Human subthalamic neurons encode speech features during listening and couple with auditory cortex

Human subthalamic neurons encode speech features during listening and couple with auditory cortex | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Human subthalamic neurons encode speech features during listening and couple with auditory cortex Yanming Zhu, Robert Richardson, Alan Bush, Matteo Vissani, Latane Bullock This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8008692/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract The basal ganglia–thalamocortical loop is widely implicated in speech production, yet its role during listening remains underexplored. Here, we show that human subthalamic nucleus (STN) single‑unit activity is modulated during auditory presentation of syllable triplets, and that a subset of neurons encodes phonetic (consonant and vowel) and sequence features prior to speech onset. Using intraoperative microelectrode recordings during deep brain stimulation implantation, ~39% of units showed significant firing‑rate modulation during the auditory epoch. Within these, we discovered “auditory‑syllable” units that carried phonetic and/or sequence information. Auditory presentation also elicited beta‑band desynchronization in STN LFPs and revealed frequency‑specific spike‑phase coupling (SPC) with cortex: auditory‑specific units coupled predominantly with superior temporal gyrus (STG) in alpha, whereas units modulated during both auditory and speech epochs preferentially coupled with sensorimotor regions in beta. Responses were observed to incidental, non‑task sounds in a subset of auditory‑specific units that support genuine auditory sensitivity. Together, these results support a role for the STN in auditory–motor integration and indicate frequency‑specific routing within human cortical–subcortical loops during listening. Biological sciences/Neuroscience/Cognitive neuroscience/Language Biological sciences/Physiology/Neurophysiology Biological sciences/Neuroscience/Sensory processing Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction The integration of auditory and motor systems enables rapid, context‑appropriate responses to sound 1,2 . While the basal ganglia have long been linked to action selection and stopping 3-6 , most human auditory perception research has emphasized cortical processing 7-11 , leaving the contribution of subcortical nodes during listening insufficiently characterized 12,13 . Likewise, intracranial investigations of human basal ganglia function have focused on production 14,15 , without yet considering activity during perception. The subthalamic nucleus (STN) may participate in rapid sensory-motor integration and motor preparation following auditory input 16 . A small, almond-shaped structure within the basal ganglia, the STN regulates motor control, emotion, and cognition. It receives monosynaptic input from the cortex via the hyperdirect pathway, and other cortical input indirectly, via the striatum and external globus pallidus (GPe), before sending glutamatergic input to the internal globus pallidus (GPi) and substantia nigra pars reticulata (SNr), which project to the ventral thalamus and onward to frontal cortical regions 17-20 . STN hyperdirect innervation traditionally has been described as originating from the frontal lobe, yet songbirds have a population of neurons in the auditory cortex with a monosynaptic projection to the songbird STN 21 , and resting-state functional MRI data suggests connections between STN and superior temporal gyrus (STG) 13 . Indeed, we recently described direct evidence in the human brain for a monosynaptic hyperdirect pathway from auditory areas of the STG to the STN 12 . We hypothesized that during listening, STN single‑unit activity would be selectively modulated by auditory features, and that STN–cortex coupling would exhibit frequency‑specific organization consistent with loop circuit‑level routing 22 . Leveraging intraoperative access during awake deep brain stimulation (DBS) lead implantation, we tested these predictions in a syllable‑triplet task designed to separate auditory and speech epochs. We found that subsets of STN neurons respond to auditory stimuli, encode syllable features, and exhibit task-specific beta desynchronization and phase-coupling with auditory and sensorimotor cortices 23,24 . These findings challenge the classical motor-centric view of the BG and position the STN as a computational hub for auditory encoding. STN single unit and ECoG recording during a syllable triplet repetition task To study the STN’s role in auditory-motor integration, we recorded single-unit STN activity during an intraoperative speech production task. Twenty-five PD patients undergoing DBS surgery completed a total of 60 task runs. Simultaneous electrocorticography (ECoG) was recorded from subdural electrode strips temporarily placed through the standard bone opening used for DBS lead implantation (Suppl Table 1) Participants listened to triplets of consonant-vowel syllables (e.g., /va/, /ti/, /su/, /ga/) via earphones and repeated them aloud. Stimuli were balanced across four consonants (/v/, /t/, /s/, /g/) and three vowels (/a/, /i/, /u/), presented at normal and high intensities (+25 dB). Figure 1A illustrates a trial with synchronized STN microelectrode recordings, sorted single-unit spikes, ECoG local field potentials (LFPs), and audio waveforms of both stimuli and responses. Participants responded consistently, with an average latency of 2.2 ± 0.3 seconds (Fig. 1B). Each trial began with a one-second baseline period, followed by an auditory presentation lasting 1.5 seconds (each syllable onset is 0.5 seconds apart), and concluded with self-paced repetition of the syllable sequence in a continuum. We recorded 205 single units across 25 participants (120 tracks, 59 sessions). Microelectrode and ECoG recording sites were normalized to MNI space via non-linear warping for group-level analysis (Fig. 1C–D) 25 . The motor subregion of the STN was defined using the DISTAL-minimal atlas, an MRI-based anatomical atlas frequently utilized in clinical and research contexts to define subregions within the STN. Microelectrode recordings predominantly sampled the motor subregion of the STN, and ECoG primarily sampled IFG, ventral pre- and post-central gyrus, supramarginal gyrus, and superior temporal gyrus. Time-frequency analyses of STN macroelectrode LFPs (Fig. 1E and Suppl. Fig. 1) revealed beta-band (13–30 Hz) desynchronization (significant reductions in beta power compared to the pre-auditory stimuli baseline, for more than 100 ms, FDR corrected one-way ANOVA) following auditory stimulus onset in 44 of 222 electrodes (20%) (Fig. 1E, right). Similarly, we analyzed single-unit recordings time-locked to auditory onset (Fig. 1F). Fifteen of 205 units (7%) exhibited a significant reduction of beta-band entrained firing (Fig. 1F, right, one-way ANOVA with FDR compared to resting state). These cells were relatively co-localized within the STN (Fig. 1G), as assessed by the average pairwise distance compared to that of random samples of recording sites in the STN (Suppl Fig. 2). Together, these findings show that STN activity is reliably modulated during listening at both the LFP and single‑unit levels, beyond baseline fluctuations (FDR‑controlled comparisons). STN single unit firing rate exhibits selective modulation during audition Hierarchical clustering based on trial-averaged activity of the auditory epoch revealed units with positive, negative, or no firing rate modulation; each line in Fig. 2A is one unit’s averaged firing rate aligned to the auditory onset. Activity was z-scored relative to baseline, and units were categorized as modulated if they differed significantly from baseline for more than 100 ms (FDR corrected). We identified nine categories, reflecting combinations of positive ( + ), negative ( - ), or no modulation ( 0 ), during auditory (a) and speech (s) epochs (Fig. 2B and Suppl Fig. 3).Figure 2B shows example units that respond selectively to audition, to speech, to both, or to neither: units are grouped in rows by their response during the auditory epoch : increased firing (a + ), no change (a 0 ), or decreased firing (a − ), and in columns by their response during the speech epoch : increased firing (s + ), no change (s 0 ), or decreased firing (s − ). Each panel shows both the spike raster and the mean firing rate (±SEM) aligned to auditory (green line) and speech (blue line) onsets. Shaded regions mark periods of significant modulation relative to baseline. For example, a + /s - denotes units that have increased FR during the auditory window and decreased FR during speech. These firing rate and spike raster plots illustrate that auditory modulation can occur independently of speech modulation, can co-occur with speech modulation in the same units, or can be absent altogether. We refer to units that were only modulated during either the auditory or speech epoch as auditory-specific (a * /s 0 ) or speech-specific (a 0 /s * ) units (asterisks represent either positive or negative significant modulation). We refer to units modulated during both auditory and speech epochs as auditory-speech (a * /s * ) units. We identified 20% (41/205) a + and 19% (39/205) a - units out of the 205 a* unites that responded with increases or decreases in firing rates, respectively, at