The ability of the brain to process information, similar to modern communication devices, is conveyed by voltage frequencies.  These frequencies are characterized by the total number of up- and down-swings or cycles in a given unit of time. Traditionally the number of cycles is measured per second and its unit is called the Hertz.  One cycle per second would be equal to 1 hertz (Hz). Cell phones convey their information in gigahertz (10 to the 9h power or 1,000,000,000) frequencies, while TV and radio convey their information in the kilohertz (10 to the 3rd power or 1000) and megahertz (10 to the 6th power or 1,000,000)  range. By way of comparison the frequencies generated by the brain are extremely slower  (0-100 Hz) and of very low amplitude.

The brainwave frequencies can be decomposed into different bandwidths according to behavioral states to which they seem to be associated. Very deep sleep is associated with so-called delta frequencies. These frequencies range from 0.5 to 4 Hz. The next higher bandwidth is called Theta (4 to 8Hz) and it is associated with daydreaming, sleep, and raw emotions.  Individuals with attention deficit hyperactivity disorder (ADHD) get “stuck” in this Theta bandwidth and have excessive Theta waves during wakefulness. Researchers have attempted to treat ADHD by having them shift their brainwaves to higher frequencies (so-called, alpha, beta or gamma frequencies). These higher frequencies are more conductive towards establishing an attentive state.

Shifting of one’s brainwave activity has usually been done by using a real-time display of brain voltage activity and having the individual self-regulate his/her brain function. As an example, a patient may sit in front of a laptop while wearing an electrode cap from which the researcher records brainwave activity. The laptop may play a movie and, if the brainwave frequencies are not in the target range, the overall viewing window may be quite small. If, on the other hand, the recorded brainwave frequency is on target, the window for the movie enlarges and can easily be watched. The job for the patient is to figure out for himself/herself, how to keep the movie window open as wide as possible for the longest period of time. This technique is called neurofeedback (NFB) and it has been recognized as a suitable tool for detecting and modulating neural plasticity due to its ability to non-invasively alter the excitability of neural circuits and inducing a short-term reorganization of associated cortical and sub-cortical neural networks in the human cortex.

Many patients with autism have attention problems. In about one-third of cases the deficit is so severe as to be called a «disorder». It therefore seems reasonable to ask ourselves whether the same neurofeedback techniques that are being used in ADHD could offer some benefit in ASD?

Several articles have reviewed the application of neurofeedback for ASD treatment and many of them provide evidence that some of the core symptoms of autism can be improved with this technique (Coben, 2013; Coben & Padolsky, 2007; Coben & Myers, 2010; Coben et al., 2010; Jarusiewicz, 2002; Kouijzer et al., 2009ab; Sokhadze et al., 2014; Wang et al., 2016). During NFB procedure, subjects are trained to enhance desired electro-cortical activity, while suppressing undesirable activity. Through the NFB training course many symptoms related to EEG abnormalities can be corrected, and improved towards normalization.

Four small randomized clinical trials (RCTs) and several reviews (Holtman et al., 2011; Hurt et al., 2014) of NFB treatment of autism have been published.  The first two RCTs (Pineda et al., 2008), focused on decreasing mu rhythm (12-14 Hz) over the sensorimotor cortex, which is considered by some authors as an EEG correlate of mirror-neuron activity associated with imitation abilities that are thought to be limited in autism (Oberman et al., 2005).  The first study randomized 8 youths (ages 7-17) with diagnosis of high-functioning autism to either NFB (N=5), mu rhythm, right hemisphere C4, 30 half-hour sessions, 3 times weekly for 10 weeks) or a sham-NFB control (N=3). Participants and parents were blinded to the treatment assignment. Compared with sham controls, NFB group significantly increased sustained attention and sensory/cognitive awareness scores on subscales of the parent-rated Autism Treatment Evaluation Checklist (ATEC, Rimland & Edelson, retrieved from http://www.autism.com). The second study examined 19 youths (ages 7-17) with rigorously diagnosed autism diagnoses in a randomized (NF=9, sham NF=10) double-blind design involving NFB training of a higher mu band (10-13 Hz). This study confirmed significant improvements in sustained attention but not in sensory-cognitive awareness and also reported significant parent rated ATEC improvements in speech/language communication, sociability, health/physical behavior subscales, and overall score. In addition, this second study found normalization of mu rhythm. Although both studies demonstrated normalization of the NFB-targeted mu rhythm as well as improvements in a variety of behaviors associated with autism, neither study showed the expected behavioral improvements in imitation. More recently the latter group continued studies using mu-rhythm suppression self-regulation in autism and published more research and review papers on this particular protocol (Datko et al., 2017; Friedrich et al., 2014, 2015; Pineda et al., 2014ab).

