participants were native speakers of Arabic.
Of these, 41 students (23.8%) were enrolled in level I1
(Intermediate) and the remaining 139 (77.2%) were in
Level I11 (High Intermediate). There were 108 males
(60%) and 72 females (40%). Ninety-six students
(53.3%) had studied in French-medium schools where
French was the first foreign language and English was
taught as a second foreign language to make it possible
for them to attend English-medium universities. The
remaining 82 students (45.6%) had studied in Englishmedium schools where English was the first foreign language; only very few of these schools offer a second foreign language, normally French. Although students in
the latter group had studied English for over 10 years,
their proficiency remained low. This low proficiency may
be explained by the fact that they came from lower-middle-class families where the parents did not necessarily
speak any foreign language fluently; moreover, their
schools tend to stress “science” subjects at the expense of
foreign language programs.
Fifty-four students had been accepted into a major in
arts, 47 in sciences, 21 in engineering, and 54 in other
fields (education, health sciences, and agriculture); 4 students did not report the major. The age of the participants
ranged from 17 to 24 years, with a mean average of 18.51
and a standard deviation of 1.15. However, as will be seen
later in the study, the number of participants was reduced
as a result of the introduction of pair-wise deletion of missing cases when applied to the statistical analyses adopted
in the study
635
FOREIGN LANGUAGE ANNALS * VOL. 33, N O . 6
Instruments
The participants’ level of motivation to study English as a
foreign language was measured by a modified version of
the Wen (1997) scale (see Appendix). This version consisted of 40 items divided into three parts and had a general internal consistency (alpha reliability) of .83 based on
estimations from the present study. Part 1 included demographic questions (6 items) regarding the participants’ age,
gender, intended field of study, level of English proficiency,
native language, and first foreign language. Part 2 consisted of three subscales (16 items) measuring the factors of
integrative motivation, instrumental motivation, and
effort. Part 3 consisted of three subscales (18 items) measuring the factors of valence, expectancy, and ability.
The integrative motivation subscale consisted of five
Likert-type, 7-point items that focused on the appreciation
of the culture, art, and literature of English and had an
internal consistency of 3 9 . Similarly, the instrumental
motivation subscale consisted of five Likert-type, 7-point
items that focused on the importance of English for obtaining employment or pursuing further education and had an
internal consistency of .60. Meanwhile, the effort subscale
consisted of six multiple-choice items that focused on the
degree of effort exerted in learning English and had an
internal consistency of .63. Responses ranged from “a” for
no effort to “d” for very high effort. Furthermore, the
valence subscale consisted of five Likert-type, 7-point
items that focused on the participants’ views of the attractiveness of English for communication, completing school
assignments, and understanding English culture and customs and had an internal consistency of .72.
Finally, the expectancy and ability subscales consisted
of six items each and had internal consistencies of .78 and
3 0 , respectively. The expectancy subscale measured the
participants’ perceptions of the probability of achieving the
objectives of speaking English fluently, developing reading
comprehension, achieving good grades, and learning about
English culture and customs. Responses ranged from “0”
for no probability to “100” for very high probability. The
ability subscale measured students’ perceptions of their
ability to achieve the above objectives. Responses ranged
from “0” for very low ability to “100” for very high ability.
The internal consistencies of the six subscales were all
based on estimations from the present study.
Instrument Administration
The researchers obtained the permission of the UOP
administration to use a whole class period (40 minutes) for
the administration of the instrument. The purpose of the
questionnaire, its structure and content, and how it was to
be filled out were explained to the students. The
researchers asked the participants to be honest in their
answers and emphasized that they had no predetermined
expectations and that no answer would reflect better on the
respondent than another answer. The participants were
also informed that the findings would be used for research
purposes only and that their individual responses would
remain anonymous.
The questionnaire was read aloud to the participants
before they recorded their responses to ensure that they
understood each item on the scale. For example, the items
that made reference to English people, culture, and customs were explained as referring to people from all
English-speaking countries and not just to the British. The
reference to “Western culture” was explained as referring
not just to English-speaking people but to Europeans as
well. These concepts were not hard for the students to
grasp, as they are familiar with the West and the outside
world as a result of the multilingual and multicultural
nature of Lebanese society and its openness to western
ideas and cultures.
Data Analysis
Six composite scores of integrative motivation, instrumental motivation, effort, expectancy, valence, and perception
of ability were computed for each respondent by adding
the scores on the subscale items that measure these variables. A total motivation score was computed by adding
the scores on the 34 items of the scale for each respondent.
