Share this article:


  • Join our comunity:

An interview with Michelle Hampson on fMRI neurofeedback: what it is, the nature of its experiments, clinical applications and the ethical issues around it

By: , Posted on: November 25, 2021

Elsevier: Michelle, you have just edited, and contributed as an author, a book on fMRI Neurofeedback.  Can you tell us what is fMRI neurofeedback and how does it work?

fMRI neurofeedback is a functional MRI technique in which participants are given feedback during a scanning session regarding how some specific, targeted aspect of their brain function is changing. The aim is to train them to control that aspect of their brain function. In an explicit training paradigm, participants are asked to actively try to control the targeted aspect of their brain function, using the feedback as a training signal. In an implicit training paradigm, participants are not told what the feedback signal reflects, but are simply rewarded (e.g. financially) when the feedback reflects a change in brain function desired by the experimenter. Studies have shown that both training approaches can be effective for altering brain activity patterns in a targeted manner and inducing specific changes in cognitive, perceptual, or emotional function.

Elsevier: Can you describe a typical explicit training experiment and an implicit one?

Well, I should be clear that although the explicit/implicit distinction in fMRI neurofeedback studies suggests a binary division in how much information participants have, in reality there is much more of a spectrum. On the far implicit end, there was study by Michal Ramot and colleagues where participants did not even know they were participating in a brain training experiment and were led to believe they were in a neuroimaging study of reward. During the scanning session, they heard auditory signals that they were told would result in positive or negative financial gain for them. However, they were unaware that these signals were based on their brain activity patterns and that they were thus being trained covertly towards the brain patterns associated with the positive reward.  On the other end of the spectrum, the explicit training end, are open label clinical studies, where the participants know what is being trained, and how it is hypothesized to affect their symptoms. An example of this is the study by Jenny Zaehringer, Christian Paret, and colleagues in which patients with borderline personality disorder were trained to control their amygdala activity to see if this translated into improvements on emotion regulation measures. The participants in this study knew that the feedback signal represented activity in an area of their brain involved in emotion processing, and understood the aim of the training was to improve their emotion regulation, so there was essentially no omission of information or deception involved. The vast majority of fMRI neurofeedback studies fall somewhere between these two. Many clinical trials in neuropsychiatric populations are explicit in the sense that the participants understand what the goal of the experimental intervention is. However, participants are often blind to whether they are receiving the experimental intervention or a control intervention, such as sham feedback. So, there is still information withheld in these so-called explicit paradigms. Similarly, on the implicit side of the spectrum, many paradigms that do not inform subjects about what is being trained in their brain, or what aspect of mental function the training is expected to alter, still do make it clear that the study involves a brain training paradigm and that the feedback signal is tracking something in their brain.  So, in short, the amount of information subjects have about the neurofeedback training protocol they receive varies across studies, but the implicit/explicit distinction is typically drawn based on whether they know what aspects of brain function and mental function are targeted by the training. 

The amount of information provided to subjects will generally depend on the aims of the study and should be balanced by ethical considerations. This issue is discussed in more detail in the chapter on Protocol Design (Chapter 3).

Elsevier: What does fMRI neurofeedback offer over and above classic neuroimaging techniques?

While classic neuroimaging is used to examine brain function, fMRI neurofeedback can be used to alter brain function in a targeted manner. This has potential clinical value, in that it may enable us to train people towards healthier brain patterns. It also has value as a basic science tool that can be used to test hypotheses regarding how specific aspects of brain function give rise to cognition, perception, and emotion. Thus, fMRI neurofeedback enables studies of mind-brain relationships that go beyond the correlational nature of classic neuroimaging studies to explore causal influences of brain function on mental function.

Let me give a concrete example that illustrates what I am saying here.  Traditional neuroimaging can be used to identify patterns associated with facial preference in the cingulate cortex. However, these studies do not clarify whether activity patterns in this region are causally influencing facial preference, or are downstream effects of it.  However, an implicit neurofeedback study that trained activity patterns in the cingulate cortex associated with facial preference demonstrated that up-regulating these patterns when viewing specific faces increased preference for those faces, and conversely that down-regulating the patterns when viewing specific faces decreased preference for those faces. I am referring here to the Shibata et al. study published in 2016.  This neurofeedback study showed that inducing specific patterns of activity in the cingulate cortex influenced facial preference and thus provided evidence that the cingulate activity patterns do not just reflect preference,  but determine it. What I am getting at is that because fMRI neurofeedback enables us to alter specific aspects of brain function and measure the resulting effects, it can be used to explore causal relationships more effectively than classic neuroimaging techniques.

