Second Language Acquisition: A Social Psychological Perspective
Abstract and Keywords
This article focuses on second language acquisition seen from a social psychological perspective. The basic premise underlying a social psychological perspective of second language acquisition is that language is a defining characteristic of the individual. It is involved in one's thoughts, self-communication, social interaction, and perception of the world. Moreover, language is a defining attribute of cultural groups. It serves to distinguish one group from another, and thus to reflect one's cultural identity. Thus, to learn a second language involves, to some extent, making part of another cultural group part of one's self, even if this is only the vocabulary, sounds, verbal forms, and so forth of that group. That is, the language is more than a symbolic system that facilitates communication among individuals; it is a defining feature of self-identity linked directly to the very social existence of the individual. The future looks bright, but continued research focusing on empirical findings represents its greatest strength.
The basic premise underlying a social psychological perspective of second language acquisition is that language is a defining characteristic of the individual.1 It is involved in one's thoughts, self-communication, social interaction, and perception of the world. Moreover, language is a defining attribute of cultural groups. It serves to distinguish one group from another, and thus to reflect one's cultural identity. Thus, to learn a second language involves, to some extent, making part of another cultural group part of one's self, even if this is only the vocabulary, sounds, verbal forms, and so forth of that group. That is, the language is more than a symbolic system that facilitates communication among individuals; it is a defining feature of self-identity linked directly to the very social existence of the individual.
A related issue that must be considered from this perspective is what constitutes learning a language. As initially formulated, the social psychological perspective viewed second language learning in terms of bilingual skill. Taking a few courses or being able to make oneself understood in the language in a halting fashion was not the criterion. The criterion was the development of a free-flowing and automatic use of the language comparable to native speakers. Thus, the social psychological perspective is concerned with the development of bilingualism, not the achievement of (p. 205) an A grade in the course or a high mark on a test. These achievements might represent stages along the way to the development of bilingual skill, but they are not the ultimate goals in the social psychological perspective. Having said this, however, it must be confessed that in those studies that assess the relationship between affective variables and achievement, most, if not all, assess proficiency in terms of grades in courses, scores on tests, ratings of oral proficiency, and so forth, and it is a testament to the validity of the social psychological perspective that it is able to explain proficiency even when it is defined in terms of these types of criteria.
Prior to the 1950s, the main determinants of success in learning a second language were thought to be intelligence and language aptitude. In his review of research on bilingualism, however, Arsenian 1945 devoted a section on social psychology of language and bilingualism in which he raised questions about the relation between language and acculturation, the role of affective factors in second language acquisition, intergroup relations and language learning, and so forth. Then, in the lead article of the inaugural issue of Language Learning, Marckwardt 1948 identified five motives that were influential in promoting the acquisition of another language. He described three of them as practical (assimilation of an ethnic minority, promotion of trade, and scientific utility), and two as nonutilitarian (self-cultural development and the maintenance of ethnic identity of a minority group). Later, Nida 1956 reported a case study of a missionary who was unable to learn the language of the group with which he was to work. He noted that the individual, who was the son of immigrant parents, had identified strongly with English-speaking America. Nida claimed that this extreme identification and an emotional block against anything “foreign” interfered with his being able to learn another language. The social psychological perspective of second language acquisition grew out of this background.
II. The Early Perspective
The initial empirical studies with a social psychological perspective were conducted by Lambert (1955, 1956a, 1956b, 1956c). His research was concerned with the assessment of bilingual dominance and the development of bilingualism. He investigated developmental changes in French and English among students differing in terms of language training—undergraduate students majoring in French, graduate students majoring in French, and native French speakers who had lived in an English-speaking country for an average of 7 years. Two of his observations are noteworthy here. One was that two of the graduate students measured dominant in French. In both cases, these students could be described as particularly motivated—one because of her career as a French teacher, and the other because of his extreme identification with France (Lambert, 1955). We can note the foundation of the integrative and instrumental dichotomy that still pervades much of the research in this field today.
