This study investigates the listening comprehension performance of Korean university students when exposed to unscripted and scripted English speech, as well as their perceptions of various linguistic features associated with each speech type. A total of 115 participants were divided into two groups, both of which had the same level of listening proficiency. The unscripted group listened to spontaneous, unscripted dialogues during the listening comprehension test, while the scripted group listened to scripted versions from which features typical of unscripted speech had been removed. Following the listening test, participants completed a questionnaire assessing their perceived level of difficulty regarding a range of features that typically distinguish the two speech types. The results revealed that the unscripted group scored significantly lower on the listening test compared to the scripted group. Additionally, the survey indicated that participants found unscripted speech significantly more challenging, especially due to factors such as pronunciation, vocabulary, and grammar. Features typical of unscripted speech, including filled pauses, false starts, and a natural speech rate, were identified as obstacles to L2 listening comprehension. Based on these findings, the study suggests implications for L2 learning materials, instructional methods, assessment practices, and teacher education programs.