If recent reports in the media are anything to go by, we now know why some people crave chocolate, why teenagers tend to be moody and – most impressively of all, perhaps – which parties voters will opt for in elections. Findings like these, we are assured, have the backing of science. We owe them to advances in brain-scanning technology that have enabled researchers to pinpoint the brain's ‘pleasure centre’ or ‘thinking area’, and so to achieve mind-reading powers that have hitherto been the stuff of fantasy. There is even some site in the brain that might be called a ‘God spot’.

It is easy to dismiss such articles on the grounds that they are forced to oversimplify, and that this makes them misrepresent the nature of the complex patterns of neuronal activation that in most cases are all the research claims to have established. Writing on this topic almost a decade ago, however, Jerry Fodor expressed exasperation at the number of column-inches being consumed by these reports for another reason: namely, that in an era when dualism is widely regarded as untenable in any case, it should matter little to us just what part of the brain the mind is held to be identical with. ‘If the mind happens in space at all, it happens somewhere north of the neck. What exactly turns on knowing how far north?’Footnote ¹ In his view, the only people with a legitimate interest in cranial co-ordinates are surgeons, lest they excise the wrong bit of brain.

To assist them in establishing the exact whereabouts of, say, brain tumours, neurologists and neurosurgeons have for more than thirty years had at their disposal such X-ray-based technology as Computed Axial Tomography (CAT). While structural imaging techniques like these have also continued to develop, in the period since then the bigger breakthrough has been with functional imaging – allowing clinicians to identify not just gross abnormalities of structure but sequences of neuronal firing associated with cognitive function. This switch in emphasis, from anatomy to activity, commended neuroimaging to the attention of psychiatry at a time when many of its leading practitioners were seeking to link their specialty more closely to the rest of medicine by recasting the major disorders treated by them as dysfunctions of the brain. To be sure, other insights and approaches borrowed from biology – including molecular biochemistry and genetic research – were incorporated into this strategy. But in terms of its potential to provide an incontestable ‘biomarker’ of disease, it was this particular branch of cognitive neuroscience that seemed to hold out most promise. Writing at the turn-of-the-century in an article surveying his profession's next 25 years, the eminent British psychiatrist R.E. Kendell anticipated this prospect with enthusiasm:

There are in fact a number of unresolved conceptual issues around both the principles of neuroimaging and its application to psychopathology. Certainly the idea of a ‘cerebroscope’, a device that would allow us quite literally to behold the workings of the mind, is one that has long stirred the imaginations of scientists and philosophers. And in the judgment of authoritative contemporary sources such as The Economist, this is precisely what the latest technology has given us: ‘It can, if you like to put it that way, see the self in action. All you need to do is to put someone inside an fMRI machine, give them a task to do and see which bits of the brain light up.’Footnote ³

The notion that, when put through a scanning machine, parts of a person's brain ‘light up’ depending on the nature of the task he has been given, is indeed a beguiling one. For then we would indeed appear to have a kind of snapshot of the mind, a polaroid of personality. Yet while the radiographic techniques underpinning CAT scans do in a sense peel away the outer body to give us a direct image of the brain (as of any other internal organ), when it comes to the supposedly still more revealing instruments of functional imaging, no area of the brain can be said to ‘light up’ for the simple reason that these devices do not in any sense take pictures. The experimenting observer ‘sees’ nothing. What he receives, rather, is a doubly indirect measurement of the object of interest. The larger philosophical problem, as we will see, concerns relating recorded brain activity to the mental processes these are taken to be associated with. But there is also the not insignificant issue of capturing brain activity in the first place.

For, notwithstanding impressions sometimes given to the contrary, the two types of functional imaging technology currently most widely-used do not directly measure the electrical activity of the brain's 100 billion or so nerve cells. To characterize them extremely crudely: Positron Emission Tomography (PET) is sensitive to fluctuations in the brain's consumption of glucose and Functional Magnetic Resonance Imaging (fMRI) to local changes in deoxyhemoglobin concentration. Each method has unique difficulties of its own. But the deficiency they share is that both are measures of cerebral blood flow and as such much less sensitive to the timing of neural activity than the electroencephalograms that they have largely eclipsed. A brain can carry out all the functions required for recognizing a face in about 300 milliseconds, for example, while its blood vessels will only begin dilating several seconds later.

