Huberman LabThe Science of Learning & Speaking Languages | Dr. Eddie Chang
CHAPTERS
- 0:00 – 4:30
Intro: Guest Background and Episode Themes
Huberman introduces neurosurgeon and neuroscientist Dr. Eddie Chang, outlining his roles at UCSF, his work on movement disorders, speech restoration, and brain–computer interfaces. He previews topics including critical periods for language, speech motor control, epilepsy, ketogenic diet, Neuralink, and their shared personal history.
- •Dr. Chang chairs neurosurgery at UCSF and treats epilepsy and speech disorders.
- •His lab pioneered communication restoration for fully locked‑in patients using brain–computer interfaces.
- •The episode will cover language learning, bilingualism, stuttering, critical periods, and future neurotechnology.
- •Huberman and Chang have been friends since age nine and once ran a childhood science club together.
- 4:30 – 16:10
Sponsors and Podcast Context
Huberman clarifies the podcast’s independence from his Stanford role and introduces sponsors related to metabolic health, sleep, and biomarkers. He explains how these tools have informed his own behavior.
- •Levels continuous glucose monitoring for understanding food, exercise, and sleep effects on blood sugar.
- •Eight Sleep cooling/heating mattress cover for sleep optimization and tracking.
- •InsideTracker for blood and DNA-based personalized health recommendations.
- •Momentous partnership for evidence-informed supplements.
- 16:10 – 23:50
Early Research: Sound, Critical Periods, and White Noise
Chang recalls his medical school research with Mike Merzenich on how early sound patterns shape auditory cortex in rodents. They discovered that masking natural sounds with white noise prolongs the auditory critical period but delays maturation, raising questions about early sound environments in humans.
- •Auditory cortex develops sensitive (critical) periods where it is highly plastic to sound patterns.
- •Rat pups raised in continuous white noise showed prolonged but immature auditory plasticity.
- •Humans initially detect many speech contrasts but tune to their native language(s) over the first years.
- •The nature and structure of early auditory input, not just genes, regulate when critical periods open and close.
- 23:50 – 35:20
White Noise for Infants: Potential Risks and Unknowns
Huberman probes whether using white-noise machines for infant sleep might impair language development. Chang explains the rodent data suggest a plausible concern but emphasizes that human studies are lacking and usage patterns differ, sharing his own choice to avoid continuous white noise for his children.
- •Parents widely use white-noise machines to help infants sleep, but this is evolutionarily unnatural.
- •Continuous white noise in rats kept auditory cortex immature, suggesting a potential developmental cost.
- •Human data on intermittent, night‑only white noise are not yet available; the question is open.
- •Chang avoided white noise for his kids and recommends using more natural structured sounds instead.
- 35:20 – 49:20
Awake Brain Surgery and Functional Brain Mapping
Chang describes awake craniotomy procedures where he stimulates exposed cortex to map language, movement, and sensation while patients talk and respond. These intraoperative observations continuously reinforce that specific cortical areas underlie speech, cognition, and emotion.
- •Patients can be awake during surgery with a sterile drape separating their face from the surgical field.
- •Electrical stimulation of motor cortex produces limb movements; stimulation of language areas can arrest speech or disrupt naming.
- •Real-time mapping allows maximal tumor or seizure-focus removal while protecting critical functions.
- •Seeing speech halted by focal stimulation illustrates in real time that the brain organ is the substrate of language and thought.
- 49:20 – 1:02:50
Emotion Circuits: Anxiety, Calm, and Seizure Misdiagnosis
Chang explains how stimulating certain regions such as orbitofrontal cortex, insula, and amygdala can modulate anxiety, calm, and disgust. He recounts a patient misdiagnosed with anxiety disorder who actually had focal amygdala seizures, underscoring how emotional states can originate in pathological brain activity.
- •Orbitofrontal cortex stimulation can reduce anxiety in already-anxious patients but may do little in calm individuals.
- •Amygdala and parts of insula stimulation can acutely evoke anxiety or disgust.
- •A young woman’s recurrent ‘panic attacks’ were ultimately traced to amygdala seizures via intracranial EEG.
- •Many people have anxiety, but only a tiny minority have seizures as the cause; careful history and EEG are key.