the auditory epoch onset (Fig. 2C). Among the a * units, five a + /s 0 (12%) and ten a - /s 0 (26%) units responded exclusively to the auditory presentation (a * /s 0 ). These unit types were spatially aggregated within the STN, where we calculated the average distance of the units within the group, then compared it to a random sample of STN recording sites’ average distance with permutations (Fig. 2D and Suppl Fig. 5). Because the task required subsequent repetition, we considered motor preparation as a potential confound. Aligning STN activity to non‑task auditory events (ambient operating‑room sounds and clinician speech) showed that two auditory‑specific (a * /s 0 ) units maintained their response (Suppl Fig. 6), suggesting genuine auditory sensitivity in at least a subset of neurons, although this control is limited in scope. Consistent with a motor integration role, beta‑desynchronization units were predominantly auditory–speech (a * /s * ) or speech‑only (a 0 /s * ) (13/15 and 2/15, respectively). (Fig. 2E). STN Units Encode Phonetic and Sequence Features During Audition We previously reported STN LFP high-gamma activity and single-unit firing rate modulation specific to phonetic and sequence features during the production of an utterance 14,15 . Here, we assessed if auditory features prior to speech onset also modulate STN single-unit firing rate, by analyzing spike-triggered averages (Fig. 3A). Linear models relating firing rates to syllable features confirmed that certain auditory-speech (a * /s * ) units encoded phonetic or sequence information (Fig. 3B–C). Leave-one-out method was used to determine which feature contributed significantly to the linear model. In this linear model, the consonants and vowels are compared to the /v/ and /a/. With these two methods, we identified five, seven, and one units that encoded vowel, consonant, and sequence features, respectively (Fig. 3D). These “auditory-syllable units” (13/205, ~6 %) were spatially clustered within the STN (Fig. 3E–F and Suppl Fig. 7). Interestingly, we found that the auditory-syllable units not only encode syllable features in the auditory epoch but also encode syllable features in the speech epoch, measured by comparing the firing rate between these features to the resting state among all these types of units. (Fig. 3G and Suppl Fig. 8). Multiple regressions using the mean firing rate at different lags (with a 200 ms window) were performed to look for syllable encoding during the auditory and speech epochs among all auditory-syllable units (Suppl Fig. 9). We found that syllable-encoding units also could encode syllable features during speech (Suppl Fig. 9). 10/12 auditory syllable encoding units encoded the same or a subset of syllables that the same unit encoded during the speech epoch, suggesting that these units support auditory-speech integration. To assess whether syllables can be decoded from STN single unit activities, we performed nested cross‑validated elastic‑net multinomial logistic regression 26 . Two decoding strategies were compared: a direct 12-class model trained to predict the full consonant–vowel combination, and a factorized approach in which separate consonant (4-way) and vowel (3-way) classifiers were trained and their output probabilities combined to recover the 12-class prediction. Consistent with the observation that most neurons encoded only a subset of phonetic features, single-unit classifiers performed near chance. In contrast, pooling features from all ten auditory-syllable units that also showed speech encoding (removing units without enough repetition) markedly improved performance, yielding modest but reliable above-chance decoding (12-way mean accuracy ≈ 0.41; factorized ≈ 0.39; chance = 1/12, Fig. 3H). These results suggest that while individual STN units carry limited information about syllable identity, the pooled population supports distributed encoding of consonant and vowel features. Auditory-specific (a * /s 0 ) units did not encode syllable features (Suppl. Fig. 18), a feature that was found in five of thirteen beta desynchronization speech units (Fig. 3I). We identified only one unit modulated by stimulus volume (Suppl Fig. 10). Thus, we found that STN units can encode speech features' phonemes (consonant and vowel), sequence order, and volume. STN connectivity with and similarities to STG activity. Given the STG’s role in human auditory processing, we compared temporal response structure across STG and STN using non‑negative matrix factorization (NMF) to ask whether similar auditory components appear across the loop circuit 27 . Previous studies have shown that NMF can separate STG responses to auditory stimuli into two types: onset and sustained 28 . Similar to this result, we found STG auditory modulation can be differentiated in these two categories (Fig. 4A and Suppl Fig. 11). Additionally, a third ECoG response type (syllable/triplet) captured the triplet nature of stimuli (Fig. 4B and Suppl Fig. 12), indicating that two components alone are not sufficient to explain the LFP dynamics in our data. Thus, we performed a percentage of variance explained analysis of the components, which showed that four components could explain ~90% of the variance in LFP modulation (Suppl Fig. 13); subsequent NMF analyses used four components, as shown in Suppl Fig. 13. We applied the same method to the STN firing rate with two components and found that the components did not recapitulate STG HG activity (Fig. 4C, sustained onset and sustained depressed type). Onset, triplet-like, and two triplet-like components (in anti-phase to each other), however, were identified when using four components (Fig. 4D). STN LFPs showed no correspondence (Suppl Fig. 14). To better appreciate the similarity of auditory modulation between the STG and STN, we applied cosine similarity matrices. STG LFP and STN firing rate modulation NMF components were correlated strongly during the auditory epoch but not during speech (Fig. 4E). STG LFP and STN firing rate modulation were not cross-correlated between the auditory epoch and the speech epochs. (Fig. 4E). To further characterize STN–cortex interactions during listening, we analyzed cortical-STN spike-phase coupling (SPC) during the auditory epoch, as described by Vissani et al. 29 . Auditory-specific (a * /s 0 ) units demonstrated strong SPC events with broad cortical areas in the alpha range but not in other frequency ranges. Auditory-speech (a * /s * ) units showed SPC with subcentral, postcentral, and STG predominantly in the beta band (Fig. 4F and Suppl Fig. 15). SPC in the high gamma bands was weak, not modulated in auditory or speech epochs 29 and did not show an STN unit type or cortical region selectivity (Fig. 4F). We noted a dissociation in the frequency bands of SPC according to the STN unit type; auditory-specific (a * /s 0 ) cells exhibited SPC predominantly in alpha, whereas auditory-speech (a * /s * ) units exhibited SPC predominantly in beta. We further explored how STN-STG SPC evolved over the course of the auditory stimuli across the unit types. We found that the SPC in the alpha band for auditory-specific (a * /s 0 ) cells was greatly reduced during the auditory epoch (Fig. 4G). For the a + /s 0 units, SPC recovered in the speech epoch, but for a - /s 0 units, despite a transient increase after the auditory epoch offset, SPC remained at very low values during speech. For auditory-speech (a * /s * ) units, there is a partial loss of beta-band SPC with sensory-motor cortical regions during auditory presentation and an almost complete loss during early speech production (Fig. 4H). Generally, SPC modulations did not overlap with their FR modulation in the temporal domain, suggesting parallel mechanisms for encoding auditory information. These results suggest the presence of an STG-STN auditory input loop and an STN-sensorimotor cortex auditory–motor transformation loop. STN macrostimulation during simultaneous ECoG recordings was performed to investigate the functional connectivity between the two structures by analyzing evoked potentials, as described previously 12 . We reanalyzed the data in Jorge et al. with specific attention to STG responses (Suppl Fig. 16). Among all effective STN stimulation locations (STN Stim ), those that induce evoked potentials in the STG (STN Stim-STG ) were significantly closer to auditory-specific (a * /s 0 ) units than expected by chance (STG-STN pair, Fig. 4I and Suppl Fig. 17). We time-aligned the average response of auditory-specific (a + /s 0 ) units, auditory-speech (a + /s * ) units, and STG ECoG auditory responding electrodes to the onset of the auditory epochs. We found that the response onset of auditory-specific units (177 ms after auditory onset) was similar to that of the STG gamma response (167 ms after auditory onset) but faster than that of auditory-speech units (374 ms after auditory onset, Fig. 4J), suggesting that auditory-specific units may receive monosynaptic input from the STG. Taken together, these data are consistent with hyperdirect projections linking auditory and sensorimotor cortices to STN that may facilitate sensory–motor integration. Discussion This study provides convergent evidence that the human STN participates in auditory–motor integration during listening. We identify two novel functional single‑unit classes: auditory‑specific neurons that respond selectively during