(p. 206) The second observation was that the developmental changes in second language acquisition appeared to involve at least two plateaus. Lambert (1956c) described the first plateau as reflecting a “vocabulary” cluster because those measures that involved vocabulary strength distinguished the undergraduate students from the other two groups. He defined the second plateau as reflecting a “cultural” cluster because these measures differentiated between the native speakers on the one hand and the undergraduate and graduate groups on the other, and involved such features as habitual word order, stereotypy of response, and so forth. He proposed, therefore, that second language acquisition entails a series of barriers to overcome, with the vocabulary barrier being the easiest and the cultural barrier the most difficult. Although Lambert did not discuss other barriers, Gardner 2007 contrasted first and second language learning, identifying four phases common to both: the elemental, consolidation, conscious expression, and automaticity and thought stages. In this sense, the true acquisition of a second language can be seen to involve the development of expert performance in that language.
Conceptualizing second language acquisition in terms of the development of expert performance has important implications. Ericsson, Krampe, and Tesch-Römer (1993) reviewed the literature on expert performance in many facets of human behavior and concluded that it is not dependent primarily on innate talent or ability, as is commonly believed. They reviewed many studies on such diverse topics as typing speed, artistic prowess, chess mastery, and so forth, and found that expert performance results from repeated practice (i.e., from consistent and persistent motivated behavior). They concluded that it requires approximately 10 years of sustained practice to develop expertise and noted that it takes about this long to acquire the vocabulary of a normal adult. If, therefore, we view second language acquisition as equivalent to the development of expert behavior, it follows that the motivation to learn the second language must be long-standing. From the social psychological perspective, learning a second language means the acquisition of near-native facility with the content and structure of the language, and near automaticity in its use both conceptually and behaviorally. Such acquisition takes time and dedicated effort for an adult to achieve! Thus, when we speak of motivation to learn a second language, we are speaking not about some transient motivation but instead about a long-term commitment to the task (with the associated effort, desire and satisfaction).
The term motivation has been variously defined in the literature (see, for example, Kleinginna and Kleinginna, 1981), largely because it has many characteristics. The individual who is motivated has a goal, wants to achieve it, exerts effort to achieve it, enjoys activities associated with it, is persistent and attentive, has expectations, makes attributions, is self-confident, and so forth (cf. Gardner, 2006). It is general in nature in that it pervades much of the individual's behavior, and it is specific to activities or goals in which it is aroused. The important point is that motivation is multifaceted, not unidimensional, and is a characteristic of the individual, not the environment. That is, regardless of whether motivation is defined as intrinsic or extrinsic, integrative or instrumental, and so forth, it is an attribute (p. 207) of the individual that displays itself in all of the characteristics listed above, and it is the degree of motivation, not the source, that influences one's degree of success in learning the language. For example, if intrinsic motivation is superior to extrinsic motivation in promoting achievement, or if integratively orientated learners are more successful than instrumentally oriented ones, it is because one is more strongly associated with all the characteristics of motivation, not that the source is different.
The first investigation of individual differences in second language acquisition from the social psychological perspective was conducted by Gardner and Lambert (1959). They studied high school students learning French as a second language in Montreal and found that two factors were associated with achievement in French. One was aptitude; the other was motivation. They concluded that the motivation was “characterized by a willingness to be like valued members of the language community” (p. 271 [italics in the original]). Since then there have been numerous studies of the relation between attitudes and motivation and achievement (e.g., Gardner, 1985; Gardner and Lambert, 1972). Clément and Gardner (2001) present a graph, based on an unpublished survey of three databases (PsycLIT, ERIC, and Linguistics and Language Behavior Abstracts) of the number of investigations published between 1985 and 1994 that dealt with individual difference variables in second language acquisition. In this 10-year period, 496 studies dealt with attitudes and 218 with motivation. Obviously, some articles referred to both attributes, and not all of them referred just to attitudes directly related to the social psychological perspective, but it is clear that there was an active interest in the field.
III. Social Psychological Models of Second Language Acquisition
The initial social psychological model was outlined by Lambert (1967, 1974), who proposed that aptitude, attitudes, orientation, and motivation promote the development of bilingual proficiency and that this can have an effect on one's self-identity. Lambert (1974) distinguished between two types of bilingualism—additive and subtractive—that reflect the effect of second language acquisition on self-identity, and linked them to different language contexts. Additive bilingualism was seen to apply to members of the majority learning the language of a minority, in which they did not lose any of their own ethnic identity but developed proficiency in the other language. Subtractive bilingualism, on the other hand, was more characteristic of minority group members who, in learning the language of the majority, ran the risk of losing some of their own cultural identity.
