November 2006. Volume 1 Issue 3
Causal Modeling – Path Analysis:
A New Trend in Research in Applied Linguistics
University of Kerman, Iran
Dr. Mina Ratsegar has been an Assistant Professor of Language and Applied Linguistics in the language department of the University of Kerman for the last 22 years. Her professional expertise lies in the area of Psycholinguistics. Her research focuses on L2 learner factors – affective, cognitive, and personality. She is currently teaching research methods, methodology, testing, and advanced writing at both B.A and M.A levels.
Article: PDF version
This article has the aim of discussing a new statistical trend in research in applied linguistics. This rather new statistical procedure is causal modeling – path analysis. The article demonstrates that causal modeling – path analysis is the best statistical option to use when the effects of a multitude of L2 learners’ variables on language achievement are investigated in one study. The proposed causal models, which are the property of causal modeling, provide a plausible explanation for the hypothesized relationships among the variables under inquiry. The causal modeling procedure and the statistical test of path analysis will reasonably manifest the causal relationship that naturally exists between the L2 learners’ variables and are otherwise overlooked with simple correlational procedures. The paper presents a historical background on the statistical procedures involved in research about L2 learners’ variables in the field of ELT. The traditional trend of linear correlation between the variables of concern can be reasonably reconsidered and replaced with the new causal modeling – path analysis procedure. The paper also presents some rudimentary information about path analysis.
In the past quarter of the century, research in the field of SLA has grown enormously, with the quantity of published research increasing annually. It is striking, however, that the main thrust of research has been towards establishing how language learners are similar and how processes of language learning are universal. That is, traditionally, the majority of the research in SLA and applied linguistics looked for phenomenon that would presumably affect all the individual language learners. In studies concerned with SLA, researchers have tried to identify universal sequences in development or common processes, such as transfer, cross-linguistic interference, and so forth that would affect everyone in the same way.
In the field of psychology two contrasting approaches to the study of human functioning have long been recognized – the experimental and the differential approaches. The former focuses on identifying structures and processes common to everyone – similarities between individuals. In contrast, the latter approach emphasizes differences between people, seeking to identify the most relevant major ways that people vary.
In the field of applied linguistics, with the researcher’s awareness of the potential impact of learners’ differences on L2 learning, an era of research with focus on the L2 learners’ variables was marked in the 1970s. Concern about the learner variables resulted in an increasing number of studies that accounted for the learner’s differences from different perspectives in ESL/EFL contexts. Learners’ differences can generally be divided into three categories: personality, cognitive, and affective.
A basic question concerned has been why all individuals with normal faculties successfully acquire their first language but meet with different degrees of success when they attempt to master a second language. The answer to this fundamental question, as the literature shows, concerns the individual L2 learner and lies in his/her personality, cognitive, and affective construct.
Focus on the language learners with a specific emphasis on the individual differences has brought about the most optimal consequences in language pedagogy. The “post method” condition (Kumaravadivelu, 1994), a new era in which the concept of “methodology” is revolutionized, is a turning point in the history of ELT. The basic premise of this new wave of interest is concerned with accounting for differences between the L2 learners (see Brown, 1993; Kumaravadivelu, 1994; Oxford, 1993). The new “post method” era in the field of ELT, nevertheless, requires a great deal of quality research about a great number of L2 learner’s variables in different L2 contexts.
The foundation of language learner variable research is that it examines attributes on which language learners vary and how such variations relate to L2 learning success (Skehan, 1991). According to Skehan four consequences follow from this fundamental approach. First, it encourages quantification of the strength of relationships between any particular learner variable (e.g. attitude) and language achievement. Second, by examining the range of influences on SLA, interesting points of contact between different single learner variables may be revealed. Third, the advantage of having a learner’s variations perspective on research is that it encourages the development of more formal models that relate learners’ variables to one another and to language acquisition. Fourth, presenting a learner variable framework manifests the multi-causal nature of language learning and also its complexity.
Linguists, psycholinguists, ELT researchers, and L2 teachers generally hold the assumption that there is a relationship between the personal traits of a language learner and his/her success in the task of L2 learning. Brown (1994) believes that inquiry about the L2 learner’s variables reveals some facts that yield insights about L2 learning.