the auditory epoch, and auditory–speech neurons modulated during both listening and production. The latter encode phonetic and sequence features across epochs, suggesting a role in transforming sensory representations into motor programs or internal monitoring. Notably, the timing of auditory‑specific responses parallels STG high‑gamma activity, preceding auditory–speech unit modulation, a pattern consistent with the presence of a hyperdirect basal ganglia pathway from auditory cortex 12 . Our finding that pooled STN single‑unit activity supports syllable‑level decoding during listening, suggests considering how best to interface with the basal‑ganglia as next generation speech neuroprosthetics and high-resolution closed‑loop brain stimulation systems are developed. While performance is modest, the presence of distributed phonetic information in STN during the auditory epoch expands the representational scope typically ascribed to this nucleus. These results also may explain why auditory encoding in the STN has been overlooked in fMRI studies. At the population level, opposing excitatory and inhibitory responses may cancel, rendering BOLD signals neutral despite rich underlying single-unit dynamics. Thus, the findings also highlight the role of single-unit electrophysiology in revealing specialized subcortical functions and encourage continued exploration of single-unit approaches to translational neuromodulation research. Beyond identifying auditory-responsive neurons, we observed parallel temporal components across STG and STN during listening, implying shared representational strategies within a cortical–subcortical loop. To synthesize these findings, we propose a schematic framework of dual cortical–subcortical loops through the STN that operate during speech listening. The first is an auditory input loop, in which auditory-specific STN units couple with STG activity predominantly in the alpha range, reflecting rapid sensory inflow from the auditory cortex to subthalamic circuits. The second is an auditory–motor transformation loop, in which auditory–speech units engage sensorimotor cortices via beta-band coupling, supporting the conversion of sensory representations into motor programs. In this framework, frequency-specific coupling mechanisms allow for parallel, functionally specific information processing within the basal ganglia–thalamocortical system. Declarations Ethics approval statement This study was approved by the University of Pittsburgh Institutional Review Board. Participant consent statement Participants consented to participate in this research and the publication of research results. References Hickok, G. Computational neuroanatomy of speech production. Nat Rev Neurosci 13 , 135-145, doi:10.1038/nrn3158 (2012). Rauschecker, J. P. Ventral and dorsal streams in the evolution of speech and language. Front Evol Neurosci 4 , 7, doi:10.3389/fnevo.2012.00007 (2012). Pasquereau, B. & Turner, R. S. A selective role for ventromedial subthalamic nucleus in inhibitory control. Elife 6 , doi:10.7554/eLife.31627 (2017). Alexander, G. E., DeLong, M. R. & Strick, P. L. Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annu Rev Neurosci 9 , 357-381, doi:10.1146/annurev.ne.09.030186.002041 (1986). Haber, S. N. Corticostriatal circuitry. Dialogues Clin Neurosci 18 , 7-21, doi:10.31887/DCNS.2016.18.1/shaber (2016). Nambu, A., Tokuno, H. & Takada, M. Functional significance of the cortico-subthalamo-pallidal 'hyperdirect' pathway. Neurosci Res 43 , 111-117, doi:10.1016/s0168-0102(02)00027-5 (2002). Mesgarani, N. & Chang, E. F. Selective cortical representation of attended speaker in multi-talker speech perception. Nature 485 , 233-236, doi:10.1038/nature11020 (2012). Overath, T., McDermott, J. H., Zarate, J. M. & Poeppel, D. The cortical analysis of speech-specific temporal structure revealed by responses to sound quilts. Nat Neurosci 18 , 903-911, doi:10.1038/nn.4021 (2015). Hickok, G. & Poeppel, D. The cortical organization of speech processing. Nat Rev Neurosci 8 , 393-402, doi:10.1038/nrn2113 (2007). Rauschecker, J. P. & Scott, S. K. Maps and streams in the auditory cortex: nonhuman primates illuminate human speech processing. Nat Neurosci 12 , 718-724, doi:10.1038/nn.2331 (2009). Chang, E. F. et al. Categorical speech representation in human superior temporal gyrus. Nat Neurosci 13 , 1428-1432, doi:10.1038/nn.2641 (2010). Jorge, A. et al. Hyperdirect connectivity of opercular speech network to the subthalamic nucleus. Cell Rep 38 , 110477, doi:10.1016/j.celrep.2022.110477 (2022). Arnold Anteraper, S. et al. Resting-State Functional Connectivity of the Subthalamic Nucleus to Limbic, Associative, and Motor Networks. Brain Connect 8 , 22-32, doi:10.1089/brain.2017.0535 (2018). Lipski, W. J. et al. Subthalamic nucleus neurons encode syllable sequence and phonetic characteristics during speech. bioRxiv , 2023.2012.2011.569290, doi:10.1101/2023.12.11.569290 (2023). Chrabaszcz, A. et al. Subthalamic nucleus and sensorimotor cortex activity during speech production. Journal of Neuroscience 39 , 2698-2708 (2019). Aron, A. R., Herz, D. M., Brown, P., Forstmann, B. U. & Zaghloul, K. Frontosubthalamic Circuits for Control of Action and Cognition. J Neurosci 36 , 11489-11495, doi:10.1523/jneurosci.2348-16.2016 (2016). Bergman, H. (MIT Press, 2021). Temel, Y., Blokland, A., Steinbusch, H. W. & Visser-Vandewalle, V. The functional role of the subthalamic nucleus in cognitive and limbic circuits. Progress in neurobiology 76 , 393-413 (2005). Haynes, W. I. & Haber, S. N. The organization of prefrontal-subthalamic inputs in primates provides an anatomical substrate for both functional specificity and integration: implications for Basal Ganglia models and deep brain stimulation. J Neurosci 33 , 4804-4814, doi:10.1523/jneurosci.4674-12.2013 (2013). Albin, R. L., Young, A. B. & Penney, J. B. The functional anatomy of basal ganglia disorders. Trends in neurosciences 12 , 366-375 (1989). Das, A. & Goldberg, J. H. Songbird subthalamic neurons project to dopaminergic midbrain and exhibit singing-related activity. Journal of Neurophysiology 127 , 373-383, doi:10.1152/jn.00254.2021 (2022). Litvak, V. et al. Resting oscillatory cortico-subthalamic connectivity in patients with Parkinson's disease. Brain 134 , 359-374, doi:10.1093/brain/awq332 (2011). Engel, A. K. & Fries, P. Beta-band oscillations--signalling the status quo? Curr Opin Neurobiol 20 , 156-165, doi:10.1016/j.conb.2010.02.015 (2010). Kühn, A. A. et al. Event-related beta desynchronization in human subthalamic nucleus correlates with motor performance. Brain 127 , 735-746, doi:10.1093/brain/awh106 (2004). Avants, B. B., Epstein, C. L., Grossman, M. & Gee, J. C. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Med Image Anal 12 , 26-41, doi:10.1016/j.media.2007.06.004 (2008). Merk, T. et al. Electrocorticography is superior to subthalamic local field potentials for movement decoding in Parkinson’s disease. eLife 11 , e75126, doi:10.7554/eLife.75126 (2022). Fogelson, N. et al. Different functional loops between cerebral cortex and the subthalmic area in Parkinson's disease. Cereb Cortex 16 , 64-75, doi:10.1093/cercor/bhi084 (2006). Hamilton, L. S., Edwards, E. & Chang, E. F. A Spatial Map of Onset and Sustained Responses to Speech in the Human Superior Temporal Gyrus. Curr Biol 28 , 1860-1871.e1864, doi:10.1016/j.cub.2018.04.033 (2018). Vissani, M. et al. Spike-phase coupling of subthalamic neurons to posterior perisylvian cortex predicts speech sound accuracy. Nature Communications 16 , 3357, doi:10.1038/s41467-025-58781-8 (2025). Chartier, J., Anumanchipalli, G. K., Johnson, K. & Chang, E. F. Encoding of Articulatory Kinematic Trajectories in Human Speech Sensorimotor Cortex. Neuron 98 , 1042-1054 e1044, doi:10.1016/j.neuron.2018.04.031 (2018). Moses, D. A. et al. Neuroprosthesis for decoding speech in a paralyzed person with anarthria. New England Journal of Medicine 385 , 217-227 (2021). Additional Declarations There is NO Competing Interest. Supplementary Files SoundinducedSTNactivitySupplv10312025.docx Human subthalamic neurons encode speech features during listening and couple with auditory cortex Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8008692","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":539462827,"identity":"382c38f7-9b86-4d51-89a3-0aaffe49332c","order_by":0,"name":"Yanming Zhu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABHUlEQVRIie2NMUsDMRTH3xFol2DXHMqdHyEuEengV7EU7pZUhG6ipdO5HM6Cfoi6OThEDu1y5daIDkoh020uVm4wl7NbTjoK5jfkvby8X/4ADsdf5ag+CHhvAAJoM0Pr+a8Kopsr0CgdspFC54vH2fvdOYTXKTuVX6+w310osrrvQ6/LqVXJjyM5yOfg3eTsZXSp4CCNmZ+qGPy0tCuCMzlIngARrpU005MIUSx0I1tSirJROloZG6VQaK/SymGbIk3KGWCtIP5Zfx55S5NC7IovTYrAhETj7dE0w1Qqb7kjYkxydWJTtgrOnlfJJAivhrcfvMoCWkTwUIp+0LsYzmzKrjAlw6Z4CeD1C7at14RTUyY/16ptz+FwOP4z36VSag36lZ48AAAAAElFTkSuQmCC","orcid":"","institution":"Massachusetts General Hospital","correspondingAuthor":true,"prefix":"","firstName":"Yanming","middleName":"","lastName":"Zhu","suffix":""},{"id":539462828,"identity":"22b35e78-7d5b-448a-b6bf-1298b12ad079","order_by":1,"name":"Robert Richardson","email":"","orcid":"https://orcid.org/0000-0003-2620-7387","institution":"Massachusetts General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Robert","middleName":"","lastName":"Richardson","suffix":""},{"id":539462829,"identity":"5adedf62-bc75-4383-9fa9-28174d24b966","order_by":2,"name":"Alan Bush","email":"","orcid":"https://orcid.org/0000-0002-0407-5750","institution":"Massachusetts General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Alan","middleName":"","lastName":"Bush","suffix":""},{"id":539462830,"identity":"77d29056-9846-4bc8-8e9d-7708268488a7","order_by":3,"name":"Matteo Vissani","email":"","orcid":"https://orcid.org/0000-0002-3908-7620","institution":"Massachusetts General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Matteo","middleName":"","lastName":"Vissani","suffix":""},{"id":539462831,"identity":"0e3ffd97-01f7-43a7-9bf4-a7998408506b","order_by":4,"name":"Latane Bullock","email":"","orcid":"","institution":"Harvard University","correspondingAuthor":false,"prefix":"","firstName":"Latane","middleName":"","lastName":"Bullock","suffix":""}],"badges":[],"createdAt":"2025-11-02 03:55:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8008692/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8008692/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":95603742,"identity":"2250c430-548c-4b5d-ad24-7868904156e2","added_by":"auto","created_at":"2025-11-11 06:33:16","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":212498,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eIntraoperative STN Single Unit and ECoG recording during auditory-speech triplet task.