The study attempts to test the following hypotheses: (1) There are significant differences in two kinds of information units (low-level and high-level information units) and the combined information units recalled by the subjects between two lecture versions; (2) There are significant differences in two kinds of information units (low-level and high-level information units) and combined information units recalled by two performance groups (Medium Performance Group vs. High Performance Group ); (3) There are interactions between language proficiency and lecture version in recalling information at different levels; (4) The learners with lower listening proficiency rely more heavily on discourse markers in recalling information at different levels than those with higher listening proficiency.The present study involves 56 students with medium listening proficiency and 59 students with high listening proficiency from a university in China. It examines how the subjects comprehend and recall information at different levels when listening to a lecture with discourse markers or without discourse markers. The detailed analysis of the data yields the following findings:(1) L2 learners listening to the lecture with discourse markers recall significantly more low-level information, more high-level information and more combined information than EFL learners exposed to the lecture without discourse markers.(2) Learners at a higher level of listening proficiency recall significantly more low-level, high-level and combined information than learners with a lower proficiency.(3) There are no statistically significant interactions between listening proficiency and lecture version for the subjects in recalling low and high-level information as well as combined information.(4) Compared with learners with higher listening proficiency, learners with lower listening proficiency rely more heavily on discourse markers in recalling low-level information and combined information. However, in terms of high-level information recall, High Performance Group rely on discourse markers just as Medium Performance Group.These findings provide additional empirical support for the beneficial effect of discourse markers and offer new evidence regarding the different extent to which the learners with different language proficiency rely on discourse markers in a spoken text.In line with the findings in the study, the teaching of the use of discourse markers is highly advocated in English listening instructions. And teachers are encouraged to raise the awareness of the role of discourse markers in listening among less able students as a way of improving their listening comprehension.Finally, future research is suggested to examine the role of discourse markers from different perspectives and further studies are called for to provide more evidence regarding the interactions between language proficiency and role of discourse markers in listening comprehension.