Yes! The MRI scanner I use can tell which bits of the brain are working harder based on changes in the oxygen level in the blood. So, you can investigate what bits of the brain are involved in doing certain tasks. However, most of our brain is really busy all the time, so to identify the interesting bits for our tasks we need to compare activation with another condition. For example, if I want to know which bits of the brain are active when a person listens to speech, I might compare this with the activity when that person listens to some other kind of sounds… So, most of my experiments are set up in this way: I compare brain activity across different tasks or sounds that are carefully chosen to answer the questions I’m interested in. We can also look at ‘networks’ of brain activity, in other words how several bits of the brain work together to do a certain task, and how this changes under different experimental conditions.
So, many of the experiments we do involve asking subtle questions under quite controlled conditions. But you might be interested in how far we can take the technique: can we read people’s thoughts? Well, actually yes. There are other research groups who are beginning to see whether we can use brain activation patterns alone to predict what a person was thinking in the scanner. At the moment, these studies have looked at pretty simple thoughts, like thinking about simple objects e.g. houses, tools. But it is likely that the techniques will get more sophisticated in the near future. This is pretty amazing, and could have positive applications – there is recent evidence that we might be able to look at brain activation patterns to communicate with someone in a vegetative state (i.e. a person who cannot communicate in any outward way like speaking or moving). However, there are also some potentially worrying applications – there are already companies who offer MRI scanning for lie detection. This has serious ethical implications. Telling lies is a really complex behaviour, and most of the published neuroimaging work on it has used very simple tasks (e.g. ‘pretend you didn’t see this playing card before’). There isn’t nearly enough evidence to show that these lie detection methods would be accurate or replicable, and what if they are used as evidence in court? It’s all a bit scary and one of the important reasons why everyone should be well informed and engaged in scientific debate!
Can I just add — the evidence that brain scans can be used to “read someone’s mind” is pretty weak. All of the examples to date use very contrived situations or extreme methods that only work on a single person. For instance, the example Carolyn mentioned regarding communicating with a person in a persistent vegetative state involved asking the person to imagine playing tennis to indicate “yes” or walking through their house to indicate “no.” They then asked the person a series of questions they already knew the answers to and found they could correctly guess the person’s response from the brain scan data *most of the timeMATOMO_URL That’s pretty impressive but all the did was figure out a way to communicate yes or no and even that wasn’t 100% reliable.
The only experiment that tried to do something more ambitious that figure a one-or-the-other choice involved scanning one person something like 100 hours to get enough background data that they could show them a word and predict what word it was from the brain scan activity. That was extremely cool because it’s a much harder problem but none of that data helped to predict what word a second person was looking at. In other words, it worked fine if you had 100 hours of baseline data per person you wanted to try it on (and they were cooperative).
So any claims that brain scanners can show what a person is thinking are probably seriously overblown. Is it possible that sometime in the future this *may* be possible? Yes. It is likely to occur within our lifetimes, no.
Joe, I totally agree that the data so far is very limited and it is clear that the techniques are not fully accurate or replicable even in the simplest tasks. However, it seems to me that an above-chance classification of object category from neural patterning alone in a basic visual imagery task can surely be described as a form of ‘thought-reading’. Nor can it be denied that even being able to extract ‘yes’ vs ‘no’ from a patient in a vegetative state in most trials is much more informative than anything you can get from their peripheral state.
What I was trying to raise in the above answer was the very serious issue that MRI technology is currently being sold as a ‘lie detector’ by companies like ‘No Lie MRI’, when the evidence that this can be used to measure complex behaviour in a accurate or replicable way is simply not there. However, a woman in India was convicted of murder last year on the basis of evidence that included an EEG ‘lie detection’ experiment. So, even though the data don’t support this application, it is already being used and it’s an interesting issue for the public to get engaged with. Part of engaging with the issues is being aware that, at least to a limited degree, very basic ‘thought reading’ *is* possible, and we need to keep an eye on the available evidence for the time when our governments think it’s a good idea to have this kind of data considered in a court of law.
Comments
Joe commented on :
Can I just add — the evidence that brain scans can be used to “read someone’s mind” is pretty weak. All of the examples to date use very contrived situations or extreme methods that only work on a single person. For instance, the example Carolyn mentioned regarding communicating with a person in a persistent vegetative state involved asking the person to imagine playing tennis to indicate “yes” or walking through their house to indicate “no.” They then asked the person a series of questions they already knew the answers to and found they could correctly guess the person’s response from the brain scan data *most of the timeMATOMO_URL That’s pretty impressive but all the did was figure out a way to communicate yes or no and even that wasn’t 100% reliable.
The only experiment that tried to do something more ambitious that figure a one-or-the-other choice involved scanning one person something like 100 hours to get enough background data that they could show them a word and predict what word it was from the brain scan activity. That was extremely cool because it’s a much harder problem but none of that data helped to predict what word a second person was looking at. In other words, it worked fine if you had 100 hours of baseline data per person you wanted to try it on (and they were cooperative).
So any claims that brain scanners can show what a person is thinking are probably seriously overblown. Is it possible that sometime in the future this *may* be possible? Yes. It is likely to occur within our lifetimes, no.
Carolyn commented on :
Joe, I totally agree that the data so far is very limited and it is clear that the techniques are not fully accurate or replicable even in the simplest tasks. However, it seems to me that an above-chance classification of object category from neural patterning alone in a basic visual imagery task can surely be described as a form of ‘thought-reading’. Nor can it be denied that even being able to extract ‘yes’ vs ‘no’ from a patient in a vegetative state in most trials is much more informative than anything you can get from their peripheral state.
What I was trying to raise in the above answer was the very serious issue that MRI technology is currently being sold as a ‘lie detector’ by companies like ‘No Lie MRI’, when the evidence that this can be used to measure complex behaviour in a accurate or replicable way is simply not there. However, a woman in India was convicted of murder last year on the basis of evidence that included an EEG ‘lie detection’ experiment. So, even though the data don’t support this application, it is already being used and it’s an interesting issue for the public to get engaged with. Part of engaging with the issues is being aware that, at least to a limited degree, very basic ‘thought reading’ *is* possible, and we need to keep an eye on the available evidence for the time when our governments think it’s a good idea to have this kind of data considered in a court of law.