1his is an author-producedPDF1he definitive, published
version is available at www.blackwell-synergy.com.
Language Learning 55:4, December 2005, pp. 613-659 Language Learners` Motivational ProIiles and Their Motivated Learning Behavior Kata Csi:er Etvs Universitv, Budapest Zoltan Drnvei Universitv of Nottingham The aim oI the present study is to deIine and describe second language (L2) learners` motivational proIiles by means oI a statistical procedure that is relatively rarely used in L2 studies: cluster analysis. To achieve this aim, 5 broad dimensions oI students` motivational and attitudi-nal dispositions toward 5 diIIerent L2s were measured and analyzed, using survey data collected Irom 8,593 13- and 14-year-old Hungarian pupils on 2 occasions. Statistical calculations yielded 4 distinct motivational groups, and we interpret this classiIication within Drnyei`s L2 Motivational SelI System. In the second part oI the study we examine the combined eIIects and interIerences oI the diIIerent motivational proIiles learners hold with regard to 2 diIIerent target languages, English and German. Individual-diIIerence (ID) research in second language (L2) studies has typically aimed at identiIying dimensions oI enduring language learner characteristics relevant to the mastery oI an L2 that are assumed to apply to everybody and on which people diIIer by degree (Drnyei, 2005). Most research eIIort in Kata Csizer, Department oI English Applied Linguistics; Zoltan Drnyei, School oI English Studies. This article was completed as part oI the OTKA T047111 project. Correspondence concerning this article should be addressed to Kata Csizer, 2030 Erd, Lakatos u. 44, Hungary. Internet: weinkatayahoo.com 613
the past has been expended on exploring the L2 impact oI iso- lated ID variables such as language aptitude, L2 motivation, or learning styles. To achieve this, researchers have usually admi- nistered selI-report questionnaires to language learners and then processed the data by means oI complex statistical procedures such as correlation and Iactor analyses, analysis oI variance, and structural equation modeling. Although studies conducted in this vein have helped us to understand the nature oI various cognitive and aIIective Iactors, past research has been less inIormative about how the diIIerent Iactors are combined in learners to achieve speciIic learner tvpes. We know Irom personality psychology that certain basic personality types or proIiles do exist: The best evidence Ior this is the validity oI the Myers-Briggs Type Indicator (MBTI), which is the most widely employed personality test in the world, designed to categorize people into distinct personality types. Peter Skehan (1991) has argued that we are likely to Iind diIIerent learner types in the L2 context as well, made up oI various abilities and personality characteristics that contribute to the successIul mastery oI an L2. This study is based on the application oI a multivariate statistical procedure, cluster analvsis, which is particularly useIul in distinguishing various learner proIiles. Even though this analytical technique can potentially oIIer both theoretical and practical insights into a wide range oI issues within L2 acquisi- tion, it has been used surprisingly rarely in past research: Even a thorough literature review will not detect more than a handIul oI relevant recent studies (cI. Kojic-Sabo & Lightbown, 1999; Skehan, 1986; Yamamori, Isoda, Hiromori, & OxIord, 2003), and as Alexander and Murphy (1999) demonstrate, the situation is no diIIerent in the broader Iield oI educational psychology. The objective oI cluster analysis is to identiIy within a given sample certain subgroupsor clustersoI participants who share similar combinations oI characteristics. The rationale Ior using the procedure in L2 studies is the observation that in spite oI the large number oI Iactors that shape L2 learning success, within a community oI L2 learners there appear to be a smaller number oI distinct subcommunities who share similar cognitive and motivational patterns. As mentioned above, indirect conIirmation Ior this in personality psychology has been provided by the MBTI, which divides people into a relatively rough grouping oI as Iew as 16 basic personality types that appears to work remarkably well (cI. Leaver, Ehrman, & Shekhtman, 2005). This would suggest that it is a realistic expectation to uncover a Iew archetypal patterns in L2-related ID Iactors as well, and cluster analysis oIIers a theoretically valid and practically sound technique Ior achieving this: It enables researchers to reassert the relative value oI various ID variables within complex conIig- urations in speciIic learner contexts, thereby allowing Ior the study oI social models oI learning and instruction (Alexander & Murphy, 1999). We must also point out, however, that similarly to Iactor analysis, cluster analysis is an exploratorv technique whereby the computer uncovers various grouping patterns based on the mathematical conIigurations Iound in the learner data. OI course, the computer cannot interpret the content oI the diIIerent variables, and thereIore in order to use the technique mean- ingIully, researchers need to be careIul to keep the analysis on sound theoretical Iooting and to substantiate the emerging learner proIiles by means oI various validating procedures. Alexander and Murphy (1999) mention one procedure that is particularly appropriate Ior substantiating results in educational contexts that involves using an external criterion variable to Iunction as an independent indicator oI cluster group diIIerences. Broadly speaking, iI the identiIied learner subcommunities perIorm signiIicantly diIIerently on the given criterion measure, this conIirms the validity oI the particular grouping solution. The Iocus in our study has been on proIiling the motivational disposition oI teenage L2 learners in Hungary, and in doing so we applied a theoretical construct that had been developed on the basis oI a consistent line oI research; Iurthermore, we also identiIied two important L2 criterion measuresthe learners` intended effort to learn the L2 and their choice of the Csi:er and Drnvei Language Learning 615 614 Jol. 55, No. 4
