How to Rewire Your Brain & Learn Faster | Dr. Michael Kilgard

Kilgard argues that the brain’s statistical learning mechanisms expect certain correlations — e.g., sound reverberation, depth from parallax, tactile feedback from movement. A child watching a flat video of someone speaking does not acquire phonemes because there is no contingent interaction. Similarly, a video game may reward fast thumb movements but does not map to full-body coordination or social nuance. He distinguishes between experiences that are ‘real’ — something a grandparent would recognize as a good use of a day — and those that are artificially constrained. He notes that even reading, if done to the exclusion of outdoor exploration, can be limiting, citing Emerson’s caution that books poorly used are among the worst things. The implication is not to ban screens, but to proportion them so that the brain maintains a rich set of reference frames.

Huberman discusses a recent episode on studying that found self-testing to be the most effective technique. Kilgard adds that in athletics and martial arts, mentally stepping through forms after practice improves performance, and that when he worked intensely in the lab, he would see those images when closing his eyes — evidence of the brain’s offline processing. Together they suggest that the period immediately after an experience is a vulnerable one; filling it with unrelated phone scrolling may interfere with consolidation. Instead, allowing the mind to drift back over the event, perhaps in a structured way, capitalizes on the plasticity window opened during the activity.

Kilgard explains that almost everyone can hear some tinnitus in a quiet environment, but it becomes pathological only when the brain locks onto it. The natural response is to investigate the sound, because in the natural world a persistent tone might signal danger. But when the sound is internally generated, this attention fuels the feedback loop. He draws an analogy to chronic pain and PTSD avoidance: the well-intentioned response makes it worse. The advice to 'ignore it' is simple but profoundly hard; it’s also the rationale behind VNS treatments that try to weaken the tinnitus representation by making other tones more salient under neuromodulator release.

Kilgard describes the paradigm: a custom handle or sensor measures force or range of motion. A computer algorithm looks back at the last few minutes of performance to determine the individual's top 10% attempts. When the patient spontaneously exceeds that threshold, the system automatically delivers a half-second VNS burst. This creates a closed-loop, personalized reinforcement that mimics the natural learning signal but with greater intensity and timing precision. In stroke, the brain is tricked into treating the therapy repetitions as critically important, overcoming the conservative bias that normally limits recovery. The therapy does not eliminate the dead neurons, but recruits surviving regions to compensate. Kilgard emphasizes that this approach is not a magic bullet; the patient must still do hundreds of hours of work, but the VNS removes the 'ceiling' that previously limited progress.

Kilgard spent years testing whether raising neuromodulators globally – with amphetamine, nicotine, fluoxetine, etc. – could replicate the precise, stimulation-induced changes. In animal experiments, repeated pairing of a tone with a drug that broadly elevated acetylcholine did not shift cortical maps. The reason, he explains, is that neurons are firing constantly; the synapse needs to know which specific events are important. Only a brief burst of neuromodulators occurring within a couple of seconds of the relevant activity (the ‘synaptic eligibility trace’) will strengthen that particular connection. Without that timing information, the brain just gets a non-specific signal that doesn’t differentiate the target pattern from background noise. This is why open-loop drugs often fail or have limited effect, and why the VNS device, which can be timely triggered, is so powerful.

Kilgard elaborates that real-world experiences contain low-level correlations that the brain expects: depth cues, peripheral motion, natural reverberation in sound, integrated touch and smell. When an experience strips away most of those (e.g., a 2D video, a swipe on a phone), the brain may fail to integrate it into the same rich predictive model. He notes that even babies are not fooled by passive language videos; they need contingent, interactive input to learn. While some virtual environments have complexity, they often lack the spontaneous, unpredictable feedback of real life. He cautions against assuming that any activity that a child or adult finds engaging is therefore beneficial; the nervous system is tuned to need certain informational patterns. This is not a categorical condemnation of technology, but a call to be mindful of the proportion of time spent in low-dimensional versus high-dimensional environments, especially during development.

Huberman proposes that friction — such as the discomfort of snorkeling for the first time, speaking a language poorly in a foreign country, or making a bagel from scratch during lockdown — amplifies the plasticity-inducing signal. Kilgard agrees and adds the importance of reflection: thinking back over the experience, replaying it mentally, even visualizing skilled movements one has already performed (as Olympic skiers do) strengthens the relevant connections offline. However, visualization works best for refining already-acquired skills, not for learning novel ones from scratch, because the brain needs real-world feedback to form the initial connections. They tie this to self-testing as a potent anti-forgetting tool, where retrieving information reinforces the memory more than re-studying. Sleep, especially REM and deep sleep, is the third pillar where the actual rewiring is consolidated. Kilgard's own childhood eye therapy and his repetitive lab work (seeing chromosome bands when closing his eyes) illustrate how intensive focus+friction engraves patterns into the brain.

Kilgard explains that in the lab, he first used deep brain stimulation to trigger acetylcholine release while playing a tone, and the adult auditory cortex rewired to favor that tone. When others substituted norepinephrine or dopamine stimulation, the same expansion of cortical territory occurred. This led to the understanding that these neuromodulators all converge on G-protein-coupled receptors that set the synaptic eligibility trace. The popular labels (acetylcholine = memory, norepinephrine = attention, dopamine = reward, serotonin = mood) are oversimplifications that don't hold at the cellular level; they all act as permissive gates for activity-dependent plasticity. However, the valence (positive or negative) may be differentially signaled by the ratio of dopamine to other modulators, but the core plasticity-gating function is shared.

Kilgard’s VNS device is a tiny chip (size of a pinky nail) implanted under the skin near the left cervical branch of the vagus nerve. It is activated wirelessly by an external coil. In therapy sessions, a computer monitors the patient’s performance (e.g., arm movement angle, finger force) and automatically triggers a half-second burst of VNS whenever the patient exceeds their own previous 10th percentile threshold. This ensures that the neuromodulator spike is time-locked to the most successful attempts. In a double-blind, placebo-controlled trial for chronic ischemic stroke published in The Lancet, 18 days of in-clinic VNS plus therapy produced significant hand/arm improvements compared to sham stimulation, and patients continued to improve at home with a magnet-based activator. Similar results have been seen in recent Nature paper on spinal cord injury. For tinnitus, the approach involves playing tones that are different from the tinnitus frequency while stimulating VNS, to weaken the over-represented tinnitus neurons and sharpen receptive fields. The technology is from MicroTransponder (University of Texas at Dallas spin-off), and Kilgard discloses his inventorship and shareholder status.

Kilgard details the cascade: high-frequency hair cell damage (from loud noise, antibiotics, aging) reduces input to the brain. Because the cortex is plastic, neurons previously tuned to high frequencies shift to respond to mid-frequencies, creating an excess of neurons tuned to the edge of the hearing loss. This imbalance can lead to positive feedback and pathological oscillations perceived as a phantom sound. The moment a person notices the sound and becomes anxious about it, the attention-driven release of neuromodulators (norepinephrine, acetylcholine) reinforces the aberrant connections, making the tinnitus louder and more intrusive. This is analogous to the amplification of chronic pain by attention. Kilgard’s practical advice: first, try hard to not pay attention to the ringing, as attention is fuel. If that fails, clinical VNS therapy aims to weaken the tinnitus representation by repeatedly pairing VNS with tones that are not the tinnitus frequency, thereby expanding other frequency representations and shrinking the tinnitus map. The initial human trial showed about 50% of participants had about 50% reduction, but a large military trial was canceled due to sequestration, so development paused.