Elsevier: Brain disease is a major health issue today. How does fMRI contribute to the challenges of tacking brain disease and what are its prospects for tackling the issue in the future?

Much of the clinical research in our field has been focused on exploring the potential of fMRI neurofeedback interventions for directly treating neuropsychiatric disorders by training people towards healthier brain patterns. The work that has been done to date is described in detail in the Clinical Applications section of the book and David Linden has provided a helpful table in his introduction to the clinical section that summarizes the current literature. A good example application is the treatment of depression.  Multiple groups have reported promising data from studies that used fMRI neurofeedback to train patients with depression towards healthier brain patterns as described in Chapter 8. One example is the paradigm used by Kym Young and colleagues where participants with depression were trained to upregulate their amygdala activity during positive autobiographical memory recall. In a double-blind study comparing subjects who received this training to a control group of subjects that were trained to upregulate another brain region, they found that the group that was trained on the amygdala had significantly greater clinical improvement. Although data from studies such as this one have shown promise, it should be emphasized that these are early clinical development trials, they are not the kind of large sample size Phase III clinical trials that are needed before a mental health intervention is considered to be an evidence-based intervention. In short, the field is young and more work is needed, but there are promising results in some of these early studies.

In addition to the potential for fMRI neurofeedback to be used directly as a clinical intervention, it can also be used to study the neural basis of clinical conditions, by training brain patterns in specific networks and seeing how that training impacts symptoms. So, here I am referring again to the basic science value of neurofeedback that was discussed before. When this is leveraged in the study of neuropsychiatric disorders or symptoms, it can be a powerful tool for furthering our understanding of the neural basis of brain dysfunction.

Elsevier: For someone setting up a clinical study using fMRI neurofeedback, what are the main issues they need to consider?

To develop a clinical fMRI neurofeedback intervention, the first step is to identify an aspect of brain function that can be measured with functional MRI that you believe has a causal and substantial impact on symptoms. Neuroimaging biomarkers identified in functional MRI studies as correlated with symptoms are often used. It’s important, however, to ensure their relationship to symptoms has a large effect size so that if you succeed in altering neural function in the desired manner then there is a reasonable possibility you will induce a change in symptoms that is substantial enough to be clinically meaningful.

In addition to identifying a promising target to train, it is also important to consider the ethical implications of the training. For many individuals with neuropsychiatric disorders, their symptoms are an integral part of their identity, so there must be a clear understanding of the aims and how the training may affect their symptoms in the consent process. The potential for adverse effects of the intervention, particularly if symptom provocation is involved, must also be carefully considered and managed.

Elsevier: Does your book give advice on how to address the ethical issues on setting up a clinical experiment?

Ethical issues associated with fMRI neurofeedback, relevant to both clinical and basic science experiments, are discussed in Chapter 14. However, the aim of this chapter is not to give advice, per se, so much as to encourage researchers to carefully think through the numerous issues of relevance, so that they understand and can appropriately communicate and manage any risks to subjects or the public that their work may involve.

Elsevier: What value does fMRI neurofeedback bring to human neuroscience?

fMRI neurofeedback can be used to alter an aspect of brain function and the downstream effects on mental function can be measured, allowing exploration of causal relationships between the brain and the mind.  As an interventional tool for changing human brain function, it is relatively safe (based on natural learning processes) and it offers both flexibility and specificity in targeting desired aspects of brain function.

Elsevier: You mentioned earlier about implicit training programs. What are the ethical issues around these?

The potential to train participants towards specific brain patterns (and thus mould their mental function) without the participants having any awareness of what is being trained raises ethical issues regarding risks to the autonomy, identity, and safety of the participants that must be carefully managed. Furthermore, if the knowledge obtained by fMRI neurofeedback studies is used to guide development of more portable and accessible neurofeedback interventions, it could contribute to the development of tools that enable widespread and subconscious social influence. This is not an immediate concern but is something for the field to be wary of.

Elsevier: Do you mean there is potential for people to be kind of brain-washed if the technology is put in the wrong hands? If so, what ethical measures need to be addressed and when and by whom?