Since then there have been a number of models presented, each changing the focus slightly and adding and subtracting elements. Gardner and Smythe 1975 (p. 208) proposed a model that retained the elements of Lambert's social psychological model but expanded it to take into account the language learning situation, distinguishing between formal and informal language learning contexts. This model, now referred to as the socio-educational model, has evolved since then (Gardner, 1985, 2000, 2007), and has become much more formal, operationalizing the concepts in terms of measures of specific attributes. Currently, it focuses on six latent constructs: language aptitude, attitudes toward the learning situation, integrativeness, motivation, language anxiety, and language achievement (characterized in terms of linguistic and nonlinguistic outcomes). Gardner and Smythe also developed and standardized the Attitude/Motivation Test Battery (AMTB) that assessed 11 relevant affective variables.2 The concept of the integrative motive is hypothesized to comprise the three constructs: attitudes toward the learning situation, integrativeness, and motivation; it can be assessed by aggregating scores on eight of the variables assessed by the AMTB. A measure of instrumental orientation is included in the measures, but it is not a construct emphasized in the model. With the exception of attitudes toward the learning situation and language anxiety, this model shares constructs with Lambert's model and, though self-identity is not explicitly identified in it, the concept of integrativeness involves the willingness to identify with the other language community.
A third model in this tradition is the social context model (Clément, 1980). It has many constructs that are similar to those in Gardner's model, with the exception of attitudes toward the learning situation, but it sometimes conceptualizes and measures them differently. A major feature of this model is that it focuses on the linguistic nature of the community, distinguishing between unicultural and multicultural communities. Moreover, it contains additional constructs such as fear of assimilation, contact with the language (which is influenced in part by the linguistic nature of the community), and self-confidence with the language. Later developments (Clément, Dörnyei, and Noels, 1994) added appraisal of the classroom environment (similar conceptually to Attitudes toward the Learning Situation in the Socio-educational model).
A fourth model was proposed by Dörnyei 1994 as the extended motivational framework. It referred to three levels of motivation—: language level, learner level, and learning situation level,—and involved 17 constructs, some of which involve more than one measure. Many of the constructs associated with the third level are different from those proposed in the preceding models, but those in the first two levels are comparable to those in at least one of the Lambert, Gardner, or Clément models.
A fifth formal model was proposed by MacIntyre, Clément, Dörnyei, and Noels (1998) referred to as a situated model of confidence and affiliation. This model, unlike those referred to above that focus on second language achievement, has a pyramidal structure comprised of six levels, with communication behavior at the top and social and individual context at the bottom. Twelve major constructs form the elements of these levels, but the ultimate criterion is the willingness to communicate in the second language, not second language achievement.
(p. 209) There are other models that could be discussed (see, e.g., Dörnyei, 2001, 2003), but the above five are most representative of the social psychological perspective because each involves the concept of motivation and some form of involvement with the other cultural community. One thing that is obvious from these models is that they involve a series of overlapping constructs, and each time a new model is proposed it tends to be more complex than its predecessor, often encompassing more variables. This tendency is valuable in that it focuses attention on more variables that might be important, but it has drawbacks too in that with greater complexity comes less parsimony, making empirical verification more difficult.
This particular approach was criticized in the 1990s as not being education friendly. For example, Crookes and Schmidt 1991 called for a new approach to the investigation of motivation, and this was augmented by a lively exchange in the 1994/1995 issues of the Modern Language Journal. There have been a number of models and approaches proposed since then (see, e.g., Dörnyei and Otto's  process model of second Language Motivation and Williams and Burden's  social constructivist model), but these are more focused on classroom learning and performance than on the development of bilingual skill. This is not intended as a criticism of these models. This type of approach is important to the area of second language education, to be sure, but the motivation involves performance in the classroom environment. One cannot argue against the importance of motivation in the classroom, but it is an open empirical question whether the motivation that might be promoted by teachers in this context eventuates in the development of bilingual skills as discussed above, or even if it is that unrelated to motivation as defined, for example, in the socio-educational model of second language acquisition.
