~
System, Vol. 23, No. 3, pp. 359 386, 1995 Copyright © 1995 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0346-251X/95 $9.50 + 0.00
Pergamon 0346-251X(95)00023-2
ADULTS' LANGUAGE LEARNING STRATEGIES IN AN INTENSIVE FOREIGN LANGUAGE PROGRAM IN THE UNITED STATES
REBECCAL. OXFORD*and MADELINEE. EHRMAN? *University of Alabama, U.S.A. ?Foreign Service Institute, U.S.A. Exploring a sample of adult language learners, this article shows the relationships between language learning strategies on the one hand and factors such as proficiency, teacher perceptions, gender, aptitude, learning style, personality type, ego boundaries, motivation, and anxiety on the other. This is an almost "optimal" group of language learners. Because of the sophisticated nature of this group of learners, statistical restrictions of range lowered the correlations, which are nevertheless significant and which point to implications for both instruction and further research.
INTRODUCTION This study explores the use of learning strategies as an important factor in the success of adult learners of foreign languages. What strategies or behaviors do adult language learners use? How are these strategies related to language learning proficiency and to other factors such as teacher perception, gender, aptitude, learning style, personality type, ego boundary, motivation, and anxiety? These are among the questions asked in this investigation. The larger project of which this study is a part (see Ehrman and Oxford, 1995) involves adults at the Foreign Service Institute (FSI, the training branch of the U.S. Department of State) learning a variety of languages intensively for 8-88 weeks. This is one of only a few language learning strategy studies involving such a wide range of factors; most studies examine just a handful of variables, not a broad or comprehensive set. One purpose of the current investigation is to detemine how these individuals' language learning strategies correlate with other factors--language proficiency and various cognitive, affective, and social traits. An ultimate purpose is eventually to provide profiles for diagnosis, prediction, counseling, and teacher training. This article is organized as follows: review of previous research, research questions, methodology, results and interpretation, and summary and implications. The students involved in the study are highly educated and motivated and are thus not completely representative of all foreign language learners. In some ways they might represent an 'optimal' foreign language learning group. The correlational information gained from this group should provide a basis for future predictive studies with the same FSI data, and it should be helpful in suggesting trends or hypotheses to be tested with different populations.
359
360
REBECCA L. OXFORD and MADEL1NE E. EHRMAN
REVIEW OF PREVIOUS RESEARCH Effective instruction in foreign or second languages requires an understanding of individual differences among learners (see, e.g., Ehrman, 1990a, in press-a; Galbraith and Gardner, 1988; Oxford, 1992b, 1993b; Oxford and Ehrman, 1993; Scarcella and Oxford, 1992; Skehan, 1989). Gradman and Hanania (1991) identified 22 learner-related factors that significantly contribute to success in foreign language learning. Here we summarize the existing research on key individual differences, knowing that relationships are typically complex.
Aptitude In the mid-sixties, educational leaders such as Lee Cronbach and Robert Glaser led a conference demonstrating that individual learners differ profoundly in exactly how they learn and in their success in a given situation (see Gagne, 1967). Since that time, much research has concentrated on aptitude-treatment interactions (ATls)--that is, the ways in which the person's aptitude interacts with the instructional mode (see Cronbach and Snow, 1977; Snow, 1989). Instructional treatment in ATI research includes such aspects as degree of structure, relative focus on verbal-symbolic vs. spatial-figural presentation, and the nature of the classroom climate, grouping, and evaluation. (While these are important, we might also suggest adding intensiveness of the instructional program.) As used in ATI research, the term aptitude includes not just cognitive aspects of general and specific ability; it also includes "personality, motivational, attitudinal, and anxiety-related aptitudes" (Snow, 1989) and may thus serve as a shorthand for an array of individual learner differences (see Oxford & Ehrman, 1993). Research has demonstrated that complex ATIs exist and that these interactions can be improved in part by teaching students to use more effective learning strategies (Snow, 1989). ATI research has not taken hold in the language instruction field, except in a few studies that have tried to place students into certain instructional methodologies--rather successfully--according to subtests of language aptitude measures (Skehan, 1989; Wesche, Edwards and Wells, 1982). Although not unsually flamed in ATI terms, the whole issue of language aptitude as a cognitive ability is extremely important, especially now when university and high school language requirements are being reinstated and increasing numbers of students of varying language ability are entering language classes. For over three decades, language aptitude has been virtually synonymous with factors measured by the Modern Language Aptitude Test (MLAT, Carroll and Sapon, 1959). This test, designed for ages 14 and above and created through factor analysis of large numbers of items, has the following components: phonemic coding ability (making soundsymbol linkages), grammatical sensitivy (recognizing grammatical functions of words in sentences), inductive language learning ability (identifying patterns by reasoning and inferring), and rote memorization (making associations between stimulus and response). These elements are combined in different ways to form five subtests: phonetic script, spelling clues, words in sentences, number learning, and paired associates. Research shows that the MLAT discriminates well at the lower levels but less well at the higher levels (Ehrman, forthcoming). It is criticized for being too analytic and thus possible less relevant to flexible, communicative classrooms than to highly structured, traditional classrooms (see Parry and Stansfield, 1990). However, it has predictive power (0.20-0.80, according to Carroll and Sapon, 1959; 0.40-0.60 according to
ADULTS' LANGUAGE LEARNING STRATEGIES
361
Skehan, 1989) in many settings, with moderate predictive power (0.50s) for intensive courses with heterogeneous students. Recent research in largely communicative classes shows the same level of predictive power (Ehrman and Oxford, 1995). Another aptitude test for adults in the military services, the Defense Language Aptitude Battery or DLAB, has just about the same predictive power as the MLAT in the DLAB's own limited setting, according to Skehan (1989). As described by Skehan (1989) and Oxford and Ehrman (1993), the Pimsleur Language Aptitude Battery (the Pimsleur) was designed for ages 13 through 18. The Pimsleur focuses on inductive language learning ability and auditory ability, without the MLAT's additional grammatical sensitivity and memorization aspects. Thus, the Pimsleur rests on a more restricted concept of language aptitude. Aptitude researchers have usually avoided predicting language achievement in distinct skill areas (e.g. listening or reading). Types of language aptitude research have included validation of existing batteries, development of new aptitude tests for specific countries, exploration of the minimal relationship between working memory and language aptitude, investigation of relationships between first and second language aptitude, discovery of strong relationships between socioeconomic status and language aptitude, and placement of students into different methodologies based on MLAT and Pimsleur subtest scores (Skehan, 1989). In the late 1980s, the U.S. government gathered language testing specialists to a conference on language aptitude measurement, resulting in a book by Parry and Stansfield (1990). Each chapter author in the Parry-Stansfield volume makes recommendations to improve language aptitude assessment. One chapter author, John B. Carroll, the chief MLAT creator, suggests only minor changes in the MLAT, but some other chapter authors call for expanding the concept of language aptitude to include many new predictors, such as general ability measures, learning strategies, learning styles, anxiety, attitude, and motivation. This brings us back to the ATI research, which came to very similar conclusions about the preferred scope of general aptitude. A suggestion at the aptitude conference was that one set of predictors might be applicable for lower levels of proficiency and a different set for higher levels of proficiency.
Learning styles, personality type, and sensory preferences Studying more than 300 university students learning several different languages (German, Russian, and French), Wildner-Bassett (1992b) found that students who often used cognitive strategies as measured by the Strategy Inventory for Language Learning, Version 5.1 for native speakers of English, significantly preferred a low-mobility learning style as assessed by the Learning Style Profile (LSP, Keefe and Monk, with Letteri, Languis and Dunn, 1989). WildnerBassett found that students who reported not using strategies to compensate for missing knowledge showed a preference for emotive-kinesthetic responses on the LSP. Use of SILL affective strategies was significantly related to the LSP's wanting to learn in a warm temperature. SILL metacognitive strategy use was negatively predicted by LSP spatial processing and verbal risktaking using regression techniques; spatial processors and risk-takers tended not to use highly structured, organized metacognitive strategies. Personality type shows strong relationships with strategy use on the SILL, Version 5.1. In an analysis of variance involving 79 language learners, teachers, and supervisors in an intensive training setting,
362
REBECCA L. OXFORD and MADELINE E. EHRMAN
Ehrman and Oxford (1989) found that those with an extraverted personality type (as measured by the Myers-Briggs Type Indicator or MBTI) were more likely than introverts to use affective and visualization strategies, while introverts more frequently than extraverts used strategies for searching for and communicating meaning. In the same study, the researchers found that intuitivetype people (interested in patterns, abstractions, and possibilities), compared with sensing-type people (present-oriented and interested in facts), used more strategies for searching for and communicating meaning, building mental models of the language, using the language for authentic communication, and managing emotions. This study showed that judging-type individuals (those who need organization and closure), compared with perceiving-type people (those who want freedom and dislike closure), more frequently used general study strategies, but perceivers compared with judgers used more strategies for searching for and communicating meaning. Sensory preferences are directly related to language learning strategy use, as shown by the research of Rossi-Le (1989). Using multivariate analysis of variance with 147 learners of English as a second language, she discovered a significant relationship between sensory preference (visual, auditory, tactile, and kinesthetic) and overall SILL (Version 7.0) strategy use and a number of other significant relationships. For example, auditory learners used memory strategies significantly more often than did students who preferred visual, tactile, and kinesthetic senses. Tactile students, compared with others, showed significantly greater use of strategies for searching for and communicating meaning and for self-management. Moreover, multiple regression results indicated that having a certain sensory preference significantly predicted the types of strategies students chose.
Learning strategies Learning strategies, the steps students take to improve their own learning, are very important to ultimate language performance. Many studies indicate that the frequency of use of language learning strategies directly relates to language performance, regardless of whether performance is measured as a course grade, a class test score, a standardized proficiency test score, a self-rating, or something else (see Oxford and Burry, 1993). The "good language learner" studies repeatedly suggested that successful learners tended to use strategies such as finding practice opportunities, guessing intelligently, using patterns, treating the language as a rule system, and communicating often in the language (see, e.g. Rubin, 1975; Naiman, FrOhlich, Stern and Todesco, 1978). Recent research suggests that there is no single strategy pattern used by effective language learners. In fact, successful learners use an array of strategies, matching those strategies to their own learning style and personality and to the demands of the task (in the context of cultural influences). Optimal learners find ways to tailor their strategy use to their individual needs and requirements; they develop combinations of strategies that work for them (Oxford, 1990). Strategy training can be very useful in improving the use of language learning strategies (Oxford, 1990; O'Malley and Chamot, 1990). According to research (Oxford, 1992/1993), effective strategy training is usually highly explicit instead of being "blind" or "implicitly embedded". Strategy training that works is generally integrated into courses rather than serving as a separate minicourse. Learners must have practice in how to use, adapt, evaluate, and transfer a strategy to new situations and tasks.
