fMRI: A Result That Could Make Neuroscientists “Gasp” In Surprise

By Neuroskeptic | May 4, 2014 5:39 am

Many fMRI studies of brain activity could be confounded by the pattern of the participants breathing.

In a new paper published in Human Brain Mapping, Dutch neuroscientists Willem Huijbers and colleagues show that peoples breathing cycle tends to get synchronized (phase-locked) with the appearance of stimuli during cognitive tasks. Because the respiratory cycle is known to affect the fMRI signal, this is a worrying result, as it suggests that some ‘brain’ activations and deactivations recorded in fMRI studies may in fact reflect breathing.

Huijbers et al used a memory task in which participants are shown words, during the fMRI scan, and later have to try to remember them. It’s well known that brain activity differs between words that are later recalled (R) and those that are forgotten (F). The difference, reflecting memory encoding, is seen in two main brain areas.

In the ventrolateral prefrontal cortex (VLPFC), activity is higher for remembered words but in the posterior default mode network (DMN), it’s lower in response to words that get recalled. The current study replicated these findings.

However, Huijbers et al’s results show that it might be breathing, not brain function, that accounts for most of the DMN’s contribution to memory. They show that participants breaths tend to phase-lock to the appearance of words, and crucially, that breathing is more tightly locked to subsequently remembered words: this graph shows that on average participants take a breath 2-4 seconds after a word appears, especially for remembered words. If there were no phase-locking, the line would be flat.

huijbers_phase_lockingWe know that the respiratory cycle affects the fMRI signal, and that the DMN is especially prone to respiratory effects (for anatomical reasons).

A second aspect of the study provides more evidence for this. At certain points during the experiment, Huijbers asked participants to hold their breath for several seconds. During this time, the stimuli words kept appearing on the screen. Clearly, if you’re holding your breath, your breathing cycle is not phase-locked to anything.

So it’s remarkable that, while memory-related activity in the VLPFC (orange) was unaffected by breath-holding, the magnitude of the posterior DMN remembered-vs-forgotten effect (blue) was much smaller (65% lower) during the breath hold:


Altogether this is pretty strong evidence that breath cycle effects could be responsible for much of the task-related DMN deactivation. Huijbers et al did find some memory-related deactivation in that region, even during breath holding, so there does seem to be a real neural effect there as well. But the majority of the signal was respiratory in origin.

Nothing here should come as a huge surprise to neuroscientists, as the impact of breathing on fMRI has been known for a long time. But many people assume that in the course of an experiment, breathing effects will cancel out, especially when the experiment uses an event-related design with different stimuli appearing every few seconds.

Surely, neuroscientists have told ourselves, breathing won’t phase-lock over such a short time-scale? After reading this paper, it’s hard to feel so confident. (Although I wonder if the presence of breath-hold periods might have encouraged subsequent phase-locking by disrupting the ‘natural’ breathing cycle?)

Huijbers et al conclude that

These findings clearly show that it is critical to incorporate respiratory measures in fMRI analyses to account for respiratory/vascular effects, even when using fast event-related fMRI designs.

Luckily, measuring breathing is pretty easy – you just need a detector belt placed around the chest. Most MRI scanners already have the ability to do this, in fact I believe that the belt comes free with the scanner in many cases. So there’s no excuse for not using them.

As to why we tend to unconsciously phase-lock our breathing to stimuli, this remains unclear. But we’ve all experienced it: we’ve all gasped in surprise, or felt a moment take our breath away. It seems that those are only the most extreme examples of a general phenomenon.
ResearchBlogging.orgHuijbers W, Pennartz CM, Beldzik E, Domagalik A, Vinck M, Hofman WF, Cabeza R, & Daselaar SM (2014). Respiration phase-locks to fast stimulus presentations: Implications for the interpretation of posterior midline “deactivations” Human Brain Mapping PMID: 24737724

CATEGORIZED UNDER: fMRI, methods, papers, select, Top Posts
  • Uncle Al

    So there’s no excuse for not using them” Grant money stops arriving. If an entire discipline depends upon white swans only, once can be certain they will enjoy Cygnus atratus hunting vacations in southern Australia.

  • Pingback: fMRI: A Result That Could Make Neuroscientists ...()

  • fMRI_today

    A few relevant papers that have also pointed to breathing issues in fMRI – solvable, but you just have to be careful:

    R. M. Birn, J. B. Diamond, M. A. Smith, P. A. Bandettini, Separating respiratory variation-related fluctuations from neuronal activity-related fluctuations in fMRI, NeuroImage 31, 1536-1548 (2006)

    R.M. Birn, K. Murphy, D. A. Handwerker, P. A. Bandettini, fMRI in the presence of task-correlated breathing variations, NeuroImage 47, 1092-1104 (2009)

    R.M. Birn, M. A. Smith, T. B. Jones, P. A. Bandettini, The respiration response function: the temporal dynamics of fMRI signal fluctuations related to changes in respiration. NeuroImage, 40, 644-654 (2008)

    • DS

      And what you mean by “solvable”? Regression of respiration waveforms? Do you think that is enough?

  • petrossa

    So how many papers from the years gone by are going to be retracted finally? fMRI just fails about every imaginable lowest level of reliability due to a mix of many causes, this latter being just one more. How much money/time has been wasted on following spurious ‘results’? This must be incalculable. But no, we’ll just go on slapping another band-aid on the holy grail. A breathing belt. Yay. Onwards to the next issue. Maybe best is just to paralyze the fMRI victims in the hope getting some kind of semi-reliable signal

    • neurd

      What a downer. So what non-invasive functional imaging modality has no cons?

