Abstract
This article explores the nature of learner beliefs, how these beliefs can change over time and how their beliefs relate to learners’ developing proficiency. It reports three studies of learner beliefs. Ellis (2002) used metaphor analysis to explore the beliefs of six beginner classroom learners of L2 German. Tanaka (2004) conducted a longitudinal study of Japanese students on a 12-week study abroad programme in New Zealand, using both quantitative and qualitative methods to show how these beliefs changed over time and the relationship between these changes and their developing proficiency. Zhong (2008) conducted a case study of a Chinese migrant learner of English in New Zealand, documenting how her beliefs changed over 10-week period and how these changes were reflected in changes in her English proficiency. These studies suggest that researchers wishing to investigate learner beliefs would do better to rely on qualitative methods such as interviews and diary studies rather than questionnaires. They also point to the situated and dynamic nature of learner belief systems and the indirect relationship between beliefs and learning.

Introduction
Language learners form ‘mini theories’ of L2 learning (Hosenfeld, 1978) which shape they way they set about the learning task. These theories are made up of beliefs about language and language learning. Clearly ‘beliefs’ constitute an individual difference variable notably different from the other individual difference factors such as language aptitude or motivation but, like these variables, beliefs influence both the process and product of learning. Also, like a number of other individual difference variables, they are dynamic and situated.

Investigating learner beliefs
Three different approaches to investigating learners’ beliefs can be distinguished (Barcelos, 2003). According to the normative approach, beliefs are seen as ‘preconceived notions, myths or misconceptions’, which can be studied by means of Likert-style questionnaires such as the Beliefs About Language Learning Inventory—BALLI (Horwitz, 1987a). The metacognitive approach views learners’ metacognitive knowledge about language learning as ‘theories in action’ (Wenden, 1999); these are examined by means of the content analysis of learner self-reports in semi-structured interviews. Finally, the contextual approach views learner beliefs as varying according to context; it involves collecting a variety of data types and diverse means of data analysis. Barcelos argued that the contextual approach is superior because rather than viewing beliefs as a ‘mental trait’, it takes into account the ‘experience-based nature of beliefs’ (p. 26). A fourth approach can also be identified—metaphor analysis (Ellis, 2002; Kramsch, 2003). This entails analysing the metaphors used by learners to describe their learning and constitutes an indirect means of identifying beliefs.
   Much of the research has been concerned with describing and classifying the types of beliefs learners hold (based on responses to questionnaires), the sources of beliefs, and the situated and dynamic nature of learners’ belief systems. Somewhat disappointingly, very few studies have examined the relationship between beliefs and language learning.

Types of learner beliefs
In an early attempt to identify the types of beliefs held by language learners, Horwitz (1987) administered the BALLI to groups of learners. Five general areas of beliefs emerged from the analysis of the responses relating to (1) the difficulty of language learning, (2) aptitude for language learning, (3) the nature of language learning, (4) learning and communication strategies, and (5) motivation and expectations. Wenden (1986, 1987) grouped the beliefs she identified in 25 adults enrolled in a part-time advanced-level class at an American university into three general categories: (1) use of the language (for example, the importance of ‘learning in a natural way’), (2) beliefs relating to learning about the language (for example, the importance of learning grammar and vocabulary), and (3) the importance of personal factors (i.e. beliefs about the feelings that facilitate or inhibit learning, self-concept, and aptitude for learning). Both of these early studies, then, identified a very similar set of learner beliefs. For example, the learners in both Horwitz’s and Wenden’s studies demonstrated beliefs about the need to study grammar. This dominant belief was also reported by Schulz (2001), who found that both Colombian learners of English in Colombia and American learners of foreign languages in the US placed great store on explicit grammar study and error correction.
   Later research attempted to classify rather than simply list types of beliefs and to link them to metacognitive knowledge (Wenden, 1999). Benson and Lor (1999), for example, distinguished higher-order ‘conceptions’ and lower-order ‘beliefs’. They defined ‘conceptions’ as ‘concerned with what the learner thinks the objects and processes of learning are’ whereas beliefs are ‘what the learner holds to be true about these objects and processes’ (p. 464). A number of studies, including that of Benson and Lor, who investigated Chinese undergraduate students at the University of Hong Kong, suggest that learners hold conceptions about what language is and how to learn and that these conceptions fall into two broad categories, which can be glossed as ‘quantitative/analytic’ and ‘qualitative/experiential’. Table 1 indicates the kinds of beliefs related to each. It should be noted that these two general conceptions are not mutually exclusive; learners can and often do hold a mixed set of beliefs. A number of studies (for example, Tanaka 2004) also suggest a third general conception—‘self-efficacy/confidence’ in language learning. This conception has more to do with how learners perceive their ability as language learners and their progress in relation to the particular context in which they are learning.