\u003c/strong\u003e \u003cstrong\u003e(A)\u003c/strong\u003e A representative recording during a single trial of the syllable triplet repetition task shows (top to bottom) the simultaneous microelectrode recording, spike activities, representative sorted single units, the stimulus, and produced audio waveforms. Participants repeated triplets of consonant-vowel syllables aloud, presented as auditory stimuli through earphones. \u003cstrong\u003e(B)\u003c/strong\u003e Response time across subjects. The orange and green shaded areas represent the auditory and speech epochs, respectively. Each blue dot is one speech onset in a trial across all participants. \u003cstrong\u003e(C)\u003c/strong\u003e Recording locations in the STN were warped to the MNI (ICBM 2009b Nlin Asym) template. D: dorsal; M: medial; P: posterior; L: lateral; I: inferior; P: posterior. \u003cstrong\u003e(D)\u003c/strong\u003e Recording locations on the cortex were warped to the MNI template. \u003cstrong\u003e(E)\u003c/strong\u003e Left: STN LFP time-frequency analysis time is locked to the auditory epoch (triplet) onset for a representative unit, showing desynchronization around the beta frequency band. Green lines indicate auditory onset, and the grey shaded areas indicate auditory epoch. Right: Desynchronization at the beta frequency band. The spectrum power is z-scored relative to the -1 to 0 seconds before the auditory onset. One-way ANOVA comparing the instantaneous power to the power of the -1 to 0 seconds before the auditory onset, with False Discovery Rate correction. \u003cstrong\u003e(F)\u003c/strong\u003eLeft: STN spikes time-frequency analysis time is locked to auditory epoch onset for a representative unit, showing desynchronization around the beta frequency band. Green lines indicate auditory onset, and the grey shaded areas indicate auditory epoch. Right: Desynchronization at the beta frequency band. The spectrum power is z-scored relative to the -1 to 0 second before the auditory onset. One-way ANOVA comparing the instantaneous power to the power of the -1 to 0 seconds before the auditory onset, with False Discovery Rate correction. \u003cstrong\u003e(G)\u003c/strong\u003eLocations of auditory-locked beta desynchronization units in MNI space.\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8008692/v1/6c1c190ea647eb9493f2bd25.jpg"},{"id":95656135,"identity":"9d95f4d5-eaee-4133-b4e2-a97f9496face","added_by":"auto","created_at":"2025-11-11 16:17:50","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":323259,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSTN single unit firing rates show selective modulation during listening.\u003c/strong\u003e \u003cstrong\u003e(A)\u003c/strong\u003e The time course of the firing rate modulation for all recorded units is hierarchically clustered by Euclidean distance with complete linkage. The grey shaded area indicates the auditory epoch. The color bar indicates the z-scored firing rate related to the -1.5 to -0.5 second prior to the auditory onset. \u003cstrong\u003e(B)\u003c/strong\u003e Example of units in different response categories: positively modulated (a\u003csup\u003e+\u003c/sup\u003e), not modulated (a\u003csup\u003e0\u003c/sup\u003e), and negatively modulated (a\u003csup\u003e-\u003c/sup\u003e) units during the auditory window are shown in the first, second, and third row, respectively. Positively modulated (s\u003csup\u003e+\u003c/sup\u003e), not modulated (s\u003csup\u003e0\u003c/sup\u003e), and negatively modulated (s\u003csup\u003e-\u003c/sup\u003e) units during speech are shown in the first, second, and third columns, respectively. Both the spike raster plot and average firing rate (Mean ± SEM) are shown here. Green vertical lines indicate auditory syllable onset, and blue vertical lines indicate speech syllable onset. Color-shaded areas indicate significant modulation compared to the resting state. Green, grey, and red firing rate and raster plot indicate a\u003csup\u003e+\u003c/sup\u003e, a\u003csup\u003e0\u003c/sup\u003e, and a\u003csup\u003e-\u003c/sup\u003e units, respectively. \u003cstrong\u003e(C)\u003c/strong\u003e Number of units in each category. Green, grey, and red firing rate and raster plot indicate a\u003csup\u003e+\u003c/sup\u003e, a\u003csup\u003e0\u003c/sup\u003e, and a\u003csup\u003e-\u003c/sup\u003e units, respectively. \u003cstrong\u003e(D)\u003c/strong\u003e Recording locations of the auditory-specific units and auditory-speech units. \u003cstrong\u003e(E)\u003c/strong\u003e Venn diagram showing the number of units positively/negatively modulated by auditory stimuli (a\u003csup\u003e+\u003c/sup\u003e or a\u003csup\u003e-\u003c/sup\u003e) or not modulated (a\u003csup\u003e0\u003c/sup\u003e) in relation to the modulation during speech production and beta desynchronization.\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8008692/v1/50498517a65ce8f0db4cc1b6.jpg"},{"id":95603739,"identity":"461e3075-814b-4254-9d4e-c5f78e6441f2","added_by":"auto","created_at":"2025-11-11 06:33:16","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":292745,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDuring listening, a subset of STN units encodes consonants, vowels, and sequence position.\u003c/strong\u003e \u003cstrong\u003e(A)\u003c/strong\u003e Examples of mean firing rate plots (Mean ± SEM) indicating consonant (top), vowel (middle), and ordinal (bottom) encoding. \u003cstrong\u003e(B)\u003c/strong\u003e A linear regression model was used to predict the mean firing rate during auditory presentation based on phonetic and sequence features of the syllable. Example of a presentation audio waveform and spectrogram (top) during a single trial and the spike density function for a simultaneously recorded STN neuron (middle). The mean firing rate within a 200-ms window centered on the consonant-vowel transition of each syllable was paired with a feature vector describing that syllable (bottom). \u003cstrong\u003e(C)\u003c/strong\u003e Examples of the coefficient of the linear regression model (left y-axis and grey bar) and partial R\u003csup\u003e2\u003c/sup\u003e with the Leave-one-out method (right y-axis and orange line, star indicates the significant feature) to determine the contribution of the feature in the linear regression model. \u003cstrong\u003e(D)\u003c/strong\u003e Summary of syllable-units’ encoded syllable feature. Green cells indicated that the auditory and speech epochs encoded the same syllable feature for this unit. Red cells indicated that the auditory and speech epochs encoded different syllable features for this unit. \u003cstrong\u003e(E)\u003c/strong\u003e Recording locations of the auditory syllable feature encoding units. \u003cstrong\u003e(F)\u003c/strong\u003e Recording locations of the speech syllable feature encoding units. \u003cstrong\u003e(G)\u003c/strong\u003e Response selectivity for all syllable feature encoding units (measures how differentiable each syllable type is at the given time point, with one-way ANOVA comparing the firing rate of each unit in response to each consonant or vowel). Grey-shaded areas indicate significantly different consonants or vowels; blue lines are consonant encoding units, and purple lines are vowel encoding units; the color is white if the p-value is \u0026gt; 0.05. \u003cstrong\u003e(H)\u003c/strong\u003e \u0026nbsp;Decoding accuracy of syllable identity from STN firing during the auditory epoch. Bars show a 12‑way elastic‑net multinomial logistic regression decoder (blue) and a factorized decoder formed by multiplying probabilities from separate consonant (4‑way) and vowel (3‑way) models (orange). Values are mean ± s.d. across 5‑fold outer cross‑validation; the dashed line indicates chance (1/12). Results are shown for each single unit (1–10) and for the combined multi‑unit population (“Combined”). \u003cstrong\u003e(I)\u003c/strong\u003e Venn diagram showing the number of auditory syllable-encoding units in relation to the beta desynchronization units.