FREQUENCY-FOLLOWING responses (FFRs) were elicited by English long vowels (female /a/ and male /e/) in a dichotic listening task. Stimuli were simultaneous and of equal duration, but differing spectra permitted unique identification of vowel components in the compound FFR. Horizontal and vertical montage FFRs were recorded with putative origins in the acoustic nerve and central brain stem, respectively. FFRs obtained during attention to each vowel showed significant effects for the voice fundamental frequency, f0, which is perceptually salient and conveys paralinguistic information such as the sex of the speaker. Amplitudes of f0 were larger when vowels were attended than when ignored. These findings provide evidence of short-latency attention effects in humans and suggest that linguistic attention may initially filter inputs based on salient paralinguistic cues.