Conception	Nature of language	Nature of language learning
Quantitative/analytic	Learning an L2 is mostly a matter of learning grammar rules. In order to speak an L2 well, it is important to learn vocabulary.	To understand the L2 it must be translated into my L1.   Memorization is a good way for me to learn an L2.
Qualitative/experiential	Learning an L2 involves learning to listen and speak in the language. To learn a language you have to pay attention to the way it is used.	It is okay to guess if you do not know a word.   If I heard a foreigner of my age speaking the L2 I would go up to that person to practise speaking.

Table 1: Types of learner beliefs (based on Benson and Lor, 1999)

The sources of learners’ beliefs
An interesting question is what determines learners’ beliefs about language learning. Little, Singleton, and Silvius (1984, reported in Little and Singleton, 1990) surveyed random samples of undergraduate and postgraduate students of foreign languages at Trinity College, Dublin. They found that ‘past experience, both of education in general and of language learning in particular, played a major role in shaping attitudes to language learning’ (1990, p. 14). For example, the students stated that they preferred to learn by production activities (repeating orally and writing) rather than through receptive activities involving listening and reading. Little and Singleton claimed that this belief reflected the general nature of the instruction they had experienced (i.e. was shaped by their instructional experiences).
   Another possibility is that beliefs are culturally determined. However, Horwitz (1999) in her review of the research into L2 beliefs concluded that there was insufficient evidence to show that learners’ beliefs varied systematically according to cultural background. It is possible, however, that learners’ beliefs are more substantially influenced by general factors such as personality and cognitive style but this remains to be shown.

The situated and dynamic nature of learner beliefs
Learner beliefs are situation specific and dynamic. Kern (1995), for example, reported changes in the beliefs of 180 students studying first-year level French at a university in the US over the course of one semester (15 weeks). He administered Horwitz’s BALLI to the students during the first and last week of the semester. Analysing the responses of 180 students, Kern reported that 35 per cent to 59 per cent of the responses changed over the 15-week period. A significant change was observed in the response to the statement ‘If you are allowed to make mistakes in the beginning, it will be hard to get rid of them later on’, with 37 per cent of the students reporting greater agreement and 15 per cent lesser agreement. This suggests that many students had become increasingly conscious of their mistakes and were having difficulty in avoiding them. The learners also changed their responses to the statement ‘Learning a foreign language is mostly a matter of learning a lot of grammar rules’, with 32 per cent showing greater agreement and 20 per cent lesser agreement.