L2 (s) to studyIor validating the obtained motivational cluster proIiles. In accordance with Skehan`s (1991) prediction that there are certain conIigurations oI learner traits that are more than the mere sum oI the cumulative contributions oI the number oI ID constituents, we will show that exploring salient patterns in the motivational disposition oI L2 learners is a IruitIul research direction that can shed meaningIul light on the reality oI L2 learning. The Second Language Motivation Construct Applied in Our Study It is universally accepted that motivation plays a vital role in academic learning in general, and this is particularly true oI the sustained process oI mastering an L2. Because oI the complex nature oI 'language, L2 motivation has been conceptualized as a multiIaceted construct that comprises a number oI more general, trait-like and more situation-speciIic, state-like components (Ior reviews, see Clement & Gardner, 2001; Drnyei, 2001; Gardner, 1985; Maclntyre, 2002; Noels, 2001; Ushioda, 2003). In a recent overview oI the Iield, Drnyei (2005) proposed a new approach to the understanding oI L2 motivation, conceived within an L2 moti- vational selI system, which attempts to integrate a number oI inIluential theoretical L2 approaches with Iindings in 'selI research in psychology. The central theme oI this new conception is the equation oI the motivational dimension that has traditionally been interpreted as 'integrativeness/integrative motivation with the 'ideal L2 selI. The latter reIers to the L2-speciIic Iacet oI one`s 'ideal selI, which is the representation oI all the attributes that a person would like to possess (e.g., hopes, aspirations, desires): II one`s ideal selI is associated with the mastery oI an L2, that is, iI the person that we would like to become is proIicient in the L2, he/she can be describedusing Gardner`s (1985) terminologyas having an 'integrative disposition. Following the work oI Higgins (1987, 1998), Drnyei (2005) has proposed the existence oI another possible selI dimension as well, the 'ought-to L2 selI, which concerns the more extrinsic (i.e., less internalized) types oI instrumental motives: This selI- guide reIers to the attributes that one believes one ought to possess (i.e., various duties, obligations, or responsibilities) and that thereIore may bear little resemblance to one`s own desires or wishes. Although ideal and ought-to selves are similar to each other in that they are both related to the attainment oI a desired end state, Higgins (1998) emphasizes that the predilections associated with the two types oI Iuture selves are motivationally distinct Irom each other: Ideal selI-guides have a promotion Iocus, concerned with hopes, aspirations, advancements, growth, and accomplishments, whereas ought selI-guides have a prevention Iocus, regulating the absence or presence oI negative outcomes, concerned with saIety, responsibilities, and obligations. Thus, Irom a 'selI perspective, L2 motivation can be seen as the desire to reduce the perceived discrepancies between the learner`s actual selI and his or her ideal and ought-to L2 selves. Db`rnyei`s L2 Motivational SelI System also contains a third major dimension oI the L2 motivation complex labeled 'L2 learning experience, which concerns executive motives related to the immediate learning environment and experience. Because oI the methodological characteristics oI our research (outlined below), this dimension has not been addressed in this study. The current investigation is a Iollow-up to Drnyei and Csizer`s (2002) and Csizer and Drnyei`s (2005) analysis oI the data obtained Irom a large-scale (N ~ 8,500) repeated stratiIied national survey oI 13- and 14-year-old schoolchildren in Hungary. In this survey we assessed attitudes toward Iive target languagesEnglish, German, French, Italian, and Russianand in order to obtain comparable measures Irom the large number oI diIIerent loca- tions Irom all over the country, we needed to Iocus on motiva- tional variables that were generalizable across various learning situations. Thus, the study reported in this article does not involve situation-speciIic motives that are rooted in the L2 learners` immediate learning environment; instead we will be 616 Language Learning Jol. 55, No. 4 617 Csi:er and Drnvei
targeting more stable and generalized motives that stem Irom a succession oI the students` past experiences in the social world. Our previous research has established that Hungarian L2 learners` generalized motives can be described well in terms oI seven components (Ior a detailed discussion oI how these dimen- sions were identiIied, see Drnyei & Csizer, 2002): The relevant Cronbach`s alpha coeIIicients, along with the number oI items making up the scale, will be provided Ior each scale and lan- guage (Ior the sake oI brevity, only the statistics Ior 1999 are listed; Ior statistics concerning the previous occasion, that is, data Ior 1993, see Drnyei & Csizer, 2002). Integrativeness, which, similarly to Gardner`s (1985) category, reIlects a general positive outlook on the L2 and its culture, to the extent that learners scoring high on this Iactor would like to communicate with and might even want to become similar to the L2 speakers (three items; English .72; German .78; French .77; Italian .79; Russian .70). Instrumentalitv, which reIers to the perceived pragmatic beneIits oI L2 proIiciency, corresponding to Gardner`s (1985) category (Iour items; English .80; German .78; French .76; Italian .78; Russian .78). Attitudes toward L2 speakers, which concerns attitudes toward having direct contact with L2 speakers and traveling to their country (three items; English/US .71; English/UK .74; German .77; French .77; Italian .77; Russian .75). Cultural interest (or 'indirect contact), which reIlects the appreciation oI cultural products associated with the particular L2 and conveyed by the media (e.g., Iilms, television programs, magazines, and popular music) (Iour items; English/ US .67; English/UK .76; German .71; French .75; Italian .77; Russian .77). Jitalitv of L2 communitv, which, Iollowing the initial concep- tualization in Giles and Byrne`s (1982) 'intergroup model, concerns the perceived importance and wealth oI the L2 communities in question (two items; English/US .62; English/UK .66; German .62; French .65; Italian .63; Russian .55). Milieu, which relates to the general perception oI the importance oI Ioreign languages in the learners` immediate environment (e.g., in the school context and in Iriends` and parents` views) (Iour items; Cronbach`s alpha .61). Linguistic self-confidence, which reIlects a conIident, anxiety-Iree belieI that the mastery oI an L2 is well within the learner`s means (three items; Cronbach`s