1 Introduction

Given that visualisations via medical imaging have tremendously increased over the last decades, the overall presence of colour-coded brain slices generated on the basis of functional imaging, that is to say, neuroimaging techniques, have for quite some now led to a notion of so-called “kinds” of brains or cognitive profiles, which, as neuroscientific research reveals, might be the predictors of “non-healthy” humans affected by neurological, neuropsychological or psychiatric syndromes or disorders (Dumit 2004). The case has been made that the use of such technological innovations of visualising brain function are to be expected in the validation of the so-called hypofrontality-hypothesis with regard to schizophrenia (Weinberger et al. 1996; Callicott et al. 2003). This assumption, of course, is not quite as neat as it sounds.

2 Technological innovations in the neurosciences

Since the 1970s morphometric neuroimaging techniques (CT, MRI) have increasingly provided important evidence of morphological changes in psychiatric disorders such as schizophrenia (McCarley et al. 1999; Shenton et al. 2001), affective disorders or drug-related diseases (Kalus et al. 2007), and therefore have become valuable tools for diagnostic purposes. This is not yet the case for functional imaging techniques (PET, SPECT, fMRI). However, position emission tomography (PET) and single-photon emission computed tomography (SPECT) nowadays play a crucial role in psychiatric research and are widely discussed in terms of their clinical implications, that is to say, not only for the diagnostic purposes but also with regard to validating the use of specific therapeutic interventions (Kuwert et al. 1998; Kalus et al. 2007). Furthermore, the use of functional magnetic resonance imaging (fMRI), which permits individual, rather than group-averaged neuroimaging data, is rapidly increasing (Braus et al. 2001; Haldane and Frangou 2006; Schneider and Fink 2007) and is therefore regarded as causing “a potential paradigm shift in psychiatric neuroimaging” (Callicott et al. 1998, 186).

PET was the first non-invasive technology to provide direct quantitative assessment of regional physiological processes in the living brain—commercial systems became available in 1978 (Taber et al. 2005). PET quickly diversified with the development of new radiotracers for a wide variety of measurements (allowing for exact measurements of cerebral metabolism (blood flow, glucose uptake) and of neurotransmitter receptor binding (dopamine-receptor uptake)) and was therefore quickly applied to studies of human cognition and of neurological and psychiatric disorders, providing insights into previously unknowable processes. In contrast to PET and SPECT-scanning, the most commonly used method in fMRI-measures does not require the administration of a contrast agent such as radiolabeled water, which has a very short half-life of only about 2 min. Hence, a bolus injection provides a snapshot of regional cerebral blood flow (rCBF) that can be repeated, if required every 10–15 min. The specific method in fMRI-measures relies on a blood-oxygen-level-dependent (BOLD) signal using the magnetic properties of oxygenated blood. Nevertheless, the metabolic and circulatory changes associated with the BOLD-signal are not yet fully understood. Actually, the correlation between the increase of blood oxygenation level, CBF, or glucose utilisation and the increase in neural activity seems to be the major point of contention among researchers of functional neuroimaging. The fundamental concept is the so-called structural and functional neurovascular coupling, which links neural processes with the hemodynamic change, offering the surrogate signal which is currently used to map “brain activity” (Logothetis and Pfeuffer 2004). Yet there are well known fundamental questions that remain from an epistemological point of view, especially when we seek to understand the pathological impairment of cognitive functions via visualising “epistemic hot spots” (Bogen 2001).

3 Imaging the brain—visualising the mind?

In highlighting some of these fundamental questions, one could rely on the astonishing hypothesis that via visualisation we could “see” our brain working and the cognitive function in question proceeding (Posner and Raichle 1995), for example, in reference to so-called activation patterns based on changes in the regional blood flow or blood oxygenation level of the brain. To see one’s own brain in action is always very impressive for individuals (Beaulieu 2002; Dumit 2004). Imaging data not only visually render pathological entities, but also tend to represent objective and concrete evidence for these psychophysical states in question.Footnote 1 First of all, this observation conflicts with the merely subjective nature of imaging data, considering inter-individual variability in brain structure and function, and, of course, individual pathology that might result in morphological and functional changes. Many factors can influence functional studies including technical factors, individual patient state, co-morbidities, medications, nicotine abuse, and with regard to PET and SPECT, tracer injection conditions. In addition to these considerations, we must also take into account that functional neuroimaging relies on the integrity of the vascular and metabolic response to neural activity. However, as neurovascular coupling seems not always be intact in diseases like Huntington’s (Montoya et al. 2006), when worse comes to worst, changes in vascular parameters may finally reflect a disturbance in neurovascular coupling. Furthermore, the assumption that visualising pathological entities intrinsically bears objective and concrete evidence for the psychophysical states in question conflicts with the epistemic status of scientific representation itself. Hence, there are at least two issues one has to consider: One is the question of how visualised entities relate to visible entities, and if visualisations bear an epistemic surplus. The other is whether visualisations do represent and if yes, what they represent.