The term brain-washing implies radically changing a person’s strongly held beliefs. This is not something fMRI neurofeedback has been demonstrated capable of to date. However, it has been shown to be capable of influencing a person’s mental function, including their perception, cognition and emotion, in a variety of ways, and the potential for that sort of influence to be misused is a concern. The questions you raise of what should be done to protect against such misuse, and when action is needed, and who should lead the effort are all very important questions. I have been struggling with these questions myself lately. I do think the fMRI neurofeedback community needs to play an active role, given that they are best positioned to understand the risks. So, I recommend we create a forum that fosters regular discussion of ethics among fMRI neurofeedback researchers and perhaps can also lead to the development of recommendations for broader action if that is deemed necessary. The next real-time functional imaging and neurofeedback (rtFIN) meeting, to be held here at Yale University in 2022, may provide a good opportunity to begin this effort.

Elsevier: What are the big challenges for the field in next few years?

The biggest challenge at present is the development of effective training paradigms. Some paradigms have proven successful at inducing the desired changes in brain patterns and others have not. There is a large parameter space to be explored in attempting to optimize neurofeedback training and a great deal of work is needed before we understand how this tool can be most effectively leveraged.

Elsevier: Can you give an example of a training program that has been successful and one that has not?

Some of the most striking examples of successful training have been basic science studies using decoded neurofeedback (DecNef) in healthy participants. DecNef is an implicit training technique, developed by Mitsuo Kawato and colleagues, that trains patterns of brain activity in specific brain areas. This approach has been used successfully to train visual perception and facial preference, to induce associative learning, to alter confidence without changing competence, and to reduce physiologic responses to feared stimuli.  

On the other hand, a number of clinical studies training patients groups have reported unsuccessful training attempts. For example, a study in treatment-seeking nicotine dependent participants run by Colleen Hanlon and colleagues was not successful in training participants to upregulate a dorsomedial prefrontal region believed to be involved in resisting craving, although they had more success training down a region of ventral anterior cingulate cortex. Similarly, a study by Ranganatha Sitaram and colleagues that attempted to train four psychopaths to upregulate the anterior insula during negative emotional imagery reported success in only one of the four participants. It is notable that this study paid participants based on their regulation success during training as the DecNef studies do. Performance-based pay is an unusual approach in a clinical study that may be helpful for amplifying learning, and could be contributing to the success in the basic science training paradigms. However, the limited success in this study of psychopaths despite their use of performance-based pay suggests there are other factors important for successful training. It is likely that some aspects of brain function are just difficult for certain populations of people to learn to control. In such cases, another aspect of symptom-relevant brain function that is more plastic may provide a better training target. Examining how brain patterns change in those who respond well to neurofeedback in current clinical studies may provide insight into other clinically relevant targets that may be easier to train and that future studies can focus on.

Elsevier: What advice would you give to researchers who want to begin using fMRI neurofeedback?

Two things come to mind:

First, check out the book! It was developed specifically as a guide for researchers entering our field. People come into our field from many different backgrounds, so the chapters that are most useful will be different for different people and it’s fine to skip around and read what is most helpful to you.

Second, I recommend you follow-up your subjects for weeks to months after neurofeedback, monitoring both the desired changes in mental function that were targeted by the intervention and also monitoring for any adverse events. We have found that effects of neurofeedback can grow over time after the training has been completed, so monitoring long-term effects ensures that you do not miss the time-point of greatest effect. Also, long-term monitoring of adverse events will be very helpful for the field to document the safety profile of different training paradigms and facilitate broader dissemination of the technique.


Ready to read this book?

fMRI Neurofeedback is now available on ScienceDirect. Or buy your own copy on the Elsevier store  and save 30% + free shipping with promo code STC30.

Connect with us on social media and stay up to date on new articles


The scientific study of the nervous system is entering a new golden age. Researchers and clinicians continue to advance the treatment of conditions such as Alzheimer’s syndrome, Parkinson’s disease, epilepsy, and traumatic brain injury. Public initiatives like the federal Brain Research Through Advancing Innovative Neurotechnologies (BRAIN) program in the United States, announced in April 2013, ensure that funding and resources will continue to be applied to this rapidly growing field. Elsevier’s journals, books, eBooks, online references, and tools are respected around the world for everything from physiology and pathology to behavioral genetics and nerve repair. Our publications are a gateway to the latest advancements in neuroscience research and leading-edge data for professionals, students, and academics alike.