Selective attention to one of two competing voices reduces, rather than enhances, the amplitude of distortion product otoacoustic emissions evoked by speech-derived stimuli at the attended voice's resolved harmonics (Steinebach and Reichenbach, 2026) - a counterintuitive sign, since it suggests that the cochlea's active amplifier boosts the attended voice less, not more. The mechanism producing this sign, its frequency dependence (in particular whether the sign reverses in more basal cochlear regions), and the neural pathway mediating the modulation (medial olivocochlear efferents versus corticofugal projections from auditory cortex) are all unidentified. The finding also stands in tension with the only prior speech-DPOAE study (Saiz-Alía et al., 2021), which reported a weak opposite-sign effect, and with a heterogeneous, inconclusive older literature on attention and otoacoustic emissions. The problem is to establish the mechanism by which selective attention to speech modulates the cochlear active process, discriminating among the candidate explanations (efferent linearization of outer-hair-cell nonlinearity, unidirectional coupling in apical cochlear mechanics, attention-dependent phase interference between emission sources, middle-ear muscle reflex).

Background

The cochlea is not a passive microphone: outer hair cells - one of the two kinds of auditory sensory cell, which mechanically amplify sound in a frequency-selective way - constitute an active process, and the brain can regulate this amplifier through the medial olivocochlear (MOC) efferent system - neural projections from the brainstem that innervate outer hair cells and reduce their amplification (Guinan, 2006). A non-invasive window on the active process is provided by otoacoustic emissions (OAEs): because the cochlea is nonlinear, stimulating the ear with two tones at frequencies f1f_1 and f2f_2 (f2>f1f_2>f_1) generates intermodulation distortion products, most strongly at the cubic frequency 2f1−f22f_1-f_2, which propagate back out of the ear and can be recorded with a sensitive microphone in the ear canal. These distortion product otoacoustic emissions (DPOAEs) report the strength of the local cochlear nonlinearity and gain.

Whether this peripheral amplifier participates in selective attention - the cocktail-party ability to follow one voice among many - is a long-standing, unsettled question. Early studies measuring OAEs during attention produced heterogeneous results of varying sign and size (Meric and Collet, 1992; Maison et al., 2001; Smith et al., 2012; Wittekindt et al., 2014; Walsh et al., 2015), and Francis et al. (2018) showed that apparent attentional modulation of emissions can be an artifact of reduced subject motion and ear-canal noise. Computational work has nevertheless proposed that frequency-specific, efferent-mediated gain control at the cochlear level could aid speech recognition in noise (Messing et al., 2009).

Saiz-Alía et al. (2021) introduced speech-DPOAEs: instead of pure tones, the eliciting stimuli are waveforms that follow the time-varying harmonics of natural speech, so that the evoked emission tracks the harmonic structure of a particular voice. Voiced speech consists of a fundamental frequency (about 100 Hz for the male and 200 Hz for the female voice used) and its harmonics; roughly the first ten harmonics are resolved - each produces a distinct excitation peak at its own place along the basilar membrane, the cochlea's tonotopic frequency-to-position map - while higher harmonics are unresolved and excite overlapping cochlear regions. A cochlear gain-control mechanism operating on resolved harmonics could therefore sharpen the neural representation of a target voice, but could not operate on unresolved components.

