Neurofeedback is a type of biofeedback that focuses on the neuronal activity of the brain. The training method is based on reward learning (operant conditioning) where a real-time feedback provided to the trainee is supposed to reinforce desired brain activity or inhibit unfavorable activity patterns.
Different mental states (for example: concentration, relaxation, creativity, distractibility, rumination, etc.) are associated with different brain activities or brain states.
Similarly, symptoms of mental or brain related health issues are associated with neuronal overarousal, underarousal, disinhibition or instability. Thus, neurofeedback tries to yield symptom relief through an improved regulation of neuronal activity.
Apart from a therapeutic approach neurofeedback is increasingly used for healthy people as well aiming at improved cognitive regulation skills according to the individual goals and needs.
There are various methods of providing feedback of the neurological activity. The most common application uses the measurement of electroencephalography (EEG) where the electrical activity of the brain is recorded by electrodes placed on the scalp. Other, less usual methods, rely on functional magnetic resonance (fMRI), functional near-infrared spectroscopy (fNIRS), or hemoencephalography biofeedback (HEG).
History
In 1898 Edward Lee Thorndike formulated the law of effect. In his work he theorized that behavior is shaped by satisfying or discomforting consequences. This set the foundation for operant conditioning.
In 1924, the German psychiatrist Hans Berger connected a couple of electrodes (small round discs of metal) to a patient's scalp and detected a small current by using a ballistic galvanometer. In his consecutive studies Berger analyzed EEGs qualitatively, but in 1932 G. Dietsch applied Fourier analysis to seven records of EEG and became the first researcher of what later is called QEEG (quantitative EEG).
In 1950 Neal E. Miller from Yale University was able to train mice to regulate their heart beat frequency. Later on, he continued his work with humans training them trough auditive feedback.
The first study to demonstrate neurofeedback was reported by Joe Kamiya in 1962. Kamiya's experiment had two parts. In the first part, a subject was asked to keep his eyes closed and when a tone sounded to say whether he thought he was in alpha. He was then told whether he was correct or wrong. Initially the subject would get about fifty percent correct, but some subjects would eventually develop the ability to better distinguish between states.
M. Barry Sterman published his work in 1967 where we trained cats to modify their EEG patterns to exhibit more of the so called sensorimotor rhythm (SMR). In his following work he discovered that the SMR trained cats where much more resistant to epileptic seizures after exposure to the convulsant chemical monomethylhydrazine than non-trained cats. 1971 he reported similar improvements with an epilepsy patient whose seizures got under control through SMR training. Joel Lubar contributed to the research of EEG-Biofeedback starting with epilepsy and later on with hyperactivity and ADHD.
Since then research on neurofeedback experienced a high ramification for different symptom clusters and performance sectors.
Neuroplasticity
In 2010, a study provided some evidence of neuroplastic changes occurring after brainwave training. Half an hour of voluntary control of brain rhythms led in this study to a lasting shift in cortical excitability and intracortical function. The authors observed that the cortical response to transcranial magnetic stimulation (TMS) was significantly enhanced after neurofeedback, persisted for at least 20-minutes, and was correlated with an EEG time-course indicative of activity-dependent plasticity
Types of Neurofeedback
The term neurofeedback is not legally protected. There are various approaches that give a feedback about neuronal activity and as such are referred to as „neurofeedback“ by their respective operators. Distinctions can be made on several levels. The first level takes into account which technology is being used (EEG, fMRI, fNIRS, HEG). Nonetheless, further distinctions are crucial even with the realm of EEG neurofeedback as different methodologies of analysis can be chosen. Some of which are backed up by a higher number of peer reviewed studies, whereas for others scientific literature is scarce or explanatory models are even still entirely missing.
