This article comes from my weekly Neurofeedback & Chill livestream, where I teach a topic and run a live neurofeedback session on my own head. This week's topic: training the left versus the right hemisphere, and what actually happens in the brain when you do it. I run Peak Brain Institute, where we map and train brains using QEEG. Audience questions from the stream are folded in here without names.
Why does the left hemisphere run faster than the right?
Start with the building blocks. The cortex is organized into mini-columns (also called micro-columns), each a stack of roughly 30,000 to 40,000 neurons plus about double that in glial cells. Picture a cluster of six-story buildings. Each building runs its own local information processing, sends short wires to its neighbors, and sends long-distance wires across town to coordinate with distant regions.
The left hemisphere has fewer of those long-distance connections than the right. Most of the neurons running between tissue chunks are inhibitory interneurons that pump the brakes and dampen information flow. With fewer of them, the left hemisphere's modules run more on their own. They run hotter, with less input, and they tend to oscillate faster.
You see this in the beta band. Functionally equivalent beta at C3 (left central) often sits in the teens, around 15 to 18 Hz, and feels good there. The same functional beta at C4 (right central) tends to run lower, around 12 to 15 Hz. The pattern holds across bands. The left is faster in alpha and faster in beta. This is the old arousal model layered onto laterality: as you move forward in the brain, things speed up; as you move back, they slow; as you move left, faster; as you move right, slower. Aging speeds things up, then slows them again at the far end.
That speed difference is the first thing to track. The left is more modular, so you can push individual modules harder. The right is more interconnected and more stabilizing.
What do the left and right hemispheres actually do?
The two sides carry different jobs. The left drives approach: reaching out, sinking your teeth in, initiating and continuing behavior, staying asleep, and for most people, language. The right shapes and supervises. It tracks whether you are paying attention, pulls you back from shiny objects, and damps down tasks that feel like too much effort.
In the frontal corners this becomes an approach-versus-avoid dance. The left frontal lobe is the character leaning in and saying come here. The right frontal corner is the character saying this is too hard, leave me alone. This maps onto Richie Davidson's frontal alpha asymmetry work: left frontal dominance correlates with an approach, positive-mood bias; right frontal dominance correlates with withdrawal, anxiety, and low mood (Davidson, 2004). Frontal midline structures, including the anterior cingulate around FZ, work as a traffic warden that holds what is important and reconciles the competing signals. When that reconciliation fails, you get stuck. Procrastination lives in that failure to decide who is in charge. I cover the mechanics of that in more depth in my work on procrastination and the brain and on anxiety circuits.
On the long-term meditation side, Cliff Saron, Alan Wallace, and colleagues studied contemplatives looking at images of suffering. Longer practice correlated with a stronger left frontal asymmetry and stronger right-hemisphere activation under intense grief. Some of those effects showed up in gamma, in the several-hundred-Hz range you cannot measure with passive scalp electrodes. Sara Lazar's work added the structural piece: cortical thickness scales with lifetime hours of meditation (Lazar et al., 2005). More practice, a thicker, more resilient cortex. For the circuit-level view of contemplative training, see my piece on the neuroscience of meditation.
How does neurofeedback train these circuits?
Neurofeedback runs on operant conditioning below conscious awareness. The game runs when your brain produces the target pattern and dims when it drifts. You do not steer the feedback consciously. Your brain gradually learns what gets applauded.
In the live session I set up two single-site protocols. On the left, C3 minus A1 (left ear): inhibit theta at 4 to 7 Hz, reward beta around 14.75 to 17.75 Hz, inhibit high-frequency beta at 24 to 36 Hz. On the right, C4 minus A2 (right ear): the same architecture, with the reward and high inhibit set lower because the right runs slower. C3 stabilizes vigilance and deep sleep. C4 supervises attention and executive function. Weakness at these sites tracks with what people loosely call ADD on the left and ADHD on the right. For the full picture of training the central strip, see my article on SMR neurofeedback and the neuroscience guide to neurofeedback for ADHD.
The auto-goal logic adjusts the thresholds every 30 seconds to sit just past where your brain currently is, then waits for your brain to flex in the right direction and rewards the movement with a beep and a piece of the picture grid dropping away. The brain hears the beep stop and start, sees the visual change, and learns. It is mostly involuntary.
What does the research say about how the brain learns from neurofeedback?
This was the core of my PhD work at UCLA. I placed electrodes at C3 and C4, put a 64-channel DC-coupled BioSemi cap over the top, and used the neurofeedback events themselves as time markers. Every beep-and-square reward got sent into the full-head record as an event. I then averaged thousands of those events into evoked potentials and watched how the brain reacted to being rewarded.
The finding: when you reward a frequency, the brain responds with an amplitude burst and a desynchronization in time at that frequency. Reward 14.75 Hz beta, and the beta bursts back. It is a clean round-trip. The signature shows up as an event-related spectral perturbation (ERSP).
I ran it as a double-blind, placebo-controlled study with about 46 UCLA students across four groups: C3 training, C4 training, and two sham groups. In sham, the software grabbed stored EEG from other subjects, scaled it to match the participant, shuffled and stitched it, then blended that non-contingent file with the participant's real EEG for the display. Coughs, blinks, and pulling a wire all behaved normally. The beeps and the game were driven by EEG that was not theirs. The sham group still got an auditory P200 from the beep, but no ERSP, no burst in the rewarded band.
Two results worth holding onto:
First, within the first 10 minutes, every single person doing real neurofeedback showed their brain reacting to the applauded frequency. Sham did not. The learning signal is immediate, even though the felt and behavioral changes take days to weeks.
Second, in later analysis with Len Tjo at Pacific Development Technology, we found that only about half of people show a within-session trend under the training electrode itself. The brain still learns and still changes later. The local training produces some of its lasting change elsewhere, in the regions the trained site connects to. Train C4, and you can get an evoked response at C4 plus a change over time at both C4 and C3.
That second point is the practical core of this whole topic. If you train one side and ignore its homotopic partner, you can build a bias instead of a balance.
Why train both hemispheres, and how does the reference matter?
A question came up on the stream about whether the bilateral effect depends on the reference point. It does. Half the signal under any electrode comes from that region; the rest is everything else in the brain, summed and lagged. So the reference ear you subtract against changes what you measure.
C4 minus A1 (right scalp minus left ear) pulls in left-hemisphere contribution and effectively trains across CZ and C3 too with that low beta. C4 minus A2 (right scalp minus right ear) brackets the right side and biases the measurement to the right. My dissertation data showed a left-hemisphere activation and change at C3 from a C4 minus A2 protocol. There is no pure localization in scalp EEG, and real-time source analysis like LORETTA adds time lag you cannot afford during live feedback. For what mapping can and cannot show, see my QEEG brain mapping guide.
I validated the laterality piece using lateralized attention testing, flashing stimuli into one visual field and requiring a same-side hand response, so each hemisphere gets tested alone. I first ran this on the split-brain patients from Roger Sperry and Joe Bogen's original work, alongside my advisor Eran Zaidel, who had been part of that research at Caltech. Those patients lack a corpus callosum, so you can ask whether hemispheric attention is intact in each side independently. It was. And the same hemispheric attention effects show up in intact brains. Mapping that attention testing against the full-head neurofeedback records showed divergent effects in each visual field from the training. That is the basis for how we tune frequencies by hemisphere at Peak Brain.
When do bilateral and intrahemispheric protocols stabilize the brain?
A provider asked why intrahemispheric protocols like C3 minus C4 or T3 minus T4 are stabilizing. Two reasons.
First, when you subtract two close-together scalp sites, you cancel out a lot of shared signal, so you get less amplitude to train. Less signal means a gentler protocol.
