Heidi Haavik

The claim

Heidi Haavik measured what an adjustment does inside the brain

Heidi Haavik took the oldest claim in chiropractic, that adjusting a spinal joint changes the nervous system, and put electrodes on it. She trained in human neurophysiology at the University of Auckland, published her first cortical study with Bernadette Murphy in 2007, and now leads research at the Centre for Chiropractic Research at the New Zealand College of Chiropractic in Auckland. Her program rests on one question. When a dysfunctional spinal segment receives a high velocity, low amplitude thrust, does anything change above the neck? Ask what that question commits you to. If nothing changes in the cortex, the adjustment is carpentry. If something does change, the spine is an input channel to the brain, and altering the channel alters processing everywhere the channel reaches.

The answer arrived in instalments. In 2007 Haavik and Murphy reported that one session of cervical adjusting reduced the amplitude of two cortical somatosensory evoked potential peaks, the parietal N20 and the frontal N30, and that the reduction lasted on average 20 minutes. In 2016 Dina Lelic and colleagues located the change using 62 channel electroencephalography with brain source modeling, and found it in the prefrontal cortex. Between and after those papers came work on reflexes, single motor units, voluntary force, stroke recovery, falls risk and bite force. One thread runs through all of it. Input from the spine shapes central processing. Change the input and the processing changes with it.

The instrument

Evoked potentials turned a philosophical claim into a measurable one

The somatosensory evoked potential is the tool that made the question answerable. Stimulate the median nerve at the wrist several hundred times, average the electroencephalographic response, and the noise cancels while the signal survives. What remains is a waveform with named peaks at fixed latencies. N11 and N13 arise at the spinal cord. P14 arises at the brainstem. N20 is the first parietal cortical arrival. N30 is a frontal component associated with sensorimotor integration across premotor, motor and subcortical circuitry. The names are only polarity and milliseconds. N30 means a negative deflection roughly 30 milliseconds after the stimulus.

This is what gives the work its force. A peak amplitude is not an opinion about whether someone feels better. It is a number produced by averaging, and it can move up, move down, or refuse to move at all. Haavik built a career on that third possibility being a genuine risk in every experiment. The same discipline sets the limit of interpretation. When an N30 amplitude falls after an adjustment, the honest statement is that this stage of processing handled the same volley differently. It is not a statement that the brain improved. Anyone who converts an amplitude change into a performance score has left the data behind.

2007

Twelve people, one session, and a cortical change lasting about twenty minutes

The founding document is a 2007 paper in Clinical Neurophysiology, volume 118, issue 2, pages 391 to 402 (Haavik Taylor and Murphy 2007). Twelve participants with a history of recurring neck stiffness or neck pain, none of them symptomatic on the day, received a single session of cervical adjusting. A separate group of twelve received passive head movement as a control. Spinal, brainstem and cortical evoked potentials were recorded before the intervention and for 30 minutes afterwards. The parietal N20 and frontal N30 amplitudes fell significantly after adjusting. Nothing moved in the control group.

The control is the whole argument. Passive head movement takes the neck through range. It loads joints, stretches muscle, fires mechanoreceptors and produces a large sensory volley of its own. It produced no cortical change. So the effect does not track gross motion of the neck. It tracks the particular afferent input generated by a thrust delivered to a segment judged dysfunctional. Force the implication. If movement alone were sufficient, every neck roll would be an adjustment and the profession would have nothing distinctive to defend. The 2007 data say otherwise, in twelve people, once.

Subclinical

She studied people who were not in pain, and that choice carries the argument

Haavik’s signature population is the subclinical group: people with a history of recurring neck complaints who report no symptoms on the day of testing. The reason is methodological, and it is the most underrated decision in her work. Pain is a confound. Adjust someone who hurts, watch them improve, and you cannot separate a change in central processing from simple analgesia or from regression to the mean. Take the pain out of the room and any measured change has to be explained some other way.

The 2011 study in the Journal of Manipulative and Physiological Therapeutics, volume 34, issue 2, pages 88 to 97 (Haavik and Murphy 2011), shows what that buys. Twenty five participants with subclinical neck pain and eighteen controls reproduced a target elbow angle using an electrogoniometer, with the head held in four positions. At baseline the controls were significantly more accurate. After cervical adjusting, the subclinical group improved in absolute error with the head neutral and in left rotation, and their variable error fell overall. The controls, who simply rested for five minutes, got worse. Read that carefully. A neck with a silent history was degrading the accuracy of an arm.

Source

The 2016 study located the change in the prefrontal cortex

In 2016 the team stopped asking whether and started asking where. Nineteen subclinical pain volunteers, nine of them male, mean age 25.6 years with a standard deviation of 3.9, attended two sessions in randomized order while 62 channel electroencephalography recorded their evoked potentials. After manipulation the N30 amplitude fell by 16.9 percent. After the control condition it did not move. Brain source modeling then decomposed the scalp signal into candidate generators. Of all the modeled sources, exactly one changed. The prefrontal source dropped by 20.2 percent (Lelic 2016).

Ask what that commits you to. The prefrontal cortex is not a joint region. It is where planning, executive control, working memory, pain modulation and motor preparation converge. A thrust delivered to a cervical segment altered how that region handled a volley arriving from a nerve at the wrist. If that is true, then the spine does more than report to the brain about itself. It contributes to the state in which the brain evaluates everything else. This is also the single result most often distorted in retelling, and the section on limits below states exactly how.

Output

Changed processing showed up as changed force

Processing is invisible. Force is not. In 2015, working with Imran Khan Niazi and Kemal Sitki Türker, Haavik recorded soleus V waves, H reflexes and maximum voluntary contraction in ten participants (Niazi and Türker 2015). After manipulation, contraction force rose by 16.05 percent and the V wave to maximum M wave ratio rose by 44.97 percent, while the H reflex threshold fell. After the control intervention force dropped by 11.35 percent. The V wave carries the weight of that result, because it indexes descending drive from above. An increase points upstream rather than to the reflex arc.

The pattern held across muscles and populations. Transcranial magnetic stimulation input to output curves showed maximum motor evoked potentials rising 54.5 percent in a thumb muscle and 44.6 percent in tibialis anterior, with no change in F waves, a spinal measure. A 2018 study shortened the cortical silent period and raised single motor unit I wave amplitude across 21 identified units. Eleven elite Taekwondo athletes gained plantar flexor force for 30 minutes and corticospinal excitability for at least 60 (Christiansen 2018). Twelve chronic stroke patients gained an average 64.2 percent in plantar flexor strength, with V waves up 54.0 percent and H reflexes unchanged (Navid 2021). Twenty eight people split between chiropractic and sham care showed maximal bite force rise 11.0 percent immediately and 13.0 percent one week later (Haavik 2018). The jaw is not the spine. That last result is the one that should make you think.

Specificity

The 2024 trial asked whether it mattered where you adjust

The obvious objection to everything above is that any thrust anywhere would do the same thing. In 2024 the group tested it directly in Scientific Reports (Niazi 2024). Ninety six adults with recurrent mild neck symptoms were randomized to receive a single instrument delivered thrust at either an upper cervical segment judged clinically relevant on assessment or a predetermined segment judged not relevant. Evoked potentials were recorded immediately before and after. The relevant group showed an N30 amplitude decrease of 16.76 percent. The non relevant group showed no significant change at all.

Force the implication and it cuts in two directions. It supports the claim that assessment carries real information, which is the claim every manual profession makes and almost none has tested this way. It also disciplines the practitioner, because if the site determines the effect then palpation, listing and clinical judgment are not ceremony. A thrust delivered without them is a different intervention with a different result. State the caution plainly. This is one trial, in one spinal region, using one instrument, and it needs replication by groups outside the collaboration before it settles anything.

Duration

Twelve weeks moved the baseline, not only the moment

Almost all of the mechanistic work is acute, and acute effects are cheap. Two studies went longer. In a 2016 randomized controlled trial, Kelly Holt and Haavik enrolled 60 community dwelling adults over 65 in Auckland and compared 12 weeks of chiropractic care against no intervention, with outcomes measured at baseline, four weeks and twelve weeks (Holt and Haavik 2016). Choice stepping reaction time improved by 119 milliseconds, with a 95 percent confidence interval of 26 to 212 milliseconds. The sound induced flash illusion improved by 13.5 percent. Ankle joint position sense improved by 0.20 degrees, and the physical component of the SF-36 quality of life score rose by 2.4 points.

The flash illusion result is the interesting one. In that test a single flash of light paired with two beeps is commonly perceived as two flashes, because the auditory stream overrides the visual. Performance depends entirely on how the brain weights competing sensory channels against each other. A 13.5 percent shift means the weighting changed. Separately, a 2017 preliminary study followed six chronic neck and upper limb pain patients through 12 weeks of care and found the dual nerve stimulation ratio for the P22 to N30 complex fell significantly, with no change across a preceding two week control period, alongside falling pain scores (Haavik 2017). Six people is not a proof. It is a direction.

The model

She renamed the lesion the central segmental motor control problem

In 2021 Haavik and seven colleagues published an invited review running from page 2675 to page 2720 of volume 121 of the European Journal of Applied Physiology (Haavik and Kumari 2021). Its purpose was vocabulary. The review replaces mechanical descriptions of the manipulable lesion with a neural one and names it a central segmental motor control problem: a spinal segment whose central neural control has degraded, which then feeds altered afferent information upward and sustains itself. The listed causes are physical injury, pain, inflammation, and acute or chronic physiological or psychological stress.

That last item reframes the whole subject. Put psychological stress in the causal list for a segmental problem and you have conceded that the lesion is a problem of regulation rather than a problem of position. A segment does not become dysfunctional only because something struck it. It becomes dysfunctional because the control system governing it has drifted, and the drift can arrive through load, through injury, through inflammation, or through a life that never lets the system settle. The review concludes that the neuromuscular changes following adjustment most likely occur through changes in supraspinal excitability. The lesion is central. The correction is central. The joint is where you reach it.