4 Visualisation as “representation”

The idea of representation has been central to scientific issues of visualisation. Recently, the concept of scientific representation as a presentation, in other words as an image of something, has been criticised: In the late 1990s Hans-Jörg Rheinberger called attention to a continuum of meaning that afflicts the concept of representation within biomedical sciences. Hence, representation is not only used as providing a substitution (hypothetical construction) or a presentation (scientific model), but also a realisation of the phenomenon itself (Rheinberger 1997a, b), that is to say, the coming-into-being of scientific entities previously not “existing” (Hacking 1983). Besides queries about how to access the cognitive function in question and how to validate these procedures, from an epistemological point of view, “seeing” in functional imaging must be regarded as technologically constructed not only because of data generation and interpretation, but also due to specific conditions of an experimental setting in general. The technological construction or reshaping of scientific phenomena, also referred to as epistemic objects via experimentation and laboratory approaches in the field of the molecular and the cognitive sciences is a widely discussed issue, especially in the history and philosophy of the life sciences and other related fields of research (Hacking 1983; Rheinberger and Hagner 1993; Rheinberger 1997a; Uttal 2001; Gallagher 2006).

Within the realm of imaging studies, designing an experiment implies choosing a specific protocol out of a set of experimental tasks or neuropsychological tests (e.g. Wisconsin card sorting test) that could be correlated with the cognitive function in question, or rather some component of this function (e.g. “set-shifting”, the ability to display flexibility in the face of changing schedules of reinforcement). This necessity evokes a classical problem in psychology, especially in cognitive psychology: not only has it always been extremely difficult to define exactly what is meant by a cognitive or mental ability, but also to transfer the phenomena in question via conceptualisation, or operationalisation to an experimental setting. William Uttal recently emphasised, that the “data base of cognitive neuroscience itself is an implicit taxonomy of mental modules and components” (Uttal 2001). Moreover, the growing number of studies using imaging techniques has already led to a proliferation of hypothetical psychological, cognitive and emotive-motivational components within the cognitive and clinical neurosciences. Last but not least one consider the question of how these components in turn relate to phenomena or entities of the life-world (Huber 2008).

5 Visualisation of individual pathology: the clinical neuroscience endeavour

Clinical studies already show or at least suggest that individual pathology may result in or just coincide with morphological and or functional changes—considering brain plasticity or side-effects of ataractics. This has to be considered especially with regard to the validity of data generated through clinical trials (Farah and Feinberg 2006). Selecting participants for study purposes has to be done properly: Defining criteria for participant inclusion (for the experimental as well as for the control group) in clinical contexts not only requires delimiting the boundaries of “healthy” subjects, but also demands enormous effort to implement different modes of screening (neurological and neurophysiological screening) in order to find “healthy” subjects at all (Dumit 2004). The aim is to optimise the signal-to-noise ratio in the resulting data; thus, validating the whole generation process of individual data, and, moreover, to make these data comparable. Therefore, initial studies using imaging technologies like PET, SPECT or fMRI are based on selections of normal age-matched controls who have been symptom-free for years, ideally have no personal or family history of psychiatric disorders or the like, and therefore could be conceptualised as “ideal subjects” or “supernormals”.Footnote 2 A related problem arises with regard to identifying homogenous subject groups of patients: Defining experimental as well as control groups solely on the basis of clinical criteria might result in quite heterogeneous subject groups (Kalus et al. 2007).

These technological preconditions raise a question about the degree to which the information generated within functional imaging in psychiatry may be compared to the data, which is produced by other fields of clinical neurosciences (i.e. neuropsychology; neurosurgery) or by the cognitive neurosciences in general (Barch 2006). Also one has to consider significant conceptual as well as epistemological issues arising with regard to transfer the specific protocols developed within these fields reliably into psychiatric research and clinical practice (Hentschel 2000). Further questions address the validity of cognitive data produced by structural and functional brain mapping in psychiatry on a patient-specific basis, considering individual psychopathology and diverse effects of ataractics (morphological and functional changes) on the one hand and also given the specific demands in psychiatry on the other—this is to say, the need for fluctuation-resistant protocols to ensure the monitoring of therapeutic interventions. To this day, the endeavour of clinical neurosciences to elucidate the aetiology of psychiatric disorders via neuroimaging devices is challenged because of methodological, epistemological, and also conceptual issues—particularly with regard to ensuring sufficient descriptions of psychiatric syndromes or disorders (Wakefield and First 2003; Pichot 2004; Berganza et al. 2005).

