fNIRS in Context: Strengths and Limits
🎯 as you read, try to
- Describe the main neuroimaging techniques and identify what each measures, its resolution, and its key constraints
- Explain the specific advantages fNIRS offers that other techniques cannot match
- Articulate the genuine limitations of fNIRS honestly and precisely
- Identify research questions or populations where fNIRS would be the preferred choice, and situations where it would not
Considering everything we have discussed so far, we are now in a position to ask the most practical question of all: given everything a researcher might want to measure, where does fNIRS genuinely fit in?
No technique is the right choice for every question. fNIRS has real strengths that make it uniquely powerful in certain contexts, and real limitations that make it the wrong tool in others. Understanding both honestly is what allows you to design studies that produce credible results.
1. The Neuroimaging Landscape
Neuroimaging techniques can be divided into two broad families based on what they actually detect. The first measures electrical or magnetic signals produced directly by neural activity: EEG records voltage fluctuations at the scalp, and MEG records the magnetic fields generated by those same currents. The second measures hemodynamic or metabolic signals that follow neural activity indirectly: fMRI, PET, and fNIRS all fall into this category. Both families measure brain function, but they are measuring fundamentally different things at fundamentally different time scales.
The table below gives a plain-language overview of the main techniques you are likely to encounter for functional imaging. It is deliberately simplified (each technique has entire textbooks devoted to it), but the key dimensions that matter for practical decisions are here.
| Technique | What it measures | Spatial resolution | Temporal resolution | Portable? | Key constraints |
|---|---|---|---|---|---|
| fMRI | mostly BOLD signal (blood oxygenation, indirect) | ~1–3 mm | ~1–2 s (HRF-limited) | No | Loud (~95 dB), very expensive, no metal implants, strict motion limits, supine position |
| EEG | Electrical potentials from neural activity (direct) | ~1–3 cm (poor) | ~1 ms (excellent) | Yes | Very sensitive to motion and muscle artifacts, poor spatial specificity, gel/cap setup time |
| MEG | Magnetic fields from neural activity (direct) | ~3–5 mm | ~1 ms (excellent) | No | Extremely expensive, requires shielded room, head must be inside fixed helmet, still rare |
| PET | Radiotracer uptake (blood flow or metabolism, indirect) | ~3–5 mm | ~30–90 s (poor) | No | Ionizing radiation, radiotracer injection, very expensive, limited repeat measurements |
| fNIRS (CW) | ΔHbO₂ and ΔHbR (hemodynamic, indirect) | ~1–2 cm | ~0.1–1 s (HRF-limited) | Yes | Cortical surface only, signal contaminated by scalp physiology, sensitive to hair and coupling |
Table 1. Simplified comparison of major non-invasive neuroimaging techniques. Resolution figures are approximate and vary considerably with system type and configuration. fNIRS values refer to standard CW systems; TD/FD systems offer additional capabilities as discussed in Lesson 1.3.
A few things are immediately apparent from this overview. fMRI and MEG offer the best spatial resolution but are entirely non-portable and come with severe environmental and participant constraints. EEG is portable and has extraordinary temporal resolution, but very poor spatial specificity. PET can measure metabolism directly, but it involves radiation and is not repeatable in the same session. fNIRS occupies a distinctive middle ground: it is the only hemodynamic technique that is genuinely portable, wearable, and silent, while offering better spatial resolution than EEG.
🎥 WATCH THIS
How Does fMRI Work?
In case you are entirely new to neuroimaging, understanding fMRI well enough to compare it meaningfully with fNIRS is important background. The video below gives an intuitive sense of the BOLD signal, what an MRI scanner requires of participants, and why spatial resolution is so much better than fNIRS.
2. What fNIRS Does Uniquely Well
The advantages of fNIRS are not marginal improvements over existing techniques. Each one opens research questions that are genuinely inaccessible with other methods.
Portability and wearability
Modern wireless fNIRS systems weigh a few hundred grams and require no fixed infrastructure. A participant can wear a fNIRS cap in a classroom, on a sports field, during a clinical rehabilitation session, or walking along a street. No other hemodynamic neuroimaging technique offers this. This portability is not merely convenient; it is the gateway to studying brain function in the contexts where behavior actually happens, rather than in the artificial environment of a scanner.
Silence
An fMRI scanner generates approximately 95 decibels (dB) of acoustic noise, roughly equivalent to a lawnmower running next to your head. This makes fMRI fundamentally incompatible with auditory paradigms involving natural sounds, music, or speech at comfortable listening levels, and it severely limits studies with infants or participants with auditory sensitivity. fNIRS systems make no acoustic noise at all. This single property has enabled an entire literature on auditory and language processing in infants and young children that would be impossible with fMRI.