To probe into the assumption of the interrelationship between learner variables and his/her success in language learning, it would be fair to hypothesize that there are some relationships, preferably causal, between a number of different variables and L2 achievements. This hypothesis has, in fact, been the focus of much research in the area of ELT. In the following section the literature on learner’s variables and L2 learning, considering the statistical procedures involved in these studies, will be reviewed.
Background on learner variables research
Literature on learner variables and L2 learning may be divided into two broad categories i.e. the simple correlational studies (bivariate and multivariate) and the more complex correlational studies called “causal modeling”. The former is focused directly on the linear relationship between some learner’s variables and L2 achievement, and the latter tests the causal relationship among some learner’s factors and L2 learning.
While, a multitude number of the studies on learner variables seem to belong to the first category, research using causal modeling to investigate the causal relationship among many learners’ factors and L2 learning seem more promising.
Majority of the studies from the first research trend are normally focused on just few variables from a domain (cognitive, affective, or personality). Gardner, Trembly, and Masgoret (1997) contend that there is a lack of research examining the relationships among cognitive, affective, and personality variables simultaneously. Similarly, Onwuegbuzie, Baily, and Daley (2000) state that only a few studies have examined the role of cognitive, affective, personality, and demographic variables concurrently. According to Onwuegbuzie et al. (2000), two studies that investigated the relationship between several classes of variables and L2 achievement are Gardner, et al. (1997) and Ehrman and Oxford (1995).
The other line of research, causal modeling, shares the characteristic of being based on causal models. Such models tend to be fairly elaborate and consider the simultaneous influence of several learners’ variables on L2 learning. Among the studies using causal models the following can be listed (Clement, 1987; Clement and Kruidnier, 1985; Ely, 1986; Fouly 1985; Gardner, 1985; Gardner and MacIntyre, 1992, 1993 Lalonde, and Pierson, 1983; Lalonde and Gardner, 1984; MacIntyre and Charos, 1996; Rastegar, 2003; Wang, 1988).
Attitudes, motivation, and anxiety are the most popular constructs included in the causal models. Gardener and his associates have, over several years, used the Attitude / Motivation Test Battery (AMTB) in studies leading to the development of the socio-educational model of language acquisition (Gardner, 1985). MacIntyre and Charos (1996) contend that while portions of the model have been, and will continue to be, updated to incorporate new research results, the basic model has consistently been replicated. For a review, see (Ehrman and Oxford, 1995; Gardner, 1985; Gardner, Trembly, and Masgoret, 1997; Rastegar, 2003).
Rastegar (2003), in extensive research on a number of language learners’ variables using causal modeling and path analysis procedures, has established a framework for the study of learners’ variable in EFL contexts. In her study, besides ‘attitudes’, ‘motivation’, and ‘FL anxiety’, variables such as ‘language ego’, ‘self-esteem’, ‘locus of control’, ‘IQ’ and ‘sholastic ability’ have also been considered in causal models using path analysis as the statistical test.
Literature on learner’s differences and L2 learning shows that studies are mostly conducted in ESL contexts. Moreover, the results of the studies are not all conclusively consistent. This reveals the fact that the field of applied linguistic benefits from more research, particularly that which utilizes causal models, to provide a more reliable and comprehensive picture of both EFL/ESL learning in different cultures.
Background on path analysis
Path analysis – a method for studying patterns of causation among set of variables – was developed by Sewall Wright. Path analysis is an important statistical tool to gain a deeper understanding of the relationship among variables. It is a method for studying the direct and indirect effects of variables hypothesized as causes of variables treated as effects. Path analysis, as Pedhauzer (1982) put it, is not a method for discovering causes, but a method applied to causal models formulated by the researcher on the basis of knowledge and theoretical considerations. Pedhauzer, then, quotes (Wright, 1934) as saying that
… the method of path coefficients is not intended to accomplish the impossible task of deducing causal relations from the values of the correlation coefficients. It is intended to combine the quantitative information given by the correlations with such qualitative information as may be at hand on causal relations to give a quantitative interpretation. (Pedhauzer, 1982, p. 580)
Path analysis is closely related to multiple regression. Regression may be considered a special case or simplest form of path analysis. Path analysis and related techniques are also called “causal modeling”. The reason for this name is that the techniques allow the testing of theoretical propositions about cause and effect without manipulating variables. However, the causal in “causal modeling” refers to an assumption of the model rather than a property of the output or consequence of the technique. That is, it is assumedsome variables are causally related, and test propositions about them using the techniques (Bryman and Cramer, 1997; Pedhazure, 1982).