alpha .48). We have Iound (Csizer and Drnyei, 2005) that the two criterion measures included in our surveythe learners` intended learning eIIort and language choice preIerencewere directly aIIected by Integrativeness (or as we relabeled it, the Ideal L2 self), which was in turn determined by two antecedent variables, Instrumentalitv and Attitudes toward L2 speakers. The contribution oI all the other variables to the criterion measures was mediated by these Iactors. As discussed brieIly earlier, in our current research we will shiIt the Iocus Irom the actual motivational variables to the learners who possess them and will Iirst examine (a) whether we can Iind distinct learner types in terms oI their motivational proIiles, and iI so, (b) how these distinct patterns aIIect motivated learning behaviors. We will then analyze whether a learner`s motivational proIile associated with a particular L2 is inIluenced by the same learner`s disposition toward other possible target languages, that is, whether dispositions toward diIIerent L2s interIere with each other and aIIect the learner`s ultimate motivation. As Iar as we know, these research topics are novel in the L2 Iield, since we Jol. 55, No. 4 618 Language Learning Csi:er and Drnvei 619 620 Language Learning Jol. 55, No. 4 are not aware oI any past research that has examined motivational types and the possible impact oI L2 interIerence on L2 motivation. Method Participants The participants in our survey were 4,765 pupils (2,377 males; 2,305 Iemales; 83 with missing gender data) in 1993 and 3,828 pupils (1,847 males; 1,907 Iemales; 74 with missing gender data) in 1999. They were all ages 13 or 14 and attended the Iinal (8th) grade oI the primary school system (see Table 1). We selected this population in 1993 because at that time this was the most mature age group in the Hungarian educational system that studied within a more or less homogenous curricular and organizational Iramework (i.e., the national primary school sys- tem): AIter the age oI 14, pupils were oIIered a range oI diIIerent secondary educational paths, some oI which concluded when the students reached 16, the upper age limit oI compulsory education. ThereIore, by sampling students Irom this cohort, we did not need to be concerned with the modiIying inIluences oI various specialized secondary school types. At the same time, these learners were in the Iinal year oI their primary school studies and were just about to make the decision about which type oI secondary education to choose Ior their Iurther studies and which Ioreign language to study during the subsequent years. This lent particular relevance and validity to our question concerning language choice. In selecting the locations oI the survey, we Iollowed a strat- iIied sampling approach, trying to sample students evenly Irom each main region and type oI settlement, while also including a balanced mixture oI places Irequented and places not Irequented by tourists. In order to ensure the compatibility oI the samples in the two surveys, almost all the 1999 locations coincided with the
1993 ones (the decrease oI the size oI the sample in 1999 is largely due to the decline in the birth rate that Hungary had been experiencing). Thus, our study is an example oI a repeated cross- sectional design, consisting oI two surveys, one conducted at the beginning oI 1993, and the other taking place during the last Iew months oI the decade. Because the target population in both phases was exactly the same (and the sampling oI the participating schools was also almost identical), the project can be considered longitudinal in nature (cI. Keeves, 1994; Menard, 1991), thereby allowing us to directly compare the results and analyze the changes that took place during the six-and-a-halI- year interval between the two phases. Instrument The questionnaire was speciIically designed Ior the purpose oI the surveys. It consisted oI 37 items, using 5-point rating scales, assessing various student attitudes toward Iive target languages (English, German, French, Italian, and Russian) and toward six L2 communities (the United States, the United Kingdom, Germany, France, Italy, and Russia), and also asking about various aspects oI the students` language learning envi- ronment and background. In order to ensure that the instrument had appropriate psychometric properties, the items we used were adopted Irom established motivation questionnaires (some oI which had been speciIically developed to be used in Hungary), with suIIicient validity and reliability coeIIicients (e.g., Clement, Drnyei, & Noels, 1994; Drnyei, 1990; Gardner, 1985). Because 21 oI the 37 items Iocused on more than one L2 or L2 community (in a grid Iormat), even this relatively short instrument yielded a total oI 139 variables. In order to make the comparison oI the obtained data possible, the same instrument was used in both phases oI the project. The main variable groups in the questionnaire were as Iollows (with the total number oI items given in parentheses): Items concerning the Iive target languages (5-point rating scales): Orientations, that is, the students` various reasons Ior learning a given language (5 items). Attitudes toward the L2 (2 items). Intended effort, that is, the amount oI eIIort the student was willing to put into learning the given language (1 item). Parents language proficiencv (2 items). Items concerning the six target language communities (5-point rating scales): Attitudes toward the L2 communitv, that is, the extent to which students Ielt positively toward the particular country and its citizens (2 items), and the international importance they attached to this community (2 items). Britain and the United States were mentioned separately to explore diIIer- ences in the evaluations oI the two communities, in spite oI their common language (reIerred to in this study, where relevant, as English/UK and English/US). Contact with the L2 and its speakers, which assessed both the quantity (2 items) and the quality (5 items) oI the contact (e.g., watching L2 TV programs, meeting tourists). Non-language-speciIic Likert scales (5-point scales): Attitudes toward L2 learning at school (1 item). Contact with foreign languages through watching satellite TV (1 item). Fear of assimilation, that is, the extent to which students believed that learning and using the Ioreign language might lead to the loss oI the native language and culture (1 item). Self-confidence in L2 learning and use (3 items). 622 Language Learning Jol. 55, No. 4 Csi:er and Drnvei 623
