Mind reading: communicating the boundaries of brain imaging15/01/2012
Neuroscientists have a responsibility to ensure that the public understand what fMRI data can tell them, otherwise the media will lead them to believe that “you love your iPhone”, as Dr Donna Rose Addis, of the Department of Psychology and the Centre for Brain Research, University of Auckland and winner of the Prime Minister’s MacDiarmid Emerging Scientist Prize 2010, explains:
Over the past few decades, there have been significant advances in magnetic resonance imaging (MRI) technology. Although the ability of MRI to obtain detailed scans of anatomy was a major advance in biomedical science, for neuroscience research it was the discovery that MRI can be used to track the function of the brain by measuring the blood oxygen levels across the brain, scientists can create visualisations of what brain regions are ‘lighting up’ during different types of cognition. It would seem we can almost ‘read the mind’.
The advances in functional MRI (fMRI) since the 1990s have been met with great enthusiasm from the scientific community – in the 16 years between 1991 and 2007, over 19,000 peer-reviewed articles reporting on fMRI research were published (Logothetis, 2008). Coupled with the increasing availability of and access to fMRI technology, this method has come to dominate brain research and led to the emergence of a new field: cognitive neuroscience. fMRI studies have investigated the neural underpinnings of every aspect of thought and emotion, from executing physical movements, language and mathematics, to the more complex and (some would argue) uniquely human abilities of remembering, imagining, and understanding the self. Although fMRI can provide an understanding of what brain regions are involved in different forms of cognition, importantly it also refines our theories about cognition and generates new ideas and hypotheses.
Despite the exciting technological innovations – and the seductive images of brain regions ‘lighting up’ with activity – there are limits to what we can understand about the brain and cognition from brain imaging. The difficulty is that with such rapid advances in hardware and analytic techniques, the actual limits of this science are always shifting. As with most areas in science, there are complex statistical analyses and assumptions that go on behind the scenes, and those utilising such technology must be well-versed experts to ensure it is used appropriately. Communication about what can and can’t be derived from MRI studies is critical so that data are not misinterpreted and pushed beyond their limits.
Reverse inference: a logical fallacy
The sin of over-interpreting MRI data usually comes about from a logical fallacy called ‘reverse inference’. Essentially, a reverse inference is when one looks at a pattern of brain activity and from that, makes conclusions about what that brain (or its owner) is thinking or feeling. That is, engaging in a form of mind reading.
In the laboratory, however, we design experiments in a way that we have some knowledge about what our participants are experiencing and then we look at what their brain does in response (a ‘forward inference’). We know what our participants are thinking from their performance on our cognitive tasks – we don’t assume it from what their brain activity looks like.
The danger of reverse inference also applies to diagnoses of cognitive or psychological disorders. Although it is tempting to think we can determine if someone has a particular disorder by ‘reading’ their brain scans, in fact the best way is usually to observe their behaviour and/ or measure their performance on relevant tests. This isn’t to say that MRI doesn’t have an important role to play in furthering our understanding of why deficits or changes in behaviour arise. MRI can also identify changes in brain activity that might be characteristic of groups of people with particular symptoms. But as yet, we cannot scan the brain function of an individual person and definitively confirm, or rule out, a diagnosis of any psychological disorder.
Reverse inference and the media
A recent case highlighted the need not only for scientists, but also the media, need to have an awareness of the boundaries of brain imaging. Martin Lindstrom, a marketing and branding expert, wrote an Opinion piece in the New York Times (Sept 30, 2011) regarding his study of brain responses to iPhones. When participants saw their iPhones, there was, he writes, a “flurry of activation in the insular cortex of the brain, which is associated with feelings of love and compassion” leading him to conclude that “they loved their iPhones.” Lindstrom didn’t actually need to spend thousands of dollars and hours of time putting people into an MRI scanner to discover whether or not they love their iPhones. He could have, of course, simply asked them. More worrying, however, is the over-interpretation of the fMRI data.
Lindstrom observed that the insular cortex (amongst a slew of other regions) was active, and based on his knowledge that some other studies have associated insular activity with feelings of love, he assumed the participants were experiencing love for their phones.
But was it love? Scientists would say not. Lindstrom’s article, “You Love Your iPhone”, resulted in an overwhelmingly negative response from the neuroscience community, including numerous postings on science blogs and a letter to the editor signed by over 40 PhDs. As noted in the letter, the insular cortex – which Lindstrom associates with “feelings of love and compassion” – is actually activated in one third of all neuroimaging studies, and is more often associated with negative than positive emotions. Moreover, the insular cortex wasn’t the only part of the brain that was activated when viewing iPhones. As with most complex states of being, a network of regions across the brain was engaged.
At the heart of the reverse inference fallacy is the idea that we can use fMRI to pinpoint a ‘love’ spot in the brain. Being able to draw such conclusions necessitates that there is a one-to-one mapping between each brain region and a particular cognitive function. Instead, the brain is a world of ‘one-to-many and many-to-one’. One brain region usually underpins a variety of cognitive functions, and each cognitive function usually engages multiple parts of the brain.
