Need we sacrifice accuracy for fluency?

System, Vol. 25, No. 2, pp. 197 201. 1997 Pergamon PII: S0346-251 X(97)00008-0 ~L 1997 Elsevier Science Ltd All rights reserved. Printed in Great Britain 0346-251 X/97 $17.00 + 0.00 NEED WE SACRIFICE ACCURACY FOR FLUENCY? M A R I A N N E NIKOLOV* and S T E P H E N K R A S H E N + *English Department, Janus Pannonius University, Pecs, Hungary *School of Education, University of Southern California, Los Angeles, California 90089-0031 Children who studied English as a foreign language in Hungary with a communication/content-based approach were compared with similar children who studied English with a form-based traditional approach. The former were slightly more accurate in their production of grammatical morphemes in an oral interview, and were more fluent, confirming that communication-based approaches do not sacrifice accuracy for fluency. ~3 1997 Elsevier Science Ltd It has been widely assumed that methods that emphasize subconscious language acquisition sacrifice accuracy for fluency. Students in these methods, it is argued, may be able to speak and understand the target language, but their output is imperfect; without grammar instruction, it is feared, their errors will become permanent. Studies with adults, however, consistently show that students in comprehensible input-based methods are as accurate as students in grammatically-based methods, and are sometimes more accurate. In addition, students in comprehensible input-based methods are always superior in tests involving communication (studies reviewed in Krashen, 1994). Here we examine whether children who have participated in a communication-based E F L program are as accurate as students in a traditional program. PROCEDURE Subjects: two groups of children served as subjects, both in Pecs, Hungary. The experimental group consisted of 15 children who received seven years of EFL classes based on a story-based syllabus that was negotiated with the children, with an emphasis on stories, tasks related to stories, songs, and games. The focus at all times was on content and not on grammatical form. These children were tested once at the end of grade 7 (age 13) and again at the end of grade 8. There was some focus on form in the year between the two tests, but this was limited to brief explanations when students asked for them, or when the teacher considered them useful, as well as some emphasis on self-correction (e.g. rewriting papers). There was no formal grammar instruction and technical terminology was not emphasized. 197 198 . M A R I A N N E N I K O L O V and STEPHEN K R A S H E N The comparison group, termed the "traditional" group consisted of 14 children who had followed a structural syllabus with a focus on form, explicit rules, and drills and exercises. They were tested in grade 8, and had the same amount of exposure to English in the classroom as the experimental group did. As noted earlier, both groups studied English in a nearly pure EFL situation, with very few opportunities to use English outside the classroom. One of the investigators (M.N.) taught the experimental class. Each class had the same teacher for their entire program. Data collection: all subjects were told that the teachers were interested in tape-recording their speech; the children were asked if there were any of them who did not want to participate. All the children volunteered to participate. Interviews were undertaken by one of the investigators (M.N.), who was well-known to both groups of children. In the interview, the children were asked to talk about themselves, to describe one day of one person they knew, and to talk about a book they had read, or a film they had seen. If the investigator felt that the child had not supplied enough examples of the past tense, the child was then asked to talk about his or her weekend. They were shown two pictures with animal characters and were asked to describe them. Most of the children said they liked the interviews and all were praised afterwards. Children were not graded on their performance, and appeared to be relaxed during the interviews, possibly because they were interviewed during time they would normally be in other classes. The interviews lasted about 10 to 15 minutes. The following grammatical structures were investigated: 1. 2. 3. 4. 5. 6. 7. 8. 9. pronoun case: (e.g. he/him; they/them; she/her) plural (short plural only, as in dogs) singular copula (as in "she is a tiger") progressive -ing (as in "he is smoking a cigarette") auxiliary (as in "she is dancing", "they are singing".) article ("a" and "the" combined) past regular (both t and d allomorphs) past irregular (e.g. "went") third person -s ("he sits") Accuracy scores were based on percentage correct in obligatory occasions, that is, whether the subject supplied the morpheme where it was required. Following Dulay and Burt, (1974), fully correct morphemes were awarded two points, lack of any morpheme no points, and misformed morphemes one point. RESULTS In Table 1 we present the results for each group. We limit the presentation to descriptive statistics because the conditions necessary for the use of inferential tests were not met. In any case, the results are extremely easy to interpret by inspection of the percentages. NEED WE SACRIFICEACCURACYFOR FLUENCY 199 Table 1. Accuracy for grammatical morphemes Item pronoun article sing. copula plural progressive irregular past regular past third pets. sing. auxiliary totals: Experimental grade 7 0.945 0.896 0.847 0.767 0.627 0.600 0.553 0.531 0.456 0.778 Experimental grade 8 (344/364) (515/576) (266/304) (112/146) (74/I 18) (65/110) (42/76) (135/254) (52/! 