Can We Increase the Power of Reading by Adding More Output and/or Correction?

Beniko Mason and Stephen Krashen
Texas Papers in Foreign Language Education (in press)

It is firmly established that free reading leads to increased second language competence (Krashen, 1993; Elley, 1991, Mason and Krashen, 1997). A recent confirmation is Tsang (1996), who reported that EFL students in an English-medium school in Hong Kong who engaged in self-selected reading for 24 weeks made significant gains in writing (overall impression, content, and language use), but those who did extra writing did not. Tsang noted, however, that the writing was done "without teacher feedback" (p. 227). Papers were "impressionistically graded, given brief positive comments" and returned to the students (p. 217).

In this study, we assume the power of reading, and seek to determine whether output practice, with and without correction, will increase its effectiveness, as it has been suggested that increasing output and form-focus may enhance the effect of comprehensible input (e.g. Swain, 1995; Schmidt, 1995).


Subjects were 104 first year female English majors, ages 18-19, in an extensive reading program in a college in Osaka, Japan. No students had been in English speaking countries, none had had extra English lessons or an English-speaking friend, and none worked in businesses that required use of English. The study was therefore done in a pure EFL situation. All were "high beginning/low intermediate" students of English (mean score of the TOEIC = 123.6 for Reading, 153.6 for Listening, out of a possible 495 for each section).

All groups engaged in extensive reading (Mason and Krashen, 1997). They were required to read about 2000 pages in two semesters (400,000 to 500,000 words), a longer duration than Tsang (1996) employed, and with far more reading. All students began reading graded readers at the lowest level, and worked up to higher levels and easy authentic reading for native young adults. Students selected books on their own according to interest and were asked to keep a record of the books they had read, pages, level of the book, and the time required to read the book. They also wrote a brief summary of each book's content and comments on their reading progress. The teacher responded with encouragement, suggestions for further reading, and answered whatever questions students had.

The three experimental groups were intact classes, and their treatments were those desired by the students themselves, based on how classes reacted to the summary writing assignment:

Group 1 ("Japanese"; n = 34) did not want to spend time writing summaries of they read; they only wrote very brief summaries in Japanese just to confirm that they had done the reading.

Group 2 ("English"; n = 34) wrote a summary, In English, for each book they read.

Group 3 ("English/correction"; n = 36) wrote a summary in English for each book they read. The summaries were corrected by a native speaker of English and rewritten by the students. Feedback was focused on the completeness of the summary as well as a few grammatical errors, eg. the third person singular, articles, and tenses. Correction was done eight times each semester.

The following measures were used:
  1. A 100 item Cloze test, written at the grade 6 level (reliability r = .86). The same test was used as a pretest (April, 1996) and posttest (December, 1996).
  2. A Writing sample: Subjects read a short story about 2000 words long, which took them about 30 minutes. They were then asked to write a summary immediately afterwards, without looking back at the story, which also took about 30 minutes. Different stories were used for the pre- and posttests: The story used for the pretest was easier, written at the 600 word level, while the story used for the posttest was at the 1600 word level. Four aspects of the writing samples were analyzed:
      (1) The total number of words written,
      (2) the total number of clauses written,
      (3) the total number of error-free clauses, as determined by the agreement of two raters, and
      (4) the total number of words in error-free clauses. Measures (3) and (4) have been found to be especially appropriate for writers at this level (Ishikawa, 1995). The writing pretest was given in April, 1996 and the posttest in January, 1997.


Table 1. Results (adjusted means)

wds in
Japanese37.7 225.944.528.9125.6
English39.1 193.836.720.9102.2

Table 1 presents post-test means, adjusted for pretest scores for that measure. Application of ANCOVA revealed no significant difference among the groups for the cloze test (F = 1.45, df = 2,92, p = .24). Differences on the writing measure were, however, significant or very close to significant (total words; F = 3.02, p = .05; total clauses, F = 5.20, p = .007; error-free clauses, F = 4.89, p = .009; words in error-free clauses, F = 2.80, p =.065).

Follow-up tests (Fisher LSD) were performed for all writing measures. In every case, the Japanese group and English/correction group significantly outperformed the English group (p < .001). Differences between the Japanese group and English/correction group were not significant.

It must be noted, however, that there was extreme variation in pretest writing measures. Even with the use of ANCOVA, this makes interpretation difficult. As presented in table 2, the English/correction group was much more fluent on the pretest, producing more words and clauses. Note that the Japanese group caught up to them and even surpassed them on the writing measures; differences, however, were not significant, as noted above, using ANCOVA. There was little difference among the groups on the cloze pretest.