Language learning milieu, that is, the extent oI the par- ents` support oI (1 item) and the Iriends` attitudes toward (1 item) L2 learning. Background questions (open-ended and multiple-choice items): Language choice. Students were asked to name three languages they were intending to learn in the next school year (1 item). Personal variables, such as the student`s sex and language learning background (7 items). As described in the introduction, the learner responses have been used to compute seven broad motivational dimensions, Iive oI them speciIic to the various target languages/ language communities examined: Integrativeness, Instrumentalitv, Jitalitv of the communitv, Attitudes toward L2 speakers, and Cultural interest. Two dimensions, Self-confidence and Milieu, were assessed in a non-L2-speciIic manner and will not be included in the current analysis because our investigation speciIically Iocuses on motivational proIiles speciIic to certain L2s and their interIerences. This exclusion is justiIiable on the basis oI the results oI Csizer and Drnyei (2005), which showed that the impact oI Self-confidence and Milieu on the criterion measures was almost entirely subsumed or mediated by the other Iive motivational dimensions. With regard to the psychometric properties oI the instrument, the mean Cronbach`s alpha internal consistency reliability coeIIicient oI the Iive scales Ior the six L2 communities in the two survey phases (i.e., 56 coeIIicients) was .70, which is acceptable Ior such short scales. Data Collection Data collection was conducted in a similar way in 1993 and 1999. On both occasions, we Iirst approached the selected schools through an oIIicial letter Irom Eb`tvos University, Budapest (which hosted the project), providing inIormation about the pur- pose oI the survey and details oI the actual administration oI the questionnaires. Once permission was granted by the principal oI a particular school, we contacted the Iorm masters oI the selected classes individually, asking Ior their cooperation. The questionnaires (in Hungarian) were Iilled in during class time; a representative oI the university was always present at the administration, providing the introduction and overseeing the procedure. Answering the questions took the students approxi- mately 20 min on average. Data Analvsis As mentioned earlier, the aim oI cluster analysis is to identiIy homogenous subgroups within a samplein our case to deIine learner groups with distinct motivational proIiles. As Kojic-Sabo and Lightbown (1999) have pointed out in discussing the use oI cluster analysis in L2 research, iI every participant in a sample could be characterized by a unique pattern, each one would Iorm a separate cluster, and there would be as many clusters as students. On the other hand, iI all students responded to the questionnaire in exactly the same way, only one cluster would emerge Irom the analysis. In actual analyses we get a meaningIul result somewhere between these two extremes. There are two main types oI cluster analysis, hierarchical and nonhierarchical clustering. In hierarchical clustering the Iirst step involves the deIinition oI each sample member as an individual cluster. Subsequent steps merge the closest clusters until one single cluster containing all sample members is arrived at. The procedure might be illustrated by a 'dendogram, which shows each step oI the process. As a result, the systematic structure oI the data is provided by showing what subgroups could be deIined. Nonhierarchical clustering, on the other hand, Iollows a diIIerent path. During the process, sample members are put into a predeIined number oI clusters. As a Iirst step, Jol. 55, No. 4 624 Language Learning Csi:er and Drnvei 625
the statistical program takes the Iirst n members oI the sample (where n equals the number oI clusters, which is deIined prior to the analysis), and sample members closest to these predeIined centers are assigned to these initial clusters. Then, on the basis oI the position oI the cluster members, new centers are identiIied, and sample members are regrouped according to these new centers. The procedure is repeated until the centers become stable, that is, they show no change aIter Iurther regrouping. Both hierarchical and nonhierarchical clustering have their advantages and disadvantages. On the one hand, hierarchical clustering is diIIicult to apply iI the sample size is too large. On the other hand, the results oI nonhierarchical clustering are highly dependent on the initial cluster centers. To avoid these limitations, clustering is usually done in two stages: First, hier- archical clustering is carried out on a smaller subsample oI the sample; in our case a 3 random subsample was selected Ior this purpose. Based on this Iirst step, the number oI clusters and their positions (i.e., the initial cluster centers) are deIined, and subsequently nonhierarchical clustering is perIormed on the whole sample by inputting the cluster centers received pre- viously; the procedure oI nonhierarchical clustering is iterated until stable cluster centers are obtained. We Iollowed this com- bined method. Results and Discussion Definition of Motivational Groups / Profiles Based on the procedure described above, the Iirst step oI the data analysis was to subject the language- and country-related multi-item scales to hierarchical clustering. The procedure pro- duced dendograms, that is, visual representations oI the steps in the solutions that showed the clusters being combined and the values oI the distance coeIIicients at each step. Based on these dendograms, it was decided to Iollow a Iour-cluster solution in the subsequent nonhierarchical clustering. This decision-making process is not unlike the one involved in Iactor analysis when researchers need to decide the Iinal number oI Iactors in the Iactor solutions: Although there are no absolute criteria, a combination oI theoretical and practical considerations can oIIer relatively good guidelines. In our case we Iollowed a straightIorward and commonsense grouping approach that resulted in similar groups Ior all the target languages. The Iour groups, as will be shown below, showed good discriminant validity with regard to the two criterion variables against which they were measured. Tables 2a through 2I describe the Iinal solutions Ior each language by indicating the means oI the clustering variables in the diIIerent cluster groups. (For a visual representation oI the clusters, see Figure 1, which presents results Ior English/ US in 1999.) As can be seen in Tables 2a-2I, the cluster groups Ior the diIIerent target languages are largely similar, with only French and Italian displaying some diIIerences Irom the common trend. The actual variable mean scores show some variation across languages, but this does not thwart the group analysis, as each language has been explored individually. Group 1 consists oI pupils scoring lower than average on all the motivational scales concerning the particular language/country; thereIore, this group has been labeled the least motivated students. Group 4, on the other hand, is the inverse oI Group 1, as it contains students who scored higher than average on each and every scale; accordingly, this group has been labeled the most motivated students. Because motivational variables and dimensions are usually intercorrelated, it was to be expected to have groups characterized by these extreme patterns. The moti- vational proIiles oI the two in-between groups displayed an intri- guing conIiguration. Instead oI a homogeneous distribution oI the variables with only their degree oI endorsement diIIering between the two groups, a more unusual pattern emerged: For some variables, Group 3 scores exceeded those oI Group 2, whereas Ior others, it was the other way around. That is, the 627 Csi:er and Drnvei Jol. 55, No. 4 626 Language Learning
Language Learning Vol. 55, No. 4
Figure 1. Visual representation oI the motivational variables in the Iour cluster groups Ior English/US (1999). two groups diIIered not only in terms oI their motivational intensity but also in terms oI the structure oI their motivational proIile. BeIore analyzing this qualitative diIIerence and discussing any L2-speciIic variations, let us examine the validity oI the grouping by substantiating the results against the criterion variables. Group Performance on the Criterion Measures As described earlier, one established way oI substantiating cluster-analytical results in educational contexts is to use external criterion variables as independent indicators oI cluster group diIIerences. Tables 3 and 4 present the mean scores oI the two criterion measures used in this studythe learners` Intended effort and Language choiceacross the Iour cluster groups. As can be seen in the tables, there is a very consistent relationship between group membership and perIormance level on the criterion measures: The higher the group number, the higher the 634 Csi:er and Drnvei 637 perIormance, with the diIIerences reaching statistical signiIicance in all but one case. Since motivation is by deIinition the antecedent oI motivated behavior, and our two criterion measures concern two key aspects oI motivated human behavior, its direction and its magnitude, this consistent relationship between the Iour motivational proIiles and the two criterion measures in all the various L2/L2 community conditions provides a strong case Ior the validity oI the clustering process. The Difference Between Groups 2 and 3 The group scores in Tables 3 and 4 unambiguously indicate that Group 3 had a higher level oI overall motivation than Group 2. The same conclusion can also be drawn Irom the Iact that the Integrativeness scores oI Group 3 were signiIicantly higher than those oI Group 2 in every condition (please recall that Csizer and Drnyei |2005| Iound that Integrativeness mediated all the other motivational components and was thereIore equated with the Ideal L2 self). However, in spite oI the overall superior motivational level oI Group 3, we Iind that Group 2 students exceed those in Group 3 on a number oI motivational dimensions. How can we interpret this seemingly contradictory situation? Let us start our analysis by taking a closer look at Table 2a, which describes the results Ior English/US, Iirst, since English is the most popular L2 in Hungary and this popularity is typically Iueled by American rather than British associations (Drnyei & Csizer, 2002). As can be seen in the table, apart Irom Integrativeness, Group 3 exceeds Group 2 in Instrumentalitv, whereas Group 2 shows superiority in terms of Attitudes toward L2 speakers, Cultural interest, and Jitalitv. In describing the L2 Motivational SelI System, Drnyei (2005) has argued that the ideal L2 selI is composed oI two complementary aspects: In our idealized image oI ourselves we want to appear personallv agreeable (associated with positive attitudes toward the L2 community and culture) and also professionallv successful (associated
with instrumental motives). This dichotomy seems to correspond to the split between Groups 2 and 3 very well: Group 2 members are superior on the aIIective side (they have more positive atti- tudes toward the L2 community and culture), whereas Group 3 members are superior on the pragmatic side (placing more emphasis oI the incentives oI L2 proIiciency). What is important Irom our point oI view is that neither group has developed a Iully Iledged salient L2 language selI (which characterizes Group 4 members), and thereIore analyzing their diIIerences can help us to understand how the selI system works. Learners in Group 3 have a powerIul sense oI the proIes- sional relevance oI L2 proIiciency, but this is not accompanied by a similarly positive regard toward the L2 culture and community. This suggests that they have not internalized their proIessional interest to the extent that it has developed into an ideal language selI. Indeed, the motivational proIile oI this group appears to be determined by the ought-to language selI, which is the counterpart oI the ideal selI in Drnyei`s (2005) system described earlier. This explains why Group 3 learners are strong on instrumental motives but much weaker on other attitudinal aspects and why their overall level oI motivation does not reach that oI members oI Group 4. Learners in Group 2 appear to be the reverse oI Group 3 members: Although they have positive attitudes toward every- thing that is associated with English, they do not see English as related to their Iuture proIessional liIe (witnessed by the low instrumental score). They simply do not think that they will need the L2, which is a common Ieature oI many language learners. Thus, although they are positively disposed toward the target language, it simply does not reach the necessary level oI relevance Ior them to develop into an ideal L2 selI. Why these learners display a lower overall level oI motivation than those in Group 3 can also be explained with the motivational selI paradigm: Although members oI neither group have a salient ideal L2 selI, those in Group 3 are at least motivated by an ought-to L2 selI. Thus, Ior English (US), the primary Ioreign language in Hungary, the cluster analysis appears to provide validation oI Drnyei`s (2005) L2 Motivational SelI System. II we look at the other languages (Tables 2b-2I), we see a largely similar pattern to the above, but Ior some variables in which the English data revealed a superiority oI Group 2, Group 3 scores actually exceed those oI Group 2. Out oI the six relevant scores Ior which the English data showed the inverse pattern Ior Groups 2 and 3 (three variables in 1993 and 1999 each), this deviation occurs Ior each language less than twice on average. For the languages that play a less important role in the Hungarian context French, Italian, and Russianthe unstable pattern can be linked to the weak and somewhat changing position oI the languages (cI. Drnyei & Csizer, 2002). However, Ior one variable (Attitudes toward L2 speakers, 1999), even English (UK) and German deviate Irom the pattern obtained with English (US). We believe that this is related to the signiIicant tendency, discussed by Drnyei and Csizer, during the 1990s Ior interethnic attitudes among Hungarians to become more negative in general. This, we believe, is due partly to some sort oI a 'disillusionment process that occurred when people realized that their heightened expectations had not been Iully met by 'joining the Iree world and partly to the increased level oI contact with Ioreign visitors, among which German-speaking tourists constituted by Iar the largest group (Ior an in-depth analysis oI the issue oI contact, see Drnyei & Csizer, 2005). The Si:e and Gender Composition of the Jarious Motivational Groups Having established the validity oI the clustering by means oI both evaluating the cluster groups against external criterion measures and discussing the cluster characteristics within a theoretical Iramework, it is interesting to look at the size and gender variations across the Iour cluster groups. Table 5 presents the size oI the cluster groups in terms oI their percentages Jol. 55, No. 4 Language Learning 638 Csi:er and Drnvei
Csi:er and Drnvei 641 oI the total sample. As could be expected, the percentages show considerable variation, but the emerging pattern is in accordance with the popularity rank order oI the Iive languages determined by Drnyei and Csizer (2002), according to which English was by Iar the most popular language in the sample, Iollowed by German, French, Italian, and Iinally Russian. Thus, we would expect English, the most popular language, to be associated with the largest most motivated students group and Russian, the least popular language, with the largest least motivated students group. As the table shows, this is indeed the case, and all the other results spread logically and evenly between these two extremes. Table 6 indicates the gender dominance in the diIIerent motivational clusters. The pattern conIirms in a visually clear manner Drnyei and Csizer`s (2002) Iinding wherein girls dis- played superior language attitudes to boys across the board. All but one oI the most motivated clusters are dominated by girls, and most oI the least motivated clusters are dominated by boys. These Iindings are in accordance with results typically reported in the literature Irom a wide variety oI learning contexts, indicating on the one hand that girls are more successIul in virtually every aspect oI language learning, and on the other hand that Ioreign language learning is increasingly seen by boys as a 'girly subject in many countries. Although some hypotheses have been proposed about the reasons Ior this general pattern, to the best oI our knowledge no comprehensive explanation has been provided to date about the marked gender variation. Target Language Interference in Language Choice and Effort The Iact that our investigation has Iocused on the learners as the Iocal unit allows us to examine the interIerence oI their various dispositions toward the various target languages. For example, it is reasonable to assume that a strong attraction to a particular language is likely to exist at the expense oI other possible target languages, but hardly any L2 motivation studies
Csi:er and Drnvei 643 have addressed such attitudinal interIerences. Because cluster analysis was carried out separately Ior each language, each student had a separate motivational group membership index with regard to each oI the target languages; that is, a learner could, Ior example, belong at the same time to the highly motivated cluster in English and to the not-at-all-motivated one in Italian. Looking at the combinations oI the various memberships can, then, answer the question as to whether it is better to be motivated to learn only a single language, or whether an overall interest in Ioreign languagesas indicated by belonging to the highly motivated groups across the boardprovides a stronger commitment. A Iurther question is how English, the indisputable world language, aIIects preIerences Ior other, less vital languages. As a Iirst step, we have divided the learners into Iive groups, depending on how many most motivated group memberships they achieved across the languages (i.e., how many times in Tables 2a- 2I they qualiIied in Group 4). In this analysis we have omitted Russian, because the learners` attitudes toward it were so negative that we could not expect the preIerence Iigures to provide a reliable contribution to the regrouping oI students according to the number of most motivated cluster memberships, and we used only one index Ior English, that associated with the United States, as this was the dominant perception in the sample. Thus, we obtained Iive groups (one group Ior each number oI memberships Irom 0 to 4, with the latter group made up oI learners who had the highest level motivational proIile in all the languages: English, German, French, and Italian) and then conducted an analysis oI variance to examine how the scores on the two criterion measures varied across the Iive groups. As can be seen in Table 7, with regard to Intended effort the picture is straightIorward: Breadth oI L2 motivation (as indicated by the number oI top group memberships) is in a direct positive rela- tionship with the intended level oI eIIort to learn all the L2s in question. That is, the more oIten a participant belonged to the most motivated group in the various languages, the higher amount oI eIIort the person wanted to invest in any oI these