There are various other technical and methodological issues around converting the streams of numerical data that are what functional neuroimaging experiments actually gather into the striking colour images – popular in scientific journals as well as newspaper articles – purporting to display activated areas of the brain. Some of these might be overcome by further innovations in scanning technology. At a more philosophical level, however, whether we choose to rely on haemodynamic indices of neural activity or employ some other means of measurement, the question what these can really tell us about cognitive function will remain. Is the concept of a ‘brain area’ (for anything) a coherent one?

So far functional neuroimaging has proceeded by taking a mental process modelled by cognitive neuropsychology which it then seeks to localize in the brain by means of what is known as the subtractive method. Bear in mind, first, that the processes it can concern itself with are limited by the cognitive activities a subject lying motionless in a scanner can carry out – speaking, listening, reading and so on. This means that in the case of the Carmelite nuns scanned for a putative ‘God spot’, for example, it was only the recalling or supposed ‘reliving’ of a mystical experience that could be measured, as opposed to the experience itself. Activity levels for the hypothesized cognitive component underlying these actions are then obtained in two conditions, a task state and a control state, with the latter being subtracted from the former to yield the location and shape of the region where this component may be said to operate.

This method has often been criticized on grounds of the assumption it must make that the task-state activity is a ‘pure insertion’ into a sequential set of processes – as opposed, for example, to a scenario in which an area of the brain active in the control state is not active in the task state. Might it be the violation of this assumption which accounts for the divergent findings characteristic of such research, with no less than 22 distinct brain regions being put forward in five separate studies to isolate a single (phonology) component?

Another concern raised has been over the inconsistency between the results of these studies and those generated by other ways of mapping cognitive function in the brain such as lesion research and single cell recording. Which method is to prevail? In a sense, it is impossible to answer this question, as there is no independent way of verifying whether the function they are attempting to localize exists in the first place. Indeed, it is the circularity implicit in all such studies that has led even the most ardent philosophical materialists to express reservations about the whole idea of brain-mapping.

The postulation of mappable cognitive components is far from being a recent phenomenon. Working in the early 19th century, the phrenologist Franz Gall may have been misguided in seeking to correlate mental functions with bumps on the skull. But thanks to the research conducted some decades later by Broca and Wernicke on language deficits and the specific neuroanatomical lesions accounting for them, a belief in cerebral localization soon revived and continues to be influential in the ‘modular’ approach to the mind taken by many cognitive scientists today. This perspective is simply a given in most neuroimaging studies, where the task under investigation is held to be brought about by patterns of neuronal activation occurring in an additive, linear fashion. The problem with such a simple, feed-forward account (which tend to fit neatly onto the input-output flow-charts presented by information-processing theories of the mind) is whether the cognitive components supposedly delimited by it really combine in this kind of causal chain. This is implicit in the term ‘activation’ and is another of the assumptions of the subtraction methodology on which these studies rely. What if, however, the task in question were to be brought about not by a sequence of single causes corresponding to anatomically distinct regions, but on the basis of a kind of reciprocal causality spread across the cognitive system as a whole? Then it might make more sense to speak of regionally specific interactions rather than activations. And it is precisely such a model of the mind, ‘connectionist’ rather than ‘classical computationalist’, which tends to find favour among cognitive scientists today.

Indeed, while media reports of the kind that riled Fodor may give the impression that the hunt is still on for the brain's ‘pleasure centre’ or ‘thinking area’, the emerging consensus among philosophically-informed neuroscientists is that attempting to map psychological phenomena in a one-to-one fashion onto underlying neural substrates is probably a misguided enterprise. So when it comes to the purported findings of functional neuroimaging, there may not in fact be ‘boxes’ (representing mental functions) for the scanner's ‘blobs’ (showing areas of the brain) to fit onto. That, at any rate, was the conclusion reached by a recent issue of Current Directions in Psychological Science devoted to the interface between psychology and neuroscience.