Steinebach and Reichenbach (2026; preprint 2025) used this paradigm in 38 human listeners: two competing audiobook voices were presented to one ear, while speech-DPOAEs evoked by resolved- and unresolved-harmonic stimuli derived from each voice were recorded from the other ear as listeners attended to one voice, the other voice, or a visual text. Speech-DPOAEs at the resolved harmonics of a voice (approximately 1-1.8 kHz) were significantly reduced when that voice was attended - the opposite of the naive expectation that attention enhances cochlear amplification of the target - while emissions evoked by the female voice's unresolved harmonics (approximately 2.3-3.5 kHz) showed no modulation. Emissions evoked by the male voice's unresolved harmonics (distortion product near 1.05 kHz) were instead significantly larger when the male voice was attended (amplitude ratio 1.5 relative to attending the female voice, i.e. +3.8 dB; p = 7x10^-7); the authors attribute this to the spectral overlap of that stimulus with the female voice's resolved-harmonic region, so that attention to the female voice spills over onto the overlapping male unresolved harmonics. An emission near 350 Hz was too noisy to measure. The sign at the resolved harmonics is counterintuitive: taken at face value, the cochlea amplifies the attended voice's resolved harmonics less. The authors list several candidate explanations, none established: (i) attention-driven MOC activation partially linearizes the outer-hair-cell nonlinearity, and because DPOAE generation depends on that nonlinearity, stronger efferent suppression yields smaller emissions even while perception improves; (ii) at apical, low frequencies (below about 4 kHz) cochlear mechanics departs from the classical picture of critical-layer absorption - the standard description in which each frequency component of the cochlear traveling wave is absorbed at the basilar-membrane place whose resonant frequency matches its own - and the active process may instead resonate and couple to the basilar membrane unidirectionally (Reichenbach and Hudspeth, 2010, 2011), so that higher gain decouples the amplifier from the basilar membrane and suppresses backward-propagating distortion products; (iii) attention could alter the relative phase of the two known emission components - the distortion source generated at the nonlinear cochlear amplification site and the coherent-reflection source produced when the forward-traveling wave scatters off intrinsic irregularities of the cochlea - causing partial cancellation at the recording site; and (iv) a weak middle-ear muscle reflex (involuntary contraction of the middle-ear muscles, evoked by sound, which stiffens the ossicular chain and reduces sound transmission) could contribute. Which mechanism operates, whether the sign of the effect reverses in more basal cochlear regions where classical mechanics holds, and which neural pathway (MOC efferents versus descending corticofugal projections from auditory cortex) carries the signal, are unknown.

Problem Statement

Establish the mechanism by which selective attention to speech modulates the human cochlear active process, as manifested in speech-derived distortion product otoacoustic emissions (speech-DPOAEs: otoacoustic emissions evoked by stimuli that follow the time-varying harmonic structure of a particular voice). Concretely, determine: (1) the origin of the effect's sign - why attending to a voice reduces, rather than enhances, speech-DPOAE amplitude at that voice's resolved harmonics (the low-order harmonics, each of which excites a distinct place along the basilar membrane) - discriminating among the candidate mechanisms (attention-driven MOC-efferent linearization of the outer-hair-cell nonlinearity; apical-cochlea mechanics with unidirectional coupling between the basilar membrane and the active process; attention-dependent phase interference between distortion and coherent-reflection emission sources; middle-ear muscle reflex activation), or identifying and validating a different mechanism; and (2) the frequency dependence of the modulation, in particular whether its sign reverses in more basal cochlear regions beyond the approximately 1-3.5 kHz bands originally tested, where classical critical-layer absorption and bidirectional basilar-membrane-to-active-process coupling are expected, together with how the modulation depends on harmonic resolvability and stimulus configuration; and (3) the neural pathway that carries the modulation - medial olivocochlear efferents versus descending corticofugal projections from auditory cortex. All three parts are required, and jointly they constitute one mechanistic question: an answer may proceed through human or animal experiments, computational modeling of the cochlea and its efferent innervation, or a combination, but it must account for the published empirical signatures, including the reduction at attended resolved harmonics, the absence of modulation at the female unresolved-harmonic emissions (approximately 2.3-3.5 kHz), the significant increase of the male unresolved-harmonic emission (approximately 1.05 kHz) when the male voice was attended - which the source attributes to spectral overlap with the female resolved-harmonic region - and the reported emission latency patterns (latencies that do not vary with attentional focus but do vary systematically with stimulus type and frequency). It must also yield a discriminating, testable prediction that distinguishes the proposed mechanism from the alternatives; a written answer argued from existing data, modeling, or a proposed experiment is sufficient, and new data collection is not required.

