People with normal hearing thresholds can nonetheless have difficulty with understanding speech in noisy backgrounds. The origins of such supra-threshold hearing deficits remain largely unclear. Previously we showed that the auditory brainstem response to running speech is modulated by selective attention, evidencing a subcortical mechanism that contributes to speech-in-noise comprehension. We observed, however, significant variation in the magnitude of the brainstem's attentional modulation between the different volunteers. Here we show that this variability relates to the ability of the subjects to understand speech in background noise. In particular, we assessed 43 young human volunteers with normal hearing thresholds for their speech-in-noise comprehension. We also recorded their auditory brainstem responses to running speech when selectively attending to one of two competing voices. To control for potential peripheral hearing deficits, and in particular for cochlear synaptopathy, we further assessed noise exposure, the temporal sensitivity threshold, the middle-ear muscle reflex, and the auditory-brainstem response to clicks in various levels of background noise. These tests did not show evidence for cochlear synaptopathy amongst the volunteers. Furthermore, we found that only the attentional modulation of the brainstem response to speech was significantly related to speech-in-noise comprehension. Our results therefore evidence an impact of topdown modulation of brainstem activity on the variability in speech-in-noise comprehension amongst the subjects.Understanding speech in noisy backgrounds such as other competing speakers is a challenging task at which humans excel 1,2 It requires the separation of different sound sources, selective attention to the target speaker, and the processing of degraded signals 3-5 Hearing impairment such as resulting from noise exposure often leads to an increase of hearing thresholds, a reduction in the information conveyed about a sound to the central auditory system, and thus to greater difficulty in understanding speech in noise 6-8 However, even listeners with normal hearing thresholds can have problems with understanding speech in noisy environments 9,10 .An extensive neural network of efferent fibers can feed information from the central auditory cortex back to the auditory brainstem and even to the cochlea 11,12 . Research on the role of these neural feedback loops for speech-in-noise listening has mostly focused on the medial olivocochlear reflex (MOCR), in which stimulation of the medial olivocochlear fibers that synapse on the outer hair cells in the cochlea reduces cochlear amplification across a wide frequency band 13 . Computational modelling as well as animal studies have shown that such reduced broad-band amplification can improve the signal-to-noise ratio of a transient signal embedded in background noise [14][15][16][17] . However, it remains debated whether the reduction of cochlear amplification through the MOCR contributes to better speech-in-noise comprehe...