Neuroscientists study the human brain using dozens of experimental techniques, each capturing a different slice of what the brain does, from the millisecond firing of individual neurons to the slow metabolic shifts that unfold over minutes. No single method reveals the whole picture. Functional MRI tracks blood flow changes linked to neural activity, EEG picks up electrical signals through the scalp, PET scans trace specific neurotransmitter systems, and newer approaches like brain organoids let researchers grow simplified models of brain tissue in the lab. Understanding what each method actually measures, and what it misses, is the key to making sense of the findings that fill science headlines.
Watching Blood Flow with fMRI
Functional magnetic resonance imaging is probably the technique most people picture when they think of brain research. Those colorful brain maps you see in news stories typically come from fMRI. But the technique does not directly measure neurons firing. Instead, it detects changes in blood oxygenation. When a brain region becomes active, local blood flow increases and the ratio of oxygenated to deoxygenated hemoglobin shifts. That shift alters the magnetic signal picked up by the scanner, producing what is called the BOLD signal. The whole chain, from neural activity to blood flow change to detectable signal, depends on a process called neurovascular coupling.1PubMed Central. Coupling mechanism and significance of the BOLD signal: a status report
This indirect measurement has consequences. The BOLD response peaks a few seconds after neurons actually fire, so fMRI has excellent spatial resolution (you can pinpoint activity to a few millimeters) but relatively poor temporal resolution. You can see where something happened in the brain, but pinning down exactly when it happened is harder. Researchers work around this by carefully designing tasks that isolate specific mental operations and by combining fMRI with faster methods like EEG.
One particularly useful extension of fMRI is pharmacological MRI, where researchers give participants a drug and then scan their brains during a cognitive task. This approach maps how different drugs modulate brain network activity, revealing how specific neurotransmitter systems contribute to cognition and how those effects differ across people with different genetic profiles or neuropsychiatric conditions.2Trends in Pharmacological Sciences. Pharmacological MRI: a new application for functional MRI
Tracking Electrical Signals Through the Scalp
Electroencephalography, or EEG, takes the opposite trade-off from fMRI. Electrodes placed on the scalp pick up the tiny electrical fields generated by large populations of neurons firing in sync. The temporal resolution is superb, capturing changes on the order of milliseconds, but the spatial resolution is blurry because the signal has to pass through skull and scalp before reaching the sensors. Researchers often focus on event-related potentials, which are characteristic voltage deflections that occur at predictable times after a stimulus, like a sound or a flash of light.
A newer cousin of EEG is magnetoencephalography, or MEG, which detects the faint magnetic fields produced by the same neural currents. Traditional MEG systems use sensors cooled to near absolute zero, which means the participant has to sit motionless inside a large, expensive helmet. Recent work with optically pumped magnetometers is pushing toward lighter, more flexible MEG systems that can be used in populations previously difficult to study, including very young infants.3PubMed Central. Pushing the boundaries of MEG based on optically pumped magnetometers towards early human life
Researchers also combine eye tracking with fMRI or EEG to get richer data. Measuring pupil diameter during a brain-scanning session, for instance, lets researchers track moment-to-moment changes in cognitive effort and arousal alongside neural activity. Pupil dilation has been linked to both working memory load and activity in brain networks that detect important events.4PubMed Central. Pupillometry tracks cognitive load and salience network activity in a working memory functional magnetic resonance imaging task
Recording Directly from Neurons
Sometimes researchers get the rare opportunity to record electrical activity directly from inside the brain. This usually happens in the context of epilepsy surgery, where patients already need electrodes implanted to locate the source of their seizures. With the patient’s consent, scientists can use those electrodes to study normal brain function as well. Hybrid depth electrodes, for example, have been designed to combine clinical seizure monitoring with microelectrode recordings that capture the activity of individual neurons.5PubMed. Chronic microelectrode investigations of normal human brain physiology using a hybrid depth electrode
This kind of invasive recording provides unmatched detail. Researchers can study how single cells respond to faces, words, or memories in a way that is simply impossible with scalp EEG or fMRI. The obvious limitation is that it can only be done in people who are already undergoing neurosurgery, which means the brain regions studied are constrained by clinical need, and the patient population is not entirely representative of healthy individuals.