ADULTS' LANGUAGE LEARNING STRATEGIES
363
Strategy classification systems are many and varied. The proliferation of strategy systems has caused problems for those researchers who believe it is important to compare results across studies. Efforts are now underway to make the learning strategy field more coherent by establishing a more rational way of defining and explaining such strategies (Oxford and Cohen, 1992).
Motivation Two motivational orientations were emphasized in the research of Gardner and Lambert (1972): (1) instrumental, for enhancing career or academic progress, and (2) integrative, for fitting in with people who speak the language natively. Crookes and Schmidt (1989) stated that language learning motivation includes seven aspects: (1) interest, (2) perception of relevance, (3) expectancy of success or failure, (4) perception of rewards, (5) overt decision to learn, (6) persistent learning behavior, and (7) high involvement. Oxford and Shearin (1994; see also Beck, 1990) have expanded the theory of language learning motivation. Some motivation is based on need for achievement, other motivation is founded on fear of failure, and still other motivation relates to fear of success. Language learning motivation will be high only if expectancy of success and value of success are high. Students must believe that the outcome is at least equal to the input (effort). Language learning motivation is influenced by students' self-efficacy and attribution of "locus of control" (to fate, society, God, or self). For optimal motivation, the goals must be clear, challenging, and reachable, and there must be feedback on goal achievement. Motivation is shaped partly by the mode of instruction: a mastery learning mode in which all students are encouraged to reach intrinsically important goals might be more motivating for many students than a norm-referenced mode or a setting in which all rewards are external. Language learning motivation helps determine the frequency with which learners use strategies. Motivation was the most significant factor influencing language learning strategy use in a study of 1200 university students (Oxford and Nyikos, 1989) and was also strongly related to learning strategy use among 107 high school students of Japanese (Oxford, Park-Oh, Ito and Sumrall, 1993a, 1993b).
Age and gender Age is often mentioned as an influence on language learning success (Singleton, 1989; Ehrman, 1987) and on choice of language learning strategies (Oxford, 1982, 1989; Oxford and Ehrman, 1993). In short, younger learners are more likely to attain fluency and native-like pronunciation through communicative practice strategies. Because of their more developed abstract thinking capabilities, older language learners often use strategies that allow them to analyze the grammatical system and to apply greater "world knowledge" to the language learning context. Advantages of language learners at different ages are attributed to: one or more critical periods for language learning, prior experience in language learning, onset of formal operations, cognitive maturity, kind of input, affective factors, and sociocultural factors. Females more often than males tend to use language learning strategies. This has been repeatedly found in recent studies of language learning strategy use around the world (Politzer, 1983; Green, 1991; Green and Oxford, 1993; Oxford, 1993a, 1993b, in press; Oxford and Nyikos, 1989; Oxford, Ehrman and Nyikos, 1988; Oxford, Park-Oh, Ito and Sumrall, 1993a, 1993b). Typically,
364
REBECCA L. OXFORD and MADELINE E. EHRMAN
males have not been shown to exceed females in the use of any general category of language learning strategies, although males have exceeded females in use of a handful of particular strategies (Bedell, 1993; Green and Oxford, 1993). Few studies have been conducted on actual foreign or second language performance of males and females, but those that exist suggest that females exhibit better listening skills than males (Larsen-Freeman and Long, 1991).
Anxiety, self-esteem, tolerance of ambiguity, and risk-taking Horwitz and Young (l 991) clearly show the potentially debilitative effects of language anxiety, a form of performance anxiety that manifests in the language learning situation. Lavine and Oxford (1990) describe symptoms of anxiety. Horwitz (1990) suggests that in the language learning environment, there is no such thing as facilitative anxiety; all anxiety in that environment is likely to be debilitative. However, Brown (1987) suggests that some degree of anxiety can indeed be facilitative for language learning, a point of view confirmed for adult intensive language students (Ehrman and Oxford, 1995). Little or no empirical research has been done linking anxiety and the use of language learning strategies, but Schmeck (1988) hypothesizes that highly anxious, extrinsically motivated students show a "shallow learning style" that is associated with strategies such as repetitive rehearsal, rote memory, and use of mnemonic devices. (In contrast, Schmeck' s "deep learning style" and "elaborative learning style" are twin aspects of what we might call "deep processing", which involves transformation of new material through a variety of analytic, synthetic, and personalizing mechanisms.) Self-esteem is a judgment of self-worth or value based on feelings of efficacy (White, 1959), with higher self-esteem related to perceptions of higher efficacy. Efficacy implies that some degree of control lies within oneself and is not strictly external (Rotter, 1966). Belief in one's own efficacy or internal control influences the amount of effort one devotes and the amount of risk one accepts. Global self-esteem is distinguished from situational self-esteem (Brown, 1987). Global self-esteem refers to the students' broad assessment of themselves as people and as learners in general. Situational self-esteem refers to the students' self-perception in particular language learning situations or in reference to attaining a particular skill. Research suggests that teachers can help learners develop positive situational self-esteem by assessing their progress realistically (Oxford, 1990) and by discounting, when possible, the areas in which students do not have peak performance (Hatter, 1986). For example, if--as is almost always the case for adult foreign or second language learners--native-like perfection in all aspects of pronunciation is an impossible aim, then teachers can help learners discount this goal and can assist them in attaining the more realistic goal of appropriate stress and intonation. Such greater realism enhances learners' situational self-esteem in language learning. Tolerance of ambiguity is acceptance of confusing situations (Ely, 1989). Students who can tolerate moderate levels of ambiguity are more likely to persist in language learning than students who cannot (Chapelle, 1983; Naiman, Frrhlich, Stem and Todesco, 1978). Ambiguity tolerance is related to the frequency of use of many kinds of learning strategies (Ehrman and Oxford, 1989, 1990; Ely, 1989). Tolerance of ambiguity is related to risk-taking ability. Those who can tolerate ambiguity are more likely to take some risks in language learning; and risk-taking is an essential for progress (Beebe, 1983; Brown, 1987; Ely, 1986; Stevick, 1976). Students who avoid risks are influenced by anticipated criticism from others or by self-criticism, and their language practice
ADULTS' LANGUAGE LEARNING STRATEGIES
365
becomes stunted. Note that both tolerance of ambiguity and risk-taking are sometimes viewed as aspects of learning style, as shown earlier; however, sometimes they are seen simply as personality variables.
Ego boundaries A new area of exploration in language learning is ego boundaries, measured by the Hartmann Boundary Questionnaire or HBQ (Hartmann, 1991, described later). This concept has been related to Guiora's (1981) concept of language ego, to cognitive flexibility, and to tolerance of ambiguity--though in fact it includes all of these. In general, the more unstructured the learning, the more important this characteristic (Ehrman, in press-a). Ehrman relates the construct of "ego boundaries" to four abilities: the ability to take in new information, the ability to hold contradictory information in mind without rejecting either of the clashing elements, the ability to avoid premature closure on incomplete information, and the ability to adapt one's existing cognitive, affective, and social schemata in the light of new information. Hartmann (1991) describes individuals with thinner boundaries as more "fluid", more cognitively and socially flexible, less guarded, more vulnerable, more open, more creative, more intuitive, more able to shift strategies easily, more accepting of new stimuli, more (seemingly) easy-going, less perfectionistic, less neat, and more prone to nightmares. In the extreme, thin-boundary people can appear overly sensitive, unreliable, or out of touch with daily realities. Many artists, writers, and psychotherapists are in the thin boundary group. In contrast, people with thicker boundaries are described as more "solid", less flexible, more guarded, less vulnerable, more closed, less creative, less intuitive, less able to shift strategies (even when doing so is obviously desirable), less willing to accept new stimuli, more hard-driving, more perfectionist, neater, and less likely to have nightmares. In the extreme, thick-boundary people can be concrete-thinking, mechanistic, and in need of specific guidelines. Many engineers and military personnel are among the thick boundary group. Hartmann (1991) found that women and younger people consistently scored thinner than men and older people. Ehrman (in press-a) discovered that in the sample from which the subsample in this study comes, females showed a slightly stronger tendency than males toward thin boundaries. She also found significant, low to moderate (0.20s-0.50s) relationships between thinner boundaries and (1) better language learning performance in speaking and reading; (2) aspects of language aptitude; (3) language anxiety; (4) personality type aspects including intuition, feeling, and perceiving; and (5) learning style aspects such as inability to discriminate well. Thus, thinboundary people might have a higher tolerance for ambiguity, but their sensitivity does not prevent anxiety. Until the current study, there were no data on the linkages between ego boundaries and the use of language learning strategies (see results section).
Cultural background Cultural background, related to ethnicity or nationality, is a key factor in language learning strategy use (Bedell, 1993). Politzer (1983) found that Hispanics and Asians differed strongly in the kinds of strategies they used for language learning. Hispanics chose more social, interactive strategies, while Asians opted for greater rote memorization--often based on their previous school experiences as dictated by their cultures. Cultural differences in learning strategies and styles are discussed by Bedell (1993), Reid (1987), and Oxford, Holloway, and Horton-Murillo (1992). In our study, the learners came from North American mainstream cultural backgrounds; readers can compare the results with the findings from studies representing other cultures in Bedell (1993).
366
REBECCA L. OXFORD and MADELINE E. EHRMAN
RESEARCH QUESTIONS The main research questions in this study are: °
2. 3. 4. 5.
What language learning strategies do adults use and how often? How are these language learning strategies related to each other? Does the use of certain language learning strategies relate to proficiency? Does strategy use correlate with teacher perceptions? Is the use of language learning strategies related to other factors? These include: • gender • educational background, previous languages, weeks of training, and language difficulty • general learning and study strategies • aptitude • learning styles • personality type • ego boundaries • motivation and anxiety.
