Download the
What's happening in MR and what it means for you? READ NOW in the latest Q4 2017 GRIT Report!

The Restless Resting State, and Why Brain Scanning is Still Valid


By Thomas Zoëga Ramsøy

Is brain scanning data invalid? A recent meta-analysis study published in the esteemed journal PNAS ( has suggested that decades of brain research studies should be questioned! An estimated 40,000 studies using the established and used method fMRI (or functional Magnetic Resonance Imaging) may be hampered by a software bug that has gone undetected for too long.

In the study, several brain scanning studies of the so-called “resting-state data” suggest that as many as 70% of “positive findings” may indeed be false. That is, 7 out of 10 reported “brain blobs” found when people are resting inside a brain scanner may in fact never be there!

This is a very interesting finding indeed. It sure reminds us of the need for replications in science, a part of scientific housecleaning that is never honoured as part of a scientist’s career. It is worth noting that the finding pertains to only one type of neuroimaging studies, fMRI (and most likely only the so-called blood oxygenation level dependent, or BOLD, fMRI), and not other methods such as electroencephalography (EEG) or magnetoencephalography (MEG). Still, the problem could be substantial, as many of our insights about the brain may be unsupported.

But is it really that bad? There is a saying that “extraordinary claims require extraordinary evidence” and here, we have a problem. Indeed, one major caveat of the study is the assumption that the resting-state data, used in this meta-analysis, can be treated as equal between individual. After all, when asking people to rest (often between more active tasks), we can assume that the brain at rest is similar for us all, correct? Instead, decades of psychological studies of the mind at rest – including mind-wandering and Task Unrelated Images and Thoughts – have shown extreme degrees of variation both within and between individuals. What follows are inspired by a yet unpublished manuscript that I have co-authored with Prof. Bernard J Baars from the Neuroscience Institute, which I hope can help put a perspective to the intended killing of fMRI.

The restless resting state

Considerable attention has been called to a putative “resting state” of the brain, observed during designated rest breaks in neuroimaging experiments. Robust brain differences have been found between task-related (TR) and task-unrelated (rest break) conditions. Some scientists speculate that TU brain activity may reflect a special state of the brain, sometimes called a “resting state,” “default mode,” or a “baseline condition”. We suggest that the explanatory use of these terms is premature. Instead, a large empirical literature points to an alternative account: people during rest breaks are reverting to their normal, spontaneous stream of thought, which is subjectively rich and self-relevant, highly variable, multimodal, often explicitly goal-directed, and probably functional. Even the word “state” may be premature, since it suggests a stable condition of the brain. Instead, some five decades of psychological studies shows TU activities to be dynamic, heterogeneous, shaped by emotional and motivational primes, and focused on current life concerns. The term “spontaneous thought” might therefore be a more accurate label for task-unrelated brain activity.

The spontaneous brain

Scientists tend to be cautious about self-reported experiences, but some facts about consciousness are as predictable as objects falling in earth gravity. The entire field of psychophysics relates precisely controlled stimuli to reliable subjective reports. Even endogenous events can be reported reliably, as in experimental studies of verbal rehearsal and visual imagery, and their brain correlates. Less well-known is almost a half century of thought-sampling studies, using real-time reports under known conditions. One of the oldest psychological demonstrations is to simply close one’s eyes and try to stop the flow of thought. We can read sources over some 26 centuries reporting how difficult that is to accomplish. The flow of spontaneous experience appears to have its own persistence and “urgency,” as William James wrote a century ago. Some five decades of systematic psychological research supports the notion that spontaneous cognition is not random, but reflects “current, personal concerns”. There is also evidence for repetitive long-term themes in spontaneous mentation, influenced by major life events, traumatic experiences, and implicit goals. In everyday life, spontaneous, apparently unstructured thinking may be the most common kind of goal-directed thought.

What happens when people go from a focused, externally instructed cognitive task to a condition that is not heavily structured by external demands?  To many subjects, designated rest breaks may be a chance to get back to a normal, spontaneous, self-relevant, and active stream of thought. In the 1960’s and 1970’s a number of studies in experimental psychology focused on internally generated images and thoughts, which showed that depriving the mind from sensory information stimulated the occurrence of internally generated experiences. TU thoughts were studied in a more elaborate and specific research program called Task Unrelated Images and Thoughts (TUITS). Here, subjects were asked to report with regular intervals the content of their thoughts. This could either be performed in laboratory settings, in which subjects were given a tasks of varying difficulty and attentional load, or they could focus on more everyday settings, where subjects were interrupted at random times during a day, and were to write down their ongoing thoughts. One of the general findings from this research was that there is a continuous shifting of attention between externally and internally generated sources of information. Furthermore, spontaneous thoughts were found to be rather repetitive and predictable, always returning to “current concerns”. The content of thought was found to become increasingly unrelated to external events as these external events become more static and predictable. In this sense, the more boring the task, the more did people spend time (during testing) on task-irrelevant thoughts.

