In brief: Students in a comprehension approach German class in which speaking was not required (but not forbidden) did much better on a test of pronunciation (imitation) than traditional students.


McCandless, P. and Winitz, H. 1986. Test of pronunication following one year of comprehension instruction in college German. Modern Language Journal 70 (4): 335-362.

Four groups of ten subjects were tested. The test: imitate five sentences spoken by native speakers of German, e.g.
"Die Erde wird zu Schlamm." (The ground will become muddy.)
Subjects' imitations (for all sentences) rated by native speakers of German on a one to five scale, where 1 = poor (mangelhaft), and 5 = excellent (ausgezeichnet).

Group N: Native speakers of German. Score = 1.28 (sd = .02).
Group CO: Controls, no previous study of German. Score = 4.40 (sd = .32).
Group CE:  Took an intensive eight unit Comprehension Approach German course during the summer, about 210 hours of classroom instruction and 30 hours listening to language tapes (The Learnables) = 240 hours total. "Their primary responsibility was to learn to listen and to understand German. They were not required to speak at any time, but they were also not told not to speak" (p. 359).
Of interest: "… instructors were told never to perform a spontaneous activity, such as erasing the blackboard, or moving a table. Instead they were to ask a student to perform the activity" (p. 359). If students did speak, errors were not corrected. Reading introduced near the end of the course.
Score = 2.50 (sd = .53).
Group T: Traditional two semester German classs, 128 hours in class, one hour per week in lab, two hours homework/week = total of 224 hours. Taught by non-native speakers. Grammar, oral drills. Most from community college.
Score = 3.10 (sd = .46).
All four groups significantly different.

Supports hypothesis that speaking not necessary, might be harmful, but
(1)     T group got less input.  It could have been the input that counted, not the fact that CE students did not speak much.
(2)     T group had non-native teachers.
(3)     Male model used for imitation test for CE, controls, but for only one T subject: female substituted because male not available.
(4)     All models were instructors in the CE class.

But results still consistent with strong delayed speaking hypothesis, even though other explanations possible.