This versatile statistical tool is specifically ideal when language learners’ variables are of concern in research. Language learners’ variables by their very nature interact with each other and have a direct and indirect effect on each other in the complicated process of L2 learning. This means that if causal modeling and path analysis is used in research on learners’ variables, a more realistic relationship between variables will be manifested and more reliable results will be attained (see Rastegar, 2003).
Path diagrams and jargon
There are customs about displays and names in path analysis. Arrows show assumed causal relations. A single-headed arrow points from cause to effect. A double-headed, curved arrow indicates that variables are merely correlated; no causal relations are assumed. The independent (X) variables are called exogenous variables. The dependent (Y) variables are called endogenous variables.
A path coefficient indicates the direct effect of a variable assumed to be a cause on another variable assumed to be an effect. Path coefficients are standardized and consequently they can be compared directly. Path coefficients may be written with two subscripts. The path from 1 to 2 is written p21 – the path to 2 from 1; note that the effect is listed first. A path analysis in which the causal flow is unidirectional (no loops or reciprocal causes) is called recursive (Bryman and Cramer, 1997; Pedhazure, 1982).
One of the specific characteristics of path analysis is that except for the constant dependent variable, the status of other variables of the study will vary between independent and intervening, depending on the structure of the model and the specific calculation related to a particular equation. In statistical analysis intervening variables – those that arrows are pointing to – are considered as the dependent variable for that particular model. For the typical path analysis there are some assumptions which are as follows.
1. All relations are linear, additive, and causal. The causal assumptions (what causes what) are shown in the path diagrams.
2. The residuals (error terms) are uncorrelated with the variables in the model and with each other.
3. The causal flow is one-way. That is, reciprocal causation between variables is ruled out.
4. The variables are measured on interval scales.
The advent of constructivist theory of learning (1980s & 1990s & early 2000) which emphasizes Roger’s humanistic views seems to have brought a new dimension – a shift in attention to the language learners. As Brown (2000, p. 156) asserts “… we have now returned to a recognition of the acute importance of individual variation, especially in the realm of education.” Generally, the research trends in SLA are harmonious with the trends in the school of thought in SLA. With awareness in the research community that concentrating on learners’ variables would be more informative, there was a major change towards considering L2 learners’ differences instead of their similarities in research. This shift in SLA research trends seems to have been conceptualized from 1980s and has continued to present.
In this regard, there appears to be consensus among ELT researchers and applied linguists that research on the language learners’ variables done in the traditional manner of considering only few variables and looking for just a linear relationship between these variables and L2 achievement will never provide comprehensive, informative, and valuable information about the nature and complexity of these relationships. Consequently, this kind of research would not, in effect, make an appropriate contribution to the field of SLA and particularly to TESL/TEFL in finding remedies for the ever-lasting problem of lack of optimal success among some L2 learners.
The research trend of causal modeling and the tendency of the researchers to consider a number of learners’ variables simultaneously in one study indicate that this line of research is well under way to making a significant contribution to the field of ELT. Future research on L2 learners’ differences should give serious attention to the process of model construction involving more L2 learners’ variables from different variable domains.
The results of Rastegar’s (2003) study clearly demonstrate that causal modeling procedures can serve some essential purposes such as variable identification, model construction, and framework design. Variable identification provides a more comprehensive understanding of L2 proficiency by identifying the best predictors of success in L2 learning and this will contribute to model construction. Model construction, in turn, contributes to framework design or framework modification. In fact, detecting predictors of success in FL learning may not be considered the ultimate aim of the research in this area, rather it is a bridge that fills the gap between variable identification and framework construction. All these steps and procedures are interrelated and seem essential for research on learners’ variables and L2 success. This important task can only be done by the use of the most appropriate statistical means, that is, causal modelling – path analysis.