Table 2. Pretest and Postest Raw Scores

 ClozeTotal Words
Japanese 30.5 (7.7)38.7 (7.1)109.0 (45.7)227.3 (67.2)
English 28.9 (8.1)39.2 (6.9)71.6 (34.3) 179.7 (39.1)
English/corr27.0 (8.6)39.4 (9.3)172.8 (86.6) 222.9 (59.8)
 Total ClausesError-free Clauses
Japanese 21.7 (9.8) 43.7 (12.3) 10.1 (5.8) 26.6 (9.9)
English 13.8 (6.8) 34.1 (7.8) 6.1 (4.6) 19.0 (6.4)
English/corr34.6 (18.2) 42.2 (11.5) 16.7 (10.5) 24.8 (9.0)
 Words in Error-free Clauses
Japanese 46.0 (26.3) 123.8 (48.3)
English 29.0 (21.2) 92.6 (32.4)
English/corr75.2 (49.5) 122.6 (48.1)
Standard deviations in parentheses.

Another way of analyzing the results is to consider efficiency. Table 3 presents the amount of time each group devoted to reading and writing in English for the extensive reading class. (All students were enrolled in 10.5 additional hours of English classes in addition to the class discussed here. These classes were the same for all students.)

Table 3. Time Devoted per Week to Reading and Writing in English

Time devoted to ReadingWritingTotal
Japanese3.35 (1.0)-3.35
English 3.7 (1.4)2.2 (.82)5.9
English/corr4.4 (1.6)2.0 (1.2)6.4
writing = extra time writing in English.

The Japanese group spent slightly less time reading, and spent far less time than the other groups overall. We estimate that for the entire academic year, the Japanese group invested about 100 hours in English, while the English group spent 180 hours and the English/correction group about 200 hours. Thus, the Japanese group got approximately the same results for about one-half the investment in time.

There was no difference among the groups for the amount read (table 4). All groups declined in the second semester, and none of the groups attained the goal of 2000 pages.

Table 4. Quantity of Reading

Amount ReadFirst Sem.Second Sem.
Japanese 965.1 (364.3)591.7 (297.1)
English 912.9 (161.8)459.8 (268.5)
English/corr 954.1 (156.8)572.4 (199.1)


There was no obvious effect of adding additional output in English or output with correction.Reading alone produced the same results, and was far more time-efficient. This result is consistent with the input hypothesis, but inconsistent with output and instruction hypotheses.

Possible objections to this conclusion include the following:

  1. Subjects may not have written enough to make a difference. This is doubtful, because subjects in the writing groups invested about 1/3 the total time in writing.
  2. Subjects may not have had enough error correction. This is possible, but our results agree with those of other studies showing no or very little effect even after substantial amounts of correction (Krashen, 1994).
  3. The same person taught all three sections (B.M.) and may have been biased toward one approach. The reading and writing, however, was done at home, and the corrections for the English/correction group were done by someone else.

It also may be the case that the English/correction group was more enthusiastic than the others: They requested correction and wrote far more on the pretest. While they outperformed the English group, it is interesting that they did not do better than those who only wrote in Japanese. A surprising result, however, is that the English/correction group did not suffer in terms of fluency when compared to the English group. Correction thus seemed to have no ill effects, even though it did not help.


We thank Mr. Julian Upton for his help in correcting student papers and Miss Mayumi Yamashita for help in calculations.


Elley, W. (1991). Acquiring literacy in a second language: The effect of book-based programs. Language Learning, 41, 375-411.

Ishikawa, S. (1995). Objective measurement of low-proficiency EFL narrative writing. Journal of Second Language Writing, 4, 51-69.

Krashen, S. (1993). The Power of Reading. Englewood, CO: Libraries Unlimited.

Krashen, S. (1994). The input hypothesis and its rivals. In N. Ellis (Ed.)

Implicit and Explicit Learning of Languages (pp. 45-77). London: Academic Press.

Mason, B. and Krashen, S. (1997). Extensive reading in English as a foreign language. System, 25, 91-102.

Schmidt, R. (Ed.) (1995). Attention and Awareness in Foreign Language Learning. Honolulu: University of Hawaii Press.

Swain, M. (1995). Three functions of output in second language learning. In Cook, G. and Seidelhofer, B. (Eds.) Principle and Practice in Applied Linguistics: Studies in Honor of H.G. Widdowson (pp. 125-144). Oxford: Oxford University Press.

Tsang, W.K. (1996). Comparing the effects of reading and writing on writing performance. Applied Linguistics, 17, 210-233.