To investigate the effects of information feedback of learning difficulties on English learning, an one-year experiment was conducted in two junior high schools, in which 704 subjects from grades 7 and 8 were sampled and divided into experimental group and control group, nearly counterbalanced for grade and sex. In the EG, the teachers were given information concerning learning difficulties before instruction, and were asked to explore actual problems by means of unitary diagnostic English tests and to correct errors immediately. While the students of the CG were distributed in normal classes without any special instruction. The criterion variables were Vocaburary-Frounciation, Grammer, Comprehension, and Listening. The control variables were intelligence, pre-achievement, achievement motivation, social economic status, and perception-toward-teacher. The results were analyzed by Finn´s multivariance analysis of covariance, stepwise multiple regression analysis, and MAPS´s canonical correslational analysis. It was found that (1) The effects of the new teaching strategy were more salient in grade 7 than in grade 8; (2) The information feedback approach was clearly in favor of Grammer learning, while in Vocaburary-Pronunciation, Comprehension, and Listening there were no consistent effects for both grades; (3) Intelligence and pre-achievement were the two major predictors for post-achievement in both grades; (4) Listening was a relatively independent canonical factor, in which intelligence, SES, and perception-toward-teacher played important roles in prediction. It suggests that the teaching strategy of information feedback is much effective in Grammer learning and when the learning difficulties are newly formed.