Stimulating the Brain to Test Causation
Imaging techniques like fMRI and EEG show that a brain region is active during a task, but they cannot prove that the region is actually necessary for the task. That distinction matters. Transcranial magnetic stimulation, or TMS, fills this gap by temporarily disrupting activity in a targeted brain area using a strong electromagnet placed against the scalp. If zapping a region during a task causes performance to drop, that is evidence the region plays a causal role, not just a correlated one.6PubMed Central. Transcranial magnetic stimulation for investigating causal brain-behavioral relationships and their time course
Researchers describe the effect of TMS as creating a brief “virtual lesion.” By varying when the stimulation is delivered relative to a task, they can also figure out the timing of a brain region’s contribution, something lesion studies in patients with brain damage cannot do. TMS can also be combined with fMRI or EEG to trace how disruption in one area ripples through connected regions, probing the brain’s functional wiring.7PubMed Central. Transcranial magnetic stimulation: studying the brain-behaviour relationship by induction of ‘virtual lesions’
Mapping the Brain’s Wiring
The brain’s white matter is made up of long-distance fiber bundles that connect distant regions. Diffusion tensor imaging, a specialized form of MRI, tracks the movement of water molecules along these fiber tracts to map structural connectivity. Because water diffuses more easily along the length of a nerve fiber than across it, researchers can reconstruct the brain’s wiring diagram in a living person.8PubMed. Diffusion-tensor MR imaging and tractography: exploring brain microstructure and connectivity
This technique has been valuable for understanding how damage to white matter tracts relates to cognitive problems. In diseases like multiple sclerosis, where lesions scatter throughout the brain’s white matter, mapping which connections are disrupted helps explain why some patients develop memory problems while others do not. A recent study found that white matter lesions disrupting connections to the hippocampus, thalamus, cerebellum, and certain cortical areas were specifically associated with cognitive impairment but not with fatigue or depression.9PubMed Central. Lesion Location and Functional Connections Reveal Cognitive Impairment Networks in Multiple Sclerosis That kind of specificity, linking a particular pattern of disconnection to a particular symptom, would be difficult to achieve without structural connectivity data.10PubMed Central. MR diffusion tensor imaging: a window into white matter integrity of the working brain
Imaging Brain Chemistry with PET
Positron emission tomography takes a fundamentally different approach from MRI-based methods. A participant is injected with a small amount of radioactive tracer that binds to a specific molecular target in the brain, such as a particular receptor or transporter. The scanner then detects where the tracer accumulates. PET has deep roots in neuroscience, starting with brain tumor and metabolism imaging and expanding over the decades to address questions about receptor density, drug occupancy, and the release of neurotransmitters like dopamine and serotonin in the living brain.11PubMed Central. PET Neurochemical Imaging Modes
One challenge with PET is that most studies report group-average maps, showing where a receptor is typically dense across a sample of people. But individuals vary. A recent study collected data on 12 different neurotransmitter receptors and transporters across seven neurotransmitter systems and found substantial person-to-person variability in receptor density, raising questions about how well group averages represent any one individual.12PubMed Central. Inter-individual variability of neurotransmitter receptor and transporter density in the human brain That variability is more than an academic concern: it may help explain why the same psychiatric medication works well for one person and not at all for another.
Controlling Neurons with Light
Optogenetics is one of the most transformative tools in modern neuroscience, though it is used almost exclusively in animals. The basic idea is to genetically modify specific neurons so they produce light-sensitive proteins. Researchers can then use pulses of light delivered through a thin fiber to turn those neurons on or off with extraordinary precision, millisecond by millisecond.13PubMed Central. Optogenetic tools for analyzing the neural circuits of behavior This level of control has enabled researchers to probe the causal role of specific cell types in everything from decision-making to fear to reward-seeking.14PubMed Central. In vivo application of optogenetics for neural circuit analysis
A complementary technique, two-photon calcium imaging, lets researchers watch neurons fire in real time in a living brain. When a neuron fires, calcium floods into the cell, and fluorescent molecules engineered into the neurons glow brighter in response. Two-photon microscopy uses infrared laser light that penetrates deeper into brain tissue than conventional light, allowing researchers to image activity in individual neurons within intact circuits.15PubMed Central. Two-photon calcium imaging of neuronal activity Early work demonstrated that this approach could resolve the activity of individual cells in a mouse’s sensory cortex during whisker stimulation.16PubMed Central. In vivo two-photon calcium imaging of neuronal networks The technique has since been extended to primates, including marmosets, where researchers have recorded spontaneous calcium transients from individual neurons in the cortex.17Cell Reports. AAV-Mediated Tetracycline-Inducible Expression System for Long-Term Two-Photon Calcium Imaging in the Marmoset Brain
Neither optogenetics nor two-photon imaging is used routinely in humans because both require genetic modification of neurons or surgical insertion of imaging windows. But the insights from animal studies profoundly shape our understanding of how human brain circuits operate.