METHODOLOGY
Sample The overall sample at the time of the current substudy consisted of 520 people, 273 (52.5%) male and 247 (47.5%) female. (Half of these participated in the learning strategy substudy reported here. Their characteristics generally matched those of the overall sample.) Most of the total group (381 or 73.3%) were from the Department of State, while 47 or 9% were from the Defense Department, 30 or 5.8% were from the U.S. Information Agency, 22 or 4.2% were from the Agency for International Development, and the rest (7.7%) were from other government agencies like the Department of Agriculture, the Department of Commerce, and the Drug Enforcement Agency. Most were employees, but a few were spouses or dependents. The participants were highly educated. Forty percent (208) had master's degrees, while another 34% (176) had bachelor's degrees. Approximately 16% (83) had either a high school diploma or a 2-year associate degree. Approximately 7% (35) had law degrees or doctorates. Three percent were unaccounted for educationally. For 99% of the sample, English was the native language. Others spoke Spanish, Cantonese, German, or Romanian natively. The group was composed of highly experienced language learners. About one-quarter (125 or 24%) had studied three or more foreign languages previously, 30% (154) studied two foreign languages previously, 31% (160) studied one foreign language previously, and only 16% (81) studied no foreign languages previously. Participants were asked to note their highest speaking and reading proficiency ratings for languages previously studied. The mean highest speaking rating was 2.5 (SD 1.2), and the mean highest reading rating was 1.84 (SD 1.1) on the Interagency Language Roundtable (ILR) 0-5 rating scale. Thus, the sample was better at speaking than reading in past language study. Sample members noted 34 languages they were studying at FSI at the time of the research. Slightly less than one-third each were learning Spanish (154 or 29.6% and French (148 or 28.5%). The other
ADULTS' LANGUAGE LEARNING STRATEGIES
367
32 languages being learned by the remaining participants were, in order of numbers, Italian, Portuguese, Chinese, Arabic, Russian, Thai, Turkish, Hebrew, German, Dutch, Urdu, Indonesian, Burmese, Polish, Romanian, Serbo-Croatian, Greek, Japanese, Korean, Czech, Danish, Hindi, Bengali, Afrikaans, Finnish, Norwegian, Bulgarian, Cantonese, Lao, Swahili, Swedish and Tagalog. Mean length of training was 20 weeks (SD 11.5), with a range of 3--44 weeks. Modes were 8 weeks, 23 weeks and 44 weeks, corresponding to the maximum duration of the FSI courses. Scheduled length of training depends on the FSI's experience with how long it takes a typical candidate to reach level 3. Western European languages are 23-week courses; Chinese, Japanese, Korean and Arabic (considered to be Category 4 languages, or hardest languages for native English speakers) are 88-week courses (one year in Washinghton and one year at an overseas FSI school); the others are 44 weeks. Eight-week students completed familiarization courses. All students in this study were in full-time intensive training.
Instruments Below are descriptions of the instruments in this study, listed in alphabetical order by title. Table 1 summarizes key aspects of the instruments used in this study. Not all subjects took all instruments; numbers taking each instrument are shown in parentheses below.
AS (N = 47). The Affective Survey (AS), which has 114 items, was developed by Madeline Ehrman and Rebecca Oxford (1991) with general ideas and in some instances adapted items from a variety of surveys by Robert Gardner, Christine Campbell, Elaine Horwitz and others. The authors recognized that no single survey covered all the important affective (emotional and motivational) areas related to language learning success, and they needed to create such a measure. The AS contains three parts: motivation (extrinsic, intrinsic, desire to use the language, and effort) beliefs about self, and anxiety (as related to public performance, language use with native speakers, making errors, comprehension, self-esteem, competition, tests, outcomes, and general comfort-discomfort with language learning). The AS also has the option of a "negativity scale", which indicates how often a person agrees with negatively worded items about motivation and anxiety. Reliability is 0.74 (Cronbach alpha) and 0.82 (standardized item alpha).
HBQ (N = 129). The Hartmann Boundary Questionnaire or HBQ (Hartmann, 1991) has 147 items and covers the following dimensions: sleep/dreams/wakefulness, unusual experiences, boundaries among thoughts/feelings/moods, impressions of childhood/adolescence/adulthood, interpersonal distance/openness/closeness, physical and emotional sensitivity, preference for neatness, preference for clear lines, opinions about children/adolescents/adults, opinions about lines of authority, opinions about boundaries among groups/peoples/nations, opinions about abstract concepts. The purpose of this questionnaire is to determine the 'thinness' or 'thickness' of a person's mental boundaries. Hartmann found that women had thinner boundaries than men. Older people scored consistently thicker than younger people. Cronbach's alpha reliability for the HBQ is 0.93, and theta reliabilities for subscales are 0.57-0.92 (Hartmann, 1991).
LASSI (N = 59). The Learning and Study Strategies Inventory or LASSI, developed by Weinstein, Palmer, and Schulte (1987), contains 90 items and contains these factors: test strategies and test preparation, attitude and interest, concentration and attention, information processing, selecting
368
REBECCA L. OXFORD and MADELINEE. EHRMAN Table 1. Characteristics of instruments N in This Study
Name, Author, Date
No. of items
Reliability
Validity
Strategy Inventory for Language Learning (SILL, Oxford, 1989)
80 for foreign language learning version as used here (50 for ESL version)
0.89-0.98 for whole measure using Cronbach alpha
High predictive validity for course grades, proficiency tests, and other language measures; concurrent validity with LASSI (0.20-0.50); ANOVA effects between MBTI learning style/personality type and SILL strategies (p < 0.05); ANOVA effects between SILL strategy use and motivation (p < 0.0001)
268
Myers-Briggs Type 126 for MBTI Form G Indicator (MBTI, Myers 290 for TDI and McCaulley, 1985) Long Form: Type Differentiation Indicator (TDI, Johnson, 1991; Sanders, 1989)
0.87 split-half for MBT1; 0.70-0.85 testretest for MBTI; approx. 0.50 internal consistency for each TDI scale
Concurrent validity of MBTI with personality, vocational preference, educational and management style measures 0.404).77; Construct validity of MBTI supported by vocational and creativity studies
137
Learning and Study Strategies Inventory (LASSI, Weinstein, Palmer and Schulte, 1987)
0.68-0.86 for subscales using Cronbach alpha; 0.72-0.85 testretest
Construct validity as a sound diagnosis tool for college/university students. Norms available
59
147
0.93 Cronbach alpha; 0.57-0.92 theta for subscales
Concurrent validity related to gender and age
129
Learning Style Profile 125 (LSP, Keefe and Monk, with Letteri, Languis and Dunn, 1989)
0.47-0.76 Cronbach alpha for subscales; 0.36--0.82 testretest
Concurrent validity of analytic subscale with GEFT 0.39; of perceptual subscales with ELSIE 0.51-0.64; of environmental subscales with LSI 0.23-0.71
135
Modern Language 146 Aptitude Test (MLAT, Carroll and Sapon, 1959)
0.92-0.97 splithalf
Predictive validity with course grades for GPA 0.20s to 0.80s. Better prediction for lower than higher levels. Prediction in 0.70s for intensive training with heterogeneous students
167
Affective Survey, (AS, Ehrman and Oxford, 1990)
114
0.73 Cronbach alpha; 0.84 standardized item alpha
Concurrent validity with proficiency, learning style, learning strategies, aptitude, etc. (0.20-0.60)
47
Teacher Ratings (Ehrman, 1990)
7
High interrater reliability
Concurrent validity with proficiency, learning style, learning strategies, aptitude, etc. (0.20-0.60)
86-90 dep. on which rating item
Hartmann Boundary Questionnaire (HBQ, Hartmann, 1991)
90
ADULTS' LANGUAGE LEARNING STRATEGIES
369
the main idea, performance anxiety, time management, self-testing and class preparation, motivation and effort, and use of support techniques and study aids. The norm group was comprised of college and university students. Subscale reliabilities (Cronbach alpha) range from 0.68 to 0.86, and test-retest reliabilities are 0.72 to 0.85 (Weinstein, 1987). This test is widely used as a diagnostic measure to help students improve their study techniques.
LSP (N = 135). The Learning Style Profile or LSP (Keefe and Monk, with Letteri, Languis and Dunn, 1989) is a composite measure with 125 items, composed of many different approaches to measuring learning style. The main subscales are cognitive skills (analytic, spatial, categorization, sequential processing, detail memory, discrimination), perceptual response (i.e. sensory preferences--visual, auditory, emotive/kinesthetic), orientations (persistence, verbal risk-taking, manipulative), study time preferences (early morning, late morning, afternoon, evening) and environmental context for learning (verbal vs spatial, posture, light, temperature, mobility and grouping). Cronbach' s alpha for the subscales ranged from 0.47 to 0.76, with an average of 0.61. Test-retest reliabilities were 0.36 to 0.82 after 10 days and somewhat lower after 30 days. Concurrent validity of the LSP's analytic subscale with the Group Embedded Figures Test (GEFT) was 0.39, p < 0.002. Concurrent validity of the perceptual response subscales of the LSP with the Edmonds Learning Style Identification Exercise (ELSIE) were 0.51 ~).64, p < 0.002. Many of the environmental context subscales of the LSP correlated with Dunn and Dunn's Learning Style Inventory, 0.23-0.71, p < 0.04~0.002. MBTI or TDI (N = 137). The Myers-Briggs Type Indicator or MBTI Form G (Myers and McCaulley, 1985) is a 126-item, forced-choice, normative, self-report questionnaire designed to reveal basic personality preferences on four scales: extraversion-introversion (whether the person obtains energy externally or internally), sensing-intuition (whether the person is concrete/sequential or abstract/random); thinking-feeling (whether the person makes decisions based on objective logic or subjective values); and judging-perceiving (whether the person needs rapid closure or prefers to keep decisions open). Internal consistency split-half reliabilities average 0.87, and test-retest reliabilities are 0.70-0.85 (Myers and McCaulley, 1985). Concurrent validity is documented with personality, vocational preference, educational style, and management style (0.40-0.77). Construct validity is supported by many studies of occupational preferences and creativity. The Type Differentiation Indicator or TDI (Johnson, 1991; Saunders, 1989) is a longer (290 items) and more intricate form of the MBTI that provides data on the following dimensions: extraversion-introversion (gregarious-intimate, enthusiastic-quiet, initiator-receptor, expressive-contained, auditory-visual); sensing-intuition (concrete-abstract, realistic-imaginative, pragmatic-intellectual, experiential-theoretical, traditional-original); thinking-feeling (critical-accepting, tough-tender, questioning-accommodating, reasonable-compassionate, logical-affective); and judging-perceiving (stress avoider-polyactive, systematic-casual, scheduled-spontaneous, planned-openended, methodogical-emergent). There are also scales for consistency of type, consistency on comfort-discomfort items, and general "strain" (discomfort). Reliability of 23 of the 27 TDI subscales is greater than 0.50, an acceptable result given the brevity of the subscales (Saunders, 1989). MLAT (N = 167). The Modern Language Aptitude Test or MLAT (Carroll and Sapon, 1959) contains 146 items and is the classic language aptitude test, published three decades ago. Its five parts are:
370
REBECCA L. OXFORD and MADELINE E. EHRMAN
I--number learning (memory, auditory alertness); II--phonetic script (association of sounds and symbols), III--spelling clues (English vocabulary, association of sounds and symbols); IV--words in sentences (grammatical structure in English); and V--paired associates (memorizing words by rote). The MLAT was correlated 0.67 with the Primary Mental Abilities Test (Wesche, Edwards and Wells, 1982)----~e latter suggesting a strong general intelligence factor operating in the MLAT. Split-half reliabilities for the MLAT are 0.92-0.97, depending on the grade or age. Validity coefficients (correlating MLAT total scores with course grades or grade point averages) for grades 9-11 are 0.28-0.78 for the regular (long) form of the MLAT; short-form reliabilities were 0.21-0.83. For college students, validity coefficients are 0.18-0.69 for the long form of the MLATand 0.21-0.68 for the short form. For adult students in intensive language programs, validity coefficients are 0.27-0.73 for the long form and 0.26-0.69 for the short form (Carroll and Sapon, 1959). Skehan (1989) shows the MLATto have moderate predictive power (0.40-0.60) in many settings.
