抄録
This paper presents 1) what limitations causal analyses have, 2) how causal analyses are conducted in English language education research in Japan, 3) what problems are seen in those causal analyses, then, 4) how the problems can be improved for further research. A causal analysis, especially an analysis according to Multiple Regression Model, is originally a powerful tool for predicting a dependent variable by some independent variables. However, when the degree of causal effect by each independent variable is focused, the problem of multi-collinearity, which is provided by correlations among dependent variables, arises. On the other hand, when stepwise method is adopted in deciding which dependent variables should be included, the problem of multi-collinearity may cause again by deleting the dependent variables which reasonably seem to contribute to independent variables. After reviewing those limitations of Multiple Regression Models, eleven articles in English language education research in Japan were reinvestigated in terms of those problems. Then, some suggestions, such as using a correlation analysis, are presented instead of regression models.