References

  1. Steinebach J., Reichenbach T. (2026). Attention to speech modulates distortion product otoacoustic emissions evoked by speech-derived stimuli in humans. Frontiers in Neuroscience 20:1756386. doi:10.3389/fnins.2026.1756386. https://doi.org/10.3389/fnins.2026.1756386
  2. Steinebach J., Reichenbach T. (2025). Attention to Speech Modulates Distortion Product Otoacoustic Emissions Evoked by Speech-derived Stimuli in Humans. bioRxiv preprint. doi:10.1101/2025.08.15.670505. https://doi.org/10.1101/2025.08.15.670505
  3. Saiz-Alía M., Miller P., Reichenbach T. (2021). Otoacoustic emissions evoked by the time-varying harmonic structure of speech. eNeuro 8:ENEURO.0428-20.2021. doi:10.1523/ENEURO.0428-20.2021. https://doi.org/10.1523/ENEURO.0428-20.2021
  4. Reichenbach T., Hudspeth A. J. (2010). A ratchet mechanism for amplification in low-frequency mammalian hearing. Proceedings of the National Academy of Sciences 107:4973-4978. doi:10.1073/pnas.0914345107. https://doi.org/10.1073/pnas.0914345107
  5. Reichenbach T., Hudspeth A. J., Shera C. A., Olson E. S. (2011). Unidirectional amplification as a mechanism for low-frequency hearing in mammals. AIP Conference Proceedings 1403:507-512 (Mechanics of Hearing 2011). doi:10.1063/1.3658139. https://doi.org/10.1063/1.3658139
  6. Reichenbach T., Hudspeth A. J. (2014). The physics of hearing: fluid mechanics and the active process of the inner ear. Reports on Progress in Physics 77:076601. doi:10.1088/0034-4885/77/7/076601. https://doi.org/10.1088/0034-4885/77/7/076601
  7. Guinan J. J. Jr. (2006). Olivocochlear efferents: anatomy, physiology, function, and the measurement of efferent effects in humans. Ear and Hearing 27:589-607. doi:10.1097/01.aud.0000240507.83072.e7. https://doi.org/10.1097/01.aud.0000240507.83072.e7
  8. Francis N. A., Zhao W., Guinan J. J. Jr. (2018). Auditory attention reduced ear-canal noise in humans by reducing subject motion, not by medial olivocochlear efferent inhibition: implications for measuring otoacoustic emissions during a behavioral task. Frontiers in Systems Neuroscience 12:42. doi:10.3389/fnsys.2018.00042. https://doi.org/10.3389/fnsys.2018.00042
  9. Stoll T. J., Vandjelovic N. D., Polonenko M. J., Li N. R. S., Lee A. K. C., Maddox R. K. (2025). The auditory brainstem response to natural speech is not affected by selective attention. PLOS Biology 23:e3003407. doi:10.1371/journal.pbio.3003407. https://doi.org/10.1371/journal.pbio.3003407
  10. Guest D. R., Farhadi A., Carney L. H. (2026). A computational model of the mammalian auditory periphery with a closed-loop medial olivocochlear reflex simulating across-channel efferent gain control. The Journal of the Acoustical Society of America 159:5445-5461. doi:10.1121/10.0044141. https://doi.org/10.1121/10.0044141
  11. Wittekindt A., Kaiser J., Abel C. (2014). Attentional modulation of the inner ear: a combined otoacoustic emission and EEG study. The Journal of Neuroscience 34:9995-10002. doi:10.1523/JNEUROSCI.4861-13.2014. https://doi.org/10.1523/JNEUROSCI.4861-13.2014
  12. Walsh K. P., Pasanen E. G., McFadden D. (2015). Changes in otoacoustic emissions during selective auditory and visual attention. The Journal of the Acoustical Society of America 137:2737-2757. doi:10.1121/1.4919350. https://doi.org/10.1121/1.4919350
  13. Smith D. W., Aouad R. K., Keil A. (2012). Cognitive task demands modulate the sensitivity of the human cochlea. Frontiers in Psychology 3:30. doi:10.3389/fpsyg.2012.00030. https://doi.org/10.3389/fpsyg.2012.00030
  14. Maison S., Micheyl C., Collet L. (2001). Influence of focused auditory attention on cochlear activity in humans. Psychophysiology 38:35-40. doi:10.1111/1469-8986.3810035. https://doi.org/10.1111/1469-8986.3810035
  15. Meric C., Collet L. (1992). Visual attention and evoked otoacoustic emissions: a slight but real effect. International Journal of Psychophysiology 12:233-235. doi:10.1016/0167-8760(92)90061-F. https://doi.org/10.1016/0167-8760(92)90061-F
  16. Riecke L., Marianu I.-A., De Martino F. (2020). Effect of auditory predictability on the human peripheral auditory system. Frontiers in Neuroscience 14:362. doi:10.3389/fnins.2020.00362. https://doi.org/10.3389/fnins.2020.00362
  17. Messing D. P., Delhorne L., Bruckert E., Braida L. D., Ghitza O. (2009). A non-linear efferent-inspired model of the auditory system; matching human confusions in stationary noise. Speech Communication 51:668-683. doi:10.1016/j.specom.2009.02.002. https://doi.org/10.1016/j.specom.2009.02.002