      • petrossa

        none, but EEG like results from a millions $$$ machine seems like a bit of a waste of money to me

  • Pingback: fMRI: A Result That Could Make Neuroscientists "Gasp" In Surprise … | the scan book()

  • BadMonkey

    If by “neurosceintists” one specifically means fMRI users who have never heard of a pneumatic belt …then still no! This is only a problem for experiments:
    1) where stimuli are presented close to the average breathing rate at rest (12-15 breaths per minute)

    2) using an event related design.

    Also, asking subjects to hold their breath is obviously going to disrupt “default mode network” activity since actively overriding autonomic responses requires effort. Duh!

    • Neuroskeptic

      This problem is limited to event-related designs, yes; but block designs suffer from another related problem which is changes in breathing rate between blocks.

      As to the fact that breath-holding will disrupt the DMN, the authors noted this but they also showed (I didn’t have space to discuss this) that – across subjects – there’s a correlation between the degree of phase-locking in breathing conditions, vs the degree of DMN suppression during breath holding.

      Which is consistent with the idea that one factor unites both things, i.e. phase-locking-driven DMN ‘deactivations’ during normal breathing.

      • BadMonkey

        One reason why nobody is worried about this is that it doesn’t really generalize to other tasks, and is only a problem if you don’t know about it. For example, if I ask subjects to press a button when they see an image on the screen, I’d be an idiot to conclude that the finger area of motor cortex reflects visual perception. So long as experimenters have the basic foresight (which I agree is too often lacking) to control for factors beyond those of interest across conditions (e.g. breathing), then there is no problem. @sander, studies of clinical populations should not be directly comparing levels
        of activation between populations, but rather differences in activation
        between conditions for different populations.Therefore they will not be affected in the way that you suggest.

  • sander

    As one of the authors I have to intervene in this discussion.

    First, Neuroskeptic wrote a very nice article about our paper. The comment about the correlation contains a slight slip up though. It should be there’s a correlation between the degree of phase-locking in
    breathing conditions, vs the degree of DMN suppression during NORMAL BREATHING. So in this analysis there are no confounds regarding breath-holding.

    Second, regarding the block designs and breathing rate, we also showed that in the same paper. Participants’ breathing rate was indeed lower during interleaving rest blocks (also included in the design) than during the normal breathing task blocks.

    We showed that this breathing rate difference also correlated with blocked task-induced “deactivations” (rest vs. normal breathing task) in the DMN. This is briefly mentioned in the discussion, but reported more elaborately in the supplemental materials including the figures.
    So, both efMRI and blocked fMRI designs suffer from the same problem I am afraid.

    • Neuroskeptic

      Thanks very much for the comment! You’re quite right, I misremembered the details of that correlation.

  • sander

    @Neuroskeptic:disqus Ah well, just a sligh slip up in a quick reply,other than that nice coverage of our article and the big fMRI monster called vascular/respiration…

    I agree with petrossa. How do we know that studies published in PNAS about autism and schizophrenic patients etc. do not just reflect arousal/respiration. How do we know that individual differences in fMRI aging studies do not reflect some degree of vascular integrity problems without checking? The key is we don’t know. Cognition, vasculature/respiration, and fMRI signal are all interrelated and difficult to tease appart.

    In older adults, the least you can do is measure brain perfusion using ASL or other techniques to account for individual differences in vascular integrity (some groups are doing that already). And the other thing is always measuring respiration during scanning, but then again I agree with DS: is that sufficient?

    • poniesinjudah

      No. The fMRI tech just sucks. It only exists because regular MRI was invented to do something very different. No one would have invented something this equivocal and statistics based from scratch. A machine in science is supposed to give you reasonably direct information. fMRI isn’t like a microscope, no. But it’s not even like culturing microbes. Culturing is a process not a machine, yes. But one that recapitulates nature. FMRI has no relationship to anything that happens in real biology. The way culturing is just growth but in a dish vs in your throat. We are told, and want to believe fMRI lets us look inside the black box of the brain but it doesn’t. That would be direct like watching microbes grow. FMRI is measurement not watching. And it is the illusion of measurement. 1) statistical 2) multiple measurements of multiple things (MR angiography is measuring tissue densities to make the pic of flowing blood) each with own degree of noise, then all combined. It’s a mess and we’re relying on non-empirical statistics to sort it out. That’s not good. On top of that the task test model is derived from ablation studies in nonhuman animals. That were very much simpler in conception. ‘This bit does movement.’ Our concepts of how to study the brain and what questions to ask has been taken over by the fMRI mode. We need new questions and we need new tech. This is not that big a deal. This happens in science.
      I like Petrossa’s “bandaid on grail” phrase. I’d add a naked emperor. And kudos to Sander, et al.

  • Pingback: Spike activity 09-05-2014 « Mind Hacks()

  • Pingback: Spike activity 09-05-2014 | NYC Neurofeedback()

  • Pingback: Spike activity 09-05-2014 | Connecticut Neurofeedback()

  • Matthew Constantinou

    Couldn’t breath-holding be considered as a specific ‘task’, which could have reduced DMN responses, rather than respiratory confounds per se?

  • sander

    @badmonkey I was thinking more about resting state than task-based studies

  • sander

    @disqus_SPifgyDbeZ:disqus: and what about emotional/pain-related studies?



No brain. No gain.

About Neuroskeptic

Neuroskeptic is a British neuroscientist who takes a skeptical look at his own field, and beyond. His blog offers a look at the latest developments in neuroscience, psychiatry and psychology through a critical lens.


See More

@Neuro_Skeptic on Twitter


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Collapse bottom bar