Descriptive statistics and product-moment correlation
coefficients were then computed for all variables in order
to determine the degree of interrelatedness among variables. In addition, we ran four multivariate analysis of variance (MANOVA) tests with the determinants of motivation
(integrative motivation, instrumental motivation, effort,
valence, expectancy, ability) as dependent variables in each
analysis and the variables of gender, proficiency level, first
foreign language, and university major as independent
variables (factors), to address the questions raised in the
study.
Results
All statistical tests used to address the questions in the present study used .05 as the minimum alpha level. This part of
the study presents descriptive statistics about the variables
as well as highlights from the intercorrelation matrix and
the results of the MANOVA analyses run in the study. The
mean scores and standard deviations on the dependent variables across the variables of gender, level, first foreign language, and university major are presented in Table 1, Table
2, Table 3, and Table 4, respectively.
Correlational Analysis
The results of the correlational analysis (summarized in
Table 5) reveal the following aspects of interest.
First, there is a very high, positive correlation between
636
N O V E M RER/DECEM KER 2000
learners’ expectancy and total motivation ( r = .91, p < . O l ) and between learners’ estimation of ability and total motivation ( r = .90, p < .01). The results also show a moderate positive correlation between expectancy and estimation of ability (r = .65, p < . O l ) . Second, the results reveal that valence has a low positive but measurable correlation with integrative motivation ( r = .46, p < . O l ) , with instrumental motivation ( r = .30, p < . O l ) , with effort ( r = .36, p < . O l ) , with expectancy (r = .49, p < . O l ) , with estimation of ability ( r = .34, p < . O l ) , and with total motivation ( u = .49, p < . O l > .
Third, the results reveal that effort has little if any correlation with estimation of ability ( r = .24, p < . O l ) , with expectancy ( r = .25, p < .Ol), and with integrative motiva- tion ( r = .17, p < .Ol). Finally, the results reveal that integrative motivation has a low, positive correlation with instrumental motivation ( r = .36, p < . O l ) , with expectancy ( r = .37, p < . O l ) , and with total motivation (r =.37,p r: .Ol). The results also reveal little if any correlation between integrative motivation and estimation of ability ( r = .25, p < . O l ) . Meanwhile, instrumental motivation was found to be unrelated to any of the determinants of motivation except to integrative motivation ( u = .36, p < . O l ) and to valence ( r = .30, p < .Ol). This suggests that instrumental motivation may not be perceived by learners as a strong motivating factor for exerting exert more effort and developing high expectancy and the ability to acquire languages other than their own. DESCRlPTlVE STATISTICS FOR RATINGS OF MOTIVATION DETERMINANTS BY GENDER Male Female (n = 65) (n = 100) Variable M SD M SD Instrumental 30.21 4.06 31.26 3.16 Integrative 21.77 4.84 22.36 4.96 Effort 17.11 3.15 18.44 2.84 Expectancy 363.80 82.70 371.84 77.31 Valence 30.80 5.96 32.47 4.96 Ability 396.60 83.82 384.76 72.88 Total 860.29 155.29 861.16 144.74 ___________ DESCRIPTIVE STATISTICS FOR RATINGS OF MOTIVATION DETERMINANTS BY LEVEL Level 11 Level 111 (n = 127) (n = 38) Variable M SD M SD Instrumental 31.34 4.10 30.40 3.64 Integrative 23.78 4.36 2 1.47 4.92 Effort 18.57 2.67 17.35 3.16 Expectancy 361.84 87.63 368.50 78.51 Valence 33.18 4.99 30.94 5.73 Ability 382.10 78.71 394.88 80.03 Total 850.84 154.05 863.56 150.82 FOREIGN LANGUAGE ANNALS * VOL.33, N O . 6 637 MANOVA Analyses perception of the valence of learning English, F(1, 166) = 3.67, p = .05. The mean effort score for femaleswas 18.44 (SD = 2.84), whereas the mean effort score for males was 17.11 (SD = 3.15). Similarly, the mean valence score for females was 32.47 (SD = 4.96), and the mean valence score for males was 30.80 (SD = 5.96). Level Scores. The results of the MANOVA scores for level 11 and level I11 participants showed that the second hypothesis is also rejected (see Table 7). The MANOVA analysis showed a significant difference between level 11 and level 111 participants in their motivation to learn English, F (6, 161) = .90, p = .01. Second, the Gender Scores. The results of the MANOVA scores for male and female participants showed that the first hypothesis is rejected (see Table 6). The MANOVA analysis revealed a significant difference between males and females in their motivation to