Top 200 terms in cognitive neuroscience according to neurosynth

Tonight I was playing around with some of the top features in neurosynth (the searchable terms with the highest number of studies containing that term). You can find the list here, just sort by the number of studies. I excluded the top 3 terms which are boring (e.g. “image”, “response”, and “time”)  and whose extremely high weights would mess up the wordle. I then created a word-cloud weighted so that the size reflects the number of studies for each term.

Here are the top 200 terms sized according to number times reported in neurosynth’s 5809 indexed fMRI studies:


Pretty neat! These are the 200 terms the neurosynth database has the most information on, and is a pretty good overview of key concepts and topics in our field! I am sure there is something useful for everyone in there 😀

Direct link to the wordle:

Wordle: neurosynth

oh BOLD where art thou? Evidence for a “mm-scale” match between intracortical and fMRI measures.

A frequently discussed problem with functional magnetic resonance imaging is that we don’t really understand how the hemodynamic ‘activations’ measured by the technique relate to actual neuronal phenomenon. This is because fMRI measures the Blood-Oxygenation-Level Dependent (BOLD) signal, a complex vascular response to neuronal activity. As such, neuroscientists can easily get worried about all sorts of non-neural contributions to the BOLD signal, such as subjects gasping for air, pulse-related motion artefacts, and other generally uninteresting effects. We can even start to worry that out in the lab, the BOLD signal may not actually measure any particular aspect of neuronal activity, but rather some overly diluted, spatially unconstrained filter that simply lacks the key information for understanding brain processes.

Given that we generally use fMRI over neurophysiological methods (e.g. M/EEG) when we want to say something about the precise spatial generators of a cognitive process, addressing these ambiguities is of utmost importance. Accordingly a variety of recent papers have utilized multi-modal techniques, for example combining optogenetics, direct recordings, and FMRI, to assess particularly which kinds of neural events contribute to alterations in the BOLD signal and it’s spatial (mis)localization. Now a paper published today in Neuroimage addresses this question by combining high resolution 7-tesla fMRI with Electrocorticography (ECoG) to determine the spatial overlap of finger-specific somatomotor representations captured by the measures. Starting from the title’s claim that “BOLD matches neuronal activity at the mm-scale”, we can already be sure this paper will generate a great deal of interest.

From Siero et al (In Press)

As shown above, the authors managed to record high resolution (1.5mm) fMRI in 2 subjects implanted with 23 x 11mm intracranial electrode arrays during a simple finger-tapping task. Motor responses from each finger were recorded and used to generate somatotopic maps of brain responses specific to each finger. This analysis was repeated in both ECoG and fMRI, which were then spatially co-registered to one another so the authors could directly compare the spatial overlap between the two methods. What they found appears at first glance, to be quite impressive:
From Siero et al (In Press)

Here you can see the color-coded t-maps for the BOLD activations to each finger (top panel, A), the differential contrast contour maps for the ECOG (middle panel, B), and the maximum activation foci for both measures with respect to the electrode grid (bottom panel, C), in two individual subjects. Comparing the spatial maps for both the index and thumb suggests a rather strong consistency both in terms of the topology of each effect and the location of their foci. Interestingly the little finger measurements seem somewhat more displaced, although similar topographic features can be seen in both. Siero and colleagues further compute the spatial correlation (Spearman’s R) across measures for each individual finger, finding an average correlation of .54, with a range between .31-.81, a moderately high degree of overlap between the measures. Finally the optimal amount of shift needed to reduce spatial difference between the measures was computed and found to be between 1-3.1 millimetres, suggesting a slight systematic bias between ECoG and fMRI foci.

Are ‘We the BOLD’ ready to breakout the champagne and get back to scanning in comfort, spatial anxieties at ease? While this is certainly a promising result, suggesting that the BOLD signal indeed captures functionally relevant neuronal parameters with reasonable spatial accuracy, it should be noted that the result is based on a very-best-case scenario, and that a considerable degree of unique spatial variance remains for the two methods. The data presented by Siero and colleagues have undergone a number of crucial pre-processing steps that are likely to influence their results: the high degree of spatial resolution, the manual removal of draining veins, the restriction of their analysis to grey-matter voxels only, and the lack of spatial smoothing all render generalizing from these results to the standard 3-tesla whole brain pipeline difficult. Indeed, even under these best-case criteria, the results still indicate up to 3mm of systematic bias in the fMRI results. Though we can be glad the bias was systematic and not random– 3mm is still quite a lot in the brain. On this point, the authors note that the stability of the bias may point towards a systematic miss-registration of the ECoG and FMRI data and/or possible rigid-body deformations introduced by the implantation of the electrodes), issues that could be addressed in future studies. Ultimately it remains to be seen whether similar reliability can be obtained for less robust paradigms than finger wagging, obtained in the standard sub-optimal imaging scenarios. But for now I’m happy to let fMRI have its day in the sun, give or take a few millimeters.

Siero, J. C. W., Hermes, D., Hoogduin, H., Luijten, P. R., Ramsey, N. F., & Petridou, N. (2014). BOLD matches neuronal activity at the mm scale: A combined 7T fMRI and ECoG study in human sensorimotor cortex. NeuroImage. doi:10.1016/j.neuroimage.2014.07.002


Effective connectivity or just plumbing? Granger Causality estimates highly reliable maps of venous drainage.

update: for an excellent response to this post, see the comment by Anil Seth at the bottom of this article. Also don’t miss the extended debate regarding the general validity of causal methods for fMRI at Russ Poldrack’s blog that followed this post. 

While the BOLD signal can be a useful measurement of brain function when used properly, the fact that it indexes blood flow rather than neural activity raises more than a few significant concerns. That is to say, when we make inferences on BOLD, we want to be sure the observed effects are causally downstream of actual neural activity, rather than the product of physiological noise such as fluctuations in breath or heart rate. This is a problem for all fMRI analyses, but is particularly tricky for resting state fMRI, where we are interested in signal fluctuations that fall in the same range as respiration and pulse. Now a new study has extended these troubles to granger causality modelling (GCM), a lag-based method for estimating causal interactions between time series, popular in the resting state literature. Just how bad is the damage?

In an article published this week in PLOS ONE, Webb and colleagues analysed over a thousand scans from the Human Connectome database, examining the reliability of GCM estimates and the proximity of the major ‘hubs’ identified by GCM with known major arteries and veins. The authors first found that GCM estimates were highly robust across participants:

Plot showing robustness of GCM estimates across 620 participants. The majority of estimated causes did not show significant differences within or between participants (black datapoints).
Plot showing robustness of GCM estimates across 620 participants. The majority of estimated causes did not show significant differences within or between participants (black datapoints).

They further report that “the largest [most robust] lags are for BOLD Granger causality differences for regions close to large veins and dural venous sinuses”. In other words, although the major ‘upstream’ and ‘downstream’ nodes estimated by GCM are highly robust across participants, regions primarily effecting other regions (e.g. causal outflow) map onto major arteries, whereas regions primarily receiving ‘inputs’  (e.g.  causal inflow) map onto veins. This pattern of ‘causation’ is very difficult to explain as anything other than a non-neural artifact, as it seems like the regions mostly ‘causing’ activity in others are exactly where you would have fresh blood coming into the brain, and regions primarily being influenced by others seem to be areas of major blood drainage. Check out the arteriogram and venogram provided by the authors:

Depiction of major arteries (top image) and veins (bottom). Not overlap with areas of greatest G-cause (below).
Depiction of major arteries (top image) and veins (bottom). Note overlap with areas of greatest G-cause (below).

Compare the above to their thresholded z-statistic map for significant granger causality; white are areas of significant g-causation overlapping with an ateriogram mask, green are significant areas overlapping with a venogram mask:

From paper:
“Figure 5. Mean Z-statistic for significant Granger causality differences to seed ROIs. Z-statistics were averaged for a given target ROI with the 264 seed ROIs to which it exhibited significantly asymmetric Granger causality relationship. Masks are overlaid for MRI arteriograms (white) and MRI venograms (green) for voxels with greater than 2 standard deviations signal intensity of in-brain voxels in averaged images from 33 (arteriogram) and 34 (venogram) subjects. Major arterial inflow and venous outflow distributions are labeled.”

It’s fairly obvious from the above that a significant proportion of the areas typically G-causing other areas overlap with arteries, whereas areas typically being g-caused by others overlap with veins. This is a serious problem for GCM of resting state fMRI, and worse, these effects were also observed for a comprehensive range of task-based fMRI data. The authors come to the grim conclusion that “Such arterial inflow and venous drainage has a highly reproducible pattern across individuals where major arterial and venous distributions are largely invariant across subjects, giving the illusion of reliable timing differences between brain regions that may be completely unrelated to actual differences in effective connectivity”. Importantly, this isn’t the first time GCM has been called into question. A related concern is the impact of spatial variation in the lag between neural activation and the BOLD response (the ‘hemodynamic response function’, HRF) across the brain. Previous work using simultaneous intracranial and BOLD recordings has shown that due to these lags, GCM can estimate a causal pattern of A then B, whereas the actual neural activity was B then A.