- 1:02:50 – 1:16:00
Epilepsy: Drugs, Surgery, and the Ketogenic Diet
The discussion turns to epilepsy, its pharmacologic management, indications for surgery, and dietary therapies. Chang outlines when drugs suffice, when surgery or stimulators are needed, and how ketogenic diets can dramatically help some patients despite incomplete mechanistic understanding.
- •Epilepsy is defined as three or more unprovoked seizures; about one-third of patients remain drug‑resistant.
- •Trying more than two to three antiseizure medications yields diminishing returns; surgery is then considered.
- •Surgery may involve resecting a seizure focus or implanting stimulators to modulate pathological activity.
- •Ketogenic diets can be life-changing for some, especially pediatric epilepsy, though not universally effective and mechanisms remain unclear.
- 1:16:00 – 1:28:00
Seizure Types and Temporal Lobe Phenomena
Huberman asks about absence and nocturnal seizures. Chang explains how different seizure types manifest depending on where and how abnormal activity spreads, using temporal lobe seizures as examples of how pathological activity hijacks memory, emotion, and olfactory systems.
- •Absence seizures involve brief lapses in awareness where a person may appear to stare blankly while standing or sitting.
- •Temporal lobe seizures can cause déjà vu, metallic taste, burning toast smell, or intense emotions without convulsions.
- •Seizure semiology often reflects the normal function of the structures involved (hippocampus, amygdala, olfactory cortex).
- •Nocturnal seizures occur preferentially during certain sleep stages; circadian and brain-state factors modulate seizure vulnerability.
- 1:28:00 – 1:38:20
Revisiting Broca’s and Wernicke’s: Historical Foundations
Chang recounts how Broca and Wernicke’s 19th-century lesion studies created the classic dissociation between articulation and comprehension. He then sets up how modern awake-surgery and intracranial recording data both support and challenge parts of this model.
- •Broca’s patient ‘Tan’ could only say “tan” but understood language; autopsy showed left frontal damage.
- •Wernicke’s patients produced fluent but meaningless ‘word salad’ and had severe comprehension deficits.
- •The traditional model: Broca’s area = production; Wernicke’s area = comprehension and lexical access.
- •Chang notes his own clinical observations often did not fit these simple mappings, prompting deeper investigation.
- 1:38:20 – 1:51:00
Modern View: Motor Cortex as a Core Speech Area
Drawing from hundreds of surgical cases, Chang explains that damage to classic ‘Broca’s area’ often spares speech, whereas injury to adjacent precentral gyrus regions that control lips, jaw, and larynx can profoundly disrupt language output. Wernicke’s-region lesions remain devastating for comprehension and lexical retrieval.
- •Many resections in the posterior inferior frontal gyrus (classic Broca’s) leave speech remarkably intact.
- •The precentral gyrus (speech motor cortex) integrates motor control with higher-level language planning.
- •Posterior temporal (Wernicke’s-like) regions are essential for understanding, naming, and fluent output; damage causes aphasia and jargon speech.
- •Textbooks oversimplify language neuroanatomy; about half of what Chang learned in medical school about the brain he now considers inaccurate or grossly simplified.
- 1:51:00 – 2:04:00
Lateralization, Handedness, and Bilingual Brains
The conversation explores how language lateralization relates to handedness and how bilingualism is represented. Chang describes high left-hemisphere dominance in right‑handers, looser patterns in left‑handers, and substantial overlap in neural circuits used for different languages.
- •Right-handers: ~99% left-hemisphere language dominance; left-handers: ~70% left-dominant, with more right or bilateral cases.
- •Anatomical gyri and sulci look similar on both sides, suggesting latent language capacity bilaterally.
- •After left-hemisphere damage, right-hemisphere regions can partially assume language roles via plasticity.
- •In bilinguals, both languages often recruit overlapping posterior temporal regions, but patterning of sequences and meanings differs.
- 2:04:00 – 2:18:00
What Speech Cortical Neurons Actually Encode
Chang delves into what neurons in Wernicke’s-region cortex respond to when people hear speech. Using high-density intracranial electrodes in epilepsy patients, he maps selectivity for phonetic features like plosives and fricatives, and introduces the concept of articulatory features as the building blocks of spoken language.
- •The ear decomposes sounds into frequency channels; cortex transforms these into speech-relevant features.
- •Primary auditory cortex is tonotopic; speech cortex can receive somewhat direct inputs and responds selectively to speech sounds.
