The present study aims to empirically and longitudinally investigate the effects of reading instructions on 14 Japanese EFL college students with respect to cohesion and coherence instruction. The students were given three reading comprehension tasks-writing protocol, the making of a story map and retrospective interview. Attempts were made to examine writing data and retrospective protocol data from the perspectives of cohesion understanding, cohesion noticing, coherence understanding and coherence noticing. The results indicate the following: (1) Cohesion understanding by and large was observed both among the experimental group and among the upper of the control group from retrospective data: (2) From cohesion noticing, the experimental group made use of meta-linguistic connective terms to explain the way they read the text, whereas the control group did not use cohesion noticing: (3) Coherence understanding was only observed in the two of the experimental groups. It seemed difficult for low level readers to read a text with knowledge of discourse structure: (4) From coherence noticing, there was a big difference between the two groups. The experimental group used connectives in the text to make it clear to construct meaning relations. The control group did not. These findings suggest that successive instructions enabled good readers to make use of information on conceptualizing textual features, and they also enabled low level readers to learn to cope with discourse texts.