IV. Research Issues
This has been a very active research field over the years with many different issues under consideration. Some of them are long-standing and were discussed in an earlier edition of this chapter (Gardner, 2002) but are still very much a part of the research literature. Among the long-standing issues, the first concerns some disagreement about which is the more important for second language acquisition: integrative or instrumental motivation. Despite the disagreement, there has been relatively little research devoted to resolving the issue. Instead, most studies concerned with this distinction focus primarily on integrative and instrumental orientations and do not assess any other aspect of motivation. In this context, there have been a number of studies focused on identifying other orientations (see, e.g., Clément and Kruidenier, 1983), generally by factor analyzing a number of different reasons for studying a language. Many of the studies do not relate the orientations to either motivation or achievement.
(p. 210) This distinction between integrative and instrumental motivation is often made in the context of a second issue—the distinction between foreign and second language learning—where it is often hypothesized that the social environment governs which type of motivation is the more dominant. For example, Oxford (1996) proposed that a foreign language is typically learned in an environment in which the language is seldom used or experienced, whereas a second language is learned in a setting in which that language is typically used by the majority of individuals for everyday communication. This distinction has been made by many others, and it has even been suggested that the motivations of individuals in the two different environments would be quite different. Thus, Dörnyei 1990 proposed that instrumental motivation may be more important for foreign language learners than for second language learners, for whom integrative motivation may be the more influential. This is a compelling hypothesis, and it is certainly reasonable to assume that the language context will influence the dynamics of language acquisition (cf. Clément, 1980); however, it is unwise to tie this distinction to the labels foreign and second language acquisition. Instead, it should be associated with an analysis of the ethnolinguistic vitality of the language in the community in terms of the relevant demographics (Giles, Bourhis, and Taylor, 1977). Simply using the labels foreign and second language as indicators of this vitality can be very misleading. For example, French and English are both considered second languages in Canada, but this is simply because they are the official languages. Some parts of Canada are relatively French/English bilingual, but many are not. For example, according to the 1996 Canada census, only 12% of the population in the province of Ontario knows French. Interestingly, this is where much of the research by Gardner has been conducted, and his research is often referred to by others as demonstrating why integrative motivation is important for the learning of a second language. Sometimes the added meaning associated with foreign and second language simply does not apply. Thus, based on these types of numbers, much of Gardner's research would be characterized as taking place in a foreign language environment.
A third issue concerns the notion of causation in the motivation—achievement sequence. That is, does motivation facilitate achievement in the second language or does achievement in the second language promote motivation? This issue has been raised by a number of researchers (see, e.g., Ellis, 1994), and it seems likely that both causative directions are applicable. Wherever individual differences are concerned, however, the question of directionality can never be answered unequivocally, regardless of the context. The only way to establish causation in a scientific sense is to randomly assign individuals to conditions, and this is not possible when the focus is on individual differences (see, e.g., Gardner, 2000). One can use structural equation modeling to test the validity of a particular “causal” model, but it is not proof of the direction of causation—rather, it is a test that the data are consistent with that model. One can test any number of models and find that they are relatively consistent with the data, but to compare one with another requires that one model be nested in the other, and this is not feasible in contrasting different directions of causation. This does not mean that the study of the correlation between these two (p. 211) classes of variables is meaningless. From the perspective of Bayesian statistics, a positive correlation between two variables indicates that given a particular score on one of the variables (say, motivation), there is a given probability that one's score on the other variable (language achievement) will be greater than some given score and vice versa. That is, there is enhanced value in knowing the nature of the correlation. To the extent that a correlation has been demonstrated in a sample of individuals, it is reasonable to expect that it exists in the population for which the sample can be considered representative. This correlation has implications in that it links the two underlying constructs. The question of which “causes” which, or if anything does at all, is indeterminate and immaterial.
