Auditory Neurophysiology publications

Pitch perception is fundamental to speech and music, yet the neural representation of pitch in auditory cortex remains poorly understood. We hypothesized that mice rely on temporal cues for pitch perception because of their wide auditory filters and high-frequency hearing. To investigate, we developed a mouse auditory filter bank model that estimates the temporal pitch salience of complex tones. We then trained mice to discriminate low vs. high fundamental frequencies (F0s) in complex tones and found that modeled temporal pitch salience was highly correlated with behavioral discrimination sensitivity. Widefield calcium imaging showed that the organization of periodotopy in auditory cortex predicts how well mice recognize temporal pitch cues. Mice trained on an F0 discrimination task had larger periodotopic maps with greater vector strengths than naïve mice, demonstrating experience-dependent expansion and refinement of temporal pitch maps in auditory cortex. Our findings establish mice as a robust model for studying the neural basis of temporal pitch perception.

Hearing is the largest modifiable mid-life risk factor for Alzheimer's disease (AD), yet its link to dementia remains unclear. We identified a neurophysiological biomarker of AD risk using the non-invasive, rapidly acquired, and clinically translatable auditory brainstem response (ABR) in normal hearing knock-in rats (Swedish familial AD risk variant to Amyloid precursor protein, AppS; male and female). Human ABRs have been proposed as a biomarker for AD and related dementias. The novel metric reported here is derived from multidimensional parametric feature extraction on the distribution statistics of repeated single-trial ABR traces. We report accurate prediction of genetic risk for AD risk in young and aged rats: AppS separate clearly from healthy humanized (AppH) in sex- and age-dependent manners. Notably, auditory learning during young adulthood shifted the AppS ABR signature towards a healthy AppH-like state that maintained over time into older age. Altogether the findings support the utility of the ABR to track disease state, progression, and effects of intervention, and point to a central neural generator of auditory dysfunction related to AD risk. ABRs could provide a very early biomarker for detection of AD risk and used to test the synergy of auditory and cognitive functions in human dementia.

Aging and noise over-exposure lead to complex mixtures of cochlear degradation that impair the structure and function of outer hair cells, inner hair cells (IHCs), and the cochlear nerve. However, IHC damage and cochlear synaptopathy (CS) remain pathologies “hidden” from the audiogram. This study aimed to identify and differentiate the physiological signatures of these two distinct pathologies using promising non-invasive assays: Envelope Following Responses (EFRs), Auditory Brainstem Response (ABRs), Wideband middle-ear reflexes (WB-MEMRs), and Distortion Product Otoacoustic Emissions (DPOAEs). We utilized chinchilla models of carboplatin-induced (CA) IHC damage (N = 4) and temporary threshold shift (TTS) noise-induced CS (N = 4) to compare the physiological signatures of each pathology. While both groups showed unchanged ABR thresholds two weeks after exposure, EFRs, ABR Wave V/I ratios, and MEMRs showed distinct effects of exposure. Despite non-elevated ABR-derived audiometric thresholds after exposure, both CA and TTS exposure resulted in severe in EFR “peakiness”, particularly for sharp, short-duty-cycle stimuli and significant elevations in ABR Wave V/I ratios. However, these findings were less-pronounced in the TTS-exposed animals. WB-MEMR amplitudes were decreased with elevated thresholds in both groups; this effect was more pronounced in the TTS group. Opposite trends in DPOAE amplitudes indicated that while both IHC damage and CS result in similar suprathreshold temporal coding deficits, effects on outer-hair-cell integrity and auditory efferent physiology may differ between the two pathologies. Future work and novel diagnostics should aim to distinguish these specific cochlear pathologies in clinical populations, or at the very least consider their overlap. Highlights A multi-metric diagnostic approach was used with chinchilla models of inner-hair-cell (IHC) damage and cochlear synaptopathy (CS). IHC damage and synaptopathy both cause suprathreshold deficits “hidden” from the audiogram. IHC damage results in more severe temporal envelope coding degradation than does synaptopathy. A combination of EFR “peakiness”, ABR Wave V/I ratio, and Wideband Middle Ear Muscle Reflex (WB-MEMR) appear to be useful measures for profiling IHC damage and CS.