The debate between ‘localizationists’ and those who prefer to regard the brain as ‘equipotential’ in its capabilities is as old – or, perhaps better said, as young – as the discipline of neuropsychology itself. From its early days and the celebrated case of Phineas Gage (who survived a severe injury to the head only to find his personality transformed), both sides in this debate have been keen to summon clinical evidence for their claims. But whereas in the past investigators had to wait for post-mortem examinations to test their hypothesis about the brain's functioning, or else take the risk of intervening in its operations by the application of various stains and tracers, today's neuroimaging technology appears to offer researchers a way of studying a ‘living brain’ without at the same time disturbing it.

The potential benefits of this technology for an understanding of psychopathology seemed obvious and – as we have already observed – some psychiatrists were quick to embrace it as marking a ‘paradigm shift’ within their field. To date, large numbers of functional neuroimaging studies have been carried out on a range of mental illnesses including depression, schizophrenia, anorexia and obsessive-compulsive disorder. However, in spite of no shortage of apparently promising findings, what has also emerged is that when applied to psychiatric disorders, neuroimaging runs up against several technical and more theoretical difficulties beyond those associated with its use in non-clinical populations.

One major such problem is defining the patient group to be studied. It is, after all, an article of faith among ‘cognitive neuropsychiatrists’ that mental disorders are essentially brain disorders. Yet even in the case of such biologically-weighted syndromes as schizophrenia, the underlying brain pathology is not known – and so cannot possibly be a basis for recruiting a group of people meant to have the disease for a brain-scanning study (not without running the risk of an obvious circularity, that is). An alternative strategy would be to target individuals displaying the symptoms of schizophrenia. This would be in keeping with the descriptive approach to diagnosis taken by more recent editions of the Diagnostic and Statistical Manual of Mental Disorders. However, since this particular manual (regarded as indispensable by most psychiatrists practicing today) specifies five characteristic symptoms of the disorder, ranging from delusions to grossly disorganized behaviour, the patient group delimited in this way would again be too heterogeneous for any study to be meaningful.

Focusing on just one set of symptoms might appear to offer a way round this difficulty. But this brings us to another set of issues raised by clinical populations. How can we be sure, for example, that the symptom in question is definitely being experienced at the time the scanning is being carried out? What about the effects of any medication the patient is likely to be on – not to mention those of a range of other factors quite possibly connected to the illness (poor diet, anxiety) any one of which might interfere with the cognitive function supposedly being captured? How are any of these potential experimental confounds to be controlled for?

Third is the most fundamental problem of all – that of interpreting the study's results. Assuming for a moment that we were to our satisfaction able to identify a piece of ‘broken brain’ implicated in a particular cognitive deficit. Should this then be seen as the cause of the mental disorder of interest to us, a consequence of that disorder or even as a compensation for it?

This last problem is one faced by all correlational design studies and is insuperable. Some of the other difficulties arising when neuroimaging is applied to psychopathology, however, might be overcome by moving away from group to single case studies. This also touches upon an important philosophical question. The reason neuroimaging needs to take into account a person's developmental trajectory is because a ‘snapshot’ of their brain at any given moment in time would never be able to render transparent the experience-derived meanings that have accumulated for that person over time. That is, even if we both respond to cats in the same way (such as by moving forward to stroke them), the complex and highly personal web of associations I have with these animals reaching back to my earliest encounters with them, makes it extremely unlikely that the pattern of neural activation in my brain set in motion by seeing one will be the same as that triggered in yours.

