Introduction
The term “reductionism” is used broadly in academic discourse about psychiatry as an effigy, often as the antithesis of the biopsychosocial model [1]. With rare exceptions – for example [2, 3] – the term is used as a perjorative, non-specifically and often, to imply naivety for example, in the case of reductive materialism [4]. It has been argued that “reductionism seems to be all but dogma in psychiatry” and this is problematic because “it is arguable that one of the central motivations for current psychiatric practice derives from a commitment to reductionism” with the consequence that treatments for psychiatric illness become “ministering the brain” with pharmacological interventions [5].
Anecdotally, at the institution I work for, each year post-graduate students submit essays on nosology, evidence-based interventions and theories of mental illness that rehearse the reductionism effigy (presumably, a perspective inherited from the people teaching them and the curriculum). I rarely find much in the way of specificity in either definition or explanation of why something is reductionist, or how this is inherently wrong. This may happen at other institutions also.
Criticism of reductionism extends to medicine more generally; in a recent paper [6] the authors argue that even in the contemporary era of adopting holistic models of patients – under the banner of “personalised medicine” – we still “reduce the patient to a collection of precise molecular sequences with detailed clinical phenotyping”. Here, “biomedical reductionism” is defined in contrast to “network medicine” that “focused on the interrelational structures emerging from complex interactions among genes, proteins, internal milieu, and external environment”. In [6], they do add some specificity by introducing reductionism in it’s historical contexts:
• In the natural sciences, reductionism pursued “explication of biologic phenomena using the principles of physics” following from [7, 8]. Subsequent debates “addressed the question of whether one science can be ’reduced’ to another or whether independent fields of knowledge display irreducible or emergent properties”
• In the medical sciences, reductionism is “often equated with older conceptions of disease as something that can be separated from the sick person and scrutinized with successively finer analytic tools”. Of note, [6] anchor the ontological approach of identifying and classifying disease using progressively smaller units of analysis in Syndenham’s 17th century nosology. Then, they trace reductionist practice through the “clinicopathological conference” (CPC) model of a patient, that progresses from subjective history of presenting complaint, through objective examination and investigations, to a pathological target such as a diseased organ, tissue, or in contemporary medicine, to the molecular level including the genome and proteome. They conclude “As a weekly ritual in hospitals and medical journals, the CPC recapitulates the diagnostic and therapeutic power as well as the objectification and alienation inherent in biomedical reductionism”
The important distinction in [6] is that of ontology versus practice. Labelling something as “reductionistic” without qualification assumes everyone agrees on the definition and that it is intellectually flawed. I am not interested in an absolute defence of reductionism – rather, trying to understand what properties or features of reductionism are so intellectually offensive and arguing that they are in fact neutral in a way the effigy does not make clear. I propose the central problem is one of definition: reductionism is assumed to be a unitary concept, rather than having different meanings implying different intellectual commitments. Without specificity in definition, people conclude reductionism has failed to capture the subtlety and complexity inherent in mental illness and there clinical consequences such as people being perceived and treated poorly.
Definitions
In technical discourse, the term reductionism requires qualification [9, 10]. Not being formally trained in philosophy (as may also be the case for many who use reductionism as a pejorative), it seems to me that philosophy of science (PoS) and mind (PoM) have distinct meanings for reductionism. These differences in definition may seem subtle (this may be the reason precise definition is largely ignored in the pejorative use) but are relevant to this discussion.
Theory Reductionism
Nagel’s theory reductionism [7, 8] appears to be held as an exemplar for reductionism in philsophy of science and mind – see for example, [11]. The proposal is: the laws of some “higher level” theory (HLT) can be explained as logical consequences of another “lower level” theory (LLT). Familiar examples being the reduction of psychology to biology; biology to chemistry, or chemistry to physics. Importantly, Nagel’s model requires that reduction is:
(i) a kind of explanation relation, which (ii) holds between two theories iff [if and only if] (iii) one of these theories is derivable from the other, (iv) with the help of bridge laws under some conditions [12]
The “some conditions” clause in Nagel’s proposal arises as a result of two kinds of reduction:
1. homogenous reduction – the terms in the HLT can be translated directly to terms in the LLT (i.e. the two theories differ only in terms of vocabulary).