6 The impact of neuroimaging on the concept of psychiatric diseases

The common practice of reducing complex phenomena to mere functional brain processes that—following metatheoretical assumptions (Fodor 1983, 2001)—are organised in modular ways does not merely have further implications for the understanding of our selves as “naturalised entities”.Footnote 3 This approach also gains immediate theoretical and practical relevance for the understanding and diagnosis of psychiatric disorders in order to look for so-called “cognitive profiles”Footnote 4 on the basis of neuroimaging devices, experimental tasks, and neuropsychological tests,Footnote 5 and in order to classify these profiles in psychopathological contexts. Given a grave methodological hiatus between various models of mental disease within psychiatric research, in particular the so-called “rift between biological and psychosocial approaches” (Garnar and Hardcastle 2004), these issues are of particular and growing interest. Referring to this fundamental difference, Marc Jeannerod proposed in 2004 to treat isolated syndromes of psychiatric, neurological or even neuropsychological patients as natural objects, regarding less further nosological questions or systematisations.Footnote 6 Actually, so-called integrationist models do become more and more important within the philosophy of psychiatry (Andreasen 1996; Garnar and Hardcastle 2004). Despite a range of different methodological levels, and neglecting a “pure reductionism”, these approaches do have an essential part in common. Currently, going beyond even an integrationist approach combining neuroscientific approaches with genetic mapping, and molecular and cell biological research, the ideal notion of “integrationism” within psychiatry is to build bridges between biological and psychosocial models. This rather pragmatic rapprochement may have special significance with regard to the description of so-called “endophenotypes”. This notion was initially coined in relation to schizophrenia in 1972, but seemingly “has experienced a renaissance in the past decade” (Robert 2007, 214). Integrationist models are also strengthened by the goal of not losing the patient being ill (the subjective perspective) out of sight, because of neglecting to access mere qualitative aspects of syndromes or disorders (Garnar and Hardcastle 2004; Robert 2007).

Currently, there is an additional alteration with regard to epistemological issues taking place: not only in neurological and psychiatric research (Price et al. 2000; Martin 2002), but also neurobiological, psychological and psychoanalytical approaches are moving closer together in terms of the technological platforms they use (Kandel 1998; Kandel 1999; Kaplan-Solms and Solms 2000; Kandel 2005). Given the growing impact of neuroimaging devices, one has to consider a kind of harmonisation of study designs and experimental protocols (cognitive and behavioural tasks, neuropsychological tests, statistical methods) within the clinical neurosciences as well as the cognitive neurosciences (Hardcastle 1999; Bechtel 2001). Such would seem to turn the purpose of imaging devices from a “clinical to a research-oriented use” (Andreasen et al. 1992), mirroring the scientific debate on the quantitative and qualitative approach to PET scans in the 1980s (Beaulieu 2002).

In terms of theoretical assumptions underlying data generation—as much the clinical as the experimental use of technological devices—one has to realise that these assumptions more often than not are neglected because of the impressive amount of data published in a very short time. Unfortunately, this is also the case concerning the experimental details as well as the specific experimental tasks or protocols in question (Fadiga 2007). However, specifically selecting the experimental task in relation to the probable outcome is mandatory for increasing the reliability of protocols and therefore should be considered as much in planning an experimental setting as in discussing generated or already published data, because the possible findings might be relevant mainly for the specific task, which is to say, their probable explanatory reach is limited to this task or setting only (Uttal 2001; Roux et al. 2003; Sinai et al. 2005; Kurthen and Schramm 2006). As a matter of fact, these epistemological concerns do affect very different fields of neuroscientific research and practice. On the one hand, these issues are of high interest with regard to uncovering complex cognitive functions via experimental operationalisation—what could for instance be seen within the neurosurgical theatre on the basis of establishing different concepts for cortical language mapping (Ojemann 1979; Ojemann et al. 1989, 2002; Picht et al. 2006). On the other hand, looking at the explanatory reach of the correlation between metabolic changes and cognitive or behavioural capacities, our knowledge about underlying processes of brain function is quite often challenged, because of conflicting data generated on the basis of neuropsychological tests as well as of brain images showing “non-corresponding” regional metabolic changes in a patient’s brain (Benton 2000; Heckers and Rauch 2001). Additionally, there are well-known queries and epistemological problems with regard to the so-called subtraction logic used for experiments of cognitive psychology as well as for the functional mapping devices that register metabolic changes via fMRI or PET (Donald 1995; Bogen 2001; Jäncke 2005).

