What Scans Reveal - and What They Don’t

In a softly lit laboratory, a volunteer lies still inside a giant doughnut - shaped machine. This is the modern ritual of peering into the living brain. Over decades, scientists have assembled an impressive arsenal of such tools to study consciousness. Let’s list some main ones and their basic powers: fMRI (Functional Magnetic Resonance Imaging), EEG (Electroencephalography), MEG (Magnetoencephalography), and intracranial recordings (electrodes directly on or in the brain, often in patients). Each sees the brain from a different angle. FMRI tracks blood flow changes, offering a crude map of neural activity with decent spatial detail (down to a few millimeters in best cases) but slow in time (lagging by seconds). EEG uses electrodes on the scalp to pick up the brain’s electrical activity directly; it’s lightning - fast (millisecond resolution) but spatially blurry (just a general sense of which region, because signals spread). MEG is similar to EEG in speed but measures magnetic fields from brain currents, which can give a bit better spatial localization than EEG. Intracranial recordings, when available (like during epilepsy surgery), are the gold standard: directly picking up actual neural firing or local field potentials at specific brain locations, with high resolution in both time and space - but obviously limited to special medical cases and small brain regions. Each tool has limits: fMRI can’t tell precisely when things happen (it’s like a long - exposure photo of brain activity), EEG/MEG can’t precisely tell where in the brain a signal originates, and electrodes inside the brain can’t cover the whole system at once.

Focus on fMRI for a moment, since it’s so prominent in media. FMRI measures BOLD signals - Blood Oxygen Level Dependent changes. Neurons, when active, gobble up oxygen and energy, so local blood flow increases to those areas, altering the ratio of oxygenated to deoxygenated blood. The MRI scanner can detect this change because oxygen - rich and oxygen - poor blood have different magnetic properties. Importantly, this is an indirect measure: it tracks blood, not neurons directly. There’s a lag - roughly 5 seconds peak delay - between neurons firing and the blood response peaking. That means when you see a brain “lighting up” in an fMRI image for a stimulus, those bright spots reflect that something was happening in that region and the blood responded afterwards. If you’re trying to claim that a certain brain activity happens before or after a person becomes conscious of something, fMRI’s lag can muddy the waters. For example, if someone reports awareness of a flash at time T, the neural activity that correlates with that might have occurred around T, but the fMRI might show peak at T+5 seconds. So careful timing claims (like “the awareness happens 300 milliseconds after stimulus”) can’t come directly from fMRI - they need EEG or other faster methods in combination. The lag also means if multiple events happen in quick succession, fMRI blobs can blur them together.

Now, despite these limitations, fMRI has achieved feats that sound like mind reading. Consider a decoding study: scientists show participants a bunch of images (say, different objects or scenes) while scanning their brains. They feed the data into a machine learning model that learns the patterns of brain activity associated with each image. Later, if the participant sees a new image, the model can guess which one (out of those learned options) they’re looking at based on the brain activity pattern alone. Some studies even reconstructed rough images of what subjects were seeing or imagining, using these patterns. So does that mean they scanned the conscious experience? Yes and no. Decoding shows that the information about the stimulus is present in the brain signals - enough to identify it. For example, an early visual area in your brain will have different patterns for a house vs. a face, and a computer can pick up on that with training. What it doesn’t show is whether you consciously experienced it. We assume in those studies people were awake and seeing things normally, so likely yes, but the decoding itself doesn’t prove the presence of subjective experience. It only demonstrates a correlation: certain brain activity corresponds to seeing a certain thing. In principle, one could decode even unconscious processing if the brain goes through some of the same steps without awareness. In fact, similar decoding methods can sometimes guess what an unconscious brain has processed (for example, in subliminal perception research, even when you aren’t aware of a quick flash, some of your brain registers what category it was). So decoding is powerful and exciting - it shows the specificity of brain patterns - but by itself it doesn’t settle whether those patterns accompanied actual conscious perception or not. For that, we need the person’s report or a marker of awareness to confirm.

This brings us to an important caution: reverse inference. If pattern A in brain = image of a cat, and we see pattern A, it’s tempting to say “the person is seeing a cat consciously.” But what if pattern A can also occur when a cat image flashed too briefly to be consciously seen? It might still activate visual cortex in a “cat - like” way without the person noticing. The safe inference direction is forward: if someone is consciously seeing a cat, certain patterns occur. The reverse - seeing the pattern, assuming conscious cat - seeing - can be wrong without context. In popular media, we often see an oversimplified reverse inference. Like “Empathy center lights up when people watch drama, so people were feeling empathy.” Maybe, or maybe that brain region does other things too, like understanding social situations generally. Without careful controls (e.g., comparing to watching something that also engages understanding but not empathy), it’s not conclusive. A serious scientist will design tasks where specific mental states are isolated and then tie brain activity to them. For example, they might have one condition where you truly empathize (seeing someone in pain) and another where you analytically judge a scenario with similar visuals but no empathy requiring, then contrast the scans. That way, if an area stands out only in the empathy case, the inference is stronger.