More recently, the study of TUITS has re-emerged in the scientific literature. A portion of this research has focused on detailed observations about the influence of TUITS on cognitive performance and a detailed examination of the intrusiveness of the task unrelated thoughts on both ongoing and later performance. Other studies have begun to couple the occurrence of TUITS to measurable physiological changes such as increased heart rate during TUITS. As with the original TUIT studies, these results confirm that increases in task difficulty, altering the attentional load, make TUITS become significantly less frequent. On the other hand, the easier the task – leading to more automatic behaviour – the more TUITS are reported by subjects. In this sense, a rest state is only at one extreme of how much attentional load is put on a subject’s mind. At the other end are highly energy- and attention-demanding tasks such as working memory 2-back or 3-back tasks. The conditions we are comparing in a resting state study are vital to our interpretation of the results.

Daydreaming was also studied in the same period as TUITS was explored. Using the “Imaginal processes inventory” it was shown that people were aware of some daydreaming every day, and that daydreams ranged from “obvious wishful thinking to elaborate and complex visions of frightening or guilty encounters”. Furthermore, factor analyses have revealed three major types of daydreaming: positive-constructive; guilty-dysphoric; and a poor attentional control pattern characterized by fleeting thoughts and an inability to focus on extended fantasy. At the same time, the test-retest reliability of daydreaming reports has been found to be high. As such, the literature on TUITS and daydreaming, both highly relevant inputs to the study of RS, is both rich in number of studies and in information about the richness of conscious content during such periods.

The non-death of fMRI findings

So what may initially seem like a failure of replication of brain scanning data, may in fact be a failure to understand the human mind. Many neuroscientists just don’t know the psychological literature enough to see that that the human mind does not “do anything” when asked to rest and relax. Indeed, as we have seen above, the mind is never at rest, and when not given an explicit task, it will defer to an inner state of “current concerns.”

So the basic assumption of this meta-analysis is flawed, and we should treat the conclusion accordingly. Analysing resting state brain activity and assuming that the results will be convergent is the same as claiming that every human mind at any time is thinking about the same things.

That said, we still replications of brain scanning studies, as we would need whether one scanning method is valid. Here, neuroscience fortunately has a vast toolbox, and now classical work from leading figures such as Nicos Logosthetis has provided clear links between fMRI studies and other measures of brain activity. For example, Logosthetis’ work has clearly shown that fMRI activation “blobs” are related to dendritic activity, rather than axonal activity. Still, there is nothing in these data to support a claim that fMRI data are invalid. Indeed, extraordinary claims have not been supported by extraordinary evidence in this case.


Interested in learning more? Sign up for Thomas Zoëga Ramsøy’s free Coursera course on neuromarketing and consumer neuroscience.

Please share...

7 responses to “The Restless Resting State, and Why Brain Scanning is Still Valid

  1. I think there are a couple of l flaws in your argument. First the brain at rest may have heterogeneous states of activity across individuals, but having a MRI is far from restful. I speak from personal experience, it is very noisy and cramped in the MRI tube. You may or may not have music playing on headphones. You may become claustrophobic. I would posit there is precious little alpha wave rhythm during a MRI. More significantly you propose that people have heterogenous restful mental states but imply that when they are aroused they snap into homogenous brain states. So fMRI works because we have homogenous brain states during a task, but we all start from different points ? I think this is a radically false assumption. Ergo, fMRI must suffer from the problem of heterogeneous “active” mental states. Assuming the brain is organized in the same across individual is manifestly false if you read some of the literature. We have cases of people with no cerebellums that are alive. A case of a french civil servant who has a tiny, flattened, cortex who lives normally. We do not know the variance in brain structure across the population as relatively few people have had MRI’s. The core dogma of MRI is localization of function in the brain. For the limbic system localization pretty is clear, but then do we reduce people to high functioning lizards ? Cortical plasticity and variations are well established, we don’t know what that the variance is in terms of “fixed locations of function”. This alone is a fundamental flaw with fMRI, let alone the type 1 error problems described in the paper which I think are valid. I don’t see how you have rebutted the paper in any way.

  2. Dear Andrew

    Thanks you for your detailed comment. I must say that I respectfully disagree with you on every single point made. So let me take this in turn.

    First, the claim is NOT that the brain at rest in an MRI is restful — even the title suggests otherwise. If I misunderstand, then we’re in agreement. If not, then I think it’s odd that the title would not imply a different stance on my behalf. The claim is that rest makes participants indulge in a variety of states, and indeed, rest-related alpha waves are not seen rising under these conditions. So let’s not make straw men.

    Yes indeed, the argument that the mind goes from a heterogeneous resting state to a homogenous active state. There are two reasons for this. First, there is the issue of experimental control. Here, “rest” can be seen as a poorly controlled experimental condition, and should not be seen as a single mental process. Experimenters basically lose control over the actual mental activity that happens in these cases. Conversely, an active task-related phase is an externally instructed and more well controlled condition. This distinction alone implies that active and passive phases are substantially different, and that implications from a poorly controlled, restless resting state hardly can be said to make valid claims for more well-controlled active task-related phases.