The other RCTs were conducted by Kouijzer and colleagues as a follow-up to their nonrandomized pilot study of 8- to 12-year-olds with PDD–NOS, in which NFB led to improved executive functioning, social communication, and atypical behavior, which was sustained for 6 months after the termination of treatment; qEEG changes included significantly reduced theta and increased beta EEG power in central and frontal brain regions. In their first RCT, Kouijzer et al. (2009ab) examined 14 youths (ages 8-12) with high-functioning autism, whose diagnoses were verified by a study psychiatrist. They were randomized to individualized qEEG-guided NFB treatment or a wait-list control. NFB involved forty 21-minute sessions, twice weekly for 20 weeks, and focused on decreasing excessive theta power in central and frontal brain areas.

The treatment gains were maintained at 6-month follow-up, and some additional treatment gains were demonstrated at 6 months for the NF group but not for the control group.  In their second small RCT, Kouijzer et al. (2013) attempted to control for some nonspecific treatment effects by comparing NFB, skin-conductance biofeedback, and a wait-list control. In this study, 38 adolescents (ages 12-18) with rigorously diagnosed autism were randomly assigned to individualized qEEG-guided NFB (N=13; forty 21-minute sessions, twice weekly for 20 weeks), skin-conductance biofeedback (N=12; identical treatment to NFB except feedback was based on skin-conductance from the index and ring fingers of the non-dominant hand); or a wait list.

More research studies should be done to understand: 1) whether children with high functioning autism can control EEG in NFB mode, 2) how EEG characteristics are changing during the training course in ASD population, and 3) what additional efforts are needed to correctly identify specific changes in EEG rhythms known to be abnormal in ASD, which EEG characteristic may serve as a valid predictor of training outcome, etc.  Another important factor is the specifics of neurofeedback protocol. There are several protocols that are prevalent in the literature and have been investigated and shown efficacious in the treatment of ADHD and were also used in ASD population (Coben, 2013; Lubar, 2003; Sichel et al., 1995; Thompson et al., 2010ab). Most of them used suppression of theta at fronto-central or central sites, enhancement of low beta (13-21 Hz) sub-band, or enhancement of sensory-motor rhythm (SMR, 12-15 Hz) at the central sites (C3, Cz, C4).  Our group’s approach has included neurofeedback training at the prefrontal topography, specifically at the midline prefrontal site. This selection of cortical topography was also determined by our prior studies on gamma oscillations in children with autism (Baruth et al., 2010; Sokhadze et al., 2009) that showed alterations of evoked and induced gamma oscillations during attention tests especially well present at the frontal topographies.

Several of the studies of NFB for autism include the use of randomization, some form of blinding, formal assessments of autism diagnoses,  neurocognitive domain assessment, and documentation of EEG changes. Future directions for research should include larger double-blind sham-controlled RCTs with follow-up, tests of the validity of the blind, use of standardized outcome measures, assessment and control of comorbidity and concomitant treatments, and monitoring/reporting of adverse effects. Regarding clinical recommendations, applying the USPSTF Level of Certainty of Research Evidence and Recommendation Grade, the quality of data-based evidence of NFB for autism was rated by Hurt et al. (2014) as “fair,” and according to the USPSTF guidelines, the clinical recommendation is to “recommend” NFB for treating autism.  However, given that the main positive results on NFB for autism derive from small studies with relatively small samples, this recommendation is again qualified with reservations and mainly relevant for families who have tried or considered conventional psychosocial treatments for autism, have the time and money and efforts to invest in NFB training course.