Brain-Computer Interfaces as Both Tool and Treatment
Brain-computer interfaces started as a way to decode neural signals and translate them into commands for external devices. For research, they provide a direct readout of how the brain encodes intentions. For people with paralysis, they offer something far more personal: a way to communicate and interact with the world again.
Early intracortical BCIs let participants with paralysis type by controlling a cursor on a screen using neural signals recorded from tiny electrode arrays implanted in the motor cortex. Performance improved dramatically as decoder algorithms advanced, with participants exceeding previous typing speeds by a factor of roughly 1.4 to 4.2 and information throughput by a factor of 2.2 to 4.0 compared with earlier systems.18eLife. High performance communication by people with paralysis using an intracortical brain-computer interface
More recently, BCIs have moved beyond cursor control to decode attempted speech directly from motor cortex activity. In one study, a participant with locked-in syndrome who had been unable to speak for over two years still had neural patterns in the speech motor cortex that could be decoded into phonemes and words well above chance.19Cell Reports. Intracortical brain-computer interfaces for decoding speech and motor intent in paralyzed individuals Perhaps the most striking demonstration to date involved a man with paralysis from ALS who used a speech-decoding BCI near-daily at home for almost two years without researchers present. He communicated over 1.9 million words at an average rate of 56 words per minute, with more than 99% word accuracy during formal testing on a 125,000-word vocabulary. He used the system to send emails, browse the internet, and maintain full-time employment.20PubMed Central. Long-term independent use of an intracortical brain-computer interface for speech and cursor control
Deep Brain Stimulation as a Window into Circuit Function
Deep brain stimulation, where thin electrodes are surgically placed in specific brain structures and deliver continuous electrical pulses, is an established treatment for Parkinson’s disease and other movement disorders. But the same electrodes that deliver therapy also double as research instruments. By recording the local electrical activity around the electrode tip, researchers can study how populations of neurons in deep brain structures behave during disease and during treatment.
A particularly informative signal is the local field potential, which reflects the synchronized activity of thousands of neurons near the electrode. In Parkinson’s disease, abnormally strong oscillations in the beta frequency band (roughly 13 to 35 Hz) have been found in the subthalamic nucleus and another structure called the globus pallidus. The strength of these beta oscillations correlates with symptom severity, and effective DBS suppresses them, with greater suppression linked to greater symptom relief.21Brain. Neurophysiological mechanisms of deep brain stimulation across spatiotemporal resolutions These oscillatory patterns are now being explored as biomarkers that could guide adaptive stimulation systems, devices that automatically adjust their output based on the brain signals they record in real time.22PubMed Central. Dissecting deep brain stimulation evoked neural activity in the basal ganglia
Growing Brains in a Dish
Brain organoids are three-dimensional clumps of tissue grown from human stem cells that self-organize into structures resembling parts of the developing brain. They contain neurons, astrocytes, and other cell types that form networks and communicate with each other.23PubMed Central. Human Brain Organoids: Development and Applications Some organoids develop enough complexity to produce oscillatory electrical activity reminiscent of what is seen in developing brain tissue.24Nature Communications. Functional neuronal circuitry and oscillatory dynamics in human brain organoids
The appeal of organoids is that they let researchers study aspects of human brain development that are otherwise inaccessible. You cannot ethically experiment on a developing human embryo’s brain, and mouse brains differ from human brains in important ways. Organoids fill that gap, allowing scientists to investigate how genetic mutations disrupt brain development, how cells communicate through signaling molecules during the formation of neural circuits, and how neurodegenerative diseases like Alzheimer’s and Parkinson’s might be modeled in human tissue.25PubMed Central. Brain organoids as models of extracellular vesicle-mediated human neural communication They are not brains, they lack blood supply, immune input, and sensory experience, but they are the closest thing researchers have to a controllable model of early human neural development.