2. heterogenous reduction, where terms in the HLT do not appear in the lower-level theory
Heterogenous reductions are of most interest: as an example, take psychology as the HLT containing terms like “belief” – these do not occur in LLTs like neuroscience. For these circumstances, bridge laws describe how terms in the LLT (e.g. states of populations of neurons) relate to terms in the HLT (the psychological term “belief”). These can be identities – e.g. “belief” in the HLT is identified with (i.e. is no more than) states of populations of neurons. It has been argued that the only logically sound and operationalizable account of bridge laws is really identity – see Section 1b of [13] – but Nagel proposed something more inclusive than strict identities, allowing empirical hypotheses to function as bridge laws [13]. So, the hypothesis that neuronal population states (LLT) co-occur with belief (HLT) could provide a bridge law, albeit one awaiting some empirical data to support it.
Ontological Reductionism
There is also ontological reductionism that asserts nature is actually composed of a parsimonious number of basic, fundamental (indivisible, atomic) entities and that all other observable entities are in fact composed of these smaller and less complex parts. On this view, a strong claim would be that beliefs are terms in a HLT (e.g. psychology) for which we have not yet discovered all the relevant components or governing laws of the LLT (e.g. neuroscience). They may turn out to be states of neuronal populations – to use Nagel’s terminology – with bridge laws that describe how discrete patterns of these states give rise to or govern these specific beliefs.
Physicalism is also an ontological-reductionist perspective [14, 15] that asserts a strong identity: mental processes or states are just states of the brain. Under this view, reports of experience – such as "I am in pain” – are just that, mere reports and they have no ontological status as e.g. higher level constructs subject to reduction.
Explanatory Reductionism
The final definition considered here is explanatory reductionism where a HLT is reduced to a LLT with respect to a set of observational data (D) – see Definition 3, pp. 13 of [16]:
• The vocabulary of HLT contains terms not in LLT
• Any part of D explainable by means of HLT is explainable by LLT
• LLT is at least as well systematised as HLT
Importantly, [16] requires that data, D, are “time stamped” to account for the possibility that future data may serve (or refute) the reduction of HLT to LLT. This requirement appears to complement Nagel’s notion of empirical hypotheses serving as bridge laws.
Canonical Example
An oft cited example of reductionism (which fits with theory and ontological reductionism) is the Boyle-Charles Law (BCL) for gases – a high-level (macroscopic) construct describing the state of a gas in terms of pressure, volume and temperature in a simple algebraic relationship. It has explanatory power and utility – say, for an engineer, who can make predictions with it. A low-level (microscopic) construct – the kinetic theory (KT) of an ideal gas – describes the behaviour of a gas by specifying the number of particles, their masses and velocities, confined in a finite volume, interacting by elastic collisions and Newtonian laws of motion. Mathematics directly allows the derivation of BCL from KT, showing that the high-level term temperature can be equated with the low-level term mean kinetic energy [10]. In this example, the BCL for gases has been reduced to the KT – the latter is, perhaps, more interesting to the physicist than the engineer.
The Reductionist ‘Toolkit’
In each of above definitions, I would argue that the central themes are:
1. compositionality: the theory, entity or phenomena in the higher level is decomposed into the usually smaller and tentatively more fundamental parts of the lower level
2. abstraction: the higher level often abstracts away from details of the lower level, providing cognitive tractability but at the sacrifice of a more complete or fundamental description
3. mechanism: explicit formulation in procedural language i.e. described by symbols and algorithms for manipulating symbols, an exemplar case being mathematical models. Often, the lower level has a more procedurally complete description than the higher level
There is nothing inherently “bad” about these intellectual tools, nor do they imply a commitment to any one reductionist standpoint:
Most contemporary reductionist as well as eliminativist positions include some commitment to materialism or physicalism - the view that the physical or material provides the fundamental reductive base. However reductionism per se is ontologically neutral. The concept of reduction entails no specific ontological positions [12]
And further:
Mechanistic explanations are constitutive or componential explanations: they explain the behavior of the mechanism as a whole in terms of the organized activities and interactions of its components [12]
But objections might be raised about their deployment and this is why definition is important – do we object to a specific commitment (theory, ontological or explanatory reductionism) or is it the intellectual tools (compositionality, abstraction, mechanism) we find questionable?
Example Applications
Diagnostic Proliferation
Assume diagnostic categories are being ‘discovered’ which have a syndromic profile of co-occuring signs and symptoms; there exists demonstrable impairment and disability and the number of people proposed to be experiencing the syndrome is significant enough to warrant attention.