646 Language Learning Jol. 55, No. 4 languages. Because we suggested earlier that membership in a most motivated group is an indication oI the salience oI one`s ideal L2 selI, this Iinding means that a highly developed ideal language selI is associated with a general interest in languages a Iinding that is in accordance with Gardner`s (1985) original conceptualization oI integrativeness, which also contains an 'interest in Ioreign languages component. The interesting question is whether the results Ior Language choice display the same pattern as those Ior Intended effort. Although the two variables are obviously linkedthat is, the extent oI one`s willingness to choose an L2 and the level oI eIIort one intends to expend on learning it are interrelatedthere is one big diIIerence between the two measures. While the responses Ior Intended effort Ior diIIerent L2s are unrelated to each other (i.e., one can assign the highest score to all the L2s), with regard to Language choice one is Iorced to state preIerences, both in real liIe (because our capacity Ior L2 learning is limited) and also in the actual survey (because students were asked to choose and rank-order three languages). Thus, the choice oI one language inevitably aIIects that oI another. The pattern in Table 8 reIlects this diIIerence in that instead oI the straightIorward positive relationship that characterized Intended effort, we Iind a more complex picture. At this point we must note that a typical Hungarian learner would master only one L2 (at best), with only the best ones having a serious go at two L2s; studying three L2s seriously is very rare, and Iour is almost unheard oI. ThereIore, Ior the sake oI clarity in this analysis we will ignore Italian, the popularity oI which is behind that oI the other three international languages (English, German, and French) and which was included in the research paradigm primarily as a control. The results Ior French in Table 8 are the closest to the straightIorward pattern obtained Ior Intended effort. This makes sense: Because most Hungarian learners want to master English and German Iirst, only the best and most motivated learners will attempt French. ThereIore, here the number oI
Csi:er and Drnvei 649 most motivated learner group memberships is relevant and cor- relates with the desire to learn French. With English and German we Iind a strikingly diIIerent pattern. For English it almost does not matter what one`s motivational proIile is like because the data reveal a very high general level oI endorsement: Even the smallest coeIIicient (Ior Group 0 in 1993) is 2.24 out oI a maximum oI 3, which actually exceeds the largest coeIIicient Iound Ior German in any oI the cells (1.87). It is no wonder, thereIore, that in 1999 there were no signiIicant between-group diIIerences concerning the English language choice results, F(4,3823) 1.8, p .125, which reIlects well the perceived 'world language status oI English: As described in Drnyei and Csizer (2002), people increasingly study English not necessarily because they are motivated to do so but because it is seen as part oI general education, similarly to reading, writing, and arithmetic. With German the surprising Iinding is that both in 1993 and 1999 the top-scoring groups Ior Language choice were Groups 0 and 4 jointly. That is, one was most likely to choose German either iI one had a very broad or a very limited interest in Ioreign languages. How can this be? We believe that this Iinding is a result oI substantial interIerence Irom World English. II a learner was not interested in language learning at all, then he or she was quite likely to give a high mark in the questionnaire Ior German, the traditional lingua Iranca in Hungary (and in Central Europe in general)this was merely a deIault rating, without implying any real meaningIul motivation. II the person had some interest in L2s, he or she was likely to mark mainly English, which was perceived to be the dominant world languagehence the lower score Ior German in the interim groups. However, iI the person was reallv interested in language learning, he/she was likely to want to learn both English and German, as these were generally accepted as the main world and regional languages, respectivelyhence the high score Ior German. In order to test this hypothesis, we have conducted a more in-depth analysis, this time Iocusing only on the German/English interIerence.
Target Language Interference in Language Choice and Effort. Focus on English and German In order to achieve a detailed understanding oI the combined eIIect and interIerence oI English and German, we computed 16 new learner groups based on the combinations oI the Iour learner proIiles Ior English and German described in Tables 2a and 2c. We then compared the two criterion measures in the 16 groups separately: Table 9 presents the results Irom 1999
(because the results Irom 1993 and 1999 were very similar, Ior the sake oI clarity we will Iocus on the latest results only). As can be seen, the trends Ior Intended effort and Language choice are almost exactly the same: Students preIer German to English only (a) iI they are motivated to learn German and they are not motivated to learn English (Groups 2-4), or (b) iI they are highlv motivated to learn German and only marginally to learn English (Groups 8-12). This clearly shows that the motivation to learn German is dependent on the motivation to learn English, whereas the latter is not aIIected by the Iormer. The above analysis Iocused on the direction oI motivation (i.e., the L2 preIerence). We also wanted to check whether the intensitv (i.e., magnitude) oI the motivated behavior showed any systematic variation across the 16 groups. As expected, an analysis oI variance with grouping as the independent variable produced highly signiIicant results in all the Iour language-criterion combinations. With regard to the Intended effort to learn English, we Iind an almost straightIorward positive linear relationship (similar to the one in Table 3): As shown by Table 10 and illustrated by Figure 2, the higher the group level oI English, the higher the eIIort scores, and the level oI German group membership does not seem to make any diIIerence. Thus, in this regard English is not aIIected by German. With regard to Language choice, the picture is somewhat diIIerent (see Table 11). The rhythmically sloping pattern oI the graph Ior this variable in Figure 2 shows some German inter- Ierence: The 'bumps in the graph coincide with the lowest German group membership level (i.e., Level 1) and the 'dips with the highest German group membership level (i.e., Level 4). Thus, even though students in general tend to express a preIer- ence Ior English over German, whether or not they like German modiIies their overall eagerness to choose English. Let us now look at the German results. Our earlier analyses have already indicated a strong interIerence oI World English when it comes to learning other languages, and indeed the Language choice scores Ior German (Table 12) conIirm this. 650 Language Learning Jol. 55, No. 4