The relationship between beliefs and learning
There have to date been very few studies of the relationship between learner beliefs and learning outcomes. Abraham and Vann (1987) found some evidence that beliefs might affect learning outcomes in a case study of two learners, Gerardo and Pedro. Both learners believed that it was important to create situations for using English outside the classroom, to practise as much as possible, and to have errors corrected. Both also believed it important to participate actively in class. Gerardo, however, believed that paying conscious attention to grammar was important, while Pedro did not and expressed a strong dislike of meta-language. Also, Gerardo thought that it was important to persevere in communicating or understanding an idea, while Pedro considered topic abandonment the best strategy in some cases. Abraham and Vann characterized Gerardo’s philosophy of language learning as ‘broad’ and Pedro’s as ‘narrow’. They suggested that this might have contributed to Gerardo’s better TOEFL score (523 versus 473) at the end of a course of instruction. Pedro, however, did better on a test of spoken English, which might suggest that different views about language learning result in different kinds of success.
   Park (1995) investigated 332 Korean university EFL students’ beliefs about language learning, their language learning strategies, and the relationships among their beliefs, strategy use, and L2 proficiency. Park found three variables predicted students TOEFL scores to some extent. One was a belief variable (i.e. beliefs about self-efficacy and social interaction) and two were strategy variables (i.e. independent/interactive strategies and metacognitive strategies). Those learners who reported having confidence in learning English and the intention of speaking to others in English tended to use English actively, especially outside the classroom, and to monitor their progress in English carefully. These behaviours were also related to improvement in L2 proficiency.
   Mori (1999) investigated the beliefs of 187 university students enrolled in Japanese at various proficiency levels in the US. She examined the relationship between epistemological beliefs (i.e. beliefs about learning in general) and beliefs about language learning and also the relationship between beliefs and L2 achievement. She found that strong beliefs in innate ability (i.e. the ability to learn is inherited and cannot be improved by effort) and in avoidance of ambiguity (i.e. the need for single, clear-cut answers) were associated with lower achievement. Learners who believed that L2 learning was easy manifested higher levels of achievement. In addition, this study showed that there were belief differences between novices and advanced learners. Advanced learners were less likely to believe in simple, unambiguous knowledge or the existence of absolute, single answers than novice learners. This study also revealed that epistemological beliefs and beliefs about language learning were for the most part unrelated. In other words, learner beliefs about language learning seemed to be task and domain specific.
   Tanaka and Ellis (2003) reported a study of a 15-week study-abroad programme for Japanese university students, examining changes in the students’ beliefs about language learning (measured by means of a questionnaire) and in their English proficiency (measured by means of the TOEFL). The results showed statistically significant changes in the students’ beliefs relating to analytic language learning, experiential language learning and especially self-efficacy/confidence during the study-abroad period. Statistically significant gains in proficiency were also reported. However, Pearson’s Product Moment correlations between the students’ responses to the Belief Questionnaire and their TOEFL scores both before and after the study abroad period were weak and generally statistically non-significant. There was also no relationship between changes in beliefs after a three-month period of study abroad and gains in proficiency.
   Overall these studies do not show a strong relationship between beliefs and learning/proficiency. However, it is perhaps not surprising that the relationship between beliefs and proficiency is weak, as the fact that learners hold a particular belief is no guarantee they will act on it; conflicts with other strongly held beliefs, situational constraints, or personal reasons may prevent them. If beliefs do impact on learning it is likely that they do so indirectly by influencing the kinds of learning strategies learners employ.

Three studies of learner beliefs
In order to further examine the nature of learner beliefs and their relationship to L2 proficiency three separate studies will now be summarised.  These studies addressed the following research questions:

1. What beliefs about language learning do L2 learners hold?
2. To what extent and in what ways do L2 learners’ beliefs about language learning change over time?
3. What is the relationship between learners’ beliefs and their developing L2 proficiency?

The first study addresses research question (1) only.  The second and third studies examined all three questions.

Study 1:  Ellis’ (2002) metaphor study of learners’ beliefs
As pointed out in the introduction, most studies of learners’ beliefs use some kind of self-report to investigate them (e.g. questionnaires or interviews). There are two problems with this approach. One is that learners may not always report their beliefs accurately (i.e. they may instead report the beliefs that they think they should hold and that the researcher wishes to hear). The second problem is that self-report assumes that learners are aware of the beliefs they hold and are able to verbalize them. While this may be the case with many beliefs it is also possible that some beliefs lie below the threshold of consciousness or cannot be easily and directly expressed. An alternative approach to investigating approach is to examine the metaphors that learners use when talking about their learning experiences. These provide an indirect means of examining their belief systems.
   Cognitive linguists like Lakoff and Johnson (1990) argue that metaphor is not ‘special’ or ‘rare’ but quite commonplace and also that metaphor is a conceptual phenomenon as well as a linguistic one. That is they reflect, how people represent the world and their experiences of it to themselves. From this perspective, therefore, conceptual metaphors can function as windows to view belief systems – people use metaphors to both construct and constrain thought. It should be noted that metaphor analysis is now an accepted tool in educational and applied linguistic enquiry (see Cameron and Low, 1999).
   The participants in Ellis’s study were 6 adult learners of German enrolled in beginner German courses in two tertiary institutions in London. Four of the learners were aged 18 years and were native English speakers. One learner was aged 25 and Spanish speaking and one was aged 20 and French speaking. The latter two were also fluent in English. The courses they enrolled in were part of their bachelor degree studies. They were highly intensive and focused to a considerable extent on mastering the grammar of English although there were also opportunities for using German communicatively (e.g. in content based lessons about Germany or through conversing with German speaking visitors to the class).
   Five learners were invited and agreed to keep a diary throughout the course. In addition, one other learner volunteered to keep a diary. They were given detailed instruction about the kinds of topics they could comment on in their diary (e.g. their attitudes to German as a language; their response to the instructional activities and to their teacher; their sense of their own progress; the learning difficulties they experienced; their motivation to learn German). However, they were not told to report their beliefs about language learning. The diaries were collected in weekly, photocopied and then returned to the learner. Altogether they kept their diaries for 10 weeks.
   The diaries were then analysed to identify and classify the metaphors they used.  This involved the following steps:

1. The metaphorical expressions in the texts were identified.
2.The source and target domains of the metaphors were identified.
3.On the basis of this analysis 'main metaphors' were identified.  These metaphors were considered to be ‘conceptual’ (i.e. to reflect ways in which the learners viewed and interpreted their world).
4.The entailments of each main metaphor were identified.
5.The key words of each conceptual metaphor were identified and a concordancing programme (Scott and Oxford University Press, 1999) employed to identify the linguistic realisations of each metaphor. These linguistic realizations were then listed under a heading for each conceptual metaphor.  Table 2 below provides examples of the key words and linguistic citations for one of conceptual metaphors (LEARNING AS A JOURNEY).
6. Raters (two applied linguists) were asked to read through the lists and to determine whether, in their view, each realization did indeed constitute a metaphor and whether it belonged to the conceptual metaphor it was listed under.

Key word	Citation example
direction	I shot off in the wrong direction.
keep up	No matter how hard I try I just seem     unable to keep up.
stuck	I find myself really stuck.
lost	I got hopelessly lost.
advancing	I feel I am advancing in German little by little.

Table 2:  key words and citation examples for the LEARNING AS A JOURNEY metaphor
The analysis yielded six main conceptual metaphors. Table 3 summarizes the six learners’ use of these metaphors. 
The most commonly used of these was LEARNING AS A JOURNEY (which corresponds closely to the LIFE AS A JOURNEY metaphor commented on in Lakoff and Johnson (1990)). The learners appeared to view learning as a kind of ‘Pilgrim’s Progress’ (i.e. involving a journey where a series of difficulties had to be overcome to achieve the final goal). This metaphor was used to refer to progress overall and to progress in particular lessons. It allowed the learners to discuss both their sense of success and failure. Interestingly, it also enabled them to express both their affective and cognitive beliefs about language. Ellis (2002) suggested that learners might be using this metaphor to distance themselves from their learning experience (i.e. it served as a kind of metacognitive strategy for evaluating their progress).
   The second main conceptual metaphor was LEARNING AS A PUZZLE.  This was the second most frequently used metaphor and was employed by all six learners.  Learners used it to address both their problems with the German course and the solutions they found. The major problem identified through the metaphor was ‘grammar’, reflecting the same pre-occupation with this aspect of learning as Kern’s (1995) learners.  Grammar was seen as a ‘puzzle’ that needed to be ‘cracked’.  Unlike the LEARNING AS A JOURNEY metaphor, this metaphor related exclusively to cognitive aspects of language learning.
   Another common main conceptual metaphor was LEARNING AS SUFFERING. The learners appeared to believe that some degree of suffering was necessary if they were to be successful in learning German. In the case of some of the learners, the suffering was intense, although it should be noted that this did not appear to have a notably adverse effect on their motivation. Rather it caused them to try harder. The source of the learners’ suffering varied; in some cases it was the teachers (e.g. one learner felt anxiety whenever the teacher addressed a question that she had to answer in front of the whole class) but in others learners it arose as a result of their sense of lack of progress or failure to understand some aspect of the language. Clearly, this metaphor relates to the affective side of learning.
   LEARNING AS A STRUGGLE was employed by just three of the learners. This metaphor positioned learners as both agents in the ‘fight’ to learn and also as ‘victims’ of the struggle to learn.
   The final conceptual metaphor was LEARNING AS WORK (i.e. that learning German was like doing a job, requiring the learners to perform certain tasks regularly and rewarding them for their efforts. There were numerous references to ‘work’ and ‘working’ but many were probably not metaphorical. The main belief reflected through this metaphor was that learning a language required self-directed effort.  However, relatively few possible entailments of the metaphor were exploited by the learners.