\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8008692/v1/3ae3b19c4c99912f746b2f5b.jpg"},{"id":95656418,"identity":"bca01fa6-c51b-40ac-890b-892af792c705","added_by":"auto","created_at":"2025-11-11 16:18:40","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":259223,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSTN and STG activity during listening share temporal components and frequency‑specific coupling.\u003c/strong\u003e (\u003cstrong\u003eA\u003c/strong\u003e) Unsupervised non-negative matrix factorization (NMF) with two components for STG high gamma activity shows two electrode types. (\u003cstrong\u003eB\u003c/strong\u003e) Scatter plot of the two-component NMF labeled with electrode types (onset: blue, sustained: orange, and syllable/triplet: green). C1 and C2 are component weights for components 1 and 2. (\u003cstrong\u003eC\u003c/strong\u003e) Time-resolved NMF with two components for STN firing rate. (\u003cstrong\u003eD\u003c/strong\u003e) Time-resolved NMF with four components for STN firing rate. (\u003cstrong\u003eE\u003c/strong\u003e) Correlation matrices with cosine similarity for STG high gamma activity with STN firing rate in auditory or speech epochs. The four correlation matrices share the same color bar on the right. (\u003cstrong\u003eF\u003c/strong\u003e) STN spike-cortical phase locking across frequency bands (alpha, beta, and high gamma) and cortical regions during auditory epochs. (\u003cstrong\u003eG\u003c/strong\u003e) Auditory-specific (a\u003csup\u003e*\u003c/sup\u003e/s0) units’ spike-STG phase locking in the alpha band in the temporal domain. (\u003cstrong\u003eH\u003c/strong\u003e) Auditory-speech (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e*\u003c/sup\u003e) units’ spike-post central gyrus phase locking in the beta band. \u003cstrong\u003e(I)\u003c/strong\u003e Spatial overlap between auditory-specific units versus the macro-stimulation site that elicited STG evoked potential. \u003cstrong\u003e(J)\u003c/strong\u003e Averaged response time course of auditory-specific units, auditory-speech units, and STG ECoG time-aligned to the auditory onset. Blue and orange shaded areas indicate time periods in which the STN units' z-scored activity was significantly different from STG ECoG activity.\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8008692/v1/a63ebdad38ca4b87b2569c8d.jpg"},{"id":96250311,"identity":"40debd6b-3a69-4082-b102-3dc6f1258776","added_by":"auto","created_at":"2025-11-19 07:38:05","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1825054,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8008692/v1/94ef0fba-23e3-49a1-871f-3c551153502d.pdf"},{"id":95603743,"identity":"009a611b-eb13-4ae7-a891-65e5a85b9258","added_by":"auto","created_at":"2025-11-11 06:33:16","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":5314385,"visible":true,"origin":"","legend":"Human subthalamic neurons encode speech features during listening and couple with auditory cortex","description":"","filename":"SoundinducedSTNactivitySupplv10312025.docx","url":"https://assets-eu.researchsquare.com/files/rs-8008692/v1/25c84066f692c6c7b58c1755.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Human subthalamic neurons encode speech features during listening and couple with auditory cortex","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe integration of auditory and motor systems enables rapid, context‑appropriate responses to sound \u003csup\u003e1,2\u003c/sup\u003e. While the basal ganglia have long been linked to action selection and stopping \u003csup\u003e3-6\u003c/sup\u003e, most human auditory perception research has emphasized cortical processing \u003csup\u003e7-11\u003c/sup\u003e, leaving the contribution of subcortical nodes during listening insufficiently characterized \u003csup\u003e12,13\u003c/sup\u003e. Likewise, intracranial investigations of human basal ganglia function have focused on production \u003csup\u003e14,15\u003c/sup\u003e, without yet considering activity during perception.\u003c/p\u003e\n\u003cp\u003eThe subthalamic nucleus (STN) may participate in rapid sensory-motor integration and motor preparation following auditory input \u003csup\u003e16\u003c/sup\u003e. A small, almond-shaped structure within the basal ganglia, the STN regulates motor control, emotion, and cognition. It receives monosynaptic input from the cortex via the hyperdirect pathway, and other cortical input indirectly, via the striatum and external globus pallidus (GPe), before sending glutamatergic input to the internal globus pallidus (GPi) and substantia nigra pars reticulata (SNr), which project to the ventral thalamus and onward to frontal cortical regions \u003csup\u003e17-20\u003c/sup\u003e. STN hyperdirect innervation traditionally has been described as originating from the frontal lobe, yet songbirds have a population of neurons in the auditory cortex with a monosynaptic projection to the songbird STN \u003csup\u003e21\u003c/sup\u003e, and resting-state functional MRI data suggests connections between STN and superior temporal gyrus (STG) \u003csup\u003e13\u003c/sup\u003e. Indeed, we recently described direct evidence in the human brain for a monosynaptic hyperdirect pathway from auditory areas of the STG to the STN \u003csup\u003e12\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe hypothesized that during listening, STN single‑unit activity would be selectively modulated by auditory features, and that STN\u0026ndash;cortex coupling would exhibit frequency‑specific organization consistent with loop circuit‑level routing \u003csup\u003e22\u003c/sup\u003e. Leveraging intraoperative access during awake deep brain stimulation (DBS) lead implantation, we tested these predictions in a syllable‑triplet task designed to separate auditory and speech epochs. We found that subsets of STN neurons respond to auditory stimuli, encode syllable features, and exhibit task-specific beta desynchronization and phase-coupling with auditory and sensorimotor cortices \u003csup\u003e23,24\u003c/sup\u003e. These findings challenge the classical motor-centric view of the BG and position the STN as a computational hub for auditory encoding.\u003c/p\u003e"},{"header":"STN single unit and ECoG recording during a syllable triplet repetition task","content":"\u003cp\u003eTo study the STN\u0026rsquo;s role in auditory-motor integration, we recorded single-unit STN activity during an intraoperative speech production task. Twenty-five PD patients undergoing DBS surgery completed a total of 60 task runs. Simultaneous electrocorticography (ECoG) was recorded from subdural electrode strips temporarily placed through the standard bone opening used for DBS lead implantation (Suppl Table 1)\u003c/p\u003e\n\u003cp\u003eParticipants listened to triplets of consonant-vowel syllables (e.g., /va/, /ti/, /su/, /ga/) via earphones and repeated them aloud. Stimuli were balanced across four consonants (/v/, /t/, /s/, /g/) and three vowels (/a/, /i/, /u/), presented at normal and high intensities (+25 dB). Figure 1A illustrates a trial with synchronized STN microelectrode recordings, sorted single-unit spikes, ECoG local field potentials (LFPs), and audio waveforms of both stimuli and responses.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eParticipants responded consistently, with an average latency of 2.2 \u0026plusmn; 0.3 seconds (Fig. 1B). Each trial began with a one-second baseline period, followed by an auditory presentation lasting 1.5 seconds (each syllable onset is 0.5 seconds apart), and concluded with self-paced repetition of the syllable sequence in a continuum. We recorded 205 single units across 25 participants (120 tracks, 59 sessions). Microelectrode and ECoG recording sites were normalized to MNI space via non-linear warping for group-level analysis (Fig. 1C\u0026ndash;D) \u003csup\u003e25\u003c/sup\u003e. The motor subregion of the STN was defined using the DISTAL-minimal atlas, an MRI-based anatomical atlas frequently utilized in clinical and research contexts to define subregions within the STN. Microelectrode recordings predominantly sampled the motor subregion of the STN, and ECoG primarily sampled IFG, ventral pre- and post-central gyrus, supramarginal gyrus, and superior temporal gyrus.\u003c/p\u003e\n\u003cp\u003eTime-frequency analyses of STN macroelectrode LFPs (Fig. 1E and Suppl. Fig. 1)\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003erevealed beta-band (13\u0026ndash;30 Hz) desynchronization (significant reductions in beta power compared to the pre-auditory stimuli baseline, for more than 100 ms, FDR corrected one-way ANOVA) following auditory stimulus onset in 44 of 222 electrodes (20%) (Fig. 1E, right).\u003c/p\u003e\n\u003cp\u003eSimilarly, we analyzed single-unit recordings time-locked to auditory onset (Fig. 1F). Fifteen of 205 units (7%) exhibited a significant reduction of beta-band entrained firing (Fig. 1F, right, one-way ANOVA with FDR compared to resting state). These cells were relatively co-localized within the STN (Fig. 1G), as assessed by the average pairwise distance compared to that of random samples of recording sites in the STN (Suppl Fig. 2). Together, these findings show that STN activity is reliably modulated during listening at both the LFP and single‑unit levels, beyond baseline fluctuations (FDR‑controlled comparisons).