Proficiency ratings (N = 126for reading, 128for speaking). At the end of training, FSI students are given proficiency assessments in ratings of 0-5 for speaking and 0-5 for reading. R-3 means reading proficiency level 3. S-2 means speaking proficiency level 2. The ratings are based on the ACTFL/ETS/FSI/ILR guidelines that have been developed over the years by government agencies with the assistance of the Educational Testing Service. These guidelines are detailed by Omaggio (1986). Reliability of the guidelines at FSI is in the range of 0.85-0.90. Reliability studies have shown that agencies have high interrater reliability for proficiency ratings within a given agency, but that the standards are not always the same at every agency; thus raters at different government agencies do not have as high an interrater reliability as raters at the same agency (Clark, 1987). Proficiency ratings are considered valid indicators of the level of language performance of an individual student. SILL (N = 268). The Strategy Inventory for Lanuage Learning or SILL (Oxford, 1989, 1990; see Oxford and Burry, 1993) is now translated into Arabic, French, Chinese, German, Japanese, Korean, Russian, Spanish, Thai and Ukrainian. It has an 80-item version for English speakers learning a foreign language and a 50-item version for learners of English as a second or foreign language. In studies worldwide, the SILL's reliability using Cronbach's alpha (corrected by the SpearmanBrown formula) is ordinarily in the range of the 0.90s. As shown earlier in the review of research, the form of the SILL used in the current study correlates in the 0.40s with learning style (LSP), and the choice of strategies on this version of the SILL is significantly influenced by personality type (MBTI) and motivation (self-ratings) in analyses of variance thus providing evidence of concurrent and construct validity. A shorter form of the SILL, for use by people learning English as a second or foreign language, has strong predictive and concurrent validity as related to language performance and sensory preference. The SILL asks students to react to a series of strategy descriptions (for example, "I make associations between new material and what I already know") in terms of how often they use the strategies (always or almost always, generally, sometimes, generally not, never or almost never). Teacher ratings (N = 86-90 depending on the rating item). Another measure was teacher ratings using a form developed by Ehrman (1990). Teachers rated students on seven items, including whether the student was an effective learner, had a high aptitude for learning, contributed more aptitude than effort (and vice versa), showed intrinsic motivation, and exhibited extrinsic
ADULTS' LANGUAGE LEARNING STRATEGIES
371
motivation. These teacher ratings were correlated with proficiency (rho = 0.23-0.38) and with many personality and affective variables (rho = 0.20-0.42), as discussed in Ehrman and Oxford (1995).
Data collection The present study is part of a much larger investigation known at FSI as the Language Learning Profiles Project. As part of this project, 268 people took the SILL, which serves as the main measure in this report. Many of these people also took a variety of other instruments, as shown in Table 1. Data analysis Descriptive statistics such as means, standard deviations, frequencies, and percentages are reported here. For correlations, the analysis of choice was Spearman' s rho, usually used for rankorder data. When used with interval or ratio data, Spearman's rho provides a more conservative result than Pearson product-moment correlation. Since some of our measures involved ordered data with uneven intervals, and other measures involved equal-interval data, we chose the most conservative correlation that could be used consistently with all the data. Note that correlational results involving two instruments necessarily involve the number of people who took both instruments and are thus much smaller than the number who completed any one of each pair. For differences between two means we conducted the usual t-tests of significance. In this study we are reporting significance at p < 0.05, but we are including only those correlations that are 0.20 or better. Factor analysis was also used for the HBQ and for the AS, as mentioned later. Detailed factor analyses of the shorter form of the SILL are provided elsewhere (Oxford and Burry, 1993).
RESULTS AND INTERPRETATION As mentioned above in the introduction, we focus on findings relative to performance at the end of training, based on the research questions listed above.
What language learning strategies do adults use and how often ? Learning strategy use is moderate overall (mean 2.44, SD 0.53) out of a possible 5. No strategy groups are reported as "always or almost always used", and none are reported as "generally not used" or "never or almost never used". The most frequently used category of strategies was compensation strategies (e.g. guessing, paraphrasing) when encountering a knowledge barrier (mean 3.16, SD 0.57), a "generally used" rating. This was followed immediately by social strategies (mean 3.15, SD 0.65) and by cognitive strategies (mean 3.10, SD 0.61); both of these strategy categories averaged in the "generally used" range. In the "sometimes used" range were the following strategy categories: metacognitive (mean 2.91, SD 0.63), memory (mean 2.56, SD .56), and affective (mean 2.34, SD 0.74). These strategy frequencies are moderate; the FSI learners typically used strategies at a medium level rather than a high level in this study. The frequencies appeared to be slightly lower than those
372
REBECCA L. OXFORD and MADELINE E. EHRMAN
found with a small sample of FSI learners, teachers, and supervisors (Ehrman and Oxford, 1989), who showed 40% of the strategies used at a high frequency and the other 60% at a medium frequency. However, compared with frequencies from other groups of foreign language learners, especially those in school, university, and military settings (see e.g. Douglas, 1992; Green, 1991; Noguchi, 1991; Oxford, 1992a), the FSI learners in the current study reported using strategies more often. Note that second language learners, who are learning a language in an environment where the language is the means of daily survival and communication, typically use more learning strategies than do foreign language learners, who are learning a language in an environment where that language is not the everyday means of communication (Oxford, 1992a). The comparatively greater level of strategy use by FSI foreign language learners might be explainable by two factors: first, they are relatively experienced language learners who already know a great deal about how to learn; and second, they are in intensive training, which is more like a second-language learning environment than is the usual foreign language classroom. How are these language learning strategies related to each other? The SILL mean was correlated with each SILL subscale at p < 0.0005, with rho correlations of 0.81 (cognitive), 0.80 (metacognitive), 0.76 (affective), 0.74 (social), 0.67 (compensation), and 0.66 (memory). These results suggest that the SILL measures a rather coherent construct: how language learners behave in their learning process. It also suggests that all subscales are related to the total SILL in a moderate to strong fashion. Significant intercorrelations (p < 0.0005) among subscales ranged from a high of 0.61 (relationship of cognitive and metacognitive strategies) to a low of 0.35 (compensation and memory), with most in the 0.40s and 0.50s. These results imply that there is a moderately strong relationship among pairs of subscales. Those that correlate only in a mild fashion which each other or with the overall mean of the instrument would seem to be more distinct in what they measure. The strong intercorrelations make us cautious about reporting results by each subscale in a way that is any more than tentative; the SILL holds together as a total instrument instead of being a set of purely distinct, mutually exclusive subscales. Nevertheless, sometimes it is useful to talk about results of the subscales, but only if one remembers that these subscales have a considerable overlap. Does the use of certain language learning strategies relate to proficiency? The use of cognitive strategies, which correlates 0.81 with the total SILL mean, was the only SILL category with a significantly positive correlation with end-of-training language proficiency ratings (rho = 0.21, p < .02). This SILL subscale addresses such as using formulas, recombining, practicing authentically, skimming, scanning, using references, taking notes, summarizing, repeating, various kinds of analyzing, looking for patterns, and adjusting understanding in light of new information. Thus, cognitive strategies on the SILL include both practice and "deep processing", which involves personalizing the information and conducting analysis, synthesis, and continuous adjustment of schemata (see Schmeck, 1988). We speculate it is not the practice component of this subscale, but rather the deep processing component, that might be more important in this correlation with proficiency. We found a lack of correlation between effort (selfreported and faculty-rated) and proficiency, which might suggest that practice, as a reflection of effort, might not be highly related to proficiency. Moreover, cognitive strategies were significantly associated with other correlates of proficiency that reflect conceptualization of new material (e.g. MBTI/TDI Intuition and its subscales, education level, and number of previous languages, all discussed below). In a factor analysis reported elsewhere (Oxford and Nyikos, 1989), the factor
ADULTS' LANGUAGE LEARNING STRATEGIES
373
most closely resembling the current set of cognitive strategies was called "Searching for and Communicating Meaning", thus showing evidence of deep processing. These results differ from a number of SILL studies, in which the frequency of use of many different kinds of strategies is significantly (and more strongly) related to language learning success (see Bedell's 1993 review and the background research in this paper). In this study, why is the correlation so low between cognitive strategy use and language proficiency, and why were there no correlations between proficiency and other kinds of strategies beyond the cognitive? One explanation might be the restricted range of proficiency outcomes and/or the restricted range of strategy use scores (moderately strong). Another possibility to be checked in later analyses is that the distribution of strategy use might be somewhat nonlinear in this study (see nonlinear findings by Phillips, 1990, 1991, but linear findings in the majority of studies reviewed by Bedell, 1993). A third and final possibility might be our speculation that the other kinds of strategies besides the cognitive deep-processing behaviors might be relatively surface-level actions that can be learned more easily, and hence are more randomly scattered among the learners. Does strategy use correlate with teacher perceptions ? SILL scores were related to teachers' ratings. Cognitive strategy use was significantly correlated with teachers' ratings of being an effective learner (rho = 0.31, p < 0.004) and having high aptitude for language learning (rho = 0.33, p < 0.002). Thus, students who used cognitive strategies frequently were noticed by their teachers and highly regarded. Faculty also rated the part played by aptitude (vs effort) in language learning performance. Students whose language performance was attributed more to aptitude were strong users of cognitive strategies (rho = 0.34, p < 0.001), while students whose language performance was attributed more to effort did not use cognitive strategies often (rho = 0.41, p < 0.005). Thus, the latter students might have tried very hard but did not seem to choose the appropriate cognitive strategies. It should be noted that effort did not necessarily impress teachers positively; in fact, the patterns of ratings showed that aptitude was much more positive to teachers, who felt that those who had to use a great deal of effort were probably lacking in aptitude. This is an almost literal example of the difference between working hard and "working smart". Is the use of language learning strategies related to gender? Only a few significant differences in strategy use appeared by gender. Our t-tests showed that compensation strategy use was linked to gender (t = 2.13, p < 0.034), with females using more compensation strategies than males. Females also score higher on overall strategy use via the average of all the SILL subscale scores (t = 1.97, p < 0.05). This pattern of gender differences was somewhat similar to that of many other studies (Oxford, 1993a, 1993b, 1993c), but the differences were not as strong or frequent in this group. We suggest that the reason for the weaker findings is that the FSI group is different from many language learning populations. The participants are largely self-selected, because only certain types of people are attracted to and are selected for the Foreign Service. Most of the males and females alike are intuitive thinking types on the MBTI. In their learning preferences, the females in the Foreign Service officer group resemble the males more than they resemble the females in the total U.S. population. The females here may therefore be considered somewhat atypical of the larger population of the U.S., while the males are less so.
