––––––––
805
––––––––

Weber State College
University of Southern California
In a study of 27 collegelevel students of Spanish, Stokes (1988) reported a positive correlation between length of residence in a Spanish speaking country and performance on a communicative test of present subjunctive use (r = .51, p<.01), and a negative but nonsignificant correlation between amount of formal study and subjunctive proficiency (r = .12, n. s.). (Average length of residence was about one year, average length of study was about two years; for full details and raw data, see Stokes, 1988.)
Stokes administered a similar test of subjunctive use after «a weeklong lesson on the use of the subjunctive» (706), and again found a strong positive correlation with length of residence in a Spanishspeaking county and test performance (r = .73, p<001). The correlation between amount of formal study and subjunctive test performance was again negative, and, this time, was at the fringes of statistical significance (r = .36, p<.10, two tails).
Stokes concluded that «amount of formal study is largely irrelevant to the actual acquisition of the subjunctive in oral expression» (707), while informal language use does contribute to the acquisition of the subjunctive.
In this study, we subject the same data to multiple regression analysis. Multiple regression is appropriate here since foreign residence and amount of formal study are themselves negatively correlated (r = .51, p< .01). Multiple regression «removes» the effect of this correlation, and allows us to determine the independent impact of each predictor (foreign residence, amount of formal study).
Multiple regression analysis of Stokes's «pretest» data is presented in Table I.
The regression coefficient for instruction in Table I predicts a gain of .0295 points on the test of subjunctive use for each quarter
regression  
predictor  coefficient  beta  tratio 
intercept  8.07  
formal study  .0295  .188  .94 ns 
foreign residence  .141  .605  3.00 p<.01 
r squared = .28 
hour of college study of Spanish. (Stokes gave subjects one point for each quarter hour of college study, eight points for each year in high school, and ten points for eight weeks of intensive missionary training). The regression coefficient for foreign residence predicts a gain of .141 on the subjunctive test for each month of foreign residence. According to the multiple regression analysis, the relationship between study and subjunctive test scores is now positive, even though the simple correlation between study and test scores was slightly negative (r = .12). However, only foreign residence is a significant predictor of performance on the subjunctive test, in agreement with Stokes's original analysis.
Betas, standardized regression coefficients, were computed to allow a comparison between the effects of the two predictors. The beta for foreign residence is much larger than the beta for formal study.
Table II presents a multiple regression analysis of «posttest» performance, that is, performance on the subjunctive test administered after one week of instruction on the subjunctive.
regression  
predictor  coefficient  beta  tratio 
intercept  8.533  
formal study  .0021  .01  .06 ns 
foreign residence  .219  .734  4.52 p<.001 
r squared = .53 
According to this analysis, residence is, once again, a
significant predictor. Instead of a negative relationship between formal study
––––––––
806
––––––––

Tables I and II also present the coefficients of determination (r squared), which indicate how well the overall repression equations account for the variation in the dependent variable. Note that adding formal study adds only .02 to the coefficient of determination in Table I (squaring the original correlation between formal study and subjunctive proficiency, .51, gives an r squared of .26) and adding formal instruction adds nothing to the coefficient of determination in Table II, confirming that formal study makes little or no contribution.
Our analysis has several limitations. First, it is likely that other predictors, such as affective variables and amount of reading, should be included. Our multiple regressions, in other words, probably suffer from «specification error». Second, our measures may be imperfect. One month of instruction in a comprehensible inputrich class may be worth more than a month of instruction in a grammarbased class. Both specification error and measurement error can affect the size of the regression coefficients, their significance, and the size of the coefficient of determination (Berry and Feldman, 1985). It should also be noted that the lowest instruction score in the data was 10, equivalent to just under one year of college language study. If instruction has the effect of bringing students to the point where informal input is at least somewhat comprehensible, the equations may not hold for lower values of instruction; the effect of length of residence may be much less when students have had less classroom background^{124}.
Nevertheless, our reanalysis confirms Stokes's original analysis, and gives us more confidence in his conclusions. When the effect of instruction is held constant, residence in a Spanishspeaking country clearly relates to proficiency on a test of subjunctive use, and when residence is held constant, amount of instruction does not relate to subjunctive proficiency.
Although formal study did not contribute to performance on the subjunctive test, this does not mean that classroom instruction has no value. Classroom instruction can help beginners become intermediates, bringing them to the point where they can improve on their own in the informal environment (Krashen, 1982). Stokes's original study and our reanalysis suggest, however, that classroom instruction does not bring students far enough along to acquire the subjunctive. Stokes's study and this reanalysis thus confirm the results of Terrell^{125}, Baycroft and Perrone (1987): the subjunctive is lateacquired^{126}.
Berry, William D. and Stanley Feldman. Multiple Regression in Practice. Beverly Hills; Sage Publications, 1985.
Krashen, Stephen D. Principles and Practice in Second Language Acquisition. New York: PrenticeHall, 1982.
Stokes, Jeffery D. «Some Factors in the Acquisition of the Present Subjunctive in Spanish». Hispania 71 (1988): 70510.
Terrell, Tracy D., Bernard Baycroft, and Charles Perrone. «The Subjunctive in Spanish Interlanguage: Accuracy and Comprehensibility». Eds. Bill VanPatten, Trisha Dvorak, and James Lee. Foreign Language Learning. A Research Perspective. New York: Newbury House Publishers, 1987: 1932.