By origin, this strategy is based on the classical “mental Chronometry” an experimental model introduced by the Dutch physiologist F·C. Donders in the late 1880s, whose studies on human cognition were based on reaction time experiments: thus, Donders isolated alleged key aspects of mental phenomena and structured them into successive functional processes, so-called “cognitive stages” (Donders 1969; Boring 1957). Donder’s strategy consists of subtracting a task state (i.e. discriminating the colour of the light) from a control state (i.e. responding to a light), and in measuring the different successive steps of a mental phenomenon. This procedure is successfully applied in the context of functional imaging studies where functional brain images obtained during a task are subtracted from functional brain images obtained during a rest condition (Posner and Raichle 1994, 1995). This method was introduced to functional imaging without the fundamental idea of Chronometry to measure the isolated stages, processes or components with respect to time. In today’s functional mapping devices, the different image is supposed to uncover only the measured metabolic change which is due to the additional processes compared to the control task. Studies using PET were the first ones that imply the subtraction logic outside of mental Chronometry or reaction time experiments within cognitive psychology. Nowadays the subtraction method is commonly used in functional neuroimaging designs (Tharin and Golby 2007). One central assumption underlying the subtraction logic is still widely criticised, the so-called additivity of conditions or cognitive components. Ideally the experimental condition consists of the processing in the baseline condition plus the processing related to the additional function in question (Donald 1995; Sidtis et al. 1999). Given this assumption and working with the rather simple linear model of a stimulus-response paradigm the application of the subtraction logic within the experimental or clinical study of complex phenomena such as language processing, or memory function apparently is confronted with diverse conceptual problems that question the validity of this methodological approach.

Scientists are not only faced with identifying key aspects that could be validated by implementing them experimentally (“operationalisation”), but also with defining a convincing baseline condition (Paller 1995; Morcom and Fletcher 2007). The additivity assumption in particular is commonly criticised when it comes to psychophysical processes that are supposed to proceed non-linearily (Donald 1995; Friston et al. 1996; Revonsuo 2001). Hence, from an epistemological point of view, it has not yet been sufficiently proved that phenomena like cognitive functions are accessible via experimentation or via imaging techniques, that is to say, if they could be considered as scientific objects at all (Erneling and Johnson 2005; Illes et al. 2006). Therefore, and this seems to be quite obvious, these conceptual issues have to be considered in advance, that is to say, before designing an experimental or clinical setting (Friston et al. 1996; Uttal 2001).

Last but not the least one has to consider the specific demands of—metaphorically speaking—seeing and understanding cognitive functioning with reference to colour-coded brain images standardised and normalised on the basis of so-called brain atlases (Beaulieu 2001). The creation of digital images of brain structure and function is not only due to these imaging devices themselves, but also to the construction of digital spaces of representation and reference: the most famous brain atlas is known as the Talairach atlas—named after the French neurosurgeon Jean Talairach who generated the first proper reference brain for stereotactic purposes in surgery in 1957 relying on the post-mortem sample of a woman in her 60s (Talairaich and Tournoux 1988). New atlases like the MNI (Montreal Neurological Institute) atlas are generated from about 300 male and female brains that have been normalised and averaged voxel by voxel (Jäncke 2005). Nowadays one is also commonly referred to so-called “probability atlases” or “population maps” which take into account the high variability of anatomical and cytoarchitectonical landmarks of the brain (Westbury et al. 1999)Footnote 7 Especially, concerning the analysis of data sets using PET and SPECT, it has long since become very important to increase the reliability of “reference brains”. Indeed, it seems imperative for functional imaging to refer to cytoarchitectonical landmarks via brain atlases for comparative reasons. From an epistemological point of view, these digital spaces of representation have to be regarded against the background of fundamental issues arising from generating and establishing scientific atlases. Lorraine Daston and Peter Galison famously have introduced them as “systematic compilations of working objects”, and therefore, as “dictionaries of the sciences of the eye” (Daston and Galison 2007, 22). Thus, one always has to imagine the possibility of overestimating the epistemic status of atlases because they intrinsically “set standards of how phenomena are to be seen and depicted” (ibid. 19). As a matter of fact, setting standards of the perception of scientific objects via brain atlases is immediately connected to processes of inscription—which commonly are conceptualised as specific, techno-based transformation procedures within the natural and the biomedical sciences that in the end create feasible images of scientific knowledge (graphs, schemata, indices, atlases, etc.), which in turn are shaped by the specific technology or experimental design that was used to generate them (Latour and Woolgar 1986; Latour 1990). Additionally, these scientific modes of writing and imaging are regarded as intrinsic, “automatic” products of inscribing bodily structures (Rheinberger 1997b; Borck 2005).Footnote 8 These concerns have to be considered especially with regard to the ambiguous shape of brain atlases because they are “both objects of knowledge and tools used in research practices” (Beaulieu 2001, 638).