The limits

Haavik’s most useful habit is refusing to overclaim

This body of work supports a narrower claim than the marketing built on top of it. Look at the sample sizes: ten, eleven, twelve, twelve, nineteen, twenty five, twenty eight. Look at the durations: twenty minutes, thirty minutes, sixty minutes. Look at the variability. A 54.5 percent rise in motor evoked potential carries a standard deviation of 93.1 percent, and a 59.5 percent rise in one electromyographic measure carries a standard deviation of 103.4 percent. When the standard deviation exceeds the mean, the group moved and the individuals inside it moved by wildly different amounts. Most of this work also comes from a small network of collaborating laboratories, which is exactly the condition under which independent replication matters most.

Three misstatements travel further than the papers. First, the claim that an adjustment raises brain function by 20 percent inverts the 2016 finding (Lelic 2016), which reported a 20.2 percent decrease in prefrontal source activity, and a change in evoked source activity is not a performance score in either direction. Second, that study used 62 channel electroencephalography with source modeling, not functional magnetic resonance imaging, and calling it brain imaging oversells its spatial resolution. Third, none of this work shows that adjusting treats disease. Haavik said so herself in a 2021 note in the Journal of Manipulative and Physiological Therapeutics, volume 44, page 511, titled Comment about Extrapolating Beyond the Data (Haavik 2021). A researcher who publishes a warning against misuse of her own field is worth reading closely.

Haavik and the model

Haavik runs the experiment the Unified Model of Tone specifies

The Unified Model of Tone predicts that a well matched input reaches the integration centers of the body and not merely the local tissue it touches. It predicts that the change arrives on the timescale of neural signaling, in seconds rather than weeks. Haavik built the instrument that can see whether this is true. Her 2007 recordings caught cortical evoked potentials changing within minutes of a cervical adjustment, with no such change after passive head movement.

The design the model names is already routine. Record heart rate variability, the phase coupling between slow and fast cortical rhythms, and cortical band synchrony immediately before a well matched input and again within minutes of it. Choose the site in advance and deliver a matched control input elsewhere. The model predicts a change in coupling and variability after the well matched input and little or none after the mismatched one. A change in coupling and variability confined to the well matched input, with the mismatched one leaving the central measures where they were, confirms the prediction.

The 2024 randomized trial is the closest anyone has come to running that design. Ninety six adults received a single upper cervical thrust at either a clinically relevant segment or a predetermined non relevant one. The relevant group showed a 16.76 percent drop in N30 amplitude and the non relevant group showed no significant change. That is the shape the prediction requires, in one cortical measure, in one trial. It is not the full design, which asks for the autonomic and coupling measures alongside the cortical one.

A second prediction sits within reach of the same equipment. A standing distortion should carry an elevated sensory gain signature, with the metabolic cost that implies, and the signature should fall when the distortion is corrected rather than when the symptom is merely quieted. A distortion carrying the gain signature confirms the model, and so does a correction that clears the signature rather than leaving it in place while the symptom quiets.

Haavik never used the word tone. She used sensorimotor integration, cortical drive, supraspinal excitability and central segmental motor control. Amplitude, latency, threshold and drive are all measurements of a readiness to respond, and readiness to respond is what tone has always meant. The measurement is hers. The generalization belongs to this site rather than to her. Every doorway which reliably changes afferent traffic is acting on the same variable. The spinal segment is one access point among many, and an unusually direct one. The prediction and the design are the model’s, and neither is proof. Read her alongside Karl Friston, whose account of prediction and free energy explains why a system fed degraded feedback drifts.

What the record shows

Haavik’s figures describe a central change, not a local one

• 2007. Parietal N20 and frontal N30 evoked potential amplitudes fell for about 20 minutes after cervical adjusting in twelve people, with no change after passive head movement (Haavik Taylor and Murphy 2007).
• 2016. The brain source localization study reported a 20.2 percent decrease in prefrontal source activity and a 16.9 percent decrease in N30 amplitude after manipulation (Lelic 2016). That is a change in how the region processed the same input, not a performance score.
• 2024. Ninety six adults received one upper cervical thrust. The clinically relevant segment produced a 16.76 percent drop in N30 amplitude and the predetermined non relevant segment produced no significant change (Niazi 2024).
• Duration. Single session effects are short. Cortical evoked potential changes averaged about 20 minutes, force gains in elite athletes lasted 30 minutes, and corticospinal excitability changes lasted at least 60 minutes.
• Twelve weeks. Chiropractic care in adults over 65 improved choice stepping reaction time by 119 milliseconds and multisensory integration by 13.5 percent compared with no intervention (Holt and Haavik 2016).
• Variability. A 54.5 percent rise in motor evoked potential carried a standard deviation of 93.1 percent, so the group moved while the individuals inside it moved by wildly different amounts.
• 2021. An invited review running from page 2675 to page 2720 of volume 121 of the European Journal of Applied Physiology renamed the manipulable lesion a central segmental motor control problem (Haavik and Kumari 2021). Its listed causes include psychological stress.