Lingo
← Back to Neuroscience

Functional MRI (fMRI)

Mapping the brain by the blood it spends — high spatial resolution bought with poor temporal resolution and an indirect signal.

What it measures

fMRI is not a direct measure of neural activity. It measures the BOLD (blood-oxygen-level-dependent) contrast: as neurons fire, they consume oxygen, local blood flow rises to compensate, and the ratio of deoxy- to oxyhaemoglobin shifts. The magnet is tuned to that shift. A task-related BOLD increase therefore reports where blood is being delivered, which is a proxy for where neurons are active — a two-step inference, not a measurement.

The other name for the effect, deoxyhaemoglobin overshoot , is a reminder of its mechanism: paramagnetic deoxyhaemoglobin shortens T2*, so more of it means less signal. The overshoot in a typical block is only a few percent, which is why the effect needs noise-averse acquisition and averaging across trials.

Acquisition and the cost of the signal

Spatial resolution of about 1–3 mm comes at the price of a repetition time (TR) of 0.5–2 s per volume: the scanner must collect each echo before the next slice of the volume, so the full brain is sampled only every TR. That is the physical floor on temporal resolution, independent of any analysis choice.

Physiological noise — respiration, cardiac pulsation, scanner drift — lives in the same low-frequency band as the BOLD effect, which is why physiological noise regression and careful design are unavoidable. Head motion between scans is a common confounder: it decorrelates signal from anatomy and inflates apparent effects, so motion must be measured and modelled, not ignored.

Design and what it can conclude

Because the vascular response lags and undershoots, event-related fMRI is analysed with general linear models over slow block or event regressors, not trial-by-trial averaging the way EEG is. That framing makes fMRI a test of where sustained engagement differs between conditions, not a millisecond timeline.

The right claim from an fMRI result is a statistical map of relative blood-flow change under a task, not a picture of a neuron. Absent a causal manipulation or a converging source measure, a BOLD difference is correlational evidence about which regions engaged, not proof that they performed the mental operation attributed to them.

References

  1. Ogawa et al. (1990) — PNAS— The original observation that deoxyhaemoglobin changes T2* and produces the BOLD contrast.
  2. Buzsáki, Ulkau, McKenzie (2006) — Cerebral Cortex— The origins of the BOLD signal: the physiology linking neural activity to local blood flow.