Skip to content

The misnomer of “functional connectivity”

February 17, 2011

Lately I’ve actually been growing frustrated with the term “functional connectivity.”  The term is used to describe an imaging strategy known as “functional connectivity MRI,” or “fcMRI.” In short, “functional connectivity MRI” is a technique in which the intensity of blood fluctuations in small regions of interest in the brain are charted over time. Then, the synchronicity of blood fluctuations for regions of interest is calculated in order to determine the degree to which those regions of the brain operate in concert with one another.

Now don’t get me wrong: I think that fcMRI is a brilliant technique. I use it in my research, and it yields indispensable information in the collective effort to understand the human brain.

That being said, I think the term “functional connectivity” is a misnomer, and I think a lot of people are assuming that “functional connectivity” tells us something about physical wiring diagrams in the brain. It does not. If anything, “functional connectivity” tells us about the software of the brain, rather than the organ’s hardware.

By analogy, let’s say that a team of sociologists is looking at the student population of a large high school, and proposes to create a diagram of “social connectivity” for the student body, in an effort to graph the cliques, clubs, and sports teams at the school. And let’s say that their technique for “social connectivity” is to look at the number of times that any two students are talking at the same time throughout the day, regardless of their location within the school. Certainly this endeavor to chart the ebb and flow of talkativeness within the student body would be interesting, and could be used to diagram lots of dynamical trends within the high school. And it is more than likely that the intensity of friendship between two students would contribute to a correlation (or an anti-correlation!) of synchronicity in their talking. But to suppose that the map of social correlation is synonymous with, or a surrogate for a diagram of “social connectivity” would be an error.

In terms of neural imaging, “functional connectivity” measures are measures of activational correlation. Period. As soon as people start assuming that the physical mapping of fiber tracts is implicit in the functional connectivity map, they are going off track. (The term “intrinsic connectivity networks” is starting to appear in the literature to describe the brain’s functional networks. I think that this only confirms and furthers the confusion.)

Now one may ask, “Isn’t this just an issue of semantics? Sure, the term ‘functional connectivity’ may be misleading. But doesn’t the confusion just create a conceptual error rather than a practical one?”

To this I would first reply that there is no such thing as a purely conceptual error. Particularly in science, all errors of concept eventually become errors of practice.

For neuroscience in particular, it is an absolute must for us to ultimately understand the underlying mechanisms that cause the functional networks to emerge from the physical wiring of the brain so that we can eventually do good translational work with the information. Gustav Deco’s lab is doing the best work that I have seen thus far in trying to mechanistically relate the physical wiring of the brain to the functional networks that emerge from the system. (I’m attaching the PDF of a 2010 SfN poster from Deco’s lab to this post. Warning: it’s dense.) As it is right now, the fcMRI data is starting to be used for markers in pathology and classification, which is great for developing new diagnostic criteria. But as far as future translational work to correct pathology and ultimately intervene at the level of brain systems–it’s going to require better understanding of the layers of regulation between the anatomical connectivity and the resultant functional networks.

A quick solution? Let’s stop using the word “connectivity” altogether to describe functional phenomena. Save the word for the anatomical stuff. Let’s instead use the more accurate and descriptive word “correlation”–the acronyms could even stay the same: fcMRI would just become “functional correlation MRI,” and the term to describe the functional networks could simply become “intrinsic correlation networks” rather than “intrinsic connectivity networks.” This might sound like nitpicking. But if it is, let’s remember that we’re constructing a scientific understanding of human nature–not a mythological one. And as is far more true of science than it is of myth, the devil is in the details.


From → Uncategorized

  1. But how will you account for the fact that studies show
    – that functional connectivity reflects actual connectivity within the brain?

    • permalink

      Great questions, ringo-ring. (And great article–thanks for linking it!)

      In the high school analogy, certainly membership in the same clique would be reflected in a temporal diagram of social activity. Social correlations would definitely correspond with social connectivity. But they’re not synonymous. And I think the differentiation is *keenly* important to keep in mind in order to anticipate the next-generation analysis of functional data, particularly with regard to the nuances of regulation and dynamical excitability.

      As the authors of the study state in their conclusions:

      …(A)lthough functional connectivity reflects structural connectivity to a large degree there is not a simple one-to-one mapping. This is most evident in the lack of MPFC–MTL tracts described above strongly supporting the intuitive notion that functional connectivity can exist in the absence of direct, monosynaptic connections. The converse, that structural connectivity can exist in the absence of functional connectivity, is less intuitive and not addressed by our data but nonetheless quite plausible. Perhaps the best example of this scenario relates to context-dependent changes in connectivity…In sum, although our data show that resting-state functional connectivity reflects structural connectivity to a large degree, each can exist without the other.

      • Thanks… Regarding the school example, I think that brain is far more interconnected and far more “serious” that school – all in all, it is not just a gathering of nerve cells in one place – these cells work together to do some task, unlike school, where everyone comes primarily to do their own business. So if there is a correlation in activity between different brain areas then there is a good reason to think these areas do some work together, i.e. functionally connected.
        An example of a team better fits here than school – if two people talk to each other simultaneously, there is a good reason to assume they’re doing some job together… and if other team members talk about unrelated things at the same time, this will likely to disrupt the function of the whole team since you won’t be able to make any sense out of this mess!…

  2. permalink

    (A)ll in all, [the brain] is not just a gathering of nerve cells in one place – these cells work together to do some task, unlike school, where everyone comes primarily to do their own business.

    Your point is well made here. An analogy of trying to discern the inner operations of a large corporation would work toward this end.

    Your comments segue into the open discussion that I introduce in the post preceding this one (“Human Connectome Project: the great controversy begins”). Namely, to what degree can we assume that resting state correlations definitively cluster into functional networks? Ultimately this will be an empirical question, as we look at the networks defined from resting state analysis, and compare them with functional topography derived from task-oriented results.

  3. I think you make good points, Michael.

  4. Andrew Poppe permalink

    I tend to agree with the main premise, but to suggest “functional correlation” as a replacement for “functional connectivity” is not necessarily better, in that statistical techniques used to determine FC often look for *nonlinear* relationships between voxels’ time courses (such as ICA). Correlation implies a linear relationship, such that two orthogonal (and thus uncorrelated) voxels may still be related temporally.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: