A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Neuroimaging comprises diverse techniques capturing brain structure (MRI), blood flow and metabolism (fMRI, PET), electrical activity (EEG, MEG), or chemistry. Each has distinct temporal and spatial resolution trade-offs: fMRI offers high spatial resolution but seconds of temporal lag; EEG provides millisecond resolution but poor localization. Interpreting neuroimaging requires understanding that correlation with cognition does not prove functional necessity—lesion studies and causal manipulations (transcranial magnetic stimulation) provide stronger evidence.
Neuroimaging is essentially a set of different "windows" into the brain, each with different glass. From your prerequisite knowledge of brain structure and functional localization, you know that different regions handle different tasks — but how do researchers actually know which region is active during which task? That's what neuroimaging answers. The core insight is that no single method is perfect; each trades off spatial resolution (how precisely you can locate activity) against temporal resolution (how quickly you can detect changes).
fMRI (functional Magnetic Resonance Imaging) exploits the BOLD signal — Blood Oxygenation Level Dependent — detecting changes in oxygenated versus deoxygenated hemoglobin. When neurons fire, local blood flow increases over the next few seconds, causing a detectable shift in the MRI signal. The payoff is excellent spatial resolution (~1–3 mm), letting you pinpoint which cortical region is active. The cost is temporal: the hemodynamic response peaks 5–6 seconds after neural activity, so fMRI cannot resolve fast cognitive events. Think of it as a photograph with sharp detail but a slow shutter speed. EEG (Electroencephalography) records electrical potentials at the scalp generated by synchronized postsynaptic activity across thousands of neurons. Its strength is millisecond temporal resolution — you can see brain responses unfold in real time during a single cognitive event. Its weakness is poor spatial resolution: electrical signals smear across the scalp through the skull and skin, making source localization mathematically ill-posed. MEG (Magnetoencephalography) records magnetic fields instead, which are less distorted by the skull and offer somewhat better localization than EEG while maintaining millisecond resolution.
PET (Positron Emission Tomography) uses radioactive tracers to measure blood flow or metabolism. It was the forerunner of fMRI for localizing function but has even worse temporal resolution (minutes per scan) and involves radiation exposure, limiting repeat measures. PET remains valuable for specific questions — measuring receptor density or neurotransmitter synthesis — that fMRI cannot address. The choice of method is never arbitrary; it follows from the research question. If you want to know *where* an effect is, use fMRI. If you want to know *when* it unfolds, use EEG or MEG. If you want to know which receptor system is involved, use PET.
The most important interpretive caution — connecting to your statistics prerequisite — is that neuroimaging establishes correlation, not causation. A region that activates during a task might merely co-occur with the real cause. True causal evidence requires either lesion studies (patients with damaged tissue who lose the function) or TMS (Transcranial Magnetic Stimulation), which temporarily disrupts a region in healthy subjects, establishing that the region is *necessary* for the function, not merely coincidentally active. Knowing when to trust localization findings and when to demand causal evidence is what separates sophisticated consumers of neuroimaging research from naive ones.