Stimulus-specific adaptation is a key feature of sensory systems, where the repeated presentation of a stimulus results in diminishing neuronal responses specifically to that stimulus, but not to a different stimulus. Cortical neurons display varying degrees of stimulus-specific adaptation, with cortical circuits shaping the dynamics of these responses. Here, we use calcium imaging and optogenetic approaches to assess the contributions of SOM and PV neurons to stimulus-specific adaptation in the mouse auditory cortex. We find opposing and complementary effects of SOM and PV neurons on the adaptation of principal neurons which manifest as context-dependent SOM neuron-mediated enhancement and PV-neuron mediated suppression of responses that depend on the temporal repetition rate of the sounds. Our results suggest that SOM and PV neurons can engage in multiple operating regimes that exist in a continuum during sensory processing.

The CUGBP Elav-like family 4 (CELF4), an RNA-binding protein, is dynamically expressed in spiral ganglion neurons (SGNs) in the cochlea over development, but how Celf4 is involved in regulating hearing functions is poorly understood. In this study, we generated a Celf4 ± mouse line and examined changes in the first two synapses along the auditory pathways. Firstly, we found that hearing functions were largely intact in Celf4 ± mice, with exception of reduced amplitude for Wave II of auditory brainstem responses (ABRs) and increased delays for Wave II and IV, both in case of 4 kHz only. Secondly, we found that counts of inner and outer hair cells (IHCs and OHCs) and SGNs remained unchanged in Celf4 ± mice, and that the number and function of ribbon synapses between IHCs and SGNs were comparable between WT and Celf4 ± mice. Lastly, function of the endbulb of Held synapse, formed between auditory nerve fibers (ANFs) of SGNs and bushy cells in the cochlear nucleus, was significantly altered in Celf4 ± mice. Specifically, synaptic vesicle release was subtly reduced, and excitability of bushy cells was significantly dampened, likely due to a hyperpolarized resting membrane potential. Furthermore, we found that spike kinetics was significantly faster in Celf4 ± bushy cells, likely caused by a larger fast-inactivating A-type potassium current (I A ) found in these cells. In conclusion, we found that Celf4 haploinsufficiency altered transmission at the endbulb of Held synapses in the cochlear nucleus in a significant and multifaceted manner, revealing the roles of Celf4 in regulating hearing functions. Graphical Abstract Title: Schematic representation of the experimental approach and major findings. Legend: In mice with Celf4 haploinsufficiency, we found that while function at hair cell ribbon synapses was largely preserved, transmission at the endbulb of Held synapse was significantly altered, including subtly reduced release of synaptic vesicles from the endbulb, dampened excitability in bushy cells, and accelerated spike kinetics caused by enhanced voltage-gated K + current.

Cochlear implants (CIs) have successfully restored hearing in more than one million patients with severe to profound hearing loss worldwide. Although CIs effectively restore speech perception in quiet environments, sound localization remains challenging for bilateral CI users, particularly their ability to utilize interaural time differences (ITDs). The majority of clinical CI processors use a coding strategy that encodes ITD information only in the envelope of electrical pulse trains rather than their pulse timing, which may contribute to the poorer spatial hearing perception of CI users. We recently demonstrated in a behavioural study on early deafened, bilaterally CI‐implanted rats that pulse timing ITDs completely dominate ITD perception, whereas sensitivity to envelope ITDs is almost negligible in comparison. Building on this, we here investigated the neurophysiological sensitivity of the inferior colliculus (IC) to envelope and pulse timing ITDs at two different pulse rates (900 and 4500 pulses/s) and three different stimulation modulations (5, 20 and 100 Hz) in CI rats with different hearing experiences. Our results indicate that IC neurons exhibit far greater sensitivity to pulse timing ITD than envelope ITD independent of pulse rate, modulation rate or hearing experience. These findings suggest that to improve binaural hearing outcomes in bilateral CI users, clinical stimulation strategies should provide informative pulse timing ITDs. image Key points Current bilateral cochlear implants mostly provide time cues of arriving sound in the envelope of stimulation rather than in the timing of pulses. A behavioural study on cochlear implanted rats has shown that pulse timing dominates binaural hearing. Here we examined the neural sensitivity to independent changes in time cues presented on the envelope or pulse timing, respectively, in rats with bilateral cochlear implants and different hearing experiences. Auditory midbrain neurons were consistently more sensitive to changes in pulse timing rather than envelope, irrespective of stimulus pulse rate, modulation rate or an animal's hearing experience.