______________________________________________________
Learner                                    Metaphors
                                    (1)        (2)        (3)        (4)        (5)
____________________________________________________
1.  Maria                      *          *          *          *          *
2.  Monique                  *          *          *          *          *
3.  Debbie                                *          *                      *
4.  Robert                     *          *          *          *          *
5.  Caroline                  *          *          *          *          *
6.  Manuel                    *          *          *          *
Totals                           61        22        14        10        18
_______________________________________________________
Key:     (1)  LEARNING AS A JOURNEY
(2) LEARNING AS A PUZZLE
(3)  LEARNING AS SUFFERING

1. LEARNING AS A PUZZLE
2. LEARNING AS WORK

Table 3:  Summary of the learners’ use of the five metaphors

What does this study show us about these learners’ beliefs about language learning? First, it reveals that many of the beliefs that learners hold relate to the problems they experience while learning. All six learners found learning German problematic – both cognitively and affectively. It also shows that these learners held beliefs that positioned them as both ‘agents’ of their own learning and as ‘patients’ who undergo experiences they could not easily control (i.e. they saw themselves as both self- and other-directed).  The study also showed that learner beliefs involve both cognitive and affective aspects of language learning. In this respect the results of the metaphor analysis do not accord very closely with the beliefs measured by belief questionnaires such as Horwitz’s BALLI, which includes no reference to ‘hardship’, ‘suffering’ or ‘long-term effort’. One reason for this might be that when asked directly about their beliefs learners do not typically respond by mentioning their affective beliefs. To tap these it may be necessary to adopt a more indirect approach such as that afforded by metaphor analysis. Overall, this study suggests that metaphor analysis is a promising tool for examining learners’ beliefs although it should also be acknowledged that the identification of metaphor remains somewhat problematic.

Study 2: Tanaka’s (2004) study of Japanese learners’ beliefs and language proficiency
The second study aimed to examine the changes in Japanese learners’ of English belief systems when they came to study in an English-speaking country in a study abroad programme. It also sought to examine the relationship between their beliefs and language proficiency. The study covered a 12 week period from the time they first arrived in New Zealand.
   A total of 132 Japanese learners of English were investigated. They were divided into two groups: (1) The New Zealand Group, which consisted of 63 Japanese students studying English in an Auckland tertiary institute for 12 weeks and (2) the Japanese Group, which consisted of 69 Japanese students who were studying English in a Japanese university in Tokyo. This second group served as a comparison group.
   Data collection involved both instruments directed at investigating the learners’ beliefs and their English language proficiency. All the learners completed a Beliefs Questionnaire consisting of 27 Likert scale items designed to measure beliefs relating to analytic learning, experiential learning and affective factors. Examples of the statements in the questionnaire are as follows:

1.  Analytic learning:
e.g.       In order to speak English well, it is important for me to learn grammar.
I would like my English teacher to correct all my mistakes.

2.  Experiential learning:
e.g.       I can learn well by speaking with others in English.
            I can learn well be listening to the radio or watching TV

3.  Affective factors:
e.g.       I am satisfied with my progress so far.
            It is possible for me not to get nervous when speaking English.