A fourth issue concerns the role of the classroom in motivation. This was initiated in part by Crookes and Schmidt 1991, who proposed that more attention should be directed to the motivation in the classroom, a stated motivation that was viewed as less stable and more susceptible to environmental characteristics (cf. Boekaerts, 1986). Those models that focus attention on motivation in the classroom (see, e.g., Dörnyei and Csizèr, 1998; Dörnyei and Otto, 1998; Williams and Burden, 1997) are concerned with the more situationally relevant elements of motivation. It is true that they might endure for the length of the course, and perhaps beyond, but the focus in the models and associated research is the effect of the classroom environment. In this respect, they do not differ that much from the research in educational psychology that applies to any school subject, and generally the findings that have been reported are comparable to those obtained with other school subjects. It is meaningful to ask, however, whether this classroom-based motivation is that different from and unrelated to the more general form of motivation characterized in a social psychological perspective. One possible hypothesis is that there is a functional relationship between the two and that there is in fact a correlation between levels of motivation in the classroom and motivation socially defined. This hypothesis was investigated recently by Bernaus, Moore, and Azevedo (2007), who adapted measures from the socio-educational model (Gardner, 1985, 2007) and the process model (Dörnyei and Otto, 1998). They found that executive motivational influences from the process model loaded on an integrative motive factor, that motivational influences on intention formation and motivational influences on intention enactment shared variance in common with Gardner's measure of interest in foreign languages, and that postactional evaluation contributed to an attitude toward the learning situation factor defined by two measures from the socio-educational model.
A fifth issue concerns the concept of integrative motivation. Is the integrative motive to be conceived in the same way as various social motives, or is it different? Some researchers appear to consider it as a trait-like characteristic of the individual that is relatively stable and long lasting (like many social motives), whereas Gardner 2005 has argued that integrative motivation is an inference made on the basis of a number of characteristics of the individual. That is, individuals can be said to be integratively motivated if they have an open willingness to take on (often linguistic) features of another community, are motivated to learn their language, and view the (p. 212) language learning situation positively (see also Gardner, 1985). As a consequence, it is reasonable to believe that features of integrative motivation can change over time. This was even demonstrated by Gardner 1985 in which significant decreases in attitudes toward the learning situation, motivation, and integrativeness (the three defining characteristics of integrative motivation in the socio-educational model of second language acquisition) were shown to take place within individuals in many settings from one year to the next. Furthermore, there was little evidence in those results that the decreases were moderated by the level of achievement (high versus low defined by a median split in grades obtained in the first year).
More recent research has considered this further. Gardner, Masgoret, Tennant, and Mihic (2004) investigated a sample of university students studying French as a second language. The students were tested with the AMTB at the beginning and end of the academic year, and the results compared for students obtaining A, B, and less than B grades at the end of the year. The results demonstrated that students who obtained A grades at the end of the year had relatively high scores on the measures of attitudes and motivation at the beginning of the year and showed very little change over time, whereas those obtaining B grades started the year with lower levels of attitudes and motivation and evidenced slight decreases, and those obtaining less than B grades began with lower levels of attitudes and motivation and showed much greater decreases. Gardner 2005 reports similar results obtained in a study conducted in Spain with secondary students learning English as a foreign language.
This leads to a sixth and final issue, namely, the role of the teacher in influencing the level of motivation in the classroom. This topic has been considered from a number of different perspectives in recent years revealing some inconsistencies in the perceptions of teachers and students and in the generalizations that might be drawn. Three studies demonstrate this.
Madrid 2002 considered the relationships between the perceptions of 18 teachers and their students regarding the power of 18 strategies to motivate students as well as the students' self-perceptions of their motivation. He found that some strategies, such as the use of audiovisual aids, were seen as motivational by many of the students whereas others, such as individual work, were not. Furthermore, there were few significant correlations between the students' and teachers' perceptions of the power of the different strategies. The highest positive correlation was for the use of audiovisual aids, whereas the highest negative correlation was for satisfying students' needs and interests. A 2 4 (Gender Grade Level) analysis of variance for each strategy revealed only 1 effect due to gender but 13 for grade level differences, indicating primarily grade level differences in the perceived power of some strategies. In addition, it was found that global motivation level varied across grade levels.
