This study investigates the listening comprehension of 388 high-intermediate listening proficiency (HILP) and low-intermediate listening proficiency (LILP) Chinese students of English as a foreign language. These students listened to a lecture, the discourse of which was (a) familiar-unmodified, (b) familiar-modified, (c) unfamiliar-unmodified, or (d) unfamiliar-modified. The modified discourse contained information redundancies and elaborations. After the lecture, the EFL subjects took a multiple-choice exam testing recognition of information presented in the lecture and general knowledge of the familiar ("Confucius and Confucianism") and unfamiliar ("The Amish People") topics. A significant interaction between speech modification (redundant vs. nonredundant speech) and listening proficiency (HILP vs. LILP) indicated that the HILP students benefited from speech modification, which entailed elaboration/redundancy of information, but the LILP students did not. A significant interaction between prior knowledge (familiar vs. unfamiliar topic) and test type (passage-independent vs. passage-dependent items) was also found. For both the HILP and LILP subjects, prior knowledge had a significant impact on subjects' memory for information contained in the passage-independent test items on the postlecture comprehension test. Those EFL subjects who listened to the familiar-topic lecture on Confucius had higher passage-independent than passage-dependent scores. There was no difference in the performance on the passage-independent and passage-dependent items of those who listened to the lecture on an unfamiliar topic (the Amish). However, the passage-independent performance of subjects who listened to the familiar topic lecture was superior to that of those who listened to the lecture on the unfamiliar topic. Subjects' performance on passage-dependent items did not differ significantly whether the familiar or unfamiliar topic was presented. Implications of the findings for assessing and teaching EFL listening comprehension are suggested.