374
REBECCA L. OXFORD and MADELINE E. EHRMAN
Is the use of language learning strategies related to educational background, number of languages previously studied, weeks of training to date, or language difficulty? People who used cognitive strategies had more previous education (rho = 0.25, p < 0.003) and had studied more languages previously (rho = 0.33, p < 0.002). People who used metacognitive strategies had more previous education (rho = 0.22, p < 0.01). Thus, cognitive and metacognitive strategy use is related to academic and linguistic experience. Those individuals who had such experience used more of these strategies. Significant correlations appeared between strategy use on the one hand and weeks of language training and difficulty of language on the other, but these were below 0.20 and wilt not be discussed. Is the use of language learning strategies related to the use of general learning and study strategies ? A volunteer subsample of 59 completed the LASSI, which addresses learning strategies applicable to any academic setting. They were above average on 9 of the 10 LASSI scales--all except use of study aids. Thus, the LASSI showed that the participants were fairly strong users of general learning and study strategies. Significant correlations ranged from a high of 0.51 to a low of 0.26. The greatest numbers of significant correlations between the SILL and the LASS1 occurred for metacognitive strategies (10 significant correlations), cognitive strategies (8), and social strategies (8). This suggests that the LASSI is strongly infused with metacognitive, cognitive, and social elements. Memory strategies correlated significantly with three LASSI scales, affective strategies with only one, and compensation strategies with none. Thus, memory, effective, and compensation aspects of strategy use are deemphasized in the LASSI. In addition, there are some LASSI subscales for which few correlations arose with the SILL. In general, these results suggest that there is moderate but significant overlap between certain aspects of the SILL and the LASSI. However, there are also major areas in which no overlap appears to exist. These two instruments are partially distinct in their purposes and their findings for language learners. Is the use of these language learning strategies related to language aptitude ? SILL-identified cognitive strategy use was significantly but weakly related (rho = 0.25) to one part of the MLAT, number learning (Part I). Thus, the intellectual leanings shown by cognitive strategy use are reflected in the associating, elaborating, and induction skills of number learning. Students who use cognitive strategies also tend to do well on this pattern-analysis and 'systematic remembering' aspect of language aptitude. However, negative links also exist between the SILL and certain aspects of aptitude as measured by the MLAT. Compensation strategies are significantly but only mildly related in a negative fashion to the MLATtotal score and to paired associates (Part V), as well as to the MLAT Index Score (all in the 0.20s). There is a negative relationship between affective strategy use and grammatical sensitivity in sentences (Part IV), rho = - 0.25. This stands to reason, because compensation and affective strategies can be used by people who experience themselves as having trouble with language learning or who are just starting out in learning a language. Some of these otherwise high-functioning people in this sample may have found that they can use compensation strategies to fill in gaps in the ability to memorize rapidly or in grammar analysis. Others may have found they can benefit from affective strategies to lower their anxiety. That is, if the students had ready access to skills manifested in the MLAT tasks, they would not need compensatory or affective strategies to the point that they would be aware of them and report high levels of use of such strategies. This finding does not mean that only low-aptitude students use such strategies, but it does suggest that proper use of such strategies might help low-aptitude students and can be taught.
ADULTS' LANGUAGE LEARNING STRATEGIES
375
Is the use of language learning strategies related to learning style ? This question was answered by examining correlations between the SILL and the LSP. Many elements of the SILL (mean, metacognitive, social, cognitive, affective, and memory) correlated moderately (rho = 0.28-0.40) with persistence on the LSP. This suggests that one of the underlying aspects of use of language learning strategies is persistence. Those who want to persist in language learning--continue until they achieve a goal--tend to use language learning strategies of various kinds. These findings are related to previous SILL results (e.g. Nyikos and Oxford, 1993; Oxford, 1989; Oxford and Nyikos, 1989) showing that high motivation is strongly related to strategy use. Persistence is a main component of motivation (Crookes and Schmidt, 1989). Use of cognitive, metacognitive, and social strategies was related to afternoon and evening study time, with rho sizes in the 0.20s. Preference for early morning study time tended to be negatively related to strategy use--a new piece of data that language teacher and program administrators might find useful. These preferences appear to be related to personality type as well (Erhman, in press-b); they might thus be an expression of a kind of personality-and-strategy nexus. Affective strategy use was related to willingness to take verbal risks (outspokenness) (rho = 0.30, p < 0.005). The self-talk found in affective strategies might represent a willingness to speak out (in a selfsupportive way) to oneself as well as to others. Categorization skill is tested by the LSP by the ability to estimate proper ranges of size, amount, and so forth. This skill is weakly and negatively related to the use of cognitive strategies (rho = -0.24, p < 0.05), although this correlation is difficult to explain. Is the use of language learning strategies related to personality type ? In educational studies, personality type--usually measured by the MBTI or its more detailed form, the TDI has often been discussed as a form of learning style (see, e.g. Carrell and Monroe, 1990; Ehrman, 1989, 1990a; Ehrman and Oxford, 1989, 1990, 1995; Lawrence, 1984). Cognitive strategy use, measured by the SILL, was related albeit mildly (rho = 0.23, p < 0.001) to a TDI preference for intellectual vs. pragmatic approach to learning new things. An intellectual approach thus appears to entail the use of learning strategies that require thinking about material being learned, rather than simply accepting it. Use of metacognitive strategies on the SILL was related at significant levels with three sub scales related to judging/perceiving on the TDI; planful (rho = 0-21, p < 0.003), systematic (rho = -0.20, p < 0.006), and methodical (rho = -0.26, p < 0.0001). The negative sign indicates that metacognitive strategy use is associated with the left or judging (closureoriented) side of the judging/perceiving continuum. Therefore, use of metacognitive strategies for language learning seems to be linked with an orderly, closure-oriented personality. Learners who reported on the SILL that they frequently used social strategies also tended to report themselves on the TDI as expressive (rho = -0.24, p < 0,001) and as realistic (rho = -0.31, p < 0.0001). The former is a form of extraversion; the latter is a form of sensing. The negative sign shows that social strategy use was related to the left (or extraverted or sensing) side of the continuum. This suggests that users of social strategies for language learning tend to be open about their thoughts and feelings and that they are realistic and down-to-earth in their learning. Likewise, the SILL mean correlated mildly with the TDI expressiveness subscale (rho = -0.20, p < 0.006). This implies that those who use learning strategies tend to be open and expressive in general.
376
REBECCA L. OXFORD and MADELINE E. EHRMAN
Is the use of language learning strategies related to ego boundaries? Higher HBQ scores indicate increasing thinness of ego boundaries. Factor analyses of the HBQ yielded two main factors. Factor I (internal) addresses such internal boundaries as those between thoughts and feelings or sleeping, waking, and dreaming. Factor II (external) describes the thinness of the interface between the individual and the outer world, e.g. low preference for neatness or clear lines of authority. The significant SILL-HBQ correlations were related to both factors but at a low level (all in the 0.20s and low 0.30s). They provide tantalizing information for further study. For example, compensation strategy use on the SILL had a wide range of significant correlations with thinner internal ego boundaries (Factor I subscales) on the HBQ. It was related to thinness of boundaries among internal states (rho = 0.31, p < 0.005), propensity for unusual experiences (rho = 0.23, p < 0.05) and dreaming (rho = 0.21, p < 0.05), and to thin boundaries overall (rho = 0.23, p < 0.05). These results strongly suggest that users of compensation strategies tend to be highly flexible, sensitive, and able to deal creatively with unusual experiences. This makes sense, because compensation strategies themselves involve coping with confusing new input in ways that allow the learning process to continue rather than be halted. Users of compensation strategies need to shift gears to handle incoming data rapidly. Metacognitive strategy use on the SILL was negatively correlated with on HBQ Factor II (extemal) subscale: lack of desire for neatness (rho = -0.28, p < 0.01). This means that metacognitive strategy users like neatness and prefer an orderly environment. This is predictable, because metacognition involves planning and organizing one's own learning. Their responses on the HBQ tended toward the 'thicker' end. It is consistent with the findings reported above for the MBTI in which metacognitive strategies were associated with judging subscales: judging is correlated with thick boundaries (Ehrman, in press-a, in press-b). Users of memory strategies as reported on the SILL were similar to metacognitive strategy users in showing slightly but significantly negative correlations with HBQ's opinions about people, groups, and nations (rho = -0.20, p < 0.05), suggesting that they tend to be more rigid in these opinions. We have found in our earlier work that many users of memory strategies for language learning score as sensing and judging types on the MBTI, thus revealing their concrete, step-by-step, clarityoriented learning style (Ehrman and Oxford, 1989, 1990). The ego boundary data support these earlier findings. Affective strategy use on the SILL was significantly related to the HBQ sleep/dreams/wakefulness dimension (rho = 0.22, p < 0.05). Affective strategy users tended to be aware of their dreams and to have nightmares; thus they scored in the area of thinner boundaries. We would expect this finding, because those who use affective strategies tend to be attuned into their emotional state, of which dreams are an important part. Users of affective strategies try to manage their feelings and motivations, and to do that, they must first be aware of these states.
Is the use of language learning strategies related to motivation and anxiety? Many relationships exist between learning strategy use on one side and motivation and anxiety on the other. The SILL mean correlated positively and significantly with many aspects of the Affective Survey or AS: total motivation (rho = 0.44, p < 0.005), intrinsic motivation (rho = 0.33, p < 0.01), and desire to use the language outside class (rho = 0.31, p < 0.05). Thus, the overall use of language
ADULTS' LANGUAGE LEARNING STRATEGIES
377
learning strategies was linked with rather strong motivation, particularly internally generated motivation, and desire to use the language outside of class. These findings support the clear association between strategy use and motivation found by Oxford and Nyikos (1990). Many parts of the AS were significantly related to cognitive strategy use on the SILL. Positive correlations appeared with anxiety about classroom speaking (rho = 0.56, p < 0.0005), intrinsic motivation (rho = 0.54, p < 0.001), total positive arousal level (rho = 0.46, p < 0.005), and positive beliefs about oneself (rho = 0.44, p < 0.005). Negative correlations were found between cognitive strategy use on the SILL and these affective elements: endorsement of negatively phrased motivation items (e.g. 'I find studying Language X depressing', rho = -0.64, p < 0.005), general discomfort about language learning (-0.32, p < 0.05), anxiety about self-esteen (rho = -0.30, p < 0.05), anxiety about competition (rho -- -0.30, p < 0.05), and anxiety about outcomes (rho = -0.30, p < 0.05). These results signify that users of cognitive strategies for language learning are a confident, positive, highly motivated lot who exhibit strong arousal. They tend to be comfortable rather than anxious with language learning, with the exception of having a degree of anxiety about speaking in class. For this sample, this type of anxiety appears to be facilitative, since it seems to correspond significantly with the use of cognitive strategies and with end-oftraining proficiency (Ehrman and Oxford, 1995). The use of metacognitive strategies was positively correlated with these aspects of the AS: intrinsic motivation (rho = 0.46, p < 0.0001), overall positive arousal level (rho = 0.35, p < 0.01), and positive belief about oneself (rho = 0.30, p < 0.05). Negative correlations appeared between metacognition and endorsement of negative motivation items (rho = -0.46, p < 0.005). Like cognitive strategy users, the users of metacognitive strategies for language learning tended to be internally motivated, self-confident and emotionally energized. Use of SILL-measured compensation strategies was positively correlated with the desire to use the language outside of class (rho = 0.30, p < 0.05) on the AS. Use of compensation strategies was negatively related to anxiety about outcomes (rho = -0.30, p < 0.05). Thus learners using compensation strategies were similar to users of cognitive and metacognitive strategies in wanting to use the language outside class and in lack of anxiety. Affective strategy use was correlated with overall positive arousal level (rho = 0.35, p < 0.05) on the AS. Thus, affective strategy users are strongly aroused, but they may or may not show specific kinds of motivation on a consistent basis. Memory strategy use was negatively correlated with anxiety about self-esteem (rho = 0.29, p < 0.05). Users of memory strategies may thus be somewhat like those who report use of the other strategy types listed above, in that they are not subject to certain kinds of anxiety. In summary, cognitive strategies are correlated with a wide range of affective and motivational variables. The SILL mean seems especially linked with various kinds of motivation. Positive beliefs about oneself as a language learner (self-confidence) are reflected in reports of high levels of use of relatively several kinds of strategies as well. Anxiety is related in a complex way to strategy use, and anxiety might occasionally be facilitative. SUMMARY AND IMPLICATIONS This section summarizes the results and offers implications in these areas: frequency of use of language learning strategies; strategy use and proficiency; persistence, motivation, and planning;
378
REBECCA L. OXFORD and MADELINE E. EHRMAN
gender differences in strategy use; teacher perceptions; facilitative anxiety; use of various kinds of strategies; and building a theory of language learning.