14) (1605/2062) 0.946 0.929 0.878 0.885 0.786 0.881 0.875 0.620 0.544 0.865 Traditional grade 8 (619/654) (970/1044) (281/320) (294/332) ( 151 / 192) (148/168) (182/208) (186/300) (81/148) (2912/3366) 0.950 0.904 0.932 0.888 0.403 0.693 0.696 0.648 0.359 0.818 (477/502) (759/840) (360/386) (174/196) (54/134) (104/150) (71 / 102) (197/304) (46/128) (2242/2742) In every case, except for the p r o n o u n , where a ceiling effect is clearly operating, there is a clear gain for the experimental g r o u p from grade 7 to 8. W h a t is crucial for the hypothesis tested in this study is that the experimental grade 8 g r o u p is slightly more accurate than the traditional grade 8 group. In addition, they produced more obligatory occasions, suggesting that they were m o r e fluent. The traditionally taught g r o u p was clearly better than the experimental eighth graders on only one item, the singular copula, and slightly better on one other, the third person singular. F o r the plural and p r o n o u n , the groups were nearly the same, but for the other five structures, the traditional g r o u p was clearly worse than the experimental group. In two cases, the traditional g r o u p performed worse than the seventh grade experimental g r o u p (progressive and auxiliary). Table 2. Individual variation Group Range in obligatory occasions Range in percent correct exp 7 exp 8 trad 8 9~212 178-340 106-356 49-97 75--95 55 93 Table 3. Comparison of grade 8 experimental and traditional students Scores 95% or above 90-94% 85-89% 80-84% 75-79% 70-74% 65-69% 60-64% 55-59% Number of students Experimental 1 4 4 1 5 0 0 0 0 Traditional 0 4 3 4 2 1 0 1 1 200 MARIANNE NIKOLOV and STEPHEN KRASHEN As presented in Tables 2 and 3, individual variation was not unreasonable. Results for each group were not due to the unusual performance of a few students, nor were a few students responsible for an unusually large share of the obligatory occasions. The dismal results for the traditional group's performance in the progressive marker, however, was partly due to the performance of two students, one of whom scored none correct out of 16 obligatory occasions and another scoring nothing out of eight. Removing their scores, however, only raises the group mean to 0.49, still well below the experimental group. SUMMARY AND CONCLUSION The experimental group participated in a program that had far more content and less focus on form than the traditional group. Up to grade eight their program was based entirely on content, and only some focus on form was introduced that year. Not unexpectedly, they produced more obligatory occasions of the grammatical morphemes studied here. They were also, however, slightly more accurate than the traditional group overall. Of the nine morphemes studied, the experimental eighth graders were superior in five and there was no difference for two. In only two of the nine was the traditional group better, an advantage of about 5.5% in one case and about 3% in another. This result confirms that there is no loss in accuracy when there is no or little emphasis on form. As in every study, there were flaws in the design. The teacher variable was not controlled; and the experimental and traditional groups each had different teachers, who stayed with them the entire time. In addition, the experimental group was taught by one of the researchers. The other teacher, however, was equally enthusiastic and firmly believed in her methodology. A second possible flaw is that the testing was done by one of the investigators. Finally, only a limited range of structures was investigated. We have, however, other reasons to suspect that the experimental group did not suffer any damage because of a lack of formal grammar teaching: at age 16, 13 of the 15 children sat for and passed a form-focussed proficiency exam in English, and four are now studying at the University as English majors. All report that they will be using other languages in their careers. 1 NOTES ~Nikolov, (1989), in a previous analysis of this data, compared the difficulty order of the grammatical morphemes in Table 1 to morpheme difficulty orders found in English as a second language (Dulay and Butt, 1974), and reported good agreement. The experimental seventh graders' order correlated very highly with Dulay and Burt's subjects (rho = 0.867 for children speaking Spanish as a first language). While the experimental and traditional eighth grade orders correlated positively, the magnitude of the correlations were somewhat more modest (rho = 0.65, 0.60, respectively). As Nikolov points out, these results suggest that all three groups were engaged in natural language acquisition, with the experimental seventh grade group engaged the most. REFERENCES Dulay, H. and Burt, M. (1974) Natural sequences in child second language acquisition. Language Learning, 24, 37-53. NEED WE SACRIFICE ACCURACY FOR FLUENCY 201 Krashen, S. (1994) The input hypothesis and its rivals. In Implicit and Explicit Learning of Language, ed. N. Ellis, pp. 45-77, Academic Press. London. Nikolov, M. (1989) English acquisition orders of three groups of Hungarian schoolchildren. British Journal oJ Language Teaching., 27, 81-82.