- •Sites are narrowly tuned to consonant classes, vowels, and articulatory maneuvers, arranged in a salt‑and‑pepper map.
- •About 12 articulatory features (specific vocal-tract movements) combine to generate all phonemes and words, enabling near-infinite expressive capacity.
- 2:18:00 – 2:35:00
Speech Production Mechanics: Larynx, Vocal Tract, and Vocalization
Chang explains how speech physically emerges from coordinated respiratory drive, laryngeal vibration, and vocal-tract shaping. He distinguishes learned speech from innate vocalizations like crying and moaning, which rely on different, more primitive circuits.
- •Speech is exhalation shaped: lungs provide airflow, larynx (vocal folds) generate voicing, and lips/tongue/jaw/pharynx sculpt consonants and vowels.
- •Male and female voice pitch differences arise largely from larynx size and vocal-fold properties (∼100 Hz vs. ∼200 Hz fundamental).
- •Vocalizations (cries, moans) persist even when speech areas are damaged, indicating separate, evolutionarily older pathways.
- •Complex feats like opera singing or freestyle rap dramatically extend the capacity of this same motor architecture.
- 2:35:00 – 2:49:00
Reading, Writing, and Dyslexia: Mapping Vision to Speech
The discussion moves to reading as a cultural overlay on speech circuitry. Chang describes how visual word forms are mapped to phonology in auditory cortex, and how disruptions in this mapping can underlie dyslexia.
- •Reading recruits a ‘visual word form area’ interfacing visual and language cortices, piggybacking on older systems.
- •Learning to read typically starts by mapping printed words to their spoken forms; expert readers can later access meaning more directly.
- •Many dyslexic individuals struggle with phonological awareness and grapheme–phoneme mapping, not just eye movements or visual tracking.
- •Remediation strategies often emphasize explicit phonics, auditory discrimination, and strengthening visual–auditory coupling.
- 2:49:00 – 2:58:00
Language Change, Dialects, and the Myth of a ‘Correct’ Way to Speak
Huberman notes how texting and informal writing styles may influence spoken language. Chang emphasizes that language and speech styles evolve naturally over time, producing dialects and new forms that can become unintelligible across groups.
- •Prescriptive ‘proper’ language taught in school does not reflect how languages naturally evolve.
- •Sound change and dialect formation are normal; isolated communities can develop distinct phonologies and vocabularies quickly.
- •The brain flexibly tracks these changes, adjusting its speech and language maps to new norms.
- •Digital communication (texts, emails) will likely continue to reshape both written and spoken forms.
- 2:58:00 – 3:08:00
Foreign Accent Syndrome and Stroke-Induced Speech Changes
Huberman raises anecdotes of people allegedly acquiring new languages after strokes. Chang dispels that myth but describes foreign accent syndrome, where strokes in speech-motor regions alter intonation and phonology so that speech sounds as though it has a different accent.
- •There is no credible evidence of strokes causing acquisition of wholly new languages.
- •Foreign accent syndrome emerges when strokes affect speech-motor control, altering prosody and phonetic patterns.
- •Patients may sound like they speak another language’s accent, though vocabulary and grammar do not change accordingly.
- •This further underscores the critical role of motor patterns in perceived language identity.
- 3:08:00 – 3:27:00
Auditory Memory and Distributed Storage of Speech
Huberman asks how memories of specific sounds and phrases are stored. Chang explains that long-term speech and motor memories are distributed across multiple cortical and subcortical regions, which is why focal resections rarely erase deeply consolidated skills or personal memories.
- •Auditory and speech memories are not stored in a single locus; they are highly distributed.
- •Patients losing Broca’s/Wernicke’s-like areas rarely forget long-held songs or motor sequences like counting.
- •Rehearsed sequences (e.g., Happy Birthday, counting) can be preserved even when spontaneous speech is severely impaired.
- •Global, severe injuries are usually required to produce profound retrograde amnesia; focal neurosurgeries seldom do.
- 3:27:00 – 4:09:00
Restoring Speech with Brain–Computer Interfaces: The BRAVO Trial
Chang describes his team’s landmark clinical work decoding attempted speech from cortical signals in a locked‑in patient named Pancho. After implanting an electrode array over speech-motor cortex, they used machine learning and language models to convert his brain activity into words and sentences on a screen.