7 Visualisation as “inscription”

Hence, to conceptualise modes of visualisation as modes of scientific representation that do not present or image natural objects but create phenomena themselves is to address issues of inscribing bodily structures via feasible images of scientific research and practice. It seems mandatory to re-evaluate the traditionally narrow representation-reference ratio with regard to digital devices of imaging. Actually, we do not have to face a mere loose “referential tie between body and image” (Schinzel 2006, 190), but rather to recognise that the representation-to-reference ratio is not valid any more. As Rheinberger puts it, “the traditional meaning of “representation” is erased” (Rheinberger 1998, 295).Footnote 9 But, how does this re-evaluation also affect epistemic issues of the visualisation of cognitive function or pathological entities? First of all, the case has been made that visualisations in contrast to the concept of the visible have to be regarded as specific products of technology. Visualisations evoke a “false presence” of real-life phenomena by establishing themselves within closed graphematic or pictorial spaces (Belting 2007). Hence, the epistemic status and explanatory reach of the so-called pictorial turn itself seems to be affected (Lüdeking 2005). Given that images become essential and integral parts of scientific research (Borck 2001), there is some urgency to clarify the question of whether pictures (here: visualisations) bear any epistemic surplus that for instance could not be achieved via mere “statistical mapping” or other scientific modes of representation. One may for instance rely on the famous rift between clinical neurosciences and other fields of neuroscientific research: The first conceptualises neuroimaging as techniques of visualisation; however, the latter redefines “scans as representation of quantitation” or “statistical maps” (Beaulieu 2002, 59).Footnote 10 This (non-picture) assumption aims at acknowledging that colour-coded brain images first of all are secondary products of data generation, manipulation and interpretation, although bearing in some respects an “aesthetic value” (Beaulieu 2000; Beaulieu 2002; Weigel 2004). To this day, the question remains unanswered, how visualisations based on digital technologies relate to pictorial versus semantic conceptions of images in general. Nevertheless, even the non-picture approach refers to an important benefit of imaging devices: The possibility of visualising your data may be of interest when it comes to sample data quality or to assessing the signal-to-noise-ratio visually (Beaulieu 2002).

8 Conclusion

Within the cognitive neurosciences functional brain mapping, also ambitiously referred to as “mind mapping,” or as “decoding mental states” (Haynes and Rees 2006) surely relies on the still very common endeavour of localising cognitive phenomena like motor or language function, memory processing or even social capacities with reference to the brain’s morphology (Zawidzki and Bechtel 2005; Hardcastle and Stewart 2005). The clinical neurosciences even tend to go one step further: In acknowledging the structural and functional data generated within the cognitive neurosciences, at this stage great interest can be observed in visualising the fundamental differences between individuals with healthy and non-healthy brains—looking for the specific features of morphological and functional changes in these brains that could be reliably correlated with neurological, neuropsychological or psychiatric syndromes or disorders. Not to the least this endeavour to establish objective “cognitive profiles” of pathological entities via neuroimaging devices has to be regarded against the background of scientific conceptions of objectivity.Footnote 11 To develop so-called “kinds of brains”, that is to say, visualised typologies that characterise pathological structures and states of the brain, and to establish “disease atlases” that reflect a patient group as a whole is to acknowledge that “atlases are always carefully (though not always explicitly) selective, in order to be objective” (Beaulieu 2001, 639). As a matter of fact, one may conclude that ‘seeing’ on the basis of functional neuroimaging does not mean that we can trust our eyes, that is to say, that the computed and colour-coded data are self-evident. This is not only because of a large number of epistemological queries arising in the field of neuroimaging as such and in the field of functional brain mapping, but also because of fundamental problems that emerge by exchanging data generated within the neuroscience laboratory and in clinical day-to-day practice. These problems—as much as the epistemic status and explanatory reach of visualisations within biomedicine—have to be considered in particular with regard to the ongoing neuroscientific reconceptualisation of psychiatric disorders, concomitant neuropsychological symptoms, or other disorders that are commonly referred to as “brain-related diseases”.