Frequency of use Candidates in FSI's intensive foreign language program were moderate, not high, users of language learning strategies in this study. When compared with subjects in other SILL studies (Oxford, 1992a), they used strategies somewhat more often than learners in other foreign language settings but less often than many second language learners, who were involved in speaking the new language for daily survival. From the point of view of increasing strategy use by learners, this implies that foreign language learners might benefit from having their learning situation emulate, insofar as possible, a second language immersion situation----complete with simulations of reallife events and activities that second language learners would naturally encounter (Crookall and Oxford, 1990), as is frequently the case in FSI classrooms. The more that a foreign language classroom can become a 'language experience' situation, the more likely the students will find the need and the will to use language learning strategies at a high level and the more rapidly they will advance toward proficiency. This does not mean that structured instruction should be totally omitted; but it does mean that lively, authentic chunks of language interaction using role-plays and simulation are also clearly needed. The fact that foreign language learners usually show somewhat lower strategy use than second language learners should not cause foreign language teachers to despair. As noted in the research review, foreign language learners whose careers hinge on their language skills and who are therefore extrinsically motivated (or who have equally strong intrinsic motivations) may develop a tremendous motivation to use language learning strategies at a high level---as high as that of secon.d language learners who are surrounded by the language constantly. Teachers should encourage learners to develop strong motivation (see Crookes and Schmidt, 1989 for ideas about how) and to use effective strategies whenever possible, with the aim of more efficient learning and better outcomes.
Strategy use and proficiency An important issue is whether language learning strategy use influences language proficiency. In this study the correlation was low but significant between cognitive strategy use and speaking proficiency (but not reading proficiency). The problem here is that because candidates tend to reach their proficiency goals, FSI has a very restricted range of proficiency outcomes. It is statistically impossible for a high correlation to appear with such a restricted range. Stronger linkages between strategy use and proficiency or achievement have appeared with groups that show a wider range of proficiency outcomes (see research review). In order to determine more precise relationships between strategy use and proficiency, researchers can work with a variety of groups--drawn from schools, universities, government agencies, and corporations--that show a broader range of proficiency outcomes. Relationships with persistence, motivation and planning One general theme that occurred repeatedly was the strong relationship between learning strategy use on the one hand and persistence (LSP), motivation (AS) and the ability to plan (MBTI/TDI judging) on the other hand. Learning strategy users were highly persistent, strongly motivated, and oriented toward systematic planning. These are good traits for learning and studying in any
ADULTS' LANGUAGE LEARNING STRATEGIES
379
subject area. They involve a combination of cognitive and affective aspects. Results of our study imply that we can foster use of language learning strategies by helping students persist in their learning, maintain and enhance their own motivation, and develop skills in systematic planning. Crookes and Schmidt (1989) and Oxford and Shearin (1994) address the first two elements, persistence in particular and motivation in general. Systematic planning is among the many metacognitive skills described by Weinstein, Goetz and Alexander (1988). Alternatively, for those who prefer the more flexible approach characterized by MBTI/TDI perceiving, we can help them maximize their adaptive skills by choosing the right strategies at the time they are needed. Gender differences in strategy use Significant strategy-related gender differences favoring females were slim but definitely present in this study, a pale reflection of larger gender differences found in other studies of language learning strategy use (Oxford, 1993a, 1993b). Teachers and researchers should keep trying to understand gender differences in strategy use when they appear; and they should understand that gender differences may often be a mask for deeper differences of personality type and career choice. Males and females should be encouraged and allowed to develop the most effective learning approaches they can, and neither should be pushed into a gender-stereotyped set of strategies. Teacher perceptions Positive teacher perceptions were correlated with student report of use of cognitive strategies. Perhaps one reason that teachers were impressed by their students' use of cognitive strategies was that these strategies were associated with personality traits of persistence, motivation, and deep processing. These are qualities that most teachers like and that usually facilitate learning. Facilitative anxiety Contrary to the concept that anxiety in language learning can only be debilitative (Horwitz, 1990), we found using the AS that self-reported anxiety about speaking the language in class had a positive relationship with cognitive strategy use on the SILL. It might be that this anxiety is not detrimental but instead somewhat facilitative, if not carried too far. Correlations of self-reported anxiety about speaking in class with end-of-training proficiency in this sample confirm this point of view (discussed in Ehrman and Oxford, 1995). Brown (1987) appears to have been right in saying that both facilitative and debilitative anxiety exist. We must learn to differentiate between the two, accounting for differences in student populations of the sort that make the FSI sample so different from Horwitz's (1990) university students, for whom speaking in class was the most debilitative form of anxiety. Use of various kinds of strategies Cognitive strategy use was positively related to teacher perceptions of FSI candidates and to actual proficiency outcomes in speaking. Students who were viewed as effective and as having a high aptitude tended to use cognitive strategies. Students who were viewed by teachers as relying more on effort than aptitude appear to have been less frequent users of cognitive strategies than those whose performance was judged as more aptitude-based. Other kinds of strategies did not play a significant role in teacher perceptions. Individuals who used cognitive strategies were better educated, had studied more languages previously, and were more persistent (LSP) than others in the FSI sample. Use of cognitive strategies on the SILL was repeatedly associated with use of general learning and study strategies on the LASSI. Cognitive strategy use was significantly linked with
380
REBECCA L. OXFORD and MADELINE E. EHRMAN
the MLAT's number learning scale, which requires analysis of input and logical remembering. On the MBTI/TDL cognitive strategy users reported themselves as more intellectual than pragmatic. Users of cognitive strategies appeared on the AS to be confident, positive, highly aroused or energized, strongly motivated, yet comfortable with language learning. Why do cognitive strategies seem to be so important? A look at the items included in this subscale of the SILL shows us that cognitive strategies include two activity types: practice and rehearsal on the one hand, and mental techniques that involve hypothesis formation and personalizing on the other. The latter set is involved with a kind of 'deep processing' that forms mental links between the new and old material and makes the new material a solid part of the learner' s own personal repertoire. Unlike the other strategy categories, which do not involve this kind of intellectual depth, cognitive strategies go deep into the learner (see Ehrman and Oxford, 1995, for more discussion on this point). The FSI sample showed a significant (though low) gender difference favoring females in the use of compensation strategies, reflecting what might be a more general trend. Compensation strategies were negatively related to some aspects of the MLAT, suggesting that high-aptitude learners might not need to use these strategies as often. Users of compensation strategies reported themselves on the HBQ as highly flexible, sensitive, and as having unusual experiences and ideas. Users of compensation strategies wanted to use the new language outside of class, were motivated, and were not too anxious about training outcomes (AS). Compensation strategy use was thus related to some helpful qualities, but it was also linked with lower language aptitude. This might suggest that lower-aptitude students feel their lack of ability and in some way strive to compensate for it. Teachers could encourage lower-aptitude students to overcome their learning obstacles by the use of compensation strategies, like guessing, predicting, choosing a familiar topic, and circumlocution ('talking around' a word). Unfortunately, many lower-aptitude students are not natural risk-takers and may be cognitively somewhat rigid. Affective strategies such as positive self-talk might help them achieve sufficient flexibility to be able to use compensation strategies. Frequent users of metacognitive strategies were persistent (LSP) and had more education than infrequent users of these strategies. Metacognitive strategy use was repeatedly related to the general learning and study strategies outside of the language field, as measured by the LASSI. These individuals were assessed on the MBTI/TDI as being methodical, planful, and systematic. Metacognitive strategy users showed on the HBQ a need for orderliness, a desire for greater personal distance, and an overall dislike of ambiguity, i.e. 'thicker' ego boundaries. On the AS, metacognitive strategy users reported being intrinsically motivated and confident. Training students in the effective use of metacognitive strategies such as planning, organizing, and evaluating might be a good idea. The more structure-oriented students would love metacognitive strategies and might already be applying them. The less structured students might feel less comfortable with these strategies but might find them useful; in fact, these strategies could provide a sense of balance for students who typically avoid structure. High-aptitude learners (on the MLAT) tended to report little use of affective strategies; perhaps they were less anxious and did not need to use affective strategies to relax or gain self-confidence. Affective strategy use was related to verbal risk-taking (LSP), persistence in learning (LSP), and thin boundaries among waking, sleeping, and dreams (HBQ). On the AS affective strategy users
ADULTS'LANGUAGELEARNINGSTRATEGIES
381
showed themselves as being energized. This study showed that affective strategies and compensation strategies were both associated with lower aptitude and with verbal risk-taking, so they appear to have some aspects in common that deserve further study. The existence of many correlates of affective strategy use in this investigation suggests that the affective realm (Oxford, 1992/1993; Oxford and Cohen, 1992; McCombs, 1988) deserves more research attention. Users of social strategies on the SILL tended to use a variety of general learning and study strategies on the LASSI. They were persistent (LSP) and both expressive and realistic (MBT1/TD1). It would be useful to uncover the hidden dynamics of social strategy use, so that these can be demonstrated and practiced in strategy training sessions. Users of memory strategies were persistent (LSP). On the AS, they were not anxious about selfesteem issues. Memory strategies did not have much impact in this study, although they were applied by learners of difficult languages to a mild degree. These scanty results for memory strategies imply that these strategies might be of minor use in FSI' s highly communicative training setting, which relies very little on the requirement to memorize a lot of material by rote. Memory strategies use some mental processes that appear superficially similar to certain cognitive strategies, but cognitive strategies bring about more deep-processing, semantic-level connections and are thus more powerful. Results of this study might imply that the focus in strategy training should clearly not be on memory strategies, except in regard to a few areas such as vocabulary learning.