- •Locked‑in syndrome results from brainstem injury or ALS disconnecting intact cognition from motor output.
- •Pancho had a brainstem stroke at 20, leaving him unable to speak; he communicated for 15 years by head‑pecking letters on a screen.
- •An implanted electrode grid over his speech-motor cortex connects via a skull port to external decoding hardware.
- •Training AI on 50 words plus a language model enabled real-time decoding of spoken attempts into English sentences.
- •His joyful giggling when he saw correct words appear disrupted decoding, highlighting both human emotion and system sensitivity.
- 4:09:00 – 4:27:00
Future of Neural Prosthetics: Avatars and Richer Communication
Building on text decoding, Chang outlines efforts to create full audiovisual communication for paralyzed patients by animating avatars that move mouths and faces according to decoded speech and emotional signals. He notes that as virtual interaction grows, enabling disabled individuals to inhabit such spaces will be crucial.
- •Nonverbal cues—facial expressions and visible mouth movements—significantly enhance intelligibility and conversational timing.
- •Chang’s lab is developing systems to animate a digital avatar’s face and speech movements from neural activity.
- •Avatar feedback may speed user learning by making the neuroprosthetic feel embodied and self‑owned.
- •These innovations parallel mainstream trends toward virtual worlds but are especially vital for people with severe disabilities.
- 4:27:00 – 4:49:00
Augmenting Normal Brains: Neuralink, Ethics, and Technological Limits
Huberman raises questions about using neural interfaces not just for restoration but for enhancement—boosting memory, communication bandwidth, or athletic performance. Chang distinguishes between long-standing human use of performance aids and the new ethical terrain of invasive neurotechnology, arguing that current tech is far below the brain’s native bandwidth.
- •Decades of BCI research preceded commercial players like Neuralink; industry is now accelerating translation.
- •Augmentation (beyond restoring deficits) raises questions about risk, access, equity, and societal impact.
- •Humans already augment with caffeine, psychoactives, education, and smartphones; neural implants are a new, more invasive tier.
- •Present-day implants record from far fewer neurons than natural systems; technological limits will cap early enhancement claims.
- 4:49:00 – 5:06:00
Stuttering: Motor Coordination and Auditory Feedback
The conversation turns to stuttering, which Chang categorizes as a speech-motor coordination disorder rather than a language deficit. Anxiety exacerbates but does not cause stuttering; therapy often focuses on initiating speech and modulating feedback loops between production and perception.
- •In stuttering, language planning is intact, but fluent articulation breaks down.
- •Stutterers have an underlying predisposition; they do not stutter constantly but in specific contexts, often worsened by anxiety.
- •Initiating speech (getting the first sound out) is frequently the hardest part.
- •Altered auditory feedback can improve or worsen stuttering, suggesting abnormal coupling between speaking and self‑hearing circuits.
- 5:06:00 – 5:23:00
State Regulation, Surgery Focus, and the OR as Sanctuary
Huberman asks how Chang manages his own cognitive state for high-stakes surgery and research. Chang describes using running and swimming primarily for mental regulation, and he views the operating room, with its intense focus and disconnection from digital distractions, as a kind of sanctuary.
- •Chang avoids caffeine due to its negative impact on his fine-motor control and cognitive steadiness.
- •Regular aerobic exercise helps his mood, focus, and interpersonal functioning; he notices decline after just a day or two off.
- •He does not allow phones or music to distract him in the OR, preferring clear communication and deep concentration.
- •Routine aspects of surgery become muscle memory, similar to running, allowing a paradoxical blend of intense focus and mental rest.
- 5:23:00
Closing Reflections and Podcast Wrap-Up
The episode closes with mutual reflections on their shared childhood and current scientific paths. Huberman highlights the translational significance of Chang’s work, and then provides standard podcast outro information on sponsors, the premium channel, and newsletter.
- •Chang and Huberman bond over their childhood bird club and lifelong curiosity about the natural world.
- •Huberman frames neurosurgeons as the ‘astronauts’ of neuroscience for their direct exploration of human cortex.
- •They agree that much about the brain remains unknown, but each patient and experiment is a step into new territory.
- •Huberman then outlines ways to support the podcast, including sponsors, premium AMAs, and the Neural Network newsletter.