Across sensory modalities, peripheral deafferentation can induce maladaptive central plasticity that amplifies sensory responses while distorting affective processing, rendering innocuous stimuli unpleasant, intrusive, or even painful. Here, we examined the affective dimension of compensatory plasticity in a mouse model of hyperacusis by introducing noise-induced focal cochlear lesions and tracking maladaptive plasticity in the lateral amygdala (LA), a key limbic hub for sensory-affective valuation. Whereas control mice habituated to neutral sounds, mice with noise-induced hearing loss (NIHL) showed sustained LA hyperresponsivity and abnormally strong temporal coupling between calcium transients and pupil dilations, an autonomic index of arousal. NIHL also disrupted discriminative auditory threat learning, producing poorly selective, non-extinguishing enhancement of LA responses and freezing to both threatening and non-threatening sounds. We posited that potentiating inhibition in the higher-order auditory cortex (HO-AC), a prominent source of auditory sensory input to the LA, could reinstate normal affective sound processing after NIHL. Brief bouts of 40-Hz optogenetic activation of HO-AC parvalbumin-expressing inhibitory neurons (PVNs) durably reversed LA sensitization, normalized pupil-indexed arousal dynamics, and reinstated discriminative auditory threat memory. These findings identify cortical inhibitory potentiation as a strategy for reversing neural, autonomic, and behavioral signatures of distorted affective sound processing after peripheral sensory injury.

Noise exposure is a known cause of hearing loss, and only a few effective preventive drugs are available. Therefore, in this study, we aimed to investigate the protective effects of brimonidine on noise‐induced inner ear hearing impairment in mice and explore its underlying mechanisms and long‐term outcomes. Mice were randomly divided into control, noise exposure, and brimonidine groups. A 62‐week follow‐up was conducted after noise exposure. Brimonidine inhibited the noise‐induced increase in inner ear glutamate concentration and downregulated inflammatory factors and immunoglobulins. Brimonidine decreased glutaminase and VGLUT2/3 expression and reduced glutamate synthesis and vesicle transport without affecting its clearance, thereby decreasing glutamate excitotoxicity and protecting synapses and spiral ganglion neurons long term. Mice exposed to noise could temporarily restore their hearing thresholds; however, their auditory function in old age remained significantly worse than those that received brimonidine‐mediated cochlear protection in youth. These findings highlight the importance of enhancing noise protection from an early age.

Background Peripheral hearing loss is associated with auditory hallucinations and increased risk for psychotic disorder, particularly in genetically vulnerable individuals. Moreover, both hearing loss and schizophrenia disrupt auditory temporal acuity, a sensitive measure of auditory brain function. Here, we used mouse models of hearing loss and schizophrenia to reveal how neural mechanisms of auditory cortical temporal acuity depend on sensory and genetic risk factors for psychosis. Methods We quantified auditory cortical temporal acuity in mice with or without hearing loss and with or without the 22q11.2 deletion — one of the strongest known genetic risk factors for schizophrenia. Cortical single-unit and population activity were recorded in awake mice (N = 23) during presentations of loudness-adjusted gap-in-noise stimuli with varying durations of silent gap, which are commonly used to assess auditory temporal acuity in humans. Results Both hearing loss and the 22q11.2 deletion disrupted auditory cortical temporal acuity, but through distinct mechanisms. Hearing loss broadly degraded temporal acuity at the level of single-unit responses and neural population activity. In contrast, the 22q11.2 deletion selectively impaired gap duration thresholds in regular-spiking (putative excitatory) but not fast-spiking (putative inhibitory) neurons. Mice with comorbid hearing loss and 22q11.2 deletion exhibited both abnormalities in auditory cortical temporal acuity. Conclusions Sensory and genetic risk factors for psychosis can disrupt auditory cortical temporal acuity via distinct mechanisms that remain partially dissociable even under comorbid conditions. These findings underscore the importance of accounting for hearing loss comorbidity when interpreting auditory cortical dysfunction in psychotic disorder.