This is because GCM acts in a relatively simple way; given two time-series (A & B), if a better estimate of the future state of B can be predicted by the past fluctation of both A and B than that provided by B alone, then A is said to G-cause B.  However, as we’ve already established, BOLD is a messy and complex signal, where neural activity is filtered through slow blood fluctuations that must be carefully mapped back onto to neural activity using deconvolution methods. Thus, what looks like A then B in BOLD, can actually be due to differences in HRF lags between regions – GCM is blind to this as it does not consider the underlying process producing the time-series. Worse, while this problem can be resolved by combining GCM (which is naïve to the underlying cause of the analysed time series) with an approach that de-convolves each voxel-wise time-series with a canonical HRF, the authors point out that such an approach would not resolve the concern raised here that granger causality largely picks up macroscopic temporal patterns in blood in- and out-flow:

“But even if an HRF were perfectly estimated at each voxel in the brain, the mechanism implied in our data is that similarly oxygenated blood arrives at variable time points in the brain independently of any neural activation and will affect lag-based directed functional connectivity measurements. Moreover, blood from one region may then propagate to other regions along the venous drainage pathways also independent of neural to vascular transduction. It is possible that the consistent asymmetries in Granger causality measured in our data may be related to differences in HRF latency in different brain regions, but we consider this less likely given the simpler explanation of blood moving from arteries to veins given the spatial distribution of our results.”

As for correcting for these effects, the authors suggest that a nuisance variable approach estimating vascular effects related to pulse, respiration, and breath-holding may be effective. However, they caution that the effects observed here (large scale blood inflow and drainage) take place over a timescale an order of magnitude slower than actual neural differences, and that this approach would need extremely precise estimates of the associated nuisance waveforms to prevent confounded connectivity estimates. For now, I’d advise readers to be critical of what can actually  be inferred from GCM until further research can be done, preferably using multi-modal methods capable of directly inferring the impact of vascular confounds on GCM estimates. Indeed, although I suppose am a bit biased, I have to ask if it wouldn’t be simpler to just use Dynamic Causal Modelling, a technique explicitly designed for estimating causal effects between BOLD timeseries, rather than a method originally designed to estimate influences between financial stocks.

References for further reading:

Friston, K. (2009). Causal modelling and brain connectivity in functional magnetic resonance imaging. PLoS biology, 7(2), e33. doi:10.1371/journal.pbio.1000033

Friston, K. (2011). Dynamic causal modeling and Granger causality Comments on: the identification of interacting networks in the brain using fMRI: model selection, causality and deconvolution. NeuroImage, 58(2), 303–5; author reply 310–1. doi:10.1016/j.neuroimage.2009.09.031

Friston, K., Moran, R., & Seth, A. K. (2013). Analysing connectivity with Granger causality and dynamic causal modelling. Current opinion in neurobiology, 23(2), 172–8. doi:10.1016/j.conb.2012.11.010

Webb, J. T., Ferguson, M. a., Nielsen, J. a., & Anderson, J. S. (2013). BOLD Granger Causality Reflects Vascular Anatomy. (P. A. Valdes-Sosa, Ed.)PLoS ONE, 8(12), e84279. doi:10.1371/journal.pone.0084279

Chang, C., Cunningham, J. P., & Glover, G. H. (2009). Influence of heart rate on the BOLD signal: the cardiac response function. NeuroImage, 44(3), 857–69. doi:10.1016/j.neuroimage.2008.09.029

Chang, C., & Glover, G. H. (2009). Relationship between respiration, end-tidal CO2, and BOLD signals in resting-state fMRI. NeuroImage, 47(4), 1381–93. doi:10.1016/j.neuroimage.2009.04.048

Lund, T. E., Madsen, K. H., Sidaros, K., Luo, W.-L., & Nichols, T. E. (2006). Non-white noise in fMRI: does modelling have an impact? Neuroimage, 29(1), 54–66.

David, O., Guillemain, I., Saillet, S., Reyt, S., Deransart, C., Segebarth, C., & Depaulis, A. (2008). Identifying neural drivers with functional MRI: an electrophysiological validation. PLoS biology, 6(12), 2683–97. doi:10.1371/journal.pbio.0060315

Update: This post continued into an extended debate on Russ Poldrack’s blog, where Anil Seth made the following (important) comment 

Hi this is Anil Seth.  What an excellent debate and I hope I can add few quick thoughts of my own since this is an issue close to my heart (no pub intended re vascular confounds).