Despite the many crucial differences a common denominator can be found in the requirement of providing feedback. Usually feedback is provided by auditory or visual input. While original feedback was provided by sounding tones according to the neurological activity many new ways have been found. It is possible to listen to music or podcasts where the volume is controlled as a feedback, for example. Often visual feedback is used in form of animations on a TV screens. Visual feedback can also be provided in combination with videos and films or even during reading tasks where the brightness of the screen represents the direct feedback. Also simple games can be used where the game itself is controlled by the brain activity. Recent development tries to incorporate virtual reality (VR) and controllers can already be used for more involved engagement with the feedback.
EEG Neurofeedback
Frequency band training / Amplitude training
Amplitude training or frequency band training (used synonymously) is the method with the largest body of scientific literature. It also represents the original method of EEG neurofeedback. The EEG signal is analyzed with respect to its frequency spectrum spit into the common frequency bands used in EEG neuroscience (Delta, Theta, Alpha, Beta, Gamma). The training involves training the amplitude of a certain frequency band on a defined location on the scalp to higher or lower values.
Depending on the training goal (for example, increasing attention and focus, reaching a calm state, reducing epileptic seizures, etc…) different positions of the active electrodes have to be selected. Additionally, the trained frequency bands and the training directions (to higher or lower amplitudes) might vary according to the training goal.
Thus, EEG wave components that are expected to be beneficial to the training goal are rewarded with positive feedback when appearing and/or increasing in amplitude. Frequency band amplitudes that are expected to be hindering are trained downwards by reinforcement through the feedback.
As an example, considering ADD/ADHD this would result in training low Beta or mid Beta frequencies in the central to frontal lobe to increase in amplitude, while simultaneously trying to reduce Theta and HighBeta amplitudes in the same region of the brain.
SCP-Training
For SCP-training (Slow Cortical Potentials, SCP) one trains the dc voltage component of the EEG signal. The application of this type of EEG neurofeedback training was mostly endorsed by the research of the group around Niels Birbaumer. The most common symptom base for SCP training is ADD/ADHD whereas SCPs also find their application in Brain-Computer-Interfaces.
Other Methods
Several other methods are known in the neurofeedback community which find their application in therapy but exhibit significantly less scientific support.
Z-Score Training
Live-Z.Score training describes training that is based on live comparison of EEG variables (power, asymmetries, coherence, phase-lag) to a normative database. The existing normative databases are a large step towards understanding general brain activity patterns. However, for the use of live-Z-Score training several major issues are prevalent.
Currently, available databases are not open to public to review the quality of the recordings.
Compared to other normative databases used in medical application (for example blood levels) the sample size for EEG databases used for neurofeedback is extremely small due to the high effort required to obtain data sets. Despite the non-negligible difficulties of obtaining large sample sizes the question about whether a generalization is possible remains unanswered. Especially, as the used databases are based solely on the American population which limits the generalization to other populations even further.
The databases provide norms to eyes open or eyes closed states. However, during live-neurofeedback people often do a mental task to control the feedback which is a state that can differ significantly from normative or „idle“ states.
Training to a „norm“ may be beneficial to low functioning people but can certainly not be used for peak performers who try to exceed the „norm“.
Coherency Training
For this method the training targets the coherency of two (or more) signals from a selected frequency band stemming from two (or more) different electrode locations. The coherence is a measure of synchronized activity of different or larger brain areas.
ILF (Infra-Low-Frequency) Training
ILF Training targets infra slow fluctuations with frequencies below 0.1 Hz where either the amplitude or the phase of the signal are trained. According to the users there is an individual „optimal reward frequency“ (ORF) that has to be found by the therapist throughout the sessions and trained. Until now, ILF training lacks a scientific explanatory model. The meaning and the effect of a personal ORF are not clear from a point of EEG neuroscience.