\u003c/p\u003e"},{"header":"STN single unit firing rate exhibits selective modulation during audition","content":"\u003cp\u003eHierarchical clustering based on trial-averaged activity of the auditory epoch revealed units with positive, negative, or no firing rate modulation; each line in Fig. 2A is one unit\u0026rsquo;s averaged firing rate aligned to the auditory onset. Activity was z-scored relative to baseline, and units were categorized as modulated if they differed significantly from baseline for more than 100 ms (FDR corrected). We identified nine categories, reflecting combinations of positive (\u003csup\u003e+\u003c/sup\u003e), negative (\u003csup\u003e-\u003c/sup\u003e), or no modulation (\u003csup\u003e0\u003c/sup\u003e), during auditory (a) and speech (s) epochs (Fig. 2B and Suppl Fig. 3).Figure 2B shows example units that respond selectively to audition, to speech, to both, or to neither: units are grouped in rows by their response \u003cem\u003eduring the auditory epoch\u003c/em\u003e: increased firing (a\u003csup\u003e+\u003c/sup\u003e), no change (a\u003csup\u003e0\u003c/sup\u003e), or decreased firing (a\u003csup\u003e\u0026minus;\u003c/sup\u003e), and in columns by their response \u003cem\u003eduring the speech epoch\u003c/em\u003e: increased firing (s\u003csup\u003e+\u003c/sup\u003e), no change (s\u003csup\u003e0\u003c/sup\u003e), or decreased firing (s\u003csup\u003e\u0026minus;\u003c/sup\u003e). Each panel shows both the spike raster and the mean firing rate (\u0026plusmn;SEM) aligned to auditory (green line) and speech (blue line) onsets. Shaded regions mark periods of significant modulation relative to baseline. For example, a\u003csup\u003e+\u003c/sup\u003e/s\u003csup\u003e-\u003c/sup\u003e denotes units that have increased FR during the auditory window and decreased FR during speech. These firing rate and spike raster plots illustrate that auditory modulation can occur independently of speech modulation, can co-occur with speech modulation in the same units, or can be absent altogether.\u003c/p\u003e\n\u003cp\u003eWe refer to units that were only modulated during either the auditory or speech epoch as auditory-specific (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e) or speech-specific (a\u003csup\u003e0\u003c/sup\u003e/s\u003csup\u003e*\u003c/sup\u003e) units (asterisks represent either positive or negative significant modulation). We refer to units modulated during both auditory and speech epochs as auditory-speech (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e*\u003c/sup\u003e) units. We identified 20% (41/205) a\u003csup\u003e+\u003c/sup\u003e and 19% (39/205) a\u003csup\u003e-\u003c/sup\u003e units out of the 205 a* unites\u0026nbsp;that responded with increases or decreases in firing rates, respectively, at the auditory epoch onset (Fig. 2C). Among the a\u003csup\u003e*\u003c/sup\u003e units, five a\u003csup\u003e+\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e (12%) and ten a\u003csup\u003e-\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e (26%) units responded exclusively to the auditory presentation (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e). These unit types were spatially aggregated within the STN, where we calculated the average distance of the units within the group, then compared it to a random sample of STN recording sites\u0026rsquo; average distance with permutations (Fig. 2D and Suppl Fig. 5).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBecause the task required subsequent repetition, we considered motor preparation as a potential confound. Aligning STN activity to non‑task auditory events (ambient operating‑room sounds and clinician speech) showed that two auditory‑specific (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e) units maintained their response (Suppl Fig. 6), suggesting genuine auditory sensitivity in at least a subset of neurons, although this control is limited in scope. Consistent with a motor integration role, beta‑desynchronization units were predominantly auditory\u0026ndash;speech (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e*\u003c/sup\u003e) or speech‑only (a\u003csup\u003e0\u003c/sup\u003e/s\u003csup\u003e*\u003c/sup\u003e) (13/15 and 2/15, respectively). (Fig. 2E).\u003c/p\u003e"},{"header":"STN Units Encode Phonetic and Sequence Features During Audition","content":"\u003cp\u003eWe previously reported STN LFP high-gamma activity and single-unit firing rate modulation specific to phonetic and sequence features during the production of an utterance \u003csup\u003e14,15\u003c/sup\u003e. Here, we assessed if auditory features prior to speech onset also modulate STN single-unit firing rate, by analyzing spike-triggered averages (Fig. 3A). Linear models relating firing rates to syllable features confirmed that certain auditory-speech (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e*\u003c/sup\u003e) units encoded phonetic or sequence information (Fig. 3B\u0026ndash;C). Leave-one-out method was used to determine which feature contributed significantly to the linear model. In this linear model, the consonants and vowels are compared to the /v/ and /a/. With these two methods, we identified five, seven, and one units that encoded vowel, consonant, and sequence features, respectively\u0026nbsp;(Fig. 3D). These \u0026ldquo;auditory-syllable units\u0026rdquo; (13/205, ~6 %) were spatially clustered within the STN (Fig. 3E\u0026ndash;F and Suppl Fig. 7).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eInterestingly, we found that the auditory-syllable units not only encode syllable features in the auditory epoch but also encode syllable features in the speech epoch, measured by comparing the firing rate between these features to the resting state among all these types of units. (Fig. 3G and Suppl Fig. 8).\u0026nbsp;Multiple regressions using the mean firing rate at different lags (with a 200 ms window) were performed to look for syllable encoding during the auditory and speech epochs among all auditory-syllable units (Suppl Fig. 9). We found that syllable-encoding units also could encode syllable features during speech (Suppl Fig. 9).\u0026nbsp;10/12 auditory syllable encoding units encoded the same or a subset of syllables that the same unit encoded during the speech epoch, suggesting that these units support auditory-speech integration.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo assess whether syllables can be decoded from STN single unit activities, we performed nested cross‑validated elastic‑net multinomial logistic regression \u003csup\u003e26\u003c/sup\u003e. Two decoding strategies were compared: a direct 12-class model trained to predict the full consonant\u0026ndash;vowel combination, and a factorized approach in which separate consonant (4-way) and vowel (3-way) classifiers were trained and their output probabilities combined to recover the 12-class prediction. Consistent with the observation that most neurons encoded only a subset of phonetic features, single-unit classifiers performed near chance. In contrast, pooling features from all ten auditory-syllable units that also showed speech encoding (removing units without enough repetition) markedly improved performance, yielding modest but reliable above-chance decoding (12-way mean accuracy\u0026nbsp;\u0026asymp;\u0026nbsp;0.41; factorized\u0026nbsp;\u0026asymp;\u0026nbsp;0.39; chance = 1/12, Fig. 3H). These results suggest that while individual STN units carry limited information about syllable identity, the pooled population supports distributed encoding of consonant and vowel features.\u003c/p\u003e\n\u003cp\u003eAuditory-specific (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e) units did not encode syllable features (Suppl. Fig. 18), a feature that was found in five of thirteen beta desynchronization speech units (Fig. 3I). We identified only one unit modulated by stimulus volume (Suppl Fig. 10). Thus, we found that STN units can encode speech features\u0026apos; phonemes (consonant and vowel), sequence order, and volume.\u003c/p\u003e"},{"header":"STN connectivity with and similarities to STG activity.","content":"\u003cp\u003eGiven the STG\u0026rsquo;s role in human auditory processing, we compared temporal response structure across STG and STN using non‑negative matrix factorization (NMF) to ask whether similar auditory components appear across the loop circuit \u003csup\u003e27\u003c/sup\u003e. Previous studies have shown that NMF can separate STG responses to auditory stimuli into two types: onset and sustained \u003csup\u003e28\u003c/sup\u003e. Similar to this result, we found STG auditory modulation can be differentiated in these two categories (Fig. 4A and Suppl Fig. 11).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAdditionally, a third ECoG response type (syllable/triplet) captured the triplet nature of stimuli (Fig. 4B and Suppl Fig. 12), indicating that two components alone are not sufficient to explain the LFP dynamics in our data. Thus, we performed a percentage of variance explained analysis of the components, which showed that four components could explain ~90% of the variance in LFP modulation (Suppl Fig. 13); subsequent NMF analyses used four components, as shown in Suppl Fig. 13.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe applied the same method to the STN firing rate with two components and found that the components did not recapitulate STG HG activity (Fig. 4C, sustained onset and sustained depressed type). Onset, triplet-like, and two triplet-like components (in anti-phase to each other), however, were identified when using four components (Fig. 4D). STN LFPs showed no correspondence (Suppl Fig. 14).