Box 1.

Basics of fMRI, EEG, and MEG.

The non-invasive nature of fMRI, EEG, and MEG, as well as their broad coverage of activity in the brain, make them useful to study biological substrates of cognition. While all these methodologies record activity from large populations of neurons, each of them has unique strengths and limitations. fMRI measures the BOLD (blood-oxygen-level-dependent) signal, which reflects the ratio of oxygenated to deoxygenated blood. Since this ratio depends on neuronal population activity, fMRI indirectly measures brain activity. It has a spatial precision of a few millimetres and records activity from all regions in the brain, that is, both cortical and sub-cortical regions. Its ability to provide a window into brain processing during cognition is however limited by its coarse temporal precision – it can only differentiate changes in brain activity occurring around 1 s or more apart. Recent developments like high-field MRI (De Martino et al., 2015) and multi-band fMRI (Todd et al., 2016) promise to further increase the temporal and/or spatial precision offered by fMRI.

Extracranial EEG measures electrical activity from neuronal populations via electrodes on the scalp, specifically post-synaptic potentials of tens of thousands of neurons firing simultaneously. Due to physics of the measurement, EEG activity predominantly reflects post-synaptic potentials of pyramidal neurons near the cortical surface, although sub-cortical contributions are also present, and secondary currents and volume conduction complicate the pattern. EEG can measure neuronal activity with fine temporal detail; this also means that activity in different frequency bands can be resolved with EEG. However, EEG suffers a coarse spatial resolution. This is because of blurring that occurs when electrical fields are propagated through regions of different conductivities (e.g. CSF and scalp) to EEG electrodes, making it difficult to infer location of active brain regions.

MEG measures magnetic induction produced by the post-synaptic electrical activity in neuronal populations measured by EEG. However, due to the different properties of magnetic and electrical fields, the activity recorded by MEG is less affected by secondary currents and more sensitive to superficial sources. At the same time, however, MEG sensors cannot detect the radial component of those currents. Like EEG, MEG can record brain activity with fine temporal and spectral resolution. Crucially though, MEG is less affected by blurring owing to different tissue types. Thus, MEG combines a high temporal resolution with a superior spatial resolution to EEG, in the order of a few centimetres (although even so, localisation is rarely certain, owing to the inverse nature of the mapping from sensors to sources). Recent developments in MEG, that is, optically pumped magnetometers (OPMs) offer the potential for MEG sensors closer to the head, which should further increase signal-to-noise ratios (Boto et al., 2017).

fMRI: functional magnetic resonance imaging; EEG: electro-encephalography; MEG: magneto-encephalography; CSF: cerebrospinal fluid.