LORETA (low resolution electromagnetic tomography analysis) Training
Normal EEG signals are restricted to the surface of the scalp. Using a high number of electrodes (19 or more) the source of certain electrical events can be localized. Similar to a tomography that renders a 3D image out of many 2D images the many EEG channels are used to create LORETA images which represent a 3D electrical activity distribution of the brain. The LORETA method can used in combination with MRI to merge structural and functional activities. It is able to provide even better temporal resolution than PET or fMRI. For the application with live-neurofeedback however, 19 channel neurofeedback and LORETA has limited scientific evidence and until now, shows no benefit over traditional 1 or 2 channel neurofeedback.
fMRI Neurofeedback
Neurofeedback can also be realized using functional magnetic resonance imaging (fMRI). This method is not in use by usual therapists or trainers due to the high cost and effort intrinsic to MRI measurements. Nonetheless, this approach finds its use in clinical research to visualize effects and changes in the brain due to neurofeedback with MRI methods. Other than EEG neurofeedback which relies on electrical signals, fMRI relies on measuring blood oxygen level dependent (BOLD) signals. Due to the MRI tube that subjects have to be in for the measurement the feedback is mostly provided via acoustic signals.
HEG Neurofeedback
Similar to fMRI, hemoencephalography (HEG) relies on signals based on changes of the hemoglobin species (oxyhemoglobin and deoxyhemoglobin). Using the species dependent optical absorption the oxygen consumption on different areas of the brain close to the scalp can be visualized. While this method is more robust against measurement artifact through movement compared to EEG recordings, it is limited to bald regions of the scalp due to interference with hair.
Discussion and Critique
Despite its growing application in the medical and performance field, there is still discussion about its effect size in the scientific literature. One major challenge is the non-standardized setting of Neurofeedback training. A large factor might be the non-standardized feedback in research. Based on operant conditioning the neurofeedback training has to provide positive feedback for favorable behavior (or in lesser cases negative feedback for unwanted activity). However, if the training person is fully indifferent about the feedback the effect might vanish. Motivation and interest are intrinsically impacting the brain’s activity which requires a feedback which effectively is evaluated as something positive for the individual such that the brain adapts the rewarded behavior.
Another challenge might be that rewarded frequency ranges can slightly vary from study to study and it is seen that specific characteristics, as for example, the alpha peak varies in position and intensity for different individuals. It’s implications on the exact frequency range that is expected to be favorable to train is not fully understood.
As neurofeedback is explained mostly based on the model of operant conditioning the sensitivity of the feedback (the difficulty to receive a reward) also plays a role. It has been shown that the desired conditioning can be reversed it the threshold values are set too easy. Other publications did not find any effect of neurofeedback apart from placebo when using automatic thresholds that update every 30 seconds in order to maintain a constant success rate of 80%. In this scenario the mechanism of operant condition can be erased. Despite these results, some neurofeedback providers still rely on fast automatic thresholds, or on too easy settings of the threshold.
Apart from the subconscious level of operant conditioning that has been proven to work even with animals neurofeedback and biofeedback can exploit beneficial effects on the conscious level. Trying different mental strategies, concentration tasks, or other similar trainings do benefit from a live feedback that explicitly show if, for example, you lost your concentration and, thus, reminds you of getting back to the task. In this scenario, neurofeedback works similar to a mirror while practicing dancing. By watching yourself in realtime you can consciously adapt your movement according to the „feedback“ of the mirror.
While there is a plethora of small studies cautiously indicating the efficacy of neurofeedback large scale, standardized studies are missing. Various details are expected to play a role, such as the type of neurofeedback, the training protocol (position and frequency bands), the feedback presented to the trainee, frequency and consistency of the training, the interaction between trainer and trainee, employed mental strategies, and many more. In the end, it is also expected that some people fall under the category of non-responders which do not yield benefits from the training. All these variables indicate that a standardized control will be similarly challenging as scientifically proving the effect of speech therapy.
See also
Further reading
- Arns M, Sterman MB (2019). Neurofeedback: How it all started. Nijmegen, The Netherlands: Brainclinics Insights. ISBN 9789083001302.
- Evans JR, Abarbanel A (1999). An introduction to quantitative EEG and Neurofeedback. San Diego: Academic Press.