acquire English as a foreign language, F(6, 161) = .91, p = .01. Second, the univariate analysis of variance showed no significant differences between males and females in their integrative motivation, instrumental motivation, expectancy, and estimation of ability. However, there was a significant difference in their effort, F(1, 166) = 7.75, p = .OO, and DESCRIPTIVE STATISTICS FOR RATINGS OF MOTIVATION DETERMINANTS BY FIRST FOREIGN LANGUAGE French (n = 90) Variable English (n = 73) M SD M SD lnstrumental 30.72 3.89 30.43 3.64 Integrative 21.74 4.89 22.93 4.89 Effort 17.42 3.13 17.82 3.04 Expectancy 369.55 76.71 363.15 86.24 Valence 31.58 5.95 31.24 5.31 Ability 394.22 80.14 386.57 77.74 Total 865.25 147.16 851.63 155.92 DESCRIPTIVE STATISTICS FOR RATINGS OF MOTIVATION DETERMINANTS BY MAJOR AT UNIVERSITY Sciences (n = 45) Arts (n = 50) Engineering Other (n = 51) (n = 19) ~~ ~~ ~~~~~~~ Variable M SD M SD M SD M SD Instrumental 31.20 3.38 30.02 3.76 29.36 5.09 31.05 3.46 Integrative 22.14 4.62 22.26 5.57 19.73 5.33 22.49 4.19 Effort 17.90 3.23 17.35 3.39 18.31 2.31 17.37 2.95 Expectancy 374.40 74.64 370.66 86.11 354.73 75.59 360.98 83.91 ~~~~~~~~~ Valence 31.46 4.71 31.35 6.07 31.94 5.40 31.37 6.26 Ability '388.40 72.29 388.66 101.59 394.73 74.86 397.25 67.44 Total 865.50 141.25 860.33 185.98 848.84 130.83 860.52 135.41 638 NOVEMBEWDECEMBER 2000 univariate analysis showed no significant differences between level I1 and level 111 participants in their instru- mental motivation, expectancy, or estimation of ability. However, there were significant differences with regard to INTERCORRELATIONS AMONG DETERMINANTS OF MOTIVATION Variable 1 2 3 4 5 6 7 1. Integrative 2. Instrumental .36** 3. Effort .17** .09 4. Valence .46** .30** .36** 5. Expectancy .37** .ll .25** .49** 6. Ability .25** .03 .24** .34** .65** 7. Total .37 ** .12 .31** .49** .91** .90** n 176 180 176 179 179 177 168 Mean 21.91 30.61 17.72 31.45 368.04 390.22 860.82 SD 4.89 3.73 3.03 5.66 80.06 80.96 152.08 ** Significant ai p c . 01 MULTIVARIATE ANALYSIS OF VARIANCE SUMMARY OF DETERMINANTS OF MOTIVATION BY GENDER Multivariate ANOVAa Source F Univariate AN OVA^ Integrative Instrumental Gender F a .91** .23 2.24 Effort Valence Expectation Ability 7.75** 3.67* .78 1.38 dk = (6, 161) bdfs = (1,166) *p c .05. **p c .01 MULTIVARIATE ANALYSIS OF VARIANCE SUMMARY OF DETERMINANTS OF MOTIVATION BY LEVEL Univariate AN OVA^ Integrative Instrumental Multivariate ANOVAa Source F Level dfs = (6. 161) hdfs = (1, 66) *p < .05. **p c .01 1 .90** F 7.16** 1.82 Effort Valence Expectation Ability 4.25* 4.27* .20 .73 FOREIGN LANGUAGE ANNALS * VOI,. 33, NO. 6 639 integrative motivation, F(1, 166) = 7.16, p = .OO; the integrative motivation mean score for level I1 was 23.78 (SD = 4.36) and for level 111was 21.47 (SD = 4.92). Furthermore, there were significant differences with regard to effort, F (1, 166) = 4.25, p = .04; the effort mean score for level I1 was 18.75 (SD = 2.67) and for level I11 was 17.35 (SD = 3.16). Finally, there were significant differences with regard to valence of the outcomes of learning English, F(1,166) = 4.27, p = .04; the valence mean score for level I1 was 33.18 (SD = 4.99) and for level I11 was 30.94 (SD = 5.37). These results indicate that students in this study with higher levels of proficiency in English have lower levels of integrative motivation. First Foreign Language Scores. The results of the MANOVA scores for first foreign language (presented in Table 8) show no significant differences between participants who had studied French as a first foreign language and those who had studied English, F(6, 159) = .97, p = .71. Therefore, the third hypothesis is accepted. University Major Scores. The MANOVA scores by university major (see Table 9) show no significant differences in the motivations of students majoring in the Arts, Sciences, Engineering, and other fields of study, F(18,474) = .89,p = .42. Therefore, the fourth hypothesis is accepted. Discussion This study set out to determine the interrelatedness of the motivational factors identified in the literature, as well as to examine the effects of gender, level of proficiency in English, first foreign language learned at school, and major of study on students’ motivation to learn EFL. The first question was about whether the determinants of motivation (integrative motivation, instrumental motivation, effort, valence, expectancy, and ability) are internally related. The results of the correlational analysis (see Table 5) indicate that students’ expectancy to achieve their goals and their perception of their ability to achieve those goals are strong determinants