Take two recent proposals as examples: selfitis and internet gaming disorder (IGD). Selfitis is characterised as a “technologically related mental health disorder” alongside others (including fear of having no mobile phone and feeling ill after searching online for symptoms of an illness) and the exploratory analysis presented in [17] used qualitative focus groups followed by development of a questionnaire that tries to describe the underlying dimensions of people’s selfie-taking behaviour. The questionnaire was applied to people who met criteria for three levels of severity of selfitis (borderline, acute, and chronic). Factor analyses were used to tease apart the latent dimensions in this data. They describe a resulting six-factor model and show, for example, that for the more severe category (labelled “chronic”) the highest mean scores were on dimensions of attention seeking, environmental enhancement and social competition.
Examined with the tools of compositionality, abstraction and mechanism we can see:
• compositionality: terms like attention-seeking and social competition might reasonably be explained as a function of underlying personality traits – the authors implicitly commit to this by proposing a number of factor-analysed dimensions less than the number of items in their questionnaire.
• abstraction: selfitis is a nominal, descriptive higher-level term for an observed behavioural phenomena that summarises (or is shorthand) for what the author’s propose is explained by other features (e.g. attention seeking) that could be more fundamental, lower-level terms (e.g. personality traits) – indeed, the term selfitis was used in the media in just this way; as a nominal shorthand for an observed behaviour
• mechanism: speculatively, we can offer that the interaction of lower-level personality traits function together so that selfitis is observed
Further, selfitis seems to mesh with an explanatory application of reductionism, but one might also apply both theory and ontological reduction to see if the concept “selfitis” should qualify as something distinct from e.g. personality traits. For theory reductionism, we require a heterogenous reduction because “selfitis” is proposed to reduce to terms like “attention seeking” or “social competition”.
From an intentionally critical viewpoint, one might say that subjecting the proposed disorder of “selfitis” to the toolkit of reductionism, we find that it can be reduced to personality traits and further, the reduction has enough explanatory power to make the higher-level little more than nominal. We might also claim that the observed higher-level phenomena (“selfitis” as selfie-taking behaviour excessive enough to cause disruption to everyday function) are then expressions of the lower-level terms (attention seeking, social competitive behaviours) and that the lower-level terms are what we should focus on. In some sense, this rehearses the Jasperian distinction of form (personality traits) and content (the expression of behaviours, in this example, the use of a new technology). We can say nothing about burden, impairment or disability (without data) but we might leverage the already-established understanding of impairment caused by the lower-level personality traits. With that data, we might even be able to support a thorough explanatory reduction.
For internet-gaming disorder, there is at least a theory reductionist commitment in it’s formulation [18] – as a ‘non-substance’ addiction disorder. Subjecting IGD to the same analysis, and using the vocabulary of the research domain criteria [19]:
• compositionality: IGD can be defined as behaviours described by the lower level vocabulary of addiction; notably compulsion and withdrawal for which there are tentative cognitive and neurobiological substrates in the positive valence (approach motivation, reward valuation, expectancy, habit) and negative valence systems (loss, frustrative non-reward)
• abstraction: the term IGD aggregates together the lower-level terms (compulsion, withdrawal and their lower-level substrates) and identifies their expression in terms of behaviours with respect to “gaming” rather than “substance use”
• mechanism: under the research domain criteria [19], there are tentative lower-level descriptors for the neurobiology underpinning the behaviour which can be described procedurally as changes in the mesolimbic system
In this analysis, IGD is the expression of a lower-level addiction syndrome in a new context (internet gaming), and with an intentionally critical purview, a theoretical, ontological and explanatory reduction suggests we currently have grounds for studying IGD with the tools available from addiction sciences.
Clinical and Scientific Reasoning
An objection to theory and ontological reductionism is that there is explanatory utility for a given higher-level over candidate lower-levels. Take the earlier example of the Boyle-Charles gas law (BCL), and its reduction to a kinetic theory (KT) of an ideal gas, where an engineer might find the BCL useful, but the physicist prefers the KT. An engineer needs to make predictions about the behaviour of gases in an application of physical theory, exploiting the abstraction represented by BCL that may be perfectly adequate in a way that the physicist finds unacceptable. By analogy, clinical psychiatry needs the engineer’s perspective to describe syndromes and predict behaviours or outcomes, the difficulties faced by people and to identify treatments. For example, assume a comprehensively reduced explanation exists for thought disorganisation – it may have little utility for the practising clinician faced with a person displaying this sign.