Molecular Studies of Post-Mortem Tissue
For all the sophistication of in vivo imaging, some questions can only be answered by examining brain tissue directly. Post-mortem studies allow researchers to analyze gene expression at the level of individual cell types, something current brain scans cannot do. Single-nucleus RNA sequencing, a technique that reads out which genes are active in individual cells from frozen brain tissue, has become a powerful tool for identifying disease-related molecular changes with cell-type precision.26PubMed. Benchmarking of single nuclei RNA-seq methods on human post-mortem brain tissue
This kind of molecular resolution is complementary to everything imaging offers. fMRI might show that a brain region is less active in people with a particular disorder, but post-mortem transcriptomics can reveal that certain neurons in that region have abnormal gene expression patterns, providing a mechanistic link between the clinical observation and the underlying biology.
Using AI to Model Brain Computations
A growing area of neuroscience research uses artificial neural networks, not as replacements for brain experiments, but as models to test ideas about how the brain processes information. Researchers have found that deep convolutional neural networks trained on image recognition tasks develop internal representations that parallel what is seen in the human visual system. Specifically, activity patterns in the lower layers of these networks, which respond to simple features like edges, resemble the signals recorded from early visual cortex, while higher layers, which respond to complex features like objects, align with later visual areas. Electrocorticography recordings have shown that gamma-band activity in human visual cortex maps onto this progression in a layer-by-layer fashion.27Communications Biology. Activations of deep convolutional neural networks are aligned with gamma band activity of human visual cortex
More advanced neural networks appear to provide even better models of brain activity. A recent study found that the most advanced deep neural networks produced feature representations that more closely predicted neural responses and showed tighter alignment with the brain’s functional hierarchy.28PubMed Central. Encoding neural representations of time-continuous stimulus-response transformations in the human brain with advanced deep neural networks The implication is that the computational strategies discovered by machine learning may overlap with the strategies the brain evolved, giving researchers a quantitative framework for asking what the brain is “computing” at each stage of processing.
Large-Scale Brain Mapping Projects
Individual experiments typically study a few dozen to a few hundred participants. Large-scale initiatives aim to change that scale. The Human Connectome Project, launched in 2010, set out to collect detailed brain imaging data from a large number of healthy young adults using improved scanning technology and then share all of it freely with the research community.29PubMed Central. The Human Connectome Project: A retrospective Similar efforts, like the UK Biobank’s brain imaging arm, have since scanned tens of thousands of participants. These datasets allow researchers to study questions about normal variability, like how brain connectivity differs between people who score high versus low on a cognitive test, that would be impossible in a single lab study.
Portable Neuroimaging and Studying Infants
One persistent challenge in brain research is that most imaging methods require participants to lie still inside bulky machines. That rules out studying brains in realistic settings, during natural social interaction, while walking, or in populations that simply will not tolerate a scanner, like newborns and toddlers. Functional near-infrared spectroscopy, or fNIRS, offers a partial solution. Like fMRI, it measures changes in blood oxygenation, but it does so using light sensors built into a lightweight cap that can be placed on someone’s head while they sit, move, or interact with other people.
The trade-off is lower spatial resolution and shallower penetration depth compared with fMRI, so fNIRS mainly picks up activity from the outer layers of the cortex. But the portability advantage is significant. In infant research, fNIRS has become increasingly popular relative to both fMRI and EEG, with a notable rise in publications on infants from birth to two years of age in recent years.30PubMed Central. Using functional near-infrared spectroscopy to study the early developing brain: future directions and new challenges For studying how the brain develops in the first months of life, having a tool that does not require sedation or perfect stillness is worth a lot of spatial precision.
Why No Single Method Is Enough
Each of these techniques illuminates a different feature of brain function, and the most convincing findings in neuroscience typically come from converging evidence across methods. fMRI can show where activity changes occur; EEG or MEG can show when; TMS can establish that the region is necessary; PET can identify the neurotransmitter system involved; diffusion imaging can map the structural connections in play; and animal optogenetics can demonstrate the causal role of specific cell types within a circuit. Organoids and post-mortem transcriptomics fill in the molecular details that no imaging modality can reach. Computational models tie these observations together into testable theories about what the brain is actually doing. The field’s progress depends less on any one breakthrough technique than on the growing ability to combine these perspectives into a coherent picture of how the brain produces perception, thought, and behavior.