So if you think you might be suffering from depression, should you or shouldn't you seek to have your head examined? Psychiatry itself is divided on the matter. What started out as almost unqualified enthusiasm within the profession for the possibilities offered by brain-scanning technology, has abated now into something closer to cautious realism. Disillusionment might be putting it too strongly – not least because, as have seen, tools like these represent for many clinicians and researchers their best hope for reuniting what they do with the rest of medicine. Making this particular branch of neuroscience the subject of a recent editorial, however, even the American Journal of Psychiatry was steadfast in its refusal to flatter such hopes: ‘Outside the realm of degenerative disorders, neuroimaging does not have any diagnostic applications in clinical psychiatry at the present time’.Footnote ⁵

All the same, as new generations of brain-scanning equipment come into use, so there arises the possibility of finding ways to make better sense of the data they yield. The key here would appear to be developing a more dynamic model of the mind, one that is sensitive to interactions not just between its anatomically distinct regions but also over time – that is across the lifespan of the person who possesses that mind.

For some philosophers, however, all attempts to use neuroimaging and similar techniques to see into a person's mind are doomed to failure, not because the brain-mapping machinery isn't up to it (yet) but because the cognitive states we are in search of do not exist on that level. They are a feature of persons, not of their organs. As Wittgenstein put it: ‘Only of a living human being and what resembles (behaves like) a living human being can one say: it has sensations; it sees; is blind; hears; is deaf; is conscious or unconscious’.Footnote ⁶ Wittgenstein's position is, of course, a contested one. It rests on a number of claims about the ascription of psychological attributes, the nature of privileged access (to one's own consciousness) and the way language works that all come together in the famous ‘private language argument’.

A thorough exposition of this argument would take us well beyond the scope of this article. But all that needs to be stressed here is the radically different view of the mind such a perspective presents to the standard cognitivist one that usually underlies neuroimaging and which accounts for several of the problems associated with it. In neuroimaging, as we have seen, the central difficulty in epistemic terms is that of inference – first the inference from cerebral blood flow to brain activity and then the inference from brain activity to cognitive function. But if we regard the brain as an inappropriate subject for psychological predicates, then these supposedly ‘inner’ processes may lose their fascination for us and the problem of inference cease to be so troubling.

In other words, Fodor may be right. In terms of the mind-body question, we tend not to think of ourselves nowadays as dualists; but our preoccupation with the purported findings of neuroimaging suggests otherwise. A genuinely non-dualist perspective would be less curious about these findings, knowing that there is no need to make hazardous inferences about that to which we have good access already. A strong case for precisely this perspective is argued for by P.M.S. Hacker and M.R. Bennett in The Philosophical Foundations of Neuroscience:

Some psychiatrists, as we have also seen, then wish to make a third inference, taking a laboratory-determined ‘biomarker’ (whether a brain scan or genetic abnormality) as proof of a mental disorder. But this step may also be regarded as questionable and for very much the same reason: what counts as ‘normal’ and ‘abnormal’ in psychopathology is based on the ways in which people behave – not according to the postulated properties of their brains.

Putting the brain in its place might be regarded as an act of heresy in an age dominated by the claims of those Raymond Tallis calls ‘neurotheologists’. But neuroscientists have themselves become aware of the extent to which they may be becoming victims of their own success: one recent study found that poor explanations for psychological phenomena accompanied by gratuitous neuroscientific references (especially those involving brain scans), tended to be believed more readily than good explanations lacking such references.

From its inception some 200 years ago, psychiatry has withstood accusations of being ‘mindless’ at certain times and ‘brainless’ at others. In recent decades, thanks to a number of scientific advances that the discipline has been able to make use of for purposes both of diagnosis and treatment, this balance has tipped towards its becoming ‘brainier’ again. Neuroimaging is one such tool. But, as we have seen, there are a number of difficulties around its use that are unlikely to be solved by further technological innovation alone.

Ultimately these are philosophical issues that psychiatry would be unwise to ignore if it is to retain its frontline position in the treatment of the mentally ill. In the words of Karl Jaspers, arguably the only figure to date to have been a genuine authority in both fields: