DICTO-PHRASE, as a test of listening comprehension, requires test takers to reproduce the gist of selected propositions of language. The purpose of the study was to discover what kinds of linguistic knowledge as well as strategies and cognitive processes candidates make use of while performing on DICTO-PHRASE. Five upper-intermediate participants were asked to think aloud while performing on the test tasks. Right after that, they answered related retrospective questions. Based on the transcribed protocols, task analysis, and the objective of the study, a coding scheme was developed for extracting common patterns of performance.The findings indicate that the participants used intelligent guessing, that is a compensatory strategy, frequently. In doing so, they drew upon topical knowledge and appealed mostly to immediate text rather than pretext or post-text. Whatsoever, they looked for more linguistic clues than discourse ones. Random guessing was also used but much less frequently with no significant gains for the test takers. One more strategy used very frequently was rereading strategy although in most cases revision was not considered.DICTO-PHRASE was found to diminish the role of STM in making responses, boosting the role of LTM due to allowing for responses made on the basis of different levels of knowledge. In fact, the procedure was shown to be capable of recognizing sufficient (comprehensible and acceptable) responses as well as exact ones as correct. Otherwise two thirds of the sufficient responses would be lost. Consequently, the validity of the test scores representing the listening comprehension construct could be expected to be higher than other measures of listening comprehension recognizing only the exact answer.