This is not the first New York Times article to cause outrage and a letter to the editor from the neuroscience community. Just prior to the US Presidential election in 2008, an opinion piece titled “This is Your Brain on Politics” (11 November 2007) provided readers with a shopping list of reverse inferences. See the word “Republican” and the brain of the American voter reveals they experience “anxiety and disgust”. See Hillary Clinton’s name and their brain activity patterns tell of their “conflict”.
Neuroscientists and the media
Why did neuroscientists respond so strongly to these cases of over-interpreted fMRI data? One reason is because it represents a misuse of fMRI technology. It suggests that some people using (or reporting on results obtained from) MRI don’t fully understand the current boundaries of the science or the ethical implications of drawing such bold conclusions. But it is also a misuse of the scientific process.
These articles report research that sounds very ‘scientific’ but in reality, this research has not been peer reviewed by experts in the field. Submitting research to peer-review is a cornerstone of the scientific process, and newspaper and magazine editors should require that any articles – even opinion pieces – are based on science that has undergone such review.
The importance of peer review standards extends beyond publishing. There are now companies selling ‘neuromarketing’ services, many of whom claim they have ‘reliable and standardised ways’ of assessing how much the market will love or hate a given product – such as iPhones – from looking at brain responses. Ask any neuroscientist and they would say this is not yet possible; while we are developing reliable methods for accurately predicting what one is thinking or feeling from the patterns of activity, this is still a way off. Of course, such companies don’t reveal their ‘secret recipe’ for their MRI scans, and so again there is no open peer review of their processes or results. It is the responsibility of scientists to bring awareness to the public about the existence and importance of the scientific process, as well as the limits of the science. Just because a brain image appears to say it, doesn’t mean it is necessarily true.
fMRI use in a legal context
Of more concern, however, are the arguments that fMRI evidence should be admissible in a legal context. Given the popular perception that MRI can be used to read minds, surely it is feasible that an MRI can detect when someone is intentionally lying? Some companies claim this can be done – such as No Lie MRI Inc., a US-based company that “provides unbiased methods for the detection of deception and other information stored in the brain”. Again, the issue of reverse inference rears its head. But so too does a raft of other issues.
In a typical fMRI study, we scan 15–30 different participants and then we put together all their brain scans to get an average picture of brain activity during a particular cognitive task. This averaging is important because everyone’s brain is slightly different in its anatomy and its functioning, and we have to cancel out any idiosyncratic fluctuations (or “noise”).
In contrast, lie detection scans inherently focus on one person – the defendant – making it impossible to know if the defendant’s brain activity shows signs of deception or if it just happens that their brain activates differently from the average. Another issue is that fMRI data can be easily corrupted. In order for lie detection scans to work, the experimenter would have to ensure the defendant is compliant and thinking about the episode in question. It would be easy, however, for a defendant to thwart the lie detection process by just thinking about random things and so ‘scrambling’ their brain activity.
As neuroethicist Martha J. Farah points out, there is the very real concern that judges and juries will treat brain images as hard and indisputable evidence – especially when presented with a hard copy of a brain scan. In reality, those final brain images are arrived at through multiple steps of data collection, processing and statistics. The quality of the images can be affected by the individual characteristics of the person being scanned (including their compliance). And the resulting images are open to (mis)interpretation. Given the potential ramifications of a wrong judgment, it is imperative that if any form of MRI data is permitted as evidence, objective experts have the chance to educate those passing judgment.
Advancing our knowledge of the brain
Although fMRI is not a mind reading device (and some argue it never will be; Logothetis, 2008), I don’t mean to imply that at its current stage of development, brain imaging cannot provide us with any useful information about the brain. MRI has revolutionised our science and significantly expanded our knowledge about the inner workings of the brain. It has sparked new hypotheses and theories that in turn have changed how we think about the mind and brain. Furthermore, our methods for human brain mapping are advancing all the time.
The recent development of pattern analysis has allowed scientists to train computers to decode some forms of information from brain activity and to statistically predict what cognitive process is most likely occurring. Not to definitively say what someone is thinking, but to infer statistically what they might be thinking. As neuroscientists, we have a responsibility to ensure that the public understand what our fMRI data can tell them. Because if the innovation in the last 20 years of brain imaging is anything to go by, before we know it Apple will have a mind reading app to tell us how much we love our iPhones.
For further information contact: firstname.lastname@example.org or visit: www.memorylab.org
If you would like to know more about the advances and limits of MRI technology, The Centre for Brain Research at The University of Auckland is holding a public MRI demonstration on 14 March 2012.
For more information visit: www.cbr.auckland.ac.nz
Farah, M.J. http://www.neuroethics.upenn.edu/index.php/penn-neuroethicsbriefi ng/brain-imaging Logothetis, N.K. (2008). Nature, 453, 869-878. Poldrack, R. http://www.russpoldrack.org/2011/10/nyt-editorial-fmri-completecrap. html Yarkoni, T. http://www.talyarkoni.org/blog/2011/10/01/the-new-york-timesblows- it-big-time-on-brain-imaging/