Safety and compatibility
NIR light is non-ionizing and entirely safe at the power levels used in fNIRS. There is no radiation exposure, no magnetic field, and no injected substance. This means fNIRS can be used repeatedly in the same participant across many sessions, used longitudinally over months or years, and used in populations where radiation or magnetic fields are contraindicated. Participants with cochlear implants, deep brain stimulators, pacemakers, or surgical metal hardware who cannot undergo MRI can be measured with fNIRS without any safety concerns.
Tolerance of movement
fMRI requires participants to remain almost completely still; even millimetre-scale head movements produce severe artifacts. fNIRS is substantially more tolerant of natural head and body movement, though not immune to motion artefacts (a topic we address in Unit 5). This relative tolerance makes fNIRS the practical choice for studying populations who cannot or do not remain still, such as infants, toddlers, people with movement disorders, or athletes.
Separate HbO₂ and HbR signals
fMRI measures a single BOLD signal that reflects a complex mixture of blood flow, blood volume, and oxygen consumption changes, which cannot be separated without additional measurements. fNIRS provides separate time courses for ΔHbO₂ and ΔHbR simultaneously. Despite potential cross-talk, these two signals carry partially independent physiological information: HbR, for example, has been suggested as a closer proxy to the BOLD signal, while HbO₂ is more sensitive to blood volume and flow changes. Having both gives researchers more physiological information than fMRI BOLD alone, even if interpreting that information requires care.
Hyperscanning: measuring multiple brains simultaneously
Perhaps the most distinctive research opportunity fNIRS enables is hyperscanning: recording brain activity from two or more people at the same time, during genuine face-to-face interaction. This is structurally impossible with MRI, as we cannot fit two people in a scanner and have them interact naturally. This is also technically very challenging with EEG due to electrical interference and high sensitivity to motion. With wireless fNIRS, two participants can sit across from each other, make eye contact, have a conversation, play a game, or teach and learn, while both their brains are recorded simultaneously. This has opened a new sub-field of social neuroscience that is simply not achievable any other way.
🎥 WATCH THIS
fNIRS Hyperscanning in Action
The video below is one among several new research questions exploring fNIRS hyperscanning setups during real social interaction. In this specific case, we see an experiment during interactions between a mother and a child.
💭 Pause and Think
Think of a specific research question about the human brain that you find interesting or that is relevant to your work. Which of the neuroimaging techniques in Table 1 would be most appropriate for answering it, and why?
Consider: what does the question require in terms of spatial precision, temporal precision, and participant freedom of movement? Does it involve a population with special constraints? Does it require a naturalistic or social setting? There is often no single right answer; the goal is to practise matching the tool to the question.
3. What fNIRS Cannot Do
No technique is without limitations, and fNIRS has some that are fundamental rather than merely technical. Understanding these clearly from the start will prevent misinterpretation of your own data and help you evaluate claims in the literature critically.
Depth: cortical surface only
NIR photons can penetrate only a few centimeters into tissue before being absorbed or scattered beyond recovery. In the head, this means fNIRS can only sample the outermost layers of the cerebral cortex (roughly 1-2 cm below the scalp). Subcortical structures such as the amygdala, hippocampus, basal ganglia, thalamus, and brainstem are entirely inaccessible. If your research question involves any of these regions, fNIRS is not the right tool, regardless of its other advantages.
Spatial resolution: limited by photon diffusion
Because photons scatter extensively through tissue, the sensitivity of a source-detector pair is spread across a broad banana-shaped volume rather than a precise voxel. The effective spatial resolution of standard CW fNIRS is approximately 1-2 centimeters, vastly coarser than the millimetre-scale resolution of fMRI. This means fNIRS cannot reliably distinguish activity in adjacent gyri, and spatially precise localisation of cognitive functions requires careful interpretation. High-density diffuse optical tomography (HD-DOT) systems using many overlapping channels can improve this substantially, but they are specialised research tools rather than standard fNIRS setups.
Scalp contamination: the extracerebral problem
The photons that travel between your source and detector do not skip directly to the cortex; instead, they pass through the scalp, skull, cerebrospinal fluid, and superficial cortex. Blood flow changes in the scalp (due to skin temperature regulation, emotional flushing, or systemic cardiovascular changes) produce hemoglobin concentration changes that are physically indistinguishable from cortical changes at the detector. This extracerebral contamination is one of the largest sources of noise in fNIRS and one of the most active areas of methodological development. Unit 5 will discuss this problem in more detail.
Hair, skin, and coupling
Optodes must make good optical contact with the scalp. Dense, curly, or dark hair can physically prevent the probe from reaching the scalp surface, or absorb NIR light before it enters tissue. Darker skin tones absorb more light at the surface, reducing signal amplitude. These are practical challenges that affect data quality and, if not addressed, can introduce systematic biases across participant groups. They are minimized with careful preparation and appropriate hardware, but they require active attention.
Indirect hemodynamic measurement
fNIRS does not measure neural firing directly; it measures the hemodynamic response that follows it, with a delay of 2-6 seconds. This limits the temporal precision with which you can infer when neural events occurred, and it means that very brief or rapidly alternating cognitive events may produce overlapping hemodynamic responses that are difficult to separate. In addition, inference of brain function with fNIRS will depend on the relationship between electric activity and hemodynamics (i.e., neurovascular coupling), which may vary in different scenarios (particularly, in clinical diseases).
|
✓ fNIRS is well suited when you need
|
✗ fNIRS is not suited when you need
|
💭 Pause and Think
A researcher wants to study fear responses in humans. They are particularly interested in the amygdala, a subcortical structure known to be central to fear processing. They propose using fNIRS because their participants are anxious and would find the MRI scanner distressing. What would you tell them?
Is there a way to make fNIRS work for this question? Or is the mismatch between the tool and the target region fundamental? Are there hybrid approaches that might help?
4. A Map of What Comes Next
You have now completed Unit 1. You have an idea of what fNIRS is, how it works at a physical and physiological level, what it measures and what it does not, and where it fits in the broader landscape of brain imaging. That is a genuine foundation; most fNIRS users go years before developing this clarity about the technique they use every day.
But there are natural questions that Unit 1 has raised without fully answering. The remaining units of this course address them in turn, each building on what came before.
|
Unit 2 How Does fNIRS Work? You know that scattered photons make path length unknown and that Beer-Lambert needs modification. Unit 2 explains the physics in more depth: absorption and scattering coefficients, the random walk of photons, the logic of the Modified Beer-Lambert Law, and why CW systems measure changes rather than absolute values. |
Unit 3 What Does the Signal Mean? You know that neural activity causes hemoglobin changes, but why? Unit 3 addresses neurovascular coupling: the physiological chain from neuron firing to blood flow change, what the hemodynamic response function looks like, and how to interpret ΔHbO₂ and ΔHbR in the context of brain function. |
|
Unit 4 Not All fNIRS Devices Are the Same You have heard CW, TD, FD, and DCS mentioned. Unit 4 explains in plain language how each works, what it can and cannot measure, and what the hardware decision means for your research. It also covers practical considerations: optode types, source-detector geometry, and what “good signal quality” actually means. |
Unit 5 Why Is the Signal Noisy? Collecting data is easy. Collecting clean data is not. Unit 5 covers the main sources of noise in fNIRS such as cardiac pulsation, breathing, slow vasomotion, motion artefacts, and scalp physiology, and introduces the strategies used to deal with each. It closes with an honest picture of why analysis choices matter as much as data collection. |
Unit 6
What Can fNIRS Do That Others Cannot?
The final unit returns to the big picture, but now with the technical foundation to appreciate what these opportunities really mean. Infant and developmental neuroscience, hyperscanning and social brain research, ecological validity and real-world measurement, and clinical applications. This is the unit where the full promise of the technique becomes clear, grounded in everything you have learned in units 1–5.
Each unit is designed to be self-contained enough that you can navigate directly to the topic most relevant to you, but they are also designed to build on each other. If you find yourself puzzled by something in Unit 3, the answer will often be in Unit 2. If a processing decision in Unit 5 seems arbitrary, the reason is usually in the physics of Units 1 and 2. The course is a coherent argument, not just a collection of topics.
🎉 Unit 1 Complete
You have covered the full conceptual foundation of fNIRS: the physics of the Optical Window, the spectroscopy of hemoglobin, the physiology of oxygenation, and the place of fNIRS in the neuroimaging landscape. These ideas will reappear, in greater depth, throughout the rest of the course.
Before moving to Unit 2, take a moment to revisit the learning objectives at the top of each lesson and check that you can address each one in your own words. If any are unclear, re-reading the relevant section with fresh eyes (or looking up one of the suggested videos) will consolidate your understanding far more effectively than pushing forward with gaps.
📌 Key Takeaways
- Neuroimaging techniques divide into those measuring electrical/magnetic signals directly (EEG, MEG) and those measuring haemodynamic signals indirectly (fMRI, PET, fNIRS). Each family has characteristic trade-offs in spatial resolution, temporal resolution, and practical constraints.
- fNIRS uniquely combines portability, silence, safety, and motion tolerance among haemodynamic techniques. These properties give access to populations and settings that fMRI and PET cannot reach: infants, clinical groups, naturalistic environments, and hyperscanning.
- fNIRS is fundamentally limited to the cortical surface (~1–2 cm depth). Subcortical structures are inaccessible. Spatial resolution (~1–2 cm) is much coarser than fMRI. The signal is contaminated by extracerebral physiology.
- Like fMRI, fNIRS measures an indirect haemodynamic signal with a 2–6 second delay relative to neural activity. Questions requiring millisecond precision require EEG or MEG instead.
- The remaining five units of this course build progressively: from the physics of light transport (Unit 2), to the neuroscience of the haemodynamic signal (Unit 3), to hardware (Unit 4), signal processing (Unit 5), and the unique opportunities fNIRS enables (Unit 6).
Society for Functional Near-Infrared Spectroscopy.
Content is for educational purposes.