The intellectual toolkit of reductionism should be considered neutral. Controversy enters when casting the products of analysis by compositionality, abstraction and mechanism into a commitment implied by say, adopting theory or ontological reductionism. I would suggest that in reality, most psychiatrists would advocate for Kendler’s position of decomposition and reassembly [3] which directly draws on Marr’s influential approach to computational theories of vision [20, 21]. Marr’s approach (to visual perception) contained multiple levels of explanation, starting with observable behaviour that provides constraints on the candidate algorithms which in turn constrain their implementation in biological substrates. The elegance of Marr’s approach is not in terms of the vocabulary, terms or specific sciences working at each level, but to exploit multiple levels of explanation to progressively decompose a complex phenomena, and use each level as constraints on the higher and lower levels. The three intellectual tools of reductionism serve the discovery and articulation of these levels.
Conclusion: Real Reductionism, Not the Effigy
My reading of “reductionism” in discourses on mental health is that practitioners of different academic disciplines want to be proprietors of the only true or fundamental theory of mental illness and that disciplines compete with each other to provide this “grand unified theory” (GUT). This leads to tribal behaviour that includes use of the reductionism effigy to summarily discredit the approach of a presumed opponent or competing tribe. I suspect that objections to reductionism can be couched as a rejection of Oppenheim and Putnam’s unification of sciences project where physics was offered as the reductive base. However, closer examination of Oppenheim and Putnam’s proposal yields:
It is not absurd to suppose that psychological laws may eventually be explained in terms of the behavior of individual neurons in the brain; that the behavior of individual cells – including neurons – may eventually be explained in terms of their biochemical constitution; and that the behavior of molecules – including the macro-molecules that make up living cells – may eventually be explained in terms of atomic physics [22]
With the caveat:
If this is achieved, then psychological laws will have, in principle, been reduced to laws of atomic physics, although it would nevertheless be hopelessly impractical to try to derive the behavior of a single human being directly from his constitution in terms of elementary particles
Under this eliminativist view, if mental illness turns out to be explicable in terms of neurophysiology, the discipline of biology “wins” and replaces other explanations (at least, until another reductive base discipline explains the biology). This is not so problematic because the reductive base discipline cannot necessarily replace the need to explain or treat signs, symptoms or distress with models that occupy ontologically higher levels of abstraction. For example, it makes no sense to use the neurobiology of G-coupled proteins, ion influx/efflux and action potential generation to explain or treat avoidance behaviours in someone suffering post-traumatic stress disorder. Complementing Oppenheim and Putnam’s observation, the gulf between levels of abstraction and the resulting cognitive complexity of the necessary compositions and mechanistic models practically exclude their use as a tool for clinical practice. By analogy, you don’t teach someone to use a word processor by showing them how to build the software themselves.
Returning to the previous analogy of ideal gases, a physicist might favour statistical mechanics (because their job is to uncover and describe the laws of nature) but an engineer might still use the BCL gas law (because it has utility in describing and predicting the behaviour of gases at a relevant level of abstraction). Psychiatry (as a clinical practice) stands across both basic science and practice (by analogy, acting as both physicist and engineer). The desire to be ultimately correct is a powerful drive in a search for a GUT, but irrelevant when your job is to work for people experiencing debilitating distress making use of the best and most relevant theories and evidence available.
The global “take home” message:
1. If your objection to something invokes an argument that it is reductionist, state which version and why that version’s commitments are wrong-headed for the case described
2. If your objection is to the intellectual toolkit, tell us why it is inappropriate or fails – show why abstraction, composition and mechanism fail to capture an understanding of the phenomenon or specific case under consideration
But please, don’t just state it’s “reductionist”, assume we will all agree, adopt an anti-reductionist stance and move on.
References
[1] G. L. Engel, “The need for a new medical model: a challenge for biomedicine,” Science, vol. 196, no. 4286, pp. 129–136, 1977. [Online]. Available: http://science.sciencemag.org/content/196/4286/129
[2] H. Karlsson and M. Kamppinen, “Biological psychiatry and reductionism. Empirical findings and philosophy.” The British journal of psychiatry : the journal of mental science, vol. 167, no. 4, pp. 434–8, oct 1995. [Online]. Available: http://www.ncbi.nlm.nih.gov/pubmed/8829708
[3] K. S. Kendler, “Explanatory Models for Psychiatric Illness,” American Journal of Psychiatry, vol. 165, no. 6, pp. 695–702, jun 2008. [Online]. Available: http://psychiatryonline.org/doi/abs/10.1176/appi.ajp.2008.07071061
[4] C. R. Cloninger, “The science of well-being: an integrated approach to mental health and its disorders.” World psychiatry : official journal of the World Psychiatric Association (WPA), vol. 5, no. 2, pp. 71–6, jun 2006. [Online]. Available: http://www.ncbi.nlm.nih.gov/pubmed/16946938
[5] I. Gold, “Reduction in psychiatry,” The Canadian Journal of Psychiatry, vol. 54, no. 8, pp. 506–512, 2009.
[6] J. A. Greene and J. Loscalzo, “Putting the patient back together-social medicine, network medicine, and the limits of reductionism.” The New England journal of medicine, vol. 377, no. 25, p. 2493, 2017.
[7] E. Nagel, “The meaning of reduction in the natural sciences,” in Science and Civilization, R. C. Stouffer, (Ed.) University of Wisconsin Press, Madison, 1949, pp. 99–135.
[8] ------, The Structure of Science: Problems in the Logic of Scientific Explantation. Routledge and Kengan Paul, 1968.
[9] I. Brigandt and A. Love, “Reductionism in biology,” in The Stanford Encyclopedia of Philosophy, spring 2017 ed., E. N. Zalta, (Ed.) Metaphysics Research Lab, Stanford University, 2017. [Online]. Available: https://plato.stanford.edu/archives/spr2017/entries/reduction-biology/
[10] F. Dizadji-Bahmani, R. Frigg, and S. Hartmann, “Who’s afraid of nagelian reduction?” Erkenntnis, vol. 73, no. 3, pp. 393–412, 2010.
[11] N. Block, “Anti-reductionism slaps back,” Noûs, vol. 31, no. s11, pp. 107–132, 1997.
[12] R. van Riel and R. Van Gulick, “Scientific reduction,” in The Stanford Encyclopedia of Philosophy, E. N. Zalta, (Ed.) Metaphysics Research Lab, Stanford University, 2016. [Online]. Available: https://plato.stanford.edu/archives/win2016/entries/scientific-reduction/
[13] A. Ney, “Reductionism,” in The Internet Encyclopedia of Philosophy, J. Fieser and B. Dowden, Eds., 2017. [Online]. Available: http://www.iep.utm.edu/
[14] J. J. Smart, “Sensations and brain processes,” The Philosophical Review, vol. 68, no. 2, pp. 141–156, 1959.
[15] U. T. Place, “Is consciousness a brain process?” in The Mind-Brain Identity Theory. plus 0.5em minus 0.4emSpringer, 1970, pp. 42–51.
[16] J. G. Kemeny and P. Oppenheim, “On reduction,” Philosophical Studies, vol. 7, no. 1, pp. 6–19, 1956.
[17] J. Balakrishnan and M. D. Griffiths, “An exploratory study of “selfitis” and the development of the selfitis behavior scale,” International Journal of Mental Health and Addiction, pp. 1–15, 2017.
[18] N. M. Petry, F. Rehbein, C.-H. Ko, and C. P. O’Brien, “Internet Gaming Disorder in the DSM-5,” Current Psychiatry Reports, vol. 17, no. 9, p. 72, sep 2015. [Online]. Available: http://link.springer.com/10.1007/s11920-015-0610-0
[19] B. N. Cuthbert and T. R. Insel, “Toward the future of psychiatric diagnosis: the seven pillars of RDoC.” BMC medicine, vol. 11, p. 126, 2013. [Online]. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3653747
[20] D. Marr and T. Poggio, “From understanding computation to understanding neural circuitry,” Massachusetts Institute of Technology, Artificial Intelligence Laboratory, A.I. Memo 357, 1976. [Online]. Available: https://dspace.mit.edu/bitstream/handle/1721.1/5782/AIM-357.pdf
[21] D. Marr, Vision: A Computational Investigation Into the Human Representation and Processing of Visual Information. WH Freeman, 1982.
[22] P. Oppenheim and H. Putnam, “Unity of science as a working hypothesis,” in Minnesota Studies in the Philosophy of Science Vol II, H. Feigl, M. Scriven, and G. Maxwell, (Eds.) University of Minnesota Press, Minneapolis, 1958.