First, back to the Webb et al paper. They indeed show that a vascular confound may affect GC-FMRI but only in the resting state and given suboptimal TR and averaging over diverse datasets.  Indeed I suspect that their autoregressive models may be poorly fit so that the results rather reflect a sort-of mental chronometry a la Menon, rather than GC per se.
In any case the more successful applications of GC-fMRI are those that compare experimental conditions or correlate GC with some behavioural variable (see e.g. Wen et al.  In these cases hemodynamic and vascular confounds may subtract out.
Interpreting findings like these means remembering that GC is a description of the data (i.e. DIRECTED FUNCTIONAL connectivity) and is not a direct claim about the underlying causal mechanism (e.g. like DCM, which is a measure of EFFECTIVE connectivity).  Therefore (model light) GC and (model heavy) DCM are to a large extent asking and answering different questions, and to set them in direct opposition is to misunderstand this basic point.  Karl, Ros Moran, and I make these points in a recent review (
Of course both methods are complex and ‘garbage in garbage out’ applies: naive application of either is likely to be misleading or worse.  Indeed the indirect nature of fMRI BOLD means that causal inference will be very hard.  But this doesn’t mean we shouldn’t try.  We need to move to network descriptions in order to get beyond the neo-phrenology of functional localization.  And so I am pleased to see recent developments in both DCM and GC for fMRI.  For the latter, with Barnett and Chorley I have shown that GC-FMRI is INVARIANT to hemodynamic convolution given fast sampling and low noise (  This counterintuitive finding defuses a major objection to GC-fMRI and has been established both in theory, and in a range of simulations of increasing biophysical detail.  With the development of low-TR multiband sequences, this means there is renewed hope for GC-fMRI in practice, especially when executed in an appropriate experimental design.  Barnett and I have also just released a major new GC software which avoids separate estimation of full and reduced AR models, avoiding a serious source of bias afflicting previous approaches (
Overall I am hopeful that we can move beyond premature rejection of promising methods on the grounds they fail when applied without appropriate data or sufficient care.  This applies to both GC and fMRI. These are hard problems but we will get there.

Mind-wandering and metacognition: variation between internal and external thought predicts improved error awareness

Yesterday I published my first paper on mind-wandering and metacognition, with Jonny Smallwood, Antoine Lutz, and collaborators. This was a fun project for me as I spent much of my PhD exhaustively reading the literature on mind-wandering and default mode activity, resulting in a lot of intense debate a my research center. When we had Jonny over as an opponent at my PhD defense, the chance to collaborate was simply too good to pass up. Mind-wandering is super interesting precisely because we do it so often. One of my favourite anecdotes comes from around the time I was arguing heavily for the role of the default mode in spontaneous cognition to some very skeptical colleagues.  The next day while waiting to cross the street, one such colleague rode up next to me on his bicycle and joked, “are you thinking about the default mode?” And indeed I was – meta-mind-wandering!

One thing that has really bothered me about much of the mind-wandering literature is how frequently it is presented as attention = good, mind-wandering = bad. Can you imagine how unpleasant it would be if we never mind-wandered? Just picture trying to solve a difficult task while being totally 100% focused. This kind of hyper-locking attention can easily become pathological, preventing us from altering course when our behaviour goes awry or when something internal needs to be adjusted. Mind-wandering serves many positive purposes, from stimulating our imaginations, to motivating us in boring situations with internal rewards (boring task… “ahhhh remember that nice mojito you had on the beach last year?”). Yet we largely see papers exploring the costs – mood deficits, cognitive control failure, and so on. In the meditation literature this has even been taken up to form the misguided idea that meditation should reduce or eliminate mind-wandering (even though there is almost zero evidence to this effect…)

Sometimes our theories end up reflecting our methodological apparatus, to the extent that they may not fully capture reality. I think this is part of what has happened with mind-wandering, which was originally defined in relation to difficult (and boring) attention tasks. Worse, mind-wandering is usually operationalized as a dichotomous state (“offtask” vs “ontask”) when a little introspection seems to strongly suggest it is much more of a fuzzy, dynamic transition between meta-cognitive and sensory processes. By studying mind-wandering just as the ‘amount’ (or mean) number of times you were “offtask”, we’re taking the stream of consciousness and acting as if the ‘depth’ at one point in the river is the entire story – but what about flow rate, tidal patterns, fishies, and all the dynamic variability that define the river? My idea was that one simple way get at this is by looking at the within-subject variability of mind-wandering, rather than just the overall mean “rate”.  In this way we could get some idea of the extent to which a person’s mind-wandering was fluctuating over time, rather than just categorising these events dichotomously.

The EAT task used in my study, with thought probes.
The EAT task used in my study, with thought probes.