Building a theory of language learning The results of this research have many practical, hands-on implications for strategy training, as we have just discussed. However, some implications also exist for the continuing evolution of a theory of language learning. As our study shows, the use of language learning strategies forms conceptual linkages that go beyond what researchers might have proposed in the past. Use of these strategies is associated in intricate and fascinating ways with many cognitive, affective, and social aspects of the individual. Our current findings can assist in the ongoing creation of a model of language learning (Ehrman, in press-a) that includes predictors such as ego boundaries, personality type, strategy use, gender, attitudes, and motivation--along with traditional measures of language aptitude such as the MLAT. The interaction of these individual characteristics (in a much expanded conception of aptitude that echoes the very broad constructs of Snow, 1989) with instructional treatment or methodology must by the subject of focused exploration in the future. No model is complete without this element of aptitude-treatment interaction. Teacher perceptions (along with a host of other institutional and environmental variables) belong in this theoretical model of language learning, but their precise role is still being defined. Appropriate measures of proficiency outcomes are crucial in this model as well. Thus, the present study brings us one step further forward toward a more comprehensive, and thus more realistic, theory of how language learning occurs. Acknowledgement--The authors thank Lucinda Hart-Gonzalesof the Foreign ServiceInstitute for her outstanding statistical work and her theoreticalcontributionsto this paper.
382
REBECCA L. OXFORD and MADELINEE. EHRMAN
REFERENCES ACKERMAN, P. L., STERNBERG, R. J. and GLASER, R. (eds) (1989) Learning and Individual Differences: Advances in Theory and Research. New York: W. H. Freeman. AU, S. Y. (1988) A critical appraisal of Gardner's socio-psychological theory of second-language (L2) learning. Language Learning 38(1), 75-100. BECK, R. C. (1990) Motivation: Theories and Principles. 3rd edn. Englewood Cliffs, NJ: Prentice Hall. BEDELL, D. (1993) Crosscultural variation in the choice of language learning strategies: a mainland Chinese investigation with comparison to previous studies. Unpublished Master's thesis, University of Alabama, Tuscaloosa, AL. BEEBE, L. M. (1983) Risk-taking and the language learner. In H. W. Seliger and M. H. Long (eds), Classroom-Oriented Research in Second Language Acquisition. Rowley, MA: Newbury House. BROWN, A. L., BRANSFORD, J. D., FERRARA, R. and CAMPIONE, J. C. (1983) Learning, remembering, and understanding. In J. H. Flavell and E. M. Markham (eds), Carmichael's Manual of Child Psychology (Vol. 1). New York: Wiley, BROWN, H. D. (1987) Principles of Language Learning and Teaching. 2nd edn. Englewood Cliffs, NJ: Prentice-Hall. CARRELL, P. L. and MONROE, L. B. (1990, Oct.) Investigating ESL students' learning styles. Paper presented at the Joint Conference of Southeast Teachers of English to Speakers of Other Languages and National Association of Foreign Student Affairs Region VI, Louisville, KY. CARROLL, J. and SAPON, S. M. (1959) Modern Language Aptitude Test. New York: Psychological Corporation. CHAMOT, A. U. and KUPPER, K. (1989) Learning strategies in foreign language instruction. Foreign Language Annals 22, 13-24. CHANG, S.-J. (1991 ) A study of language learning behaviors of Chinese students at the University of Georgia and the relation of those behaviors to oral proficiency and other factors. Unpublished doctoral dissertation, University of Georgia, Athens, GA. CHAPELLE, C. A. (1983) The relationship between ambiguity tolerance and success in acquiring English as a second language in adult learners. Unpublished doctoral dissertation, University of Illinois, Champaign-Urbana, IL. CLARKE, J. L. D. (1987) Comparability Study of the Oral Proficiency Interview. Washington, DC: Center for Applied Linguistics. COHEN, A. D. (1990). Language Learning: Insights for Learners, Teachers, and Researchers. New York: Newbury House/Harper & Row. Now Boston: Heinle & Heinle. CORRALES, O. and CALL, M. E. (1989). At a loss for words: The use of communication strategies to convey lexical meaning. Foreign Language Annals, 22, 227-240. CRONBACH, L. J. and SNOW, R. E. (1977). Aptitudes and Instructional Methods: A Handbook for Research on Interactions. New York: Irvington. CROOKALL, D. and OXFORD, R. L. (1990). Simulation, gaming, and language learning. New York: Harper & Row. Now Boston: Heinle & Heinle. CROOKES, R. and SCHMIDT, R. (1989) Motivation: reopening the research agenda. University of Hawaii Working Papers in ESL 8, 217-256. DANSEREAU, D. (1988) Cooperative learning strategies. In C. E. Weinstein, E. T. Goetz and P. A. Alexander (eds), Learning and Study Strategies: Issues in Assessment, Instruction, and Evaluation. New York: Academic Press. DAVIS, E. and ABAS, H. (1991) Second Language Learning Strategies Utilized by Some Members of Language Departments at Four lnstitutions--Sulawesi, Indonesia. Summer Institute for Linguistics, Sulawesi Indonesia. DOUGLAS, M. (1992) Kanji learning strategies of learners of Japanese as a foreign language. Unpublished doctoral dissertation, University of Southern California, Los Angeles. EHRMAN, M. E. (1987) Recent research on aging and verbal processing in normal adults. Unpublished manuscript, Foreign Service Institute, Arlington, VA. EHRMAN, M. E. (1989) Ants and grasshoppers, badgers and butterflies: quantitative and qualitative investigation of adult language learning styles and strategies. Unpublished doctoral dissertation, The Union Institute, Cincinnati, OH. EHRMAN, M. E. (1990a) Owls and Doves: Cognition, personality, and learning success. In J. E. Alatis (ed.), Linguistics, Language Teaching, and Language Acquisition: The Interdependence of Theory, Practice and Research. Washington, DC: Georgetown University Press. EHRMAN, M. E. (1990b) Teacher Rating Forms. Arlington, VA: Foreign Service Institute.
ADULTS' LANGUAGELEARNINGSTRATEGIES
383
EHRMAN, M. E. (1993a) A study of extremes: Strongest and weakest learners in intensive language training. Presented at the American Council on the Teaching of Foreign Languages annual meeting. EHRMAN, M. E. (1993b) Analyses from the Language Learning Profiles Project. Unpublished manuscript, Foreign Service Institute, Arlington, VA. EHRMAN, M. E. (in press-a) Ego boundaries revisited: Toward a model of personality and learning. Paper presented at the Georgetown Round Table on Languages and Linguistics, Washington, DC. Forthcoming in J. E. Alatis (ed.) Strategic Interaction and Language Acquisition: Theory, Practice, and Research. Washington, DC: Georgetown University Press. EHRMAN, M. E. (in press-b) The Type Differentiation Indicator and adult language learning success. Journal of Psychological Type. EHRMAN, M. E. (1993) A study of extremes: Strongest and weakest learners in intensive language training. Presented at the American Council on the Teaching of Foreign Languages annual meeting. EHRMAN, M. E. (1995) Weakest and strongest learners in intensive language learning: a study of extremes. In C. A. Klee (ed.), Faces in a Crowd: Individual Learners in Multisection Programs. Boston: Heinle & Heinle. EHRMAN, M. E. and OXFORD, R. L. (1989) Effects of sex differences, career choice, and psychological type on adults' language learning strategies. Modern Language Journal 73, 1-13. EHRMAN, M. E. and OXFORD, R. L. (1990) Adult language learning styles and strategies in an intensive training setting. Modern Language Journal 74(3), 311-327. EHRMAN, M. E. and OXFORD, R. L. (1991 ) Affective Survey. Arlington, VA: Foreign Service Institute. EHRMAN, M. E. and OXFORD, R. L. (1995) Cognition plus: Correlates of language learning success. Modern Language Journal 79, 67-89. ELY, C. (1986) An analysis of discomfort, risk-taking, sociability, and motivation in the L2 classroom. Language Learning 36(1), 1-25. ELY, C. (1989) Tolerance of ambiguity and use of second language learning strategies. Modern Language Journal 22(5 ), 437-445. GAGNE, R. (ed.) (1967) Learning and Individual Differences, Columbus, OH: Merrill. GALBRAITH, V. and GARDNER, R. C. (1988) Individual Difference Correlates of Second-Language Achievement: An Annotated Bibliography. London: University of Western Ontario. GARDNER, R. C. (1985) Social Psychology and Second Language Learning: The Role of Attitudes and Motivation. London: Edward Arnold. GARDNER, R. C. and LAMBERT, W. (1972) Attitudes and Motivation in Second Language Learning. Rowley, MA: Newbury House. GRADMAN, H. and HANANIA, E. (1991) Language learning background factors and ESL proficiency. Modern Language Journal 75, 39-51. GREEN, J. M. (1991) Language learning strategies of Puerto Rican university students. Paper presented at the annual meeting of Puerto Rico Teachers of English to Speakers of Other Languages, San Juan, PR. GREEN, J. M. and OXFORD, R. L. (1993, April) Learning strategies: Patterns of use by gender and proficiency. Paper presented at the annual meeting of Teachers of English to Speakers of Other Languages, Atlanta. GUIORA, A. Z. (1981) Language, personality, and culture, or the Whorfian hypothesis revisited. In M. Hines and W. Rutherford (eds), On TESOL '81. Detroit, MI: Teachers of English to Speakers of Other Languages. HARTER, S. (1986). Feeling good about yourself isn't enough. Psychology Today 8(2), 2-3. HARTMANN, E, (1991) Boundaries in the Mind: A New Psychology of Personality. New York: Basic Books. HORWITZ, E. K. (1990) Attending to the affective domain in the foreign language classroom. In S. S. Magnan (ed.), Shifting the Instructional Focus to the Learner (pp. 15-33). Middlebury, VT; Northeast Conference on the Teaching of Foreign Languages. HORWITZ, E. K. and YOUNG, D. J. (1991) Language Learning Anxiety: From Theory and Research to Classroom Implications. Englewood Cliffs, MJ: Prentice-Hall. HUANG, X.-H. (1984) An investigation of learning strategies in oral communication that Chinese EFL learners in China employ. Unpublished Master's thesis, Chinese University of Hong Kong. JOHNSON, D. A. (1991, July) The five-factor model of personality, the NBO-PI, and the Type Differentiation Indicator. paper presented at the Ninth International Conference of the Association for Psychological Type, Richmond, VA. JOHNSON, D. A. and SAUNDERS, D. R. (1990) Conf'Lrmatoryfactor analysis of the Myers-Briggs Type Indicator-Expanded analysis report. Educational Psychology Measurement 50, 561-571. KEEFE, J. W. and MONK, J. S. (1986) Learning Style Profile Examiner's Manual. Reston, VA: National Association of Secondary School Principals.