Second, you are training the relationship between homotopic regions rather than a single local amplitude. That can break up hypercoherence and build the beta resonance I found between matched sites. It works well for the central and temporal strips, which are corticothalamic hubs that spend communication energy across the thalamus and corpus callosum. I would not run F3 minus F4 the same way; the frontal tissue does not behave like that. C3 minus C4 and T3 minus T4 tap genuine cross-hemisphere channels.
You can extend this with sum protocols. In a two-channel alpha-theta sum, you add the raw signals before measuring the alpha and theta in the combined waveform. If the two sides come into phase, the rewarded amplitude doubles; if they go 180 degrees out of phase, it cancels. So the reward itself trains the two regions toward coherent communication. The dual protocols do something similar.
The most powerful version is a regional bilateral activation. Stack frontal sites with central sites in sequence, for example a frontal dual followed by C3 minus C4, to create a stronger regional frontal activation than either piece alone. Sequenced and segmented protocols like these, including Jay Gunkelman-style beta resets and the alpha sweeps I have built, combine hemispheric communication across circuits. Run them after you have balanced the individual resources on each side, then layer in bilateral and sum protocols to train the communication between them.
On CZ at the vertex: training CZ minus A1 gives a broad, gentle SMR effect because CZ sits in the middle and pulls roughly equal contribution from both sides. It functions more as a global central-strip enhancement than a strict bilateral protocol.
What about the frontal mood sites?
A question came in about F4 and F8 with an A2 reference for decision-making. F4 and F8 sit over right dorsolateral prefrontal cortex; F8 also sits over the insula, which is deeper. We treat these as mood, anxiety, and anger sites. You typically relax them by bringing up alpha and bringing down very fast beta, which tends to feel calming, especially at F8 and the insula region. On the left you would raise beta instead.
If those protocols smooth decision-making, the likely mechanism is one of two things. Either anxiety was blocking action (excess theta and fast beta producing dread and overwhelm) and you cleared it, or you incidentally reduced theta over the frontocentral strip. FC4 carries some of the frontal mood character and some of the central executive character, and reducing theta there helps decision-making. Dissolving right-sided resistance can lubricate action, because the right is the side that avoids.
One anatomical point worth keeping: the C sites are frontal lobe. The central strip is the back of the frontal lobe, and it runs down into the thalamus. C3, C4, and CZ are the mind-body connection, the sensorimotor strip where cortex meets the thalamic relay. That is part of why training there is so productive for both attention and sleep.
A note on the right hemisphere and altered states
The stream closed on Jill Bolte Taylor's left-hemisphere stroke and her account of feeling oneness. My read is that the bleed pushed on the medial temporal lobe and produced something like a temporal-lobe-epilepsy experience. A bleed floods tissue with glutamate, and the brain handles iron and blood poorly, so irritation and discharge in the medial temporal lobe is plausible. Medial temporal stimulation has a well-reported tendency to produce profound oneness experiences, sometimes called the god spot, and there are reasonable hypotheses linking temporal lobe epilepsy to classic ecstatic visions.
Julian Jaynes's The Origin of Consciousness in the Breakdown of the Bicameral Mind is worth reading alongside this. Hemispheric language laterality may be relatively recent in evolutionary terms, and non-dominant-hemisphere language, your other hemisphere speaking to you, may sit underneath some early religious experience. Whether you read a temporal lobe event as a seizure or as a genuine spiritual door is a separate question, and I stay agnostic on it.
Putting it together
The training logic runs in a clear order. Read each side's resources from the QEEG. Account for the left running faster and more modular, the right running slower and more interconnected. Train the specialized tissue, and keep the homotopic partner in mind, because a chunk of the lasting change shows up in the regions the trained site connects to. Balance each side's individual resources first, then combine them with bilateral, sum, dual, or segmented protocols to train the communication between them.
If you want to see your own hemispheric pattern before training anything, the place to start is a brain map. We run QEEG mapping at Peak Brain offices and remotely, and you can read more about whether neurofeedback is legitimate and what the anxiety research shows if you want the evidence base first.