Male bushcrickets produce complex species-specific songs to attract mating partners. Since the neuronal processing underlying song-pattern detection is not fully understood, we analysed extracellularly recorded song-induced spike responses and local field potentials across the first stage of auditory processing in the prothoracic ganglion of Mecopoda elongata. We analysed the influence of the temporal structure and of any change in amplitude of spiking and local field potentials. We presented the complete chirp and compared the neuronal response pattern with those responses evoked individually by each pulse of the chirp. Our data revealed: (i) the complete chirp leads to a transient summation in local field potentials over the time course of the chirp, and (ii) significantly fewer action potentials occur during the chirp presentation than to the individually presented pulses. Our data provide insights into the question of how an acoustical signal can be used to extract those features from sensory information that are important for auditory behaviour, such as chorus formation, species recognition, or phonotaxis.

In the auditory cortex, neural responses to stimuli inside receptive fields (RFs) can be further facilitated by behavioral demands, such as attending to a spatial location. It is less clear how off-RF stimuli modulate neural responses and contribute to behavioral tasks. Our recent study revealed a particular form of location-specific facilitation evoked by repeated stimulation from an off-RF location, suggesting behavioral modulation of spatial RFs. To further explore this question, we trained marmosets to attend to sound locations that were either inside or outside the RFs of auditory cortical neurons. The majority of neurons showed increased firing rates at target locations inside their RFs. Interestingly, this increase also occurred outside the RFs, sometimes exceeding the responses at the RF center during passive listening. This task-related off-RF facilitation was much more common in the caudal area than in the rostral area and the primary auditory cortex. A normalization model reproduced the off-RF facilitation using widespread suppression. The model’s prediction was confirmed by experimental observations of widespread reductions in firing rate and hyperpolarized membrane potentials for off-RF stimuli. These results suggest that behavioral task demands recruit a broader range of neurons than those that are responsive to a target sound in the passive state.

Our brains constantly filter incoming sounds to understand our environment. While extensively studied in adults, how this ability develops across childhood remains unclear. We recorded intracranial brain activity from 54 participants aged 4-21 while they watched movie clips containing simultaneous speech and music. We used deep neural networks to separate the mixed audio into isolated speech and music streams, then built encoding models to determine which stream best predicted neural responses in the auditory cortex. Although participants heard only the original mixture with no instruction to attend to either stream, higher-order auditory regions including the superior temporal gyrus (STG), superior temporal sulcus (STS) and middle temporal gyrus (MTG), responded preferentially to speech. This speech-bias strengthened with age in STG, suggesting that this region progressively sharpens its representation of socially relevant sound across development. These findings indicate that speech prioritization in the developing brain emerges automatically, without directed attention.

The acoustic flutter is processed through complementary monotonic rate coding and cannot be modulated by other acoustic parameters in the auditory cortex (AC). However, it remains unclear how the inferior colliculus (IC) encodes acoustic flutter, especially when changing other acoustic parameters. Here, we recorded IC neural activity in response to acoustic flutter and determined the existence of conjunctive processing between repetition rate and other acoustic parameters. We found that most IC neurons also encode the repetition rate at the flutter range through complementary monotonic rate coding. In addition, although the acoustic parameters did not change their monotonicity, most IC neurons encode both repetition rate and other acoustic parameters, different from the flutter processing in AC. Thus, complementary monotonic rate coding for acoustic flutter was widespread in the auditory system; however, coding specificity for repetition rate increased from IC to AC, and the capacity for conjunctive coding with other acoustic parameters decreased.

Age-related hearing loss (ARHL) is a prevalent neurodegenerative condition characterized by the progressive loss of spiral ganglion neurons (SGNs). Although oxidative stress is recognized as a central pathogenic driver of ARHL, the precise molecular triggers that initiate and amplify SGN damage remain elusive. Here, we investigated the role of the Transient Receptor Potential Vanilloid 2 (TRPV2) channel in ARHL. We found that TRPV2 expression was significantly upregulated in the SGNs of aged mice, which was associated with elevated oxidative stress. Pharmacological activation of TRPV2 in 6-month-old mice (a pre-senescent stage with preserved baseline hearing) accelerated the onset of high-frequency hearing damage, as evidenced by auditory brainstem response (ABR) measurements. Consistently, TRPV2 activation exacerbated oxidative damage (assessed by 4-HNE staining) and increased apoptotic cell death (detected via TUNEL) in the SGN population. In primary SGN cultures, TRPV2 overexpression aggravated oxidative stress, whereas TRPV2 knockdown in SH-SY5Y cells (a human neuroblastoma cell line) markedly mitigated the oxidative injury as reflected by reduced 4-HNE. Our findings establish that the age-related upregulation of TRPV2 sensitizes SGNs to oxidative stress, thereby promoting neuronal damage and exacerbating ARHL. This work highlights TRPV2 as a promising therapeutic target for intervening in the progression of ARHL.