Bryman, A. & Cramer, D. (1997). Quantitative data analysis with SPSS for Windows. London and New York: Routledge.
Brown, H. D. (1993). Requiem for methods. Journal of Intensive English Studies, 7, 1-13
Brown, H. D. (1994). Principles of language learning and teaching. (3rd ed.). Englewood Cliffs. N.J: Prentice Hall.
Brown, H. D. (2000). Principles of language learning and teaching. (4th ed.). Englewood Cliffs, N.J: Prentice Hall.
Clement, R. (1987). Second language proficiency and acculturation: An investigation of effects of language status and individual characteristics. Journal of Language and
Social Psychology, 5, 271- 290.
Clement, R. & Kruidenier, B. G. (1985). Aptitude, attitude and motivation in second language proficiency: A test of Clement’s model. Journal of language and Social Psychology, 4, 21-37.
Ehrman, M. E. & Oxford, R.L. (1995). Cognition plus: Correlates of language learning success The Modern Language Journal, 79, 67-89.
Ely, C. M. (1986). An analysis of discomfort, risk-taking, sociability, and motivation in the L2 classroom. Language Learning, 36, 1-25.
Fouly, K. (1985). A confirmatory multivariate study of the nature of second language proficiency and its relationships to learner variables. Unpublished Ph.D. dissertation, University of Illinois, Urbana.
Gardner, R. C. (1985). Social psychology and second language learning: The role of attitudes and motivation. London: Edward Arnold.
Gardner, R.C., & Lalonde, R. N. & Pierson, R. (1983). The socio-educational model of second language acquisition: An investigation using LISREL causal modeling. Journal of Language and Social Psychology, 2, 51-65.
Gardner, R. C., & MacIntyre, P. D. (1992). A student’s contributions to second language learning: Part I. Cognitive variables. Language Teaching, 25, 211- 220.
Gardner, R. C., & MacIntyre, P.D. (1993a). On the measurement of affective variables in second language learning. Language Learning, 43, 157-194.
Gardner, R. C., Tremblay, P. F., & Masgoret, A. M. (1997). Toward a full model of second language learning: An empirical investigation. Modern Language Journal, 81, 344-362.
Kumaravadivelu, B. (1994). The post method condition: (E)merging strategies for second/foreign language teaching. TESOL Quarterly, 28, 27-49.
Lalonde, R. N., & Gardner, R. C. (1984). Investigating a casual model of second language acquisition: Where does personality fit? Canadian Journal of Behavioral Science, 16, 224-237.
Lalonde, R. N., & Pierson, R. (1983). The socio-educational model of second language acquisition: An investigation using LISREL causal modeling. Journal of Language and Social Psychology, 2, 1-15.
MacIntyre, P.D., & Charos, C. (1996). Personality, attitude, and affect as predictors of second language communication. Journal of Language and Social Psychology, 15, 3-18.
Onwuegbuzie, A. J., Baily, P., & Daley, C.E. (2000). Cognitive, affective, personality, and demographic predictors of foreign-language achievement. Journal of Educational Research, 94, 3-16.
Oxford, R. L. (1993). Individual differences among your students: Why a single method can’t work? Journal of Intensive English Studies, 7, 27-43.
Pedhazure, E. J. (1982). Multiple regression in behavioral research: Explanation and prediction (2nd ed.). CBS College Publishing New York, N.Y.
Rastegar, M. (2003). Affective, cognitive, and personality predictors of foreign-language proficiency among Iranian EFL learners. Ph.D. Dissertation. Shiraz, Iran.
Skehan, P. (1991). Individual differences in second language learning. Studies in Second Language Learning, 13, 275-298.
Wang, L. S. (1988). A comparative analysis of cognitive achievement and psychological orientation among language minority groups: a LISREL approach. Ph.D. Dissertation, University of Illinois, Urbana.