    Second, there is the issue of homogeneity in responses. This is an emerging and growing field in psychology and neuroscience: the study in how even small samples can predict population responses, simply because the variation in task related activity is so small. The prime example is horror movies, in which a scary moment makes everybody respond with a shock response. Some show stronger responses than others, but the response found in the brain is pretty homogenous. Such homogenous responses are highly predictive of population responses. There are indeed individual differences to this, and aspects such as neurochemical and genetic differences account for some of this variation (see one of my earlier papers here:

    Therefore, we should see the mind as having a range of different states, from the very homogenous to the most heterogeneous (both within and between individuals). The brain is organised very much like a fingerprint: the overall structure is the same but with individual variations (although the fingerprint does not change over time, as the brain does).

    I must also argue fundamentally against the antiquated straw man account you are making on behalf of neuroscience, as if it is arguing that it holds a neo-phrenological account of a 1:1 relation between mind, function and brain region. Indeed, your example of cerebellar lesion or ablation is true: a person CAN exist normally without it. However, such ablations HAVE to occur in early youth, allowing the brain to restructure accordingly. The same goes for the rare cases of cortical thinning and other cases, in which people seemingly have a thin slice of brain, half a brain and so on. Indeed, these people mostly seem to be well functioning. I would still contend that a full neuropsychological profile would not show them to be well functioning on all counts. Also, it is well known that these patients do show significantly earlier signs of age-related mental disorders and degeneration of function. So I am not sure we can ever call these “normal” — it’s more a sensationalistic claim than anything else. During my years as a clinical neuropsychologist I have seen far too many cases in which 1) highly specific lesions lead to specific deficits, 2) global injuries/diseases lead to a diffuse aetiology, and 3) cortical thinning and related issues ALWAYS can be shown to have functional disorders.

    You argue that we do not know the variation in function. This is inherently false. Even my own fMRI PhD thesis work from 2008 mentioned this only in passing, and demonstrated variation in function as something needed to be dealt with. Prominent researchers such as Karl Friston has employed the classical biological language here to determine structure-function relationships, and distinguish between the terms: degeneracy, redundancy, and pluripotentiality, all which describe the complex relationship between brain and mind.

    So fMRI and neuroimaging in general has moved well beyond the “blobology” you refer to. This was criticised a decade ago, which in neuroscience terms is a millennium ago. Today, we study networks that support function, and are not married to the notion that a single structure equates 100% to a single function, or that a single function is tied to a single structure alone.

    So I find the criticism raised here ill-informed, archaic and too simple. Just as I found the original article to ignore decade-old well established observations — even century-old, if you take William James into account, or millennia-old if you take the Upanishads into account). And that I have to take issues with 😉

  3. This, from the article’s abstract, is what first caught my eye:

    “Functional MRI (fMRI) is 25 years old, yet surprisingly its most common statistical methods have not been validated using real data.”

    That aside, here is a discussion that may be of interest:

    Thomas (or anyone) can join the fray is they wish as far as I know.

    FYI, I have no commercial interests in neuroimaging whatsoever. I work with all sorts of data and the way neuroscience most affects my business is through potentially better quality ad pre-testing that I can incorporate into Market Response Modeling (aka Marketing Mix Modeling).

  4. For me the main problem remains the same: we do not have any proven data as a link between the measured activities and the content of the mental processes (or making two steps forward: the attitudes and the behavior). Even your argument with the horror movie shows very well that we should know the details of the perception and reception, though they show the same activities in certain parts of the brain some people would be excited, others horrified and some of them disgusted – according to their interpretation.
    (Moreover, perception is not a static process either, it involves the previous experiences, the cultural coding, etc.)
    So having an fMRI result at hand in an ad-test (in the best case if the algorithm is good and statistically correct 🙂 ) will say merely there was some activities in certain lobes nothing more.

    1. Dear József
      I beg to differ. In my comments earlier on this discussion I’ve provided a lot of links and references that, with a single click, should take you to documentation against the view you hold.
      What you make is an assertion that is no longer grounded in the science center we have.. It’s been a criticism of neuroscience that at one time was valid, but that’s at least 15 years ago now.
      The references and links I’ve provided should demonstrate otherwise.
      You may well hold a view that the brain is a completely mushy substance that has no specialised regions and that finding increase activity in one part of the brain has no causal relationship with thought and action. But that would be an assertion and view that is antiquated and completely misguided.
      If anything we also analyse neuroscience data as fluctuations and not as static “blobs in the brain”. The criticism raised in this discussion more than anything imply a call for more information and updates from neuroscience. The science behind this discipline has moved well beyond the questions raised here. So I understand that there’s much work to do, but suffice to say that the criticisms raised so far in this discussion can be answered easily, and can even be countered by a question to the effect of “if the brain is not causing the mind, what is?” – – arguing for a dualism between mind and body is as antiquated as ever, but it seems the arguments come from such a position and an antagonism towards anything with a smack of biology.

Join the conversation

Sign up to our newsletter

Don't miss out...
Get great insights content delivered straight to your inbox.
I agree to receive emails with insights-related content from GreenBook.

You can manage your email preferences or unsubscribe at any time. GreenBook protects your privacy under the General Data Protection Regulation.