Reconstructing speech from neural recordings is crucial for understanding human speech coding and developing brain-computer interfaces (BCIs). However, existing methods trade off acoustic richness (pitch, prosody) for linguistic intelligibility (words, phonemes). To overcome this limitation, we propose a dual-path framework to concurrently decode acoustic and linguistic representations. The acoustic pathway uses a long-short term memory (LSTM) decoder and a high-fidelity generative adversarial network (HiFi-GAN) to reconstruct spectrotemporal features. The linguistic pathway employs a transformer adaptor and text-to-speech (TTS) generator for word tokens. These two pathways merge via voice cloning to combine both acoustic and linguistic validity. Using only 20 min of electrocorticography (ECoG) data per human subject, our approach achieves highly intelligible synthesized speech (mean opinion score = 4.0/5.0, word error rate = 18.9%). Our dual-path framework reconstructs natural and intelligible speech from ECoG, resolving the acoustic-linguistic trade-off.

BACKGROUND: Noise-induced hearing loss (NIHL) has been traditionally attributed to outer hair cell (OHC) damage, but synaptic injury at inner hair cell (IHC)–spiral ganglion neuron (SGN) junctions may better explain persistent functional deficits. OBJECTIVE: To test whether IHC ribbon loss and functional synapse loss scale with pathological noise intensity and to define their longitudinal relationship to hearing function. METHODS: A rodent NIHL model was exposed to 100 dB SPL for 2 h or 110 dB SPL for 4 h. Cochleae were immunolabeled for CtBP2 (presynaptic ribbons) and GluA2 (postsynaptic AMPA receptor puncta). Putative functional synapses were quantified as closely apposed CtBP2/GluA2 pairs. Hearing function was assessed longitudinally and correlated with synaptic measures. RESULTS: Both exposure conditions produced graded reductions in IHC ribbon counts and CtBP2/GluA2 pairs, with greater losses after 110 dB/4 h. The number of CtBP2/GluA2 apposed pairs closely tracked preserved hearing function, providing a stronger association with functional outcome than ribbon counts alone. Ribbon loss remained largely stable across post-exposure intervals, whereas functional synapse loss was disproportionately larger at early time points, consistent with transient pre–post misalignment. Over time, CtBP2/GluA2 pairing partially recovered toward levels predicted by surviving ribbons, suggesting re-alignment or reconnection of a subset of remaining synaptic elements. CONCLUSION: Pathological noise intensity drives exposure-dependent IHC synaptopathy in addition to OHC injury. Functional synapse counts (CtBP2/GluA2 pairs) sensitively reflect stored hearing function and indicate a potential post-exposure window for synaptic re-establishment.

The clinical manifestations of mood and psychotic disorders encompass phasic and transdiagnostic features. The complexity of these symptoms may hamper the development of biomarkers for these diagnostic categories. In the present study, machine learning was employed to cluster their transdiagnostic clinical manifestations as a means of developing state-dependent biomarkers with electroencephalography (EEG). This data-driven clustering classified the items of multiple symptom scales into three symptom domains, which subsequently stratified patients with major depressive, bipolar, and schizophrenia spectrum disorders into distinct symptom state groups. The specific brain activity profiles of each stratum were then characterized by resting-state and auditory steady-state EEG paradigms. The resting-state readings of patients in manic states revealed significantly increased gamma and beta oscillations, whereas 40-Hz and 20-Hz auditory steady-state responses (ASSRs) showed no significant differences. These findings suggest that manic states are associated with heightened high-frequency oscillatory activity, represented by gamma and beta oscillations, without a concomitant increase in sensory-evoked information processing precision. Further investigations of excessive gamma and beta oscillations may facilitate the development of state-dependent biomarkers for the assessment, diagnosis, and treatment of manic state symptomatology in clinical psychiatry.