Iocus on motivational learner types and proIiles is a IruitIul one and that cluster analysis is an eIIective statistical procedure to apply in this context. We have uncovered Iour broad motiva- tional proIiles that characterized learners regardless oI the spe- ciIic target language or the time oI the survey. The Iirst group consisted oI the least motivated learners, who were basically not interested in Ioreign languages, cultures, and language learning. The other extreme was Group 4, the most motivated learners, who showed a generally high disposition across all the motiva- tional dimensions. We argued that these latter learners have successIully developed a salient ideal L2 selI, which was also associated with an interest in Ioreign languages in general. The two interim groups showed an intriguing proIile diIIerence: Whereas Group 2 Ieatured more positive attitudes toward the L2 culture and community, Group 3 members were superior on instrumental aspects. This pattern was interpreted within Drnyei`s (2005) L2 Motivational SelI System: We argued that the reason why the learners in these groups had not developed a strong ideal L2 selI was (a) in the case oI Group 2, a lack oI a proIessional Iuture relevance oI the L2, and (b) in the case oI Jol. 55, No. 4 Language Learning Jol. 55, No. 4 Language Learning 654 Csi:er and Drnvei 655
motives that are rooted in the L2 learners` immediate learning environment, but only more stable and generalized motives. This is why only two main components oI Db`rnyei`s (2005) L2 Motivational SelI System were identiIied: the ideal L2 selI and the ought-to L2 selI; the third component, the L2 learning experience, was not addressed in this study. Revised version accepted 27 April 2005 ReIerences Alexander, P. A., & Murphy, P. K. (1999). Learner proIiles: Valuing indi- vidual diIIerences within classroom communities. In P. L. Ackerman, P. C. Kyllonen, & R. D. Roberts (Eds.), Learning and individual differences. Process, trait, and content determinants (pp. 412-431). Washington, DC: American Psychological Association. Clement, R., Drnyei, Z., & Noels, K. A. (1994). Motivation, selI-conIidence, and group cohesion in the Ioreign language classroom. Language Learning, 44, 417-448. Clement, R., & Gardner, R. C. (2001). Second language mastery. In H. Giles & W. P. Robinson (Eds.), The new handbook of language and social psvchologv (2nd ed., pp. 489-504). London: Wiley. Csizer, K., & Drnyei, Z. (2005). The internal structure oI language learning motivation and its relationship with language choice and learning eIIort. Modern Language Journal, 89, 19-36. Drnyei, Z. (1990). Conceptualizing motivation in Ioreign-language learning. Language Learning, 40, 45-78. Drnyei, Z. (2001). Teaching and researching motivation. London: Longman. Drnyei, Z. (2005). The psvchologv of the language learner. Individual differences in second language acquisition. Mahwah, NJ: Erlbaum. Drnyei, Z., & Csizer, K. (2002). Motivational dynamics in second language acquisition: Results oI a longitudinal nationwide survey. Applied Linguistics, 23, 421-462. Drnyei, Z., & Csizer, K. (2005). The eIIects oI intercultural contact and tourism on language attitudes and language learning motivation. Journal of Language and Social Psvchologv, 24(4), 1-31. Gardner, R. C. (1985). Social psvchologv and second language learning. The role of attitudes and motivation. London: Arnold. Jol. 55, No. 4 Language Learning 658 Giles, H., & Byrne, J. L. (1982). An intergroup approach to second language acquisition. Journal of Multilingual and Multicultural Development, 3, 17-40. Higgins, E. T. (1987). SelI-discrepancy: A theory oI relating selI and aIIect. Psvchological Review, 94, 319-340. Higgins, E. T. (1998). Promotion and prevention: Regulatory Iocus as a motivational principle. Advances in Experimental Social Psvchologv, 30, 1-46. Keeves, J. P. (1994). Longitudinal research methods. In T. Husen & T. N. Postlethwaite (Eds.), The international encvclopedia of education (2nd ed., pp. 3512-3524). OxIord: Pergamon. Kojic-Sabo, I., & Lightbown, P. M. (1999). Students` approaches to vocabu- lary learning and their relationship to success. Modern Language Journal, 83, 176-192. Leaver, B. L., Ehrman, M. E., & Shekhtman, B. (2005). Achieving success in second language acquisition. Cambridge: Cambridge University Press. Maclntyre, P. D. (2002). Motivation, anxiety and emotion in second language acquisition. In P. Robinson (Ed.), Individual differences in second language acquisition (pp. 45-68). Amsterdam: Benjamins. Menard, S. (1991). Longitudinal research. Newbury Park, CA: Sage. Noels, K. A. (2001). New orientations in language learning motivation: Toward a contextual model oI intrinsic, extrinsic, and integrative orien- tations and motivation. In Z. Drnyei & R. Schmidt (Eds.), Motivation and second language acquisition (pp. 43-68). Honolulu: University oI Hawai`i Press. Skehan, P. (1986). Cluster analysis and the identiIication oI learner types. In V. Cook (Ed.), Experimental approaches to second language acquisition (pp. 81-94). OxIord: Pergamon. Skehan, P. (1991). Individual diIIerences in second language learning. Studies in Second Language Acquisition, 13, 275-298. Ushioda, E. (2003). Motivation as a socially mediated process. In D. Little, J. Ridley, & E. Ushioda (Eds.), Learner autonomv in the foreign language classroom. Teacher, learner, curriculum, assessment (pp. 90-102). Dublin, UK: Authentik. Yamamori, K., Isoda, T., Hiromori, T., & OxIord, R. (2003). Using cluster analysis to uncover L2 learner diIIerences in strategy use, will to learn, and achievement over time. International Review of Applied Linguistics, 41, 381-409.