   More qualitative information about the learners’ beliefs was collected by means of an interview of selected learners at the end of the 12 weeks and a diary that five students kept about their English learning experiences. The learners’ language proficiency was measured at the beginning and the end of the 12 week period by means of the listening and grammar sections of the Oxford Placement Test and  by an oral narrative task which was recorded and transcribed and then analyzed in terms of fluency, complexity and accuracy.
   No statistically significant changes in beliefs were evident in the questionnaire responses – the learners differed in the direction of the change in their responses with the result that positive and negative shifts cancelled each other out. The interview and diary data were more revealing. Most of the students expressed dissatisfaction with their English proficiency at the beginning of the study, which they attributed to the poor English language education they had received in Japan. Their views about learning English changed in a number of ways over the 12 weeks. In particular, a number of the learners became more balanced (i.e. they identified the need for both experiential and analytic approaches). This was reflected in their changing attitudes to grammar. Initially they were opposed to grammar, a reaction to the grammar-dominated lessons they had experienced in Japan. However, as time passed they came to realise that grammar was important if they were to express themselves effectively. They also became more realistic learners in that they realised that living in an English-speaking country did not lead to automatic proficiency. They came to see that learning English was a long and difficult process. Finally, they recognized over time that they could not just rely on the language lessons they received and saw the importance of their own efforts and aptitude for learning English.
   The development in proficiency was not entirely as expected. The NZ group did manifest greater gains in the Oxford Placement Test scores than the Japan group.  However, although the NZ showed a significant gain in fluency this was not statistically greater than that observed in the Japan group. In other words, there was no evidence that living in New Zealand for 12 weeks led to greater improvement in fluency than learning English in communicatively oriented classrooms in Japan.
   Overall the relationships between beliefs (as measured by the questionnaire) and the proficiency measures were very weak. The NZ Japanese students who reinforced their beliefs relating to experiential learning during the study abroad tended to advance more in general proficiency but a similar advantage was not observed in speaking ability. The changes in beliefs relating to analytic learning and affective states were not related to either general proficiency or speaking ability.
   One clear effect of the NZ experience concerned the attributions that learners gave for their perceived success or failure in learning. They switched from naïve optimism about learning in a natural environment to recognizing the importance of their own efforts. In other words, the study abroad situation afforded them experiences that enabled them to evaluate their own progress more effectively and make changes in the way they approached learning English. However, there were considerable individual differences among the students. Not all were able to ‘learn from failure’ by adjusting their approach to learning – some just gave up.
   Tanaka’s study points to the limitations of questionnaires as a means of investigating learners’ beliefs. It was clear from the qualitative data that a number of changes did take place in the learners’ belief systems yet these were not captured by their responses to the questionnaire. His study also suggests that the relationship between learner beliefs and their developing proficiency may not be a strong one.  This is an issue we will return to in the conclusion to this article.

Study 3: Zhong’s (2008) study of a migrant ESL learner
This situated case study investigated the beliefs of one Chinese learner of English living in Auckland over a 10-week period. It aimed to examine the developments that occurred in this learner’s beliefs and the relationship between her beliefs and changes in her language proficiency.
   The learner was Lin (a pseudonym), a 26 year old Chinese-speaking migrant who was living in Auckland, New Zealand. She had been living in New Zealand about 6 months when the study commenced and had enrolled as a fulltime student in an intermediate course for migrant learners in a tertiary college.
   As in Tanaka’s study, data relating to both beliefs and language proficiency were collected. Two interviews (one at the beginning and one at the end of the 10 week period) were conducted to elicit information about her beliefs.  In an attempt to how the learner behaved in a classroom context, three observations were carried out, each complemented with a stimulated recall session where Lin was invited to comment on her behaviour in different episodes taken from the lessons.  Her language proficiency was measured by means of the Oxford Placement Test and Nation’s vocabulary level tests, both of which were administered at the beginning and end of the 10 week period.  Also, Lin completed an oral narrative twice. This was recorded and transcribed and then analysed to provide measures her fluency, complexity and accuracy.
   The main change in Lin’s beliefs concerned self-efficacy – Lin gained in both her confidence to learn English and in her ability to manage her own learning. There were also a number of other changes. She became less convinced in the value of rote learning. She came to see the value of working in pairs and groups with other students. She placed less emphasis on the importance of being corrected. She broadened her belief in the importance of ‘using English’ to include not just practising words in contrived sentences but in actual communication.         Overall, Lin developed a strong belief in the importance of self-direction and in using language and learning experientially. However, overall most of Lin’s beliefs remained unchanged – e.g. her belief in the importance of vocabulary.
   Zhong used the classroom observations and information gleaned from the interviews to consider the learning strategies that Lin employed. Notable strategies included keeping a vocabulary book, always sitting next to non-Chinese students in class, seeking out opportunities to communicate in English outside the classroom and constantly monitoring her own progress. These constituted a mixture of cognitive, social and metacognitive strategies.
   The most notable change in Lin’s proficiency was evident in vocabulary – she showed considerable gains in the 3,000 and 5,000 levels. She also made a sizable gain (12%) in the Oxford Placement Test. Like Tanaka’s Japanese learners, Lin showed a clear gain in fluency when performing the oral narrative but there was no change in complexity and the accuracy of her spoken English actually decreased.
   The study suggests a relationship between Lin’s changing beliefs and her developing proficiency. The greater importance she attached to communicating is reflected in the increase in fluency. However, this involved a trade-off with accuracy, which clearly became less important to her over time. Her belief in the importance of vocabulary was also reflected in large gains in this aspect of language. Methodologically this study again points to the value of collecting qualitative data to investigate learner beliefs. It also suggests that to understand how beliefs are related to developing proficiency it is helpful to examine the learner’s choice of learning strategies and how these do or do not reflect beliefs. In the case of Lin there was a close relationship between her beliefs and the actions she carried out to learn English.

Conclusion              
The three studies reported in this article suggest a number of important points about the nature of learner beliefs, methods for researching them and their relationship to language learning. It is clear that learners hold beliefs not just about the cognitive aspects of language and language learning (i.e. whether a language is best learned analytically or experientially) but also about affective aspects (e.g. how to manage their emotional response to their learning experiences). In particular, they hold beliefs about their own self-efficacy (i.e. how well equipped they are personally to succeed in learning the language). The distinction between cognitive and affective beliefs is an important one because it would seem that although changes in cognitive beliefs can occur (e.g. Lin changed her views about the best way to practise vocabulary) changes in affective beliefs are common. As a result of their learning experiences, learners may express greater or lesser confidence in their ability to succeed. 
   Learners’ beliefs about language learning derive from a variety of sources – their past experience, both of education in general and of language learning in particular, their cultural background and their personality (which, in particular, may affect self-efficacy beliefs). But, as these three studies have shown, learners’ beliefs are both situated and dynamic. They change as a product of new situational experiences and, in particular, the attributions that learners make for their successes and failures.  Learners who engage deeply and seriously with language learning come to recognize that learning is a slow and difficult process, involves adopting a variety of analytic and experiential strategies, and, crucially, that it depends more on them than the teacher or instructional context.
   Only one of the three studies (Tanaka’s) made use of a beliefs questionnaire. The results obtained from this were somewhat disappointing. In particular, the questionnaire failed to show any changes in beliefs over time. Ellis’ study demonstrated the value of examining learners’ beliefs indirectly through metaphor analysis. Such an analysis revealed how central affective beliefs (e.g. ‘learning involves suffering’) were to the beginner learners that Ellis investigated. Both Tanaka’s and Zhong’s studies made use of a variety of qualitative data collection methods (interviews, diaries and stimulated recall) where learners self-reported their beliefs. These methods provided clear evidence of both the dynamic and situated nature of learner beliefs. Overall these studies point to a limitation in what Barcelos (2003) referred to as the normative approach to investigating beliefs and to the need for a more contextual approach. There is also need to examine the extent to which learners act on their beliefs (i.e. the learning strategies that they employ). However, only one of the three studies (Zhong’s) attempted this, showing a clear relationship between beliefs and actions in the learner this study investigated.
   Overall, however, the relationship between beliefs and learning/proficiency evident in these studies does not emerge as very strong. The relationship is necessarily an indirect one. That is beliefs do not have a direct effect on language learning but are mediated by the actions that learners perform. Thus, the strength of the relationship depends on the extent to which individual learners are able or are prepared to act on their beliefs - conflicts between beliefs, situational constraints, cultural background (see Schulz 2001) or personal reasons may prevent them. Thus, in the case of some learners (such as the learner Zhong studied) beliefs can have a clear effect on learning because the learners are able and prepared to act on their beliefs. In the case of other learners (such as some of the Japanese learners in Tanaka’s study), beliefs have little effect on learning because learners do not engage in learning activities compatible with them. What has emerged as especially important in these studies is the extent to which learners develop beliefs related to self-efficacy and self-directed learning as these govern the extent to which they are prepared to work on opportunities for learning in their particular learning context. 
   Finally, a few comments about learner beliefs and teaching. If beliefs influence the actions that learners perform to learn an L2, they cannot be ignored by teachers. Little learning is likely if there is a mismatch between the teacher’s and the students’ belief systems. This suggests the need for teachers to make their own beliefs about language learning explicit, to find out about their students’ beliefs, to help their students become aware of and to evaluate their own beliefs and to address any mismatch in their and their students’ belief systems.

References
Abraham, R. and Vann, R. (1987). Strategies of two language learners: A case study. In A. Wenden and J. Rubin (Eds.), Learner strategies in language learning (pp. 85-102). Englewood Cliffs, N.J.: Prentice Hall.

Barcelos, A. (2003). Researching beliefs about SLA: A critical review. In P. Kalaja and A. Barcelos (Eds.). Beliefs about SLA: New research Approaches (pp. 7-33). Dordrecht: Kluwer.

Benson, P. and Lor, W. (1999). Conceptions of language and language learning. System, 27, 459–72.

Cameron, L. and Low, G. (eds.). (1999). Researching and applying metaphor.  Cambridge: Cambridge Applied Linguistics.

Ellis, R. (2002). A metaphorical analysis of learner beliefs. In P. Burmeister, T. Piske and A. Rohde (Eds.), An integrated view of language development: Papers in honor of Henning Wode. Trier, Germany: Wissenschaftlicher Verlag.

Horwitz, E. (1987). Surveying student beliefs about language learning. In A. Wenden and J. Rubin (Eds.), Learner strategies in language learning (pp. 119-132). New York: Prentice Hall.

Horwitz, E. (1999). Cultural and situation influences on foreign language learners’ beliefs about language learning: A review of BALLI studies. System, 27, 557–76.

Hosenfeld, C. (1978). Students’ mini-theories of second language learning. Association Bulletin, 29, 2.

Kern, R. (1995). Students’ and teachers’ beliefs about language learning. Foreign Language Annals, 28, 71–91.

Lakoff, G. and Johnson, M. (1980). Metaphors we live by. Chicago: University of  Chicago Press.

Little, D. and Singleton, D. (1990). Cognitive style and learning approach. In R. Duda and P. Riley (Eds.), Learning styles (pp. 11-19). Nancy, France: University of Nancy.

Little, D., Singleton, D. and Silvius, W. (1984). Learning second languages in Ireland: Experience, attitudes and needs. Dublin: Trinity College, Centre for Language and Communication Studies.

Mori, Y. (1999). Epistemological beliefs and language learning beliefs: What do language learners believe about their learning. Language Learning, 49, 377–415.

Park, G. (1995). Language learning strategies and beliefs about language learning of university students learning English in Korea. Unpublished PhD dissertation, University of Texas at Austin.

Scott, M. and Oxford University Press. (1997). Wordsmith Tools Manual. Oxford: Oxford University Press.

Schulz, R. (2001). Cultural differences in student and teacher perceptions concerning the role of grammar instruction. The Modern Language Journal, 85, 244–58.

Tanaka, K. (2004). Changes in Japanese students’ beliefs about language learning and English language proficiency in a study-abroad context. Unpublished PhD thesis, University of Auckland.

Tanaka, K. and Ellis, R. (2003). Study abroad, language proficiency, and learner beliefs about language learning. JALT Journal, 25, 63–85.

Wenden, A. (1986). What do second language learners know about their language learning? A second look at retrospective accounts. Applied Linguistics, 7, 186–201.

Wenden, A. (1987). How to be a successful learner: Insights and prescriptions from L2 learners. In A. Wenden and J. Rubin (Eds.), Learner strategies in language learning (pp. 103-117). Englewood Cliffs, N.J.: Prentice Hall.

Wenden, A. (1999). An introduction. System, 27, 435–41.

Zhong, M. (2008).  Report of a pilot study of the beliefs of one migrant learner of English. Unpublished paper, Department of Applied language Studies and Linguistics, University of Auckland.