Auditory areas on the superior temporal plane and lateral convexity are key initial stages of speech processing in the human cortex, representing acoustic and phonetic attributes in a temporally precise manner. More complex representations in auditory-related cortex along the ventral and dorsal processing streams and prefrontal cortex are associated with perception and action. In this study, we used intracranial electroencephalography (iEEG) to clarify where and how activity leading to perceptually driven behavioral events emerges. Participants were patients undergoing iEEG monitoring for medically intractable epilepsy. Stimuli were monosyllabic words, and participants pressed a button in response to a semantic target category. Significant high gamma activity after stimulus onset and immediately prior to motor response defined stimulus- and behavior-related activity patterns, respectively. The stimulus-related pattern was more common than behavior-related throughout the cortical auditory hierarchy as well as sensorimotor cortex. Behavior-related activity was sparsely represented, with the highest prevalence in the prefrontal cortex and a more limited representation in anterior temporal and parieto-occipital cortex. Hemispheric asymmetries included a higher prevalence of stimulus-related activity in the right sensorimotor cortex and a higher prevalence of the behavior-related pattern in the left prefrontal cortex. Faster behavioral responses were associated with greater stimulus-locked high gamma power in non-core auditory, prefrontal, and premotor cortex. Results reveal the cortical distribution of sensory stimulus-driven responses and activity time-locked to behavior and provide insights into neural substrates of speech perception.

Background/Objectives: Blast-induced hearing loss (BIHL) is a major concern, particularly for military personnel, and is linked to impaired auditory neuron survival and synaptic plasticity. This study investigates the potential of the TrkB agonist 7,8-dihydroxyflavone (7,8-DHF) to reduce the severity of BIHL and promote recovery in a mouse model. Methods: Eight-week-old male C57BL/6J mice were used. A custom-built, compressed air-driven system utilizing a modified paintball apparatus was employed to deliver controlled unilateral double blasts (~22 psi exposure pressure) to the left ear. The blasts were administered 30 min apart. Immediately following the second blast, mice received either 7,8-DHF (10 mg/kg) or vehicle (10% DMSO) via intraperitoneal injection. Auditory brainstem responses (ABRs) were measured in both ears at baseline (pre-blast) and at several post-exposure time points. Results: The consecutive blast exposure induced a significant elevation in ABR thresholds, indicative of hearing loss, in both the ipsilateral (exposed) and contralateral (unexposed) ears of vehicle-treated mice. Notably, mice treated with 7,8-DHF demonstrated a marked improvement in hearing recovery compared to the vehicle group. Significant reductions in ABR thresholds were observed in the ipsilateral ear at 4 weeks post-blast (p < 0.0001) and in the contralateral ear as early as 1-week post-blast (p = 0.0236). However, the recovery was partial, with ABR thresholds plateauing after 4 weeks. Conclusions: A controlled blast model demonstrates that systemic administration of the TrkB agonist 7,8-DHF exerts a protective effect, partially restoring auditory function after blast injury. This supports the therapeutic potential of targeting the BDNF-TrkB signaling pathway for managing BIHL.

Harmonicity is a property of complex sounds such as vocalizations or music, but it remains unclear how harmonicity is processed in the auditory cortex (ACtx). Subregions of ACtx are thought to process harmonic stimuli differently. Selective responses to sound features in ACtx emerge hierarchically from primary ACtx (A1) L4 and secondary ACtx (A2) layer (L)2/3, which is believed to be the most responsive to harmonic sounds. Since harmonic stacks can range from 2 to >10 components, being more similar to naturalistic vocalizations, harmonic sensitivity might also arise hierarchically across layers and areas. We studied responses to harmonic stacks of 2–10 frequencies across A1 L4, A1 L2/3, and A2 L2/3 in adult male and female mice using in vivo two-photon microscopy. We found harmonic-sensitive neurons (HNs) responding only to harmonic stacks but not to individual frequencies in all areas at similar proportions. HNs showed highly nonlinear spectral integration of harmonic frequencies that decreased as the harmonic stacks became more complex. Specifically, onset-biased HNs showed greater nonlinearity than offset-biased HNs only in A1 L4. Moreover, HNs in A1 L4 exhibited higher signal correlation than A2 L2/3. Sound-responsive neurons in A1 L4 have the weakest noise correlation compared with A1 L2/3 and A2 L2/3. Together, harmonic sensitivity is not a unique feature of A2 L2/3 but is already established in A1 L4, where neurons robustly encode harmonic sounds through sparse connections.