This article investigates the perception of vocal manifestations of excessive sleepiness. Although previous efforts have demonstrated that naive listeners are able to estimate behavioral sleepiness, we aim to assess in this study this ability on both subjective (medical questionnaires) and physiological (polysomnography) measurements of sleepiness. We asked 71 naive French-speaking listeners to annotate a subset of the Multiple Sleep Latency Test corpus, each listener participating in two annotation sessions, using one among three annotation tools, with two different annotation paradigms (with or without reference). Based on these data, we then evaluated their ability to correctly annotate subjective or physiological sleepiness depending on the annotation tool or the paradigm of the test they undertook. We also measured the interaction between their performance and their characteristics, as well as the correlation between the listeners' performances and speakers' characteristics.