To do this, we combined a classical meta-cognitive response inhibition paradigm, the “error awareness task” (pictured above), with standard interleaved “thought-probes” asking participants to rate on a scale of 1-7 the “subjective frequency” of task-unrelated thoughts in the task interval prior to the probe.  We then examined the relationship between the ability to perform the task or “stop accuracy” and each participant’s mean task-unrelated thought (TUT). Here we expected to replicate the well-established relationship between TUTs and attention decrements (after all, it’s difficult to inhibit your behaviour if you are thinking about the hunky babe you saw at the beach last year!). We further examined if the standard deviation of TUT (TUT variability) within each participant would predict error monitoring, reflecting a relationship between metacognition and increased fluctuation between internal and external cognition (after all, isn’t that kind of the point of metacognition?). Of course for specificity and completeness, we conducted each multiple regression analysis with the contra-variable as control predictors. Here is the key finding from the paper:

Regression analysis of TUT, TUT variability, stop accuracy, and error awareness.
Regression analysis of TUT, TUT variability, stop accuracy, and error awareness.

As you can see in the bottom right, we clearly replicated the relationship of increased overall TUT predicting poorer stop performance. Individuals who report an overall high intensity/frequency of mind-wandering unsurprisingly commit more errors. What was really interesting, however, was that the more variable a participants’ mind-wandering, the greater error-monitoring capacity (top left). This suggests that individuals who show more fluctuation between internally and externally oriented attention may be able to better enjoy the benefits of mind-wandering while simultaneously limiting its costs. Of course, these are only individual differences (i.e. correlations) and should be treated as highly preliminary. It is possible for example that participants who use more of the TUT scale have higher meta-cognitive ability in general, rather than the two variables being causally linked in the way we suggest.  We are careful to raise these and other limitations in the paper, but I do think this finding is a nice first step.

To ‘probe’ a bit further we looked at the BOLD responses to correct stops, and the parametric correlation of task-related BOLD with the TUT ratings:

Activations during correct stop trials.
Activations during correct stop trials.
Deactivations to stop trials (blue) and parametric correlation with TUT reports (red)
Deactivations to stop trials (blue) and parametric correlation with TUT reports (red)

As you can see, correct stop trials elicit a rather canonical activation pattern on the motor-inhibition and salience networks, with concurrent deactivations in visual cortex and the default mode network (second figure, blue blobs). I think of this pattern a bit like when the brain receives the ‘stop signal’ it goes, (a la Picard): “FULL STOP, MAIN VIEWER OFF, FIRE THE PHOTON TORPEDOS!”, launching into full response recovery mode. Interestingly, while we replicated the finding of medial-prefrontal co-variation with TUTS (second figure, red blob), this area was substantially more rostral than the stop-related deactivations, supporting previous findings of some degree of functional segregation between the inhibitory and mind-wandering related components of the DMN.

Finally, when examining the Aware > Unaware errors contrast, we replicated the typical salience network activations (mid-cingulate and anterior insula). Interestingly we also found strong bilateral activations in an area of the inferior parietal cortex also considered to be a part of the default mode. This finding further strengthens the link between mind-wandering and metacognition, indicating that the salience and default mode network may work in concert during conscious error awareness:

Activations to Aware > Unaware errors contrast.
Activations to Aware > Unaware errors contrast.

In all, this was a very valuable and fun study for me. As a PhD student being able to replicate the function of classic “executive, salience, and default mode” ‘resting state’ networks with a basic task was a great experience, helping me place some confidence in these labels.  I was also able to combine a classical behavioral metacognition task with some introspective thought probes, and show that they do indeed contain valuable information about task performance and related brain processes. Importantly though, we showed that the ‘content’ of the mind-wandering reports doesn’t tell the whole story of spontaneous cognition. In the future I would like to explore this idea further, perhaps by taking a time series approach to probe the dynamics of mind-wandering, using a simple continuous feedback device that participants could use throughout an experiment. In the affect literature such devices have been used to probe the dynamics of valence-arousal when participants view naturalistic movies, and I believe such an approach could reveal even greater granularity in how the experience of mind-wandering (and it’s fluctuation) interacts with cognition. Our findings suggest that the relationship between mind-wandering and task performance may be more nuanced than mere antagonism, an important finding I hope to explore in future research.

Citation: Allen M, Smallwood J, Christensen J, Gramm D, Rasmussen B, Jensen CG, Roepstorff A and Lutz A (2013) The balanced mind: the variability of task-unrelated thoughts predicts error monitoringFront. Hum. Neurosci7:743. doi: 10.3389/fnhum.2013.00743