The relation between speech intelligibility and stimulus level typically takes the form of a non-decreasing function, for both normal-hearing listeners and listeners with sensorineural hearing loss (SNHL). A performance decrease at high presentation levels – so-called rollover (RO) in the performance-intensity function – has traditionally been interpreted as a sign of retro-cochlear hearing loss. In some recent studies based on speech stimuli presented at high levels, RO was also observed in young listeners with normal audiograms, possibly reflecting cochlear synaptopathy. Overall, RO measurements could therefore be a useful tool for characterizing suprathreshold hearing abilities in different listener groups. While RO has been observed in studies that employed monosyllabic words or low-context sentences, reports of RO in measurements performed with high-context sentences are lacking. Here, we hypothesized that RO in the performance-intensity function is related to the amount of context information available in the employed speech material. To test this, 22 young adults with normal audiograms and without any self-reported hearing problems were tested at two presentation levels: 80 and 95 dB SPL (broadband). Three speech materials were used: monosyllabic words, low-context sentences and high-context sentences. To avoid ceiling effects and upward spread of masking, all measurements were performed in stationary speech-shaped noise with the stimuli bandpass-filtered from 1.4 to 7.4 kHz . RO effects were found in the speech scores collected with the two low-context materials but not with the high-context material. Overall, this suggests that low-context sentence materials allow for sensitive RO measurements in individual listeners.