Discerning words and consequently listening comprehension has been an impediment for Iranian EFL learners. Part of this problem has been due to the bottom-up processing. The present study was aimed at investigating the correct parsing and segmentation of of words in in the blurred connected speech. There were eighty students majoring in English Literature at the University of Kurdistan participating in this study. After administering a TOEFL test comprising listening comprehension, reading, and grammar, 60 students were picked out. Then, they were divided two to two experimental and controlled groups. Their homogeneity was obtained through running a t-test. The experimental group received instructions of correct segmentation from from the book "American Accent Training", and the control group were exposed to the traditional oral input. At the end of the term two versions of the oral text, one with correct parsing and the other with dislodged segmentation were read aloud by an educated Canadian native speaker at a speech rate of 230 words a minute. They were then given two tests, namely, MCQT and a cloze test. The result revealed that the displacement of the boundaries did not affect the listening comprehension of the experimental group on the MCQT significantly. Although the subjects in the experimental group performed better on the cloze test, this difference was not statistically significant. The implication being that the fast blurred connected speech affected the performance of the two groups so greatly that correct segmentation did not make up for this kind of natural connected speech with features of fast rate, liaison, assimilation, and omission.
