Don't Assume that fMRI and MEG Will Give You Comparable Results

Accessibility: Intermediate/Advanced

There are three common methods of studying brain function in normal human populations: fMRI, MEG, an EEG. There is surprisingly little crosstalk between the techniques, mostly due to practical issues.For better or worse, labs tend to specialize in one technology.

It's often assumed that the relationship with techniques is straightforward, that it's simple to map results from one technique onto another. However, a recent study by Johanna Vartianen and colleagues suggests otherwise.



The group wanted to study reading using all three brain techniques. Participants performed the same experimental paradigm twice: once with simultaneous EEG and fMRI, and once with simultaneous EEG and MEG. Participants saw words, pseudowords, consonant strings, and symbol strings, and words embedded in noise. Their task was to detect immediate repetitions. The EEG results from the two sessions were comparable, so the researchers went on to compare the fMRI and MEG activation patterns for the experiment.

To summarize, activation patterns between MEG and fMRI did not show a straightforward relationship. In some regions, the two techniques showed the same pattern. For example, in the occipital lobe, both MEG and fMRI measures had more activation to noisy words than other types of stimuli.

If you look at the occipitaltemporal lobe however, the two techniques had opposite results. MEG showed more activation to real letters than symbols, while FMRI showed more activation to symbols then letters.

In the left frontal cortex the two regions had completely different patterns. FMRI activation was higher for words and pseudowords than symbols and noisy words. The MEG results showed no difference at all between stimulus types.

I guess this is one of these results that you don't see going in, but in hindsight make you hit yourself over the head. FMRI and MEG measure very different things, so it’s entirely possible that results would come out differently. FMRI measures cerebral blood flow on a timescale of several seconds, while MEG measures synchronous electrical activation with millisecond resolution. So ( as the authors suggest) non-synchronous activity may be lost in MEG. Meanwhile, fMRI picks up average activity over a longer time period and may miss short-term activity.

Interestingly, the authers mentioned that previous MEG results for the visual word form area were fairly robust to task differences, while fMRI results do seem to vary with task. Now I don't know the MEG literature well, but they're certainly right about the fMRI literature. In that case, I wonder what it is about the MEG that makes its results relatively task independent. Is it the better temporal resolution? Perhaps MEG analyses focus on early, bottom up processing, which may be relatively task independent?


Vartiainen J, Liljeström M, Koskinen M, Renvall H, & Salmelin R (2011). Functional magnetic resonance imaging blood oxygenation level-dependent signal and magnetoencephalography evoked responses yield different neural functionality in reading. The Journal of neuroscience : the official journal of the Society for Neuroscience, 31 (3), 1048-58 PMID: 21248130