We also have to consider what brain scans cannot currently settle. Many provocative questions remain open. For one, just because we find a brain pattern associated with consciousness doesn’t mean we know if it’s essential or just a byproduct. A scan might show a certain network active whenever someone reports an experience. But is that network causing the consciousness or is it a consequence? For example, some theories propose a “global workspace” ignition in prefrontal and parietal areas when something becomes consciously known. We see that in many scans as a P3 wave or late widespread activation. But some argue that might be more about the act of reporting or remembering the stimulus, not consciousness itself. Brain scans alone can’t easily tell the difference between the experience and the later processes around the experience (like reflecting on it). Another thing scans can’t yet do: peek at the quality of an experience with certainty. If two people both show activation in, say, the fusiform face area when looking at an image, we assume they both see “a face.” But what about the vividness or the nuance of it? If one person’s conscious experience was especially sharp or emotionally resonant, we have correlates like maybe stronger activation or connectivity, but we can’t measure the qualitative aspect directly. And as of now, no scan can definitively tell whether consciousness is present or absent in borderline cases without other evidence. For example, if someone is in a vegetative state (eyes open but not outwardly responsive), a scan might show some brain activity to pain or words. But whether that indicates actual conscious awareness or just reflex processing is heavily debated and usually needs clever active tests (like asking the patient to imagine something to signal yes or no).

Consider scenarios where behavior is minimal: deep sleep, coma, anesthesia - the scans can show “low activity” or “different connectivity,” but whether a tiny lingering island of consciousness remains in some of those conditions can’t be certain from just a passive scan. We also can’t currently use scans to read someone’s specific thoughts or experiences arbitrarily. We need a predefined model; they’re not universal mind - readers. A brain pattern we haven’t seen before we can’t just decode from scratch. And critically, even if we map all neural correlates, the scan doesn’t by itself tell us why those go with consciousness - it’s descriptive.

So what do we do to strengthen claims about consciousness with these tools? The answer is converging evidence. We don’t rely on one measure alone; we combine approaches. For instance, say you have a patient who is minimally conscious (occasionally shows signs of awareness). You might give an auditory stimulus and see a certain brain response (maybe a P3 wave on EEG). On its own, that wave hints they recognized the odd sound as different, maybe a sign of awareness. But you’d bolster that by also doing a behavioral test - maybe ask them to follow a command in the scanner via imagining something (like the famous tennis imagery vs spatial imagery test with fMRI: one type of imagination lights the motor cortex, another lights spatial memory areas, used as yes/no signals). If the patient can willfully do that, it supports that the EEG wave wasn’t just an automatic blip but part of conscious perception. Add to that maybe a perturbation - use TMS (transcranial magnetic stimulation) to zap the brain and see if it responds with a complex pattern (something we’ll talk about soon). If it does, that’s another independent sign of conscious - level complexity. When behavioral, EEG, fMRI, and other measures all line up, you have a far stronger case. It’s like multiple witnesses to the same event.

A simple way researchers design more convincing studies is with within - subject contrasts. For example, show a stimulus in two conditions: one where the person becomes conscious of it and one where they don’t. Masking is a classic technique: you show a quick image and then mask it with a jumble so sometimes people don’t consciously see the quick image even though it hit the eyes. Then you compare the brain activity on trials where they say “I saw it” vs “I saw nothing,” keeping the stimulus same type and intensity. This way the difference in brain data is more likely to reflect consciousness itself, not just the stimulus. If in such a setup you find a signature (like a certain EEG wave only present when they saw it), that’s a candidate marker of consciousness. Then you can test that marker in other contexts.

Finally, to not overclaim from scans, scientists often emphasize what the technology cannot infer, clarifying limits. For instance: “This fMRI pattern suggests the patient understood the command, but we can’t be sure they experienced it like a normal awake person; they may lack full self - awareness.” Recognizing limits is part of the standard of evidence.

In summary, brain scans and technologies give us correlations and constraints. They reveal what parts of the brain are active when and how they communicate, which is essential for any theory of consciousness to respect. But they’re not magical truth serums. They’re only as good as our experimental design and our interpretation. They reveal a lot about brain processes - sometimes surprising things like hidden awareness in unresponsive patients or the distinct activity associated with conscious vs unconscious perception. Yet, they don’t directly reveal the presence of the feeling itself without interpretation. That requires trust in reports or clever active tests.

With that in mind, some of the most illuminating evidence about consciousness comes from cases where the usual link between brain and behavior breaks down. These cases are nature’s (or medicine’s) experiments that tease apart aspects of consciousness. Let’s explore a few famous ones - from split brains to blindsight - to see what they show and how they caution us.