of their motivation. This finding suggests that these two factors (estimation of ability and expectancy) should be used as variables in future research on motivation and second/foreign language learning. Furthermore, these results indicate that learners need to feel that their study program goals are attainable and that they are capable of achieving them in order to increase MULTIVARIATE ANALYSIS OF VARIANCE SUMMARY OF DETERMINANTS OF MOTIVATION BY FIRST FOREIGN LANGUAGE Multivariate ANOVAO Source F Univariate AN OVA^ Integrative Instrumental Effort Valence Expectation Ability First Foreign Language F .67 .14 .06 .12 .97 1.12 .06 adfs = (6, 159) bdfs = (1,164) * p c .05. **p < .01 I MULTIVARIATE ANALYSIS OF VARIANCE SUMMARY OF DETERMINANTS OF MOTIVATION BY MAJOR AT UNIVERSITY I ~ Univariate AN OVA^ Integrative Instrumental Multivariate ANOVAa Source F Major a dfs = (18,474) bdk = (3,161) *p c .05. **p c .01 .89 F 1.60 1.73 ~ ~ ~ ~ ~ ~~ Effort Valence Expectation Ability .66 .05 .40 .13 NOVEMKEIUDECEMBER 2000 640 their motivation. Along similar lines, the study goals should be made attractive to students in order to motivate them to learn a language other than their own. The pedagogical implications of these results underscore the importance of affective factors in language acquisition and the need for setting up developmentally appropriate programs and specifying interesting and relevant goals for these programs. On the other hand, the results revealed a low but measurable correlation among all the determinants of motivation, except for instrumental motivation. This finding sheds some light on the traditional dichotomy between integrative and instrumental motivation and the role of each one in learning a foreign language. It is clear, for example, that integrative motivation correlates positively with all the other variables, including instrumental motivation, and with total motivation, a finding that could explain why it is identified in the literature (Gardner and Lambert 1959; Gardner 1985; Dornyei 1990) as a major indicator of positive orientation towards the target language that could lead to better achievement. Instrumental motivation was found to correlate only with integrative motivation and valence; it did not correlate with effort, expectancy, perception of ability, or total motivation. In other words, the presence of instrumental motivation in students may not mean that they will in fact exert more effort or develop high expectancy or high ability with regard to EFL learning; however, instrumental motivation may contribute to the enhancement of integrative motivation and the development of positive affective orientation towards particular learning outcomes. This finding of the limited effect of instrumental motivation in the Lebanese EFL context is in contrast to the findings of Kruidenier and Clement (1986) and Belmechri and Hummel (1998), who found instrumental motivation determined by an instrumental orientation to be generally prominent in foreign language contexts. The second question related to gender differences in motivation. Results of the MANOVA analysis indicated that there were differences between males and females in their motivation to learn English. More specifically, the univariate analysis indicated that females were more likely than males to report that they would exert effort in their learning and to perceive the goals of learning English in a positive manner. This corroborates the findings of Sung and Padilla (1998), Karsenti and Thibert (1994), and Zammit (1993), who reported similar results; however, they contradict the findings of Suleiman (1993), who reported negative attitudes to learning English among female Arab EFL learners. One possible explanation for the negative attitudes of females in Suleiman’s study towards English is that these attitudes could be context-specific and may vary due to sociocultural factors. In fact, Zughoul and Taminian (1984), who conducted their study in a similar context, found that Arab university students in general tended to have negative attitudes towards English, which they perceived as a threat to Arabic - their native language and the language of the Koran. In their study, there were no statistically significant differences between males and females in their attitudes towards the use of English as a medium of instruction, towards the effect of using English on Arab identity, and towards the utility of English compared with Arabic. However, when comparing means of responses, the authors concluded that females tended to have a more open attitude towards English than did males. The third question raised...