August 11, 2000 (Preliminary Report)

3.  Basic Design of the Experiment
To test the above hypothesis, the subjects of the experiment (See Section 4 for more details) were divided into two groups of equal size, Experiment Group and Control Group. Both groups were given the same letter-writing assignments from Tests 1 to 7. The former group wrote the assignments making full use of the BLC Concordancer (except Test 1 [1]), whereas the latter group did not use the Concordancer in writing the assignments. All the assignments submitted by the subjects in both groups were checked for errors directly related to articles and prepositions, and their numbers compared to see whether there are any significant differences in the numbers of errors they made.

The initial assumption was that the subjects in the Experiment Group would make less errors in all the assignments (i.e., Tests 2 through 7) consistently than those in the Control Group. If this was found to the case, the reason could only be attributed to the use of the BLC Concordancer by this group, which is the main controlling factor of the experiment as shown the following table which summarizes the major conditions of the experiment for the Control and Experiment Groups.

An additional factor that might influence the result of experiment is the difference in the types of feedback. A number of previous reseach, however, indicate that the difference in the types of instructor's feedback does not make any significant influence on the overall quality of students' writing (Robb et al. 1986; Hattori et al. 1990; Kanatani et al. 1993; Shizuka 1996; Oikawa and Takayama 2000). These research clearly show that it's not what kind of feedback or how much of it we give to our students, but how our feedback is accepted, attended to and acted upon by the students that makes real difference.

In the current experiment, therefore, a particular attempt was made that the instructor's feedback given to the Experiment Group was always accompanied by, or comprised of, specific instructions for follow-up activities involving the use of the BLC Concordancer (See the Sample Feedback in Section 5), so that the subjects are given the opportunity to redress the problems at hand by themselves rather than being deprived of that opportunity by being given the right "answers" by the instructor in the form of overt error correction. This means that this factor is in effect subsumed into the main controlling factor of the experiemnet, i.e., the use of the Concordancer. Such being the case, we will not consider the effect, if any at all, of the difference in the types of feedback.

[1] Test 1 was used to measure statistical homogeneity of the two groups (i.e., to see whether they belong to the same mother population); therefore, the requirement to use the Concordancer was not applied to Test 1.