\u003c/p\u003e\n\u003cp\u003eTo better appreciate the similarity of auditory modulation between the STG and STN, we applied cosine similarity matrices. STG LFP and STN firing rate modulation NMF components were correlated strongly during the auditory epoch but not during speech (Fig. 4E). STG LFP and STN firing rate modulation were not cross-correlated between the auditory epoch and the speech epochs. (Fig. 4E).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo further characterize STN\u0026ndash;cortex interactions during listening, we analyzed cortical-STN spike-phase coupling (SPC) during the auditory epoch, as described by Vissani et al. \u003csup\u003e29\u003c/sup\u003e. Auditory-specific (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e) units demonstrated strong SPC events with broad cortical areas in the alpha range but not in other frequency ranges. Auditory-speech (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e*\u003c/sup\u003e) units showed SPC with subcentral, postcentral, and STG predominantly in the beta band (Fig. 4F and Suppl Fig. 15).\u0026nbsp;SPC in the high gamma bands was weak, not modulated in auditory or speech epochs \u003csup\u003e29\u003c/sup\u003e and did not show an STN unit type or cortical region selectivity (Fig. 4F). We noted a dissociation in the frequency bands of SPC according to the STN unit type; auditory-specific (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e) cells exhibited SPC predominantly in alpha, whereas auditory-speech (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e*\u003c/sup\u003e) units exhibited SPC predominantly in beta.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe further explored how STN-STG SPC evolved over the course of the auditory stimuli across the unit types. We found that the SPC in the alpha band for auditory-specific (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e) cells was greatly reduced during the auditory epoch (Fig. 4G). For the a\u003csup\u003e+\u003c/sup\u003e/s\u003csup\u003e0\u0026nbsp;\u003c/sup\u003eunits, SPC recovered in the speech epoch, but for a\u003csup\u003e-\u003c/sup\u003e/s\u003csup\u003e0\u0026nbsp;\u003c/sup\u003eunits, despite a transient increase after the auditory epoch offset, SPC remained at very low values during speech. For auditory-speech (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e*\u003c/sup\u003e) units, there is a partial loss of beta-band SPC with sensory-motor cortical regions during auditory presentation and an almost complete loss during early speech production (Fig. 4H). Generally, SPC modulations did not overlap with their FR modulation in the temporal domain, suggesting parallel mechanisms for encoding auditory information. These results suggest the presence of an STG-STN auditory input loop and an STN-sensorimotor cortex auditory\u0026ndash;motor transformation loop.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSTN macrostimulation during simultaneous ECoG recordings was performed to investigate the functional connectivity between the two structures by analyzing evoked potentials, as described previously \u003csup\u003e12\u003c/sup\u003e. We reanalyzed the data in Jorge et al. with specific attention to STG responses (Suppl Fig. 16). Among all effective STN stimulation locations (STN\u003csup\u003eStim\u003c/sup\u003e), those that induce evoked potentials in the STG (STN\u003csup\u003eStim-STG\u003c/sup\u003e) were significantly closer to auditory-specific (a\u003csup\u003e*\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e) units than expected by chance (STG-STN pair, Fig. 4I and Suppl Fig. 17).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe time-aligned the average response of auditory-specific (a\u003csup\u003e+\u003c/sup\u003e/s\u003csup\u003e0\u003c/sup\u003e) units, auditory-speech (a\u003csup\u003e+\u003c/sup\u003e/s\u003csup\u003e*\u003c/sup\u003e) units, and STG ECoG auditory responding electrodes to the onset of the auditory epochs. We found that the response onset of auditory-specific units (177 ms after auditory onset) was similar to that of the STG gamma response (167 ms after auditory onset) but faster than that of auditory-speech units (374 ms after auditory onset, Fig. 4J), suggesting that auditory-specific units may receive monosynaptic input from the STG.\u003c/p\u003e\n\u003cp\u003eTaken together, these data are consistent with hyperdirect projections linking auditory and sensorimotor cortices to STN that may facilitate sensory\u0026ndash;motor integration.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study provides convergent evidence that the human STN participates in auditory\u0026ndash;motor integration during listening. We identify two novel functional single‑unit classes: auditory‑specific neurons that respond selectively during the auditory epoch, and auditory\u0026ndash;speech neurons modulated during both listening and production. The latter encode phonetic and sequence features across epochs, suggesting a role in transforming sensory representations into motor programs or internal monitoring. Notably, the timing of auditory‑specific responses parallels STG high‑gamma activity, preceding auditory\u0026ndash;speech unit modulation, a pattern consistent with the presence of a hyperdirect basal ganglia pathway from auditory cortex \u003csup\u003e12\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eOur finding that pooled STN single‑unit activity supports syllable‑level decoding during listening, suggests considering how best to interface with the basal‑ganglia as next generation speech neuroprosthetics and high-resolution closed‑loop brain stimulation systems are developed. While performance is modest, the presence of distributed phonetic information in STN during the auditory epoch expands the representational scope typically ascribed to this nucleus. These results also may explain why auditory encoding in the STN has been overlooked in fMRI studies. At the population level, opposing excitatory and inhibitory responses may cancel, rendering BOLD signals neutral despite rich underlying single-unit dynamics. Thus, the findings also highlight the role of single-unit electrophysiology in revealing specialized subcortical functions and encourage continued exploration of single-unit approaches to translational neuromodulation research.\u003c/p\u003e\n\u003cp\u003eBeyond identifying auditory-responsive neurons, we observed parallel temporal components across STG and STN during listening, implying shared representational strategies within a cortical\u0026ndash;subcortical loop. To synthesize these findings, we propose a schematic framework of dual cortical\u0026ndash;subcortical loops through the STN that operate during speech listening. The first is an auditory input loop, in which auditory-specific STN units couple with STG activity predominantly in the alpha range, reflecting rapid sensory inflow from the auditory cortex to subthalamic circuits. The second is an auditory\u0026ndash;motor transformation loop, in which auditory\u0026ndash;speech units engage sensorimotor cortices via beta-band coupling, supporting the conversion of sensory representations into motor programs. In this framework, frequency-specific coupling mechanisms allow for parallel, functionally specific information processing within the basal ganglia\u0026ndash;thalamocortical system.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthics approval statement This study was approved by the University of Pittsburgh Institutional Review Board.\u003c/p\u003e\n \u003cp\u003eParticipant consent statement Participants consented to participate in this research and the publication of research results.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eHickok, G. Computational neuroanatomy of speech production. \u003cem\u003eNat Rev Neurosci\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 135-145, doi:10.1038/nrn3158 (2012).\u003c/li\u003e\n\u003cli\u003eRauschecker, J. P. Ventral and dorsal streams in the evolution of speech and language. \u003cem\u003eFront Evol Neurosci\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 7, doi:10.3389/fnevo.2012.00007 (2012).\u003c/li\u003e\n\u003cli\u003ePasquereau, B. \u0026amp; Turner, R. S. A selective role for ventromedial subthalamic nucleus in inhibitory control. \u003cem\u003eElife\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, doi:10.7554/eLife.31627 (2017).\u003c/li\u003e\n\u003cli\u003eAlexander, G. E., DeLong, M. R. \u0026amp; Strick, P. L. Parallel organization of functionally segregated circuits linking basal ganglia and cortex. \u003cem\u003eAnnu Rev Neurosci\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 357-381, doi:10.1146/annurev.ne.09.030186.002041 (1986).\u003c/li\u003e\n\u003cli\u003eHaber, S. N. Corticostriatal circuitry. \u003cem\u003eDialogues Clin Neurosci\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 7-21, doi:10.31887/DCNS.2016.18.1/shaber (2016).\u003c/li\u003e\n\u003cli\u003eNambu, A., Tokuno, H. \u0026amp; Takada, M. Functional significance of the cortico-subthalamo-pallidal \u0026apos;hyperdirect\u0026apos; pathway. \u003cem\u003eNeurosci Res\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, 111-117, doi:10.1016/s0168-0102(02)00027-5 (2002).