384
REBECCA L. OXFORDand MADELINEE. EHRMAN
KEEFE, J. W. and MONK, J. S., with LEq"FERI, C. A., LANGUIS, M. and DUNN, R, (1989) Learning Style Profile. Reston, VA: National Association of Secondary School Principals. LARSEN-FREEMAN, D. and LONG, M. (1991 ) An Introduction to Second Language Acquisition Research, Oxford: Oxford University Press. LAVINE, R. and OXFORD, R. (1990, Dec.) Addressing affective issues in the second and foreign language classroom. Paper presented at the annual meeting of the Modem Language Association, Chicago. LAWRENCE, G. (1984) A synthesis of learning style research involving the MBTI. Journal of Psychological Type, 2-15. McCOMBS, B. L. (1988) Motivational skills training: combining metacognitive, cognitive, and affective learning strategies. In C. E. Weinstein, E. T. Goetz and P. A. Alexander (eds) Learning and Study Strategies: Issues in Assessment, Instruction, and Evaluation (pp. 141-169). New York: Academic Press. MULLINS, P. (1992) Successful English language learning strategies of students enrolled at the Faculty of Arts, Chulalongkom University, Bangkok, Thailand. Unpublished doctoral dissertation, United States International University, San Diego, CA. MYERS, I. B. and McCAULLEY, M. H. (1985) A Guide to the Use of the Myers-Briggs Type Indicator. Palo Alto, CA: Consulting Psychologists Press. NAIMAN, N., FROHLICH, M. and TODESCO, A. (1975) The good second language learner. TESL Talk 6, 68-76. NAIMAN, N., FRI3HLICH, M., STERN, H. and TODESCO, A. (1978) The Good Language Learner. Research in Education Series, No. 7. Toronto: Ontario Institute for Studies in Education. NOGUCHI, T. (1991) Review of language learning strategy research and its implications. Unpublished bachelor's thesis, Tottori University, Tottori, Japan. NYIKOS, M. and OXFORD, R. (1992) A factor analytic study of language-learning strategy use: Interpretations from information-processing theory and social psychology. Modern Language Journal 77(1), 11-22. O'BRIEN, L. (1990) Learning Channel Preference Checklist. Rockville, MD: Specific Diagnostic Studies. OMAGGIO, A. C. (1986) Teaching Language in Context." A Proficiency-Oriented Approach to Language Instruction. Boston: Heinle & Heinle. O'MALLEY, J. M. and CHAMOT, A. U. (1990) Learning Strategies in Second Language Acquisition. Cambridge: Cambridge University Press. O'MARA, F. and LEVI', J. (1990) Foreign Language Learning and Retention: Interim Results of the Language Skill Change Project. Reston, VA: Advanced Technology. OXFORD, R, L. (1982) Research on language loss: A review with implications for language learning. Modern Language Journal 66, 160-169. OXFORD, R. L. (1989) Use of language learning strategies: A synthesis of studies with implications for strategy training. System 17(2), 235-247. OXFORD, R. L. (1989) Strategy Inventory for Language Learning. Alexandria, VA: Oxford Associates. OXFORD, R. L. (1990) Language Learning Strategies: What Every Teacher Should Know. New York: Newbury House/Harper & Row. Now Boston: Heinle & Heinle. OXFORD, R. L. (1992/1993) Language learning strategies in a nutshell. TESOL Journal 2(2), 18-22. OXFORD, R. L. (1992a) The Story of the SILL: Evolution, Use Reliability, and Validity of the Strategy Inventory for Language Learning. Tuscaloosa, AL: University of Alabama. OXFORD, R. L. (1992b) Who are our students? A synthesis of foreign and second language research on individual differences. TESL Canada Journal 9(2), 30-49. OXFORD, R. L. (1993a) Instructional implications of gender differences in language learning styles and strategies. Applied Language Learning 4(1-2), 65-94. OXFORD, R. L. (1993b) La difference continue...: Gender differences in second/foreign language learning styles and strategies. In J. Sutherland (ed.), Exploring gender (pp. 140-147), Englewoods Cliffs, NJ: Prentice-Hall. OXFORD, R. L. (1993c) Research on second language learning strategies. In W. Grabe (ed.), Annual Review of Applied Linguistics (pp. 174-187). Cambridge, UK: Cambridge University Press. OXFORD, R. L. and BURRY, J. A. (1993, April) Evolution, norming, and psychometric testing of Oxford's Strategy Inventory for Language Learning. Paper presented at the National Council on Measurement in Education, Atlanta, GA. OXFORD, R. L. and COHEN, A. D. (1992) Language learning strategies. Crucial issues in concept and classification. Applied Language Learning 3(1-2), 1-35. OXFORD, R. L. and CROOKALL, D. (1989) Language learning strategies: Methods, findings, and instructional implications. Modern Language Journal 73, 404-419.
ADULTS' LANGUAGELEARNING STRATEGIES
385
OXFORD, R. L. and EHRMAN, M. E. (1993) Second language research on individual differences. In W. Grabe (ed.),
Annual Review of Applied Linguistics (pp. 188-205). Cambridge, UK: Cambridge University Press. OXFORD, R., EHRMAN, M. E. and LAVINE, R. Z. (1991). Style wars: teacher-student style conflicts in the language classroom. In S. S. Magnan (ed.), Challenges in the 1990s for College Foreign Language Programs (pp. 1-25). Boston: Heinle & Heinle. OXFORD, R. L., EHRMAN, M. E. and NYIKOS, M. (1988) Vive la diff6rence? Reflections on sex differences in use of language learning strategies. Foreign Language Annals 21(4), 321-329. OXFORD, R. L., HOLLAWAY, M. and HORTON-MURILLO, D. (1992) Language learning styles: research and practical implications in the multicultural tertiary ESL/EFL classroom. System 20(4), 439-456. OXFORD, R. and NYIKOS, M. (1989) Variables affecting choice of language learning strategies by university students.
Modern Language Journal 73(3), 291-300. OXFORD, R., PARK-OH, Y., ITO, S. and SUMRALL, M. (1993a) Factors affecting achievement in a satellite-delivered Japanese language program. American Journal of Distance Education 7(1), 10-25. OXFORD, R., PARK-OH, Y., ITO, S. and SUMRALL, M. (1993b) Learning a language by satellite television: what influences student achievement? System 21(1), 31-48. OXFORD, R. and SHEAR/N, J. (1994) Broadening the theoretical framework of language learning motivation. Modern
Language Journal 78, 12-28. PARRY, T. S. and STANSFIELD, C. (eds) (1990) Language Aptitude Reconsidered. Englewood Cliffs, N J: PrenticeHall. PHILLIPS, V. (1990) English as a second language learner strategies of adult Asian students using the Strategy Inventory for Language Learning. Unpublished doctoral dissertation, San Francisco State University, San Francisco. PHILLIPS, V. (1991, Nov.) A look at learner strategy use and ESL proficiency. CATESOL Journal, 57-67. REID, J. M. (1987). The learning style preferences of ESL students. TESOL Quarterly 21, 87-111. ROSSI-LE, L. (1989) Perceptual learning style differences and their relationship to language learning strategies in adult students of English as a second language. Unpublished doctoral dissertation, Drake University, Des Moines, IA. ROq~FER, J. B. (1966) Generalized expectancies for internal vs. external control of reinforcement. Psychological
Monograph 80, Whole No. 609. RUB/N, J. (1975). What the 'good language learner' can teach us. TESOL Quarterly 9, 41-51. SAUNDERS, D. (1989) Type Differentiation Indicator Manual: A Scoring System for Form J of the Myers-Briggs Type Indicator. Palo Alto, CA: Consulting Psychologists Press. SCARCELLA, R. C. and OXFORD, R. L. (1992) The Tapestry of Language Learning: The Individual in the Communicative Classroom. Boston: Heinle & Heinle. SCHMECK, R. (1988) Individual differences and learning strategies. In C. Weinstein, E. Goetz and P. Alexander (eds),
Learning and Study Strategies: Issues in Assessment, Instruction, and Evaluation (pp. 171-191). New York: Academic Press. SINGLETON, D. (1989) Language Acquisition: The Age Factor. Clevedon, Avon: Multilingual Matters. SKEHAN, P. (1989) Individual Differences in Second Language Learning. London: Edward Arnold. SNOW, R. (1989) Aptitude-treatment interaction as a framework for research on individual differences in learning. In P. L. Ackerman, R. J. Steinberg and R. Glaser (eds), Learning and Individual Differences: Advances in Theory and Research (pp. 11-59). New York: W. H. Freeman. STEVICK, E. W. (1976) Memory, Meaning, and Method: Some Psychological Perspectives on Language Learning. Rowley, MA: Newbury House. TOUBA, N. ( 1992, Feb.) Language Learning Strategies of Egyptian Student Teachers of English. Paper presented at the Twelfth National Symposium on English Teaching in Egypt, Nasr City, Cairo, Egypt. VANN, R. and ABRAHAM, R. (1989) Strategies of unsuccessful language learners. Paper presented at the annual meeting of Teachers of English to Speakers of Other Languages, San Francisco, CA. WATANABE, Y. (1990) External variables affecting language learning strategies of Japanese ESL learners: Effects of entrance examination, years spent at college/university, and staying overseas. Unpublished master's thesis, Lancaster University, Lancaster, UK. WEINSTEIN, C. (1987) LASS1 User's Manual for Those Administering the Learning and Study Strategies Inventory. Clearwater, FL: H & H Publishers. WEINSTEIN, C. E., GOETZ, E. T. and ALEXANDER, P. A. (eds) (1988) Learning and Study Strategies: Issues in Assessment, Instruction, and Evaluation. New York: Academic Press. WEINSTEIN, C., PALMER, D. R. and SCHULTE, A. (1987) Learning and Study Strategies Inventory. Clearwater, FL: H & H Publishers.
386
REBECCA L. OXFORD and MADELINEE. EHRMAN
WEN, Q. and JOHNSON, R. K. (1991) Language learning approaches and outcomes: A study of tertiary English majors in China. Paper presented at the Sixth International Conference of the Institute of Language Education, Hong Kong. WENDEN, A. L. ( 1991) Learner Strategiesfor Learner Autonomy: Planning and Implementing Learner Trainingfor Language Learners. Englewood Cliffs, NJ: Prentice-Hall. WENDEN, A. and RUBIN, J. (eds) (1987). Learner Strategies in Language Learning. Englewood Cliffs, NJ: PrenticeHall. WESCHE, M., EDWARDS, H. and WELLS, W. (1982) Foreign language aptitude and intelligence. Applied Psycholinguistics 3, 127-140. WHITE, R. H. (1959) Motivation reconsidered. PsychologyReview 66(5), 291-333. WILDNER-BASSETT, M. (1992, Nov.) Relationships among language learning strategies, personality types, and learning styles. Paper presented at the annual meeting of the American Council on the Teaching of Foreign Languages, Chicago. YANG, N-D. (1992) Second language learners' beliefs about language learning and their use of language learning strategies. Unpublished doctoral dissertation, University of Texas at Austin.