In another study, Guilloteaux and Dörnyei 2008 investigated 25 strategies in a sample of 40 classrooms in English as a foreign language. However, rather than having teachers and students rate the use of specific strategies, the behavior of both the teachers and the students with respect to these strategies were observed in a time sampling scheme in terms of a set of categories. In addition, a teacher evaluation scale was administered immediately after each class, and students completed three (p. 213) Likert scales: attitudes toward the course, linguistic self-confidence, and classroom anxiety. The classroom observations produced two aggregate variables, teacher's motivational practice and learners' motivated behavior, whereas the aggregate of the Likert scales was used as a measure of motivation. The results demonstrated significant correlations among the three measures, with the highest correlation between the two classroom observation measures, and lower and comparable correlations with the Likert measure of motivation.
Bernaus and Gardner (2008) investigated the effects of teachers' and students' perceptions of the use of 12 innovative and 14 traditional teaching strategies on students' attitudes, motivation, and English achievement in 31 classes of English as a foreign language. The results demonstrated significant correlations between teachers' and students' perceptions for 10 of the traditional strategies and 6 of the innovative ones. Separate aggregate scores for traditional and innovative strategies were computed based on teachers' perceptions, class perceptions, and individual student perceptions, and these were correlated with English achievement and the six composite scores on the mini-AMTB. None of the correlations based on teacher perceptions were significant, but when class perceptions were investigated, significant correlations were obtained between traditional strategy use and integrativeness, attitudes toward the learning situation, motivation, instrumental orientation, and parental encouragement, and between innovative strategy use and attitudes toward the learning situation and motivation. When the student was the unit of analysis, significant positive correlations were obtained with all of the composite measures except language anxiety with both the traditional and innovative strategies. The only significant correlation with the measure of English achievement was with traditional strategy use and this correlation was negative. That is, students who rated the use of traditional strategies as being used frequently obtained lower English scores.
VI. Future Directions
In the previous edition of this chapter (Gardner, 2002) we focused on five issues that seemed to be important in future considerations:
• Employing meta analysis to investigate the consistency of findings
• The use of laboratory studies as analogues to test hypotheses about process
• The investigation of immigrant populations learning a second language
• The study of both societal and educational environmental factors in second language learning
• The validity of the various models
To some extent, some of these issues have been dealt with in the interim, and I have been able to integrate them into the preceding paragraphs, and some have not been considered to any great extent.
(p. 214) In this edition, I will discuss four issues that appear necessary to strengthen the scientific foundation for the investigation of factors that influence second language acquisition. Although I will elaborate on them using research from the socio-educational model of second language acquisition, all of the points apply equally to the other models.
As proposed in the previous edition (Gardner, 2002), there is considerable need for further research to assess the validity of the various models that have been, and are continuing to be, developed and especially to clearly articulate the differences between them. It is often the case that the various models use somewhat different concepts and measures but do not make predictions that are in disagreement, so to some extent the results of many studies can be seen to apply to all the models. It is necessary, however, to obtain more information making use of measures that are specific to the various models. It is unlikely that discrepant results will be obtained, but clarifying the role of these other variables would be an important advance. To elaborate, a study by Tremblay and Gardner 1995 included measures of attributions and goal salience along with many of the measures generally used in the context of the socio-educational model. The results demonstrated how these variables could be incorporated into the model, but there was no indication that they appreciably improved prediction of measures of achievement.
This gives rise to the second issue, namely, the need to clarify concepts in this research area and eliminate the use of multiple definitions for a given construct. For example, Gardner 1985 operationally defined integrativeness as the sum of scores on three scales from the AMTB—in other words, attitudes toward French Canadians, ratings of an integrative orientation—and interest in foreign languages and concluded that it “thus reflects attitudinal reactions toward the cultural aspects of language learning” (p. 93). The operational definition has been employed in the bulk of his research since, though it is the case that recently he has chosen to define integrativeness more generally. For example, Gardner (2006) states “individuals who are high in integrativenss do not focus on their own identity but instead are willing and able to take on features of another language group (if only just the language) as part of their own behavioral repertoire” (p. 247). More recently, Gardner has referred to integrativeness as cultural openness, pointing out that concepts such as Kraemer's (1993) social/political attitudes and Yashima's (2002) inercultural posture are comparable to integrativeness because they also focus on the intercultural basis of language acquisition. Cultural openness incorporates all three conceptions. Other researchers, however, have defined the concept of integrativeness differently, varying all the way from attitudes toward the language group to perceptions of the ideal self as a language learner. Of course, researchers can define concepts in ways that are most meaningful to themselves and their research, but using the same labels for clearly different operational definitions can cause confusion in communication. It would seem that, at a minimum, various operational definitions of the same construct should yield measures that are correlated with one another.
The third issue that I believe is important to the advancement of research in the area of second language acquisition is the proper use of data analytic procedures to answer specific questions. Thus, the correlation coefficient is an indicator of the (p. 215) relationship between two variables, but a regression coefficient in multiple correlation is not. It is an indicator of the extent to which a variable adds to prediction given the other variables in the equation (i.e., increases the correlation of the composite with the criterion variable). Structural equation modeling is a powerful analytic procedure that permits one to determine how well a predefined model fits the data, but it does not “prove” the model. There are an infinite number of models that might fit the data as well or better for the particular data set. Finally, there is a relatively new data analytic procedure, hierarchical linear modeling (Raudenbush and Bryk, 2002), which could be very instructive in studies of individual differences when the data are derived from students in different classes. It is comparable to multiple regression except that in addition to determining the coefficients in the model, it also permits an assessment of the variability of these coefficients in the different classes as well as the relationship between these coefficients and characteristics of the classes themselves (see, e.g., Bernaus and Gardner, 2008). Thus, it is a procedure that allows for a data analytic linking of individual difference variables and classroom characteristics, and represents an exciting new frontier.
The fourth, and I believe the most important, issue is the need for improved measurement in our research. As noted previously, there are a large number of constructs that have been hypothesized as important for learning a second language. Moreover, the majority of these variables are measured through the use of verbal report, often in the form of questionnaires. Too often these questionnaires are made up of items written for a particular study, and a common procedure in this area is to factor analyze the correlations among the items to identify common clusters of items. These are then aggregated to produce scales, and often these are correlated with other variables to assess their validity. This is useful in the initial stages of research, but if new scales are continually developed in this way without cross-validation, one must be cautious of interpretation. Scales formed in this way will evidence artificially high estimates of internal consistency reliability because of sampling error that contributes to the factor structure, and these sampling errors can artificially inflate validity coefficients. At a minimum, such results should be cross-validated with an independent sample to determine whether the scales maintain their internal consistency reliability as well as their discriminant and convergent validity with other constructs in the area of research. Only when measurement improves will advances be made in this research area.
The value of employing reliable and valid measurement can be illustrated with an example involving measures associated with the socio-educational model of second language acquisition. Earlier it was noted that some researchers have claimed that the research associated with the socio-educational model of second language acquisition was applicable to Canada because it was a bilingual country and that different results would be obtained in different settings, particularly if the language being learned was a foreign as opposed to a second language. Studies purporting to support this conjecture generally made use of measures that were not the same as those in the AMTB. When such research is conducted using the AMTB, the results do not support this conjecture. Gardner (2006) presents results obtained with eight samples involving (p. 216) students in two different grade levels studying English as a foreign language in four European countries (Croatia, Poland, Romania, and Spain). The results demonstrated that the internal consistency reliability coefficients of the 12 scales from the AMTB were consistent across all eight samples and similar to those obtained in Canadian samples. Furthermore, factor analyses of the 12 scales in each country revealed structures that were very similar to each other and comparable to those obtained in Canada. A table of 54 correlations assessing factor congruence in the eight samples revealed that all but 4 of the correlations were.90 or higher. Furthermore, when the predictive validity of final grades in English by the six constructs was investigated, it was found that significant correlations were obtained in all eight samples for the measures of integrativeness, motivation, and language anxiety, and that these results were shown to be comparable to those obtained in Canada. Thus, when uniform measures are employed, there is considerable stability in the results obtained.
Attention to these four issues will, I believe, be beneficial. As can be seen, this is an active area of research with great potential. The future looks bright, but continued research focusing on empirical findings to support or clarify theoretical models and hypotheses represents its greatest strength.