\u003c/li\u003e\n\u003cli\u003eMesgarani, N. \u0026amp; Chang, E. F. Selective cortical representation of attended speaker in multi-talker speech perception. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e485\u003c/strong\u003e, 233-236, doi:10.1038/nature11020 (2012).\u003c/li\u003e\n\u003cli\u003eOverath, T., McDermott, J. H., Zarate, J. M. \u0026amp; Poeppel, D. The cortical analysis of speech-specific temporal structure revealed by responses to sound quilts. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 903-911, doi:10.1038/nn.4021 (2015).\u003c/li\u003e\n\u003cli\u003eHickok, G. \u0026amp; Poeppel, D. The cortical organization of speech processing. \u003cem\u003eNat Rev Neurosci\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 393-402, doi:10.1038/nrn2113 (2007).\u003c/li\u003e\n\u003cli\u003eRauschecker, J. P. \u0026amp; Scott, S. K. Maps and streams in the auditory cortex: nonhuman primates illuminate human speech processing. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 718-724, doi:10.1038/nn.2331 (2009).\u003c/li\u003e\n\u003cli\u003eChang, E. F.\u003cem\u003e et al.\u003c/em\u003e Categorical speech representation in human superior temporal gyrus. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1428-1432, doi:10.1038/nn.2641 (2010).\u003c/li\u003e\n\u003cli\u003eJorge, A.\u003cem\u003e et al.\u003c/em\u003e Hyperdirect connectivity of opercular speech network to the subthalamic nucleus. \u003cem\u003eCell Rep\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 110477, doi:10.1016/j.celrep.2022.110477 (2022).\u003c/li\u003e\n\u003cli\u003eArnold Anteraper, S.\u003cem\u003e et al.\u003c/em\u003e Resting-State Functional Connectivity of the Subthalamic Nucleus to Limbic, Associative, and Motor Networks. \u003cem\u003eBrain Connect\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 22-32, doi:10.1089/brain.2017.0535 (2018).\u003c/li\u003e\n\u003cli\u003eLipski, W. J.\u003cem\u003e et al.\u003c/em\u003e Subthalamic nucleus neurons encode syllable sequence and phonetic characteristics during speech. \u003cem\u003ebioRxiv\u003c/em\u003e, 2023.2012.2011.569290, doi:10.1101/2023.12.11.569290 (2023).\u003c/li\u003e\n\u003cli\u003eChrabaszcz, A.\u003cem\u003e et al.\u003c/em\u003e Subthalamic nucleus and sensorimotor cortex activity during speech production. \u003cem\u003eJournal of Neuroscience\u003c/em\u003e \u003cstrong\u003e39\u003c/strong\u003e, 2698-2708 (2019).\u003c/li\u003e\n\u003cli\u003eAron, A. R., Herz, D. M., Brown, P., Forstmann, B. U. \u0026amp; Zaghloul, K. Frontosubthalamic Circuits for Control of Action and Cognition. \u003cem\u003eJ Neurosci\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 11489-11495, doi:10.1523/jneurosci.2348-16.2016 (2016).\u003c/li\u003e\n\u003cli\u003eBergman, H. (MIT Press, 2021).\u003c/li\u003e\n\u003cli\u003eTemel, Y., Blokland, A., Steinbusch, H. W. \u0026amp; Visser-Vandewalle, V. The functional role of the subthalamic nucleus in cognitive and limbic circuits. \u003cem\u003eProgress in neurobiology\u003c/em\u003e \u003cstrong\u003e76\u003c/strong\u003e, 393-413 (2005).\u003c/li\u003e\n\u003cli\u003eHaynes, W. I. \u0026amp; Haber, S. N. The organization of prefrontal-subthalamic inputs in primates provides an anatomical substrate for both functional specificity and integration: implications for Basal Ganglia models and deep brain stimulation. \u003cem\u003eJ Neurosci\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 4804-4814, doi:10.1523/jneurosci.4674-12.2013 (2013).\u003c/li\u003e\n\u003cli\u003eAlbin, R. L., Young, A. B. \u0026amp; Penney, J. B. The functional anatomy of basal ganglia disorders. \u003cem\u003eTrends in neurosciences\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 366-375 (1989).\u003c/li\u003e\n\u003cli\u003eDas, A. \u0026amp; Goldberg, J. H. Songbird subthalamic neurons project to dopaminergic midbrain and exhibit singing-related activity. \u003cem\u003eJournal of Neurophysiology\u003c/em\u003e \u003cstrong\u003e127\u003c/strong\u003e, 373-383, doi:10.1152/jn.00254.2021 (2022).\u003c/li\u003e\n\u003cli\u003eLitvak, V.\u003cem\u003e et al.\u003c/em\u003e Resting oscillatory cortico-subthalamic connectivity in patients with Parkinson\u0026apos;s disease. \u003cem\u003eBrain\u003c/em\u003e \u003cstrong\u003e134\u003c/strong\u003e, 359-374, doi:10.1093/brain/awq332 (2011).\u003c/li\u003e\n\u003cli\u003eEngel, A. K. \u0026amp; Fries, P. Beta-band oscillations--signalling the status quo? \u003cem\u003eCurr Opin Neurobiol\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 156-165, doi:10.1016/j.conb.2010.02.015 (2010).\u003c/li\u003e\n\u003cli\u003eK\u0026uuml;hn, A. A.\u003cem\u003e et al.\u003c/em\u003e Event-related beta desynchronization in human subthalamic nucleus correlates with motor performance. \u003cem\u003eBrain\u003c/em\u003e \u003cstrong\u003e127\u003c/strong\u003e, 735-746, doi:10.1093/brain/awh106 (2004).\u003c/li\u003e\n\u003cli\u003eAvants, B. B., Epstein, C. L., Grossman, M. \u0026amp; Gee, J. C. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. \u003cem\u003eMed Image Anal\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 26-41, doi:10.1016/j.media.2007.06.004 (2008).\u003c/li\u003e\n\u003cli\u003eMerk, T.\u003cem\u003e et al.\u003c/em\u003e Electrocorticography is superior to subthalamic local field potentials for movement decoding in Parkinson\u0026rsquo;s disease. \u003cem\u003eeLife\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, e75126, doi:10.7554/eLife.75126 (2022).\u003c/li\u003e\n\u003cli\u003eFogelson, N.\u003cem\u003e et al.\u003c/em\u003e Different functional loops between cerebral cortex and the subthalmic area in Parkinson\u0026apos;s disease. \u003cem\u003eCereb Cortex\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 64-75, doi:10.1093/cercor/bhi084 (2006).\u003c/li\u003e\n\u003cli\u003eHamilton, L. S., Edwards, E. \u0026amp; Chang, E. F. A Spatial Map of Onset and Sustained Responses to Speech in the Human Superior Temporal Gyrus. \u003cem\u003eCurr Biol\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 1860-1871.e1864, doi:10.1016/j.cub.2018.04.033 (2018).\u003c/li\u003e\n\u003cli\u003eVissani, M.\u003cem\u003e et al.\u003c/em\u003e Spike-phase coupling of subthalamic neurons to posterior perisylvian cortex predicts speech sound accuracy. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 3357, doi:10.1038/s41467-025-58781-8 (2025).\u003c/li\u003e\n\u003cli\u003eChartier, J., Anumanchipalli, G. K., Johnson, K. \u0026amp; Chang, E. F. Encoding of Articulatory Kinematic Trajectories in Human Speech Sensorimotor Cortex. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e98\u003c/strong\u003e, 1042-1054 e1044, doi:10.1016/j.neuron.2018.04.031 (2018).\u003c/li\u003e\n\u003cli\u003eMoses, D. A.\u003cem\u003e et al.\u003c/em\u003e Neuroprosthesis for decoding speech in a paralyzed person with anarthria. \u003cem\u003eNew England Journal of Medicine\u003c/em\u003e \u003cstrong\u003e385\u003c/strong\u003e, 217-227 (2021).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8008692/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8008692/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The basal ganglia–thalamocortical loop is widely implicated in speech production, yet its role during listening remains underexplored. Here, we show that human subthalamic nucleus (STN) single‑unit activity is modulated during auditory presentation of syllable triplets, and that a subset of neurons encodes phonetic (consonant and vowel) and sequence features prior to speech onset. Using intraoperative microelectrode recordings during deep brain stimulation implantation, ~39% of units showed significant firing‑rate modulation during the auditory epoch. Within these, we discovered “auditory‑syllable” units that carried phonetic and/or sequence information. Auditory presentation also elicited beta‑band desynchronization in STN LFPs and revealed frequency‑specific spike‑phase coupling (SPC) with cortex: auditory‑specific units coupled predominantly with superior temporal gyrus (STG) in alpha, whereas units modulated during both auditory and speech epochs preferentially coupled with sensorimotor regions in beta. Responses were observed to incidental, non‑task sounds in a subset of auditory‑specific units that support genuine auditory sensitivity. Together, these results support a role for the STN in auditory–motor integration and indicate frequency‑specific routing within human cortical–subcortical loops during listening.","manuscriptTitle":"Human subthalamic neurons encode speech features during listening and couple with auditory cortex","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-11 06:33:11","doi":"10.21203/rs.3.rs-8008692/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"5deaae1b-a552-42af-9001-4aa3840f9b0d","owner":[],"postedDate":"November 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":57374406,"name":"Biological sciences/Neuroscience/Cognitive neuroscience/Language"},{"id":57374407,"name":"Biological sciences/Physiology/Neurophysiology"},{"id":57374408,"name":"Biological sciences/Neuroscience/Sensory processing"}],"tags":[],"updatedAt":"2025-12-03T07:56:19+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-11 06:33:11","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8008692","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8008692","identity":"rs-8008692","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}