The problem of reductionism in educational theory: Complexity,causality,values

Introduction

This article draws on the concept of reductionism to examine inadequate theorisation and mechanistic causal assumptions in education which result in a loss of complexity, openness and values. This seems particularly appropriate at present for England and similar systems, where reductionist approaches have become hegemonic in many aspects of schooling. In recent years, a strident ideological campaign, led by government ministers, has sought to reconceptualise and reconfigure what counts as educational achievement, high-quality teaching, social justice and research evidence. This involves not only the depoliticisation of politics and policy (Harvey, 2010: 218–219), but also a shift towards a technical discourse when considering curricular and pedagogical issues and practices (Shapiro, 1990: 13; Clarke, 2012).

A critique of reductionism is important, therefore, to protect against misleading simplifications. Although the concept is relatively familiar in the natural and social sciences (see Bock and Goode, 1998; Sayer, 2010), it has rarely been highlighted in the field of education, where its use has generally either been applied to very specific issues or used as a vague term of abuse (some honourable but rare exceptions are referred to below, for example, Apple, 1984; Beck, 2007). For that reason, this article attempts to map different forms of reductionism, drawing on debates in the natural and social sciences. Further, to ground this in a coherent ontology and epistemology, it refers to theories of causality, stratification and emergence in Critical Realism. It moves on to reflect on important manifestations of reductionism in education, involving pedagogical theory, poverty and underachievement, school evaluation and change, and evidence-based practice.

Reductionism affects policy and administrative systems as well as related research paradigms, but goes right down to fundamental assumptions about learning and knowledge. In practical terms, reductionism is disempowering because it denies to professionals, as well as policymakers, the understanding they need to operate skilfully and with agency in challenging situations. Reductionism, in various forms, depends too much on surface readings of reality, reinforcing ‘common-sense’ assumptions. It obscures, distorts or simply fails to perceive the power dimension of what it studies. This results in ideological misrepresentation, which disempowers practitioners.

The concept of reductionism

Reductionism takes many forms but generally signifies a loss of complexity which hinders an adequate understanding of reality. In its classic usage, it refers to the inappropriate deployment of one scientific discipline to explain matters which require a different branch of science. This is described, metaphorically, as using too ‘low’ a science – for example, trying to explain biological phenomena mainly through chemistry, or complex social situations such as war in terms of aggressive animal instincts.

Reductionism is not, however, limited to this idea of a ‘hierarchy’ of sciences. More generally, it can also refer to the ‘omission of important co-determinants of a multi-causal situation’ (Sayer, 2010: 34), or the choice of an inappropriate perspective or conceptual framework. It is for this reason that Rose (2005: 95) insists on ontological unity but epistemological pluralism. Using the example of a frog jumping into a pond (Rose, 2005: 100–113), it is perfectly valid to study the biochemical properties of its muscles, but this will not explain why now. Its reaction to a nearby snake requires other explanations, including intelligent evasion or instinctive reactions rewarded by natural selection. The argument goes back a long way; indeed, it is prefigured by Aristotle’s model of four complementary types of causality: material, formal, efficient and final causes. Drawing on and moving beyond Sayer (2010), I will attempt in this section to map out the various directions which reductionism can take.

It may be, however, that we are also dealing with something in the zeitgeist – a certain haste in finding fast answers and quick policy fixes, along with the eclipsing of philosophical discussion in public life. Dennett speaks of a ‘greedy reductionism’:

In their eagerness for a bargain, in their zeal to explain too much too fast, scientists and philosophers … underestimate the complexities, trying to skip whole layers or levels of theory in their rush to fasten everything securely and neatly to the foundation. (Dennett, 1995: 82)

The critical realist challenge: causality, stratification, emergence

The preceding section has looked at various types of reductionist simplification, resulting in inadequate understandings of reality. This section aims to develop and deepen the argument by contrasting critical realist approaches to ontology and epistemology with an empiricism derived ultimately from David Hume. Although Humean empiricism is often tacit and unacknowledged, it is inherent in much experimental and statistical work in education and other social sciences. The critique of empiricism is not, of course, an argument against the need to empirically test out theories, or an invitation to engage in anti-realist or anti-materialist belief systems, but rather a challenge to inadequate versions of materialism or realism.

The empiricist error begins with Hume’s sceptical argument that however many times one billiard ball hits another and the second ball moves, we still cannot assert a causal relationship. In other words, scientific laws can only express regularity, since we cannot get hold of the causal mechanism. Critical realists such as Bhaskar (1978) turn this round by distinguishing (1) the real, (2) the actual and (3) the phenomenal. They argue that causal forces (the real) do not always actualise and, even when they do, they may not be easily visible. ‘Causal laws must be analysed as tendencies, which may be possessed unexercised and exercised unrealized, just as they may of course be realized unperceived (or undetected) by anyone’ (Bhaskar, 2011: 16). In open systems, forces may be real (level 1) and formulable as scientific laws without resulting in regularity (levels 2 and 3), since other forces are at work or the environment is unconducive.

Critical realism, as a metatheory of both natural and social reality – and consequently natural and social sciences – directly challenges the Humean emphasis on regularity. In Sayer’s words:

The conventional impulse to prove causation by gathering data on regularities, repeated occurrences, is therefore misguided: at best these might suggest where to look for candidates for causal mechanisms. What causes something to happen has nothing to do with the number of times we observe it happening. (Sayer, 2000: 14)

There is, as Archer (1995: 34) argues, an association between empiricism and individualism: empiricist methods favour individualism, or at least the activities of small groups because they are directly visible, unlike social structures. Popper (1945: 98), for example, insisted that ‘all social phenomena, and especially the functioning of social institutions, should be understood as resulting from the decisions, actions, attitudes etc. of human individuals’. Crucially, reducing the social world to the actions and decisions of individuals or aggregates of individuals amounts to a failure to see how social structures, albeit often not directly visible, are causally efficacious. Reductionism, in this case, fails to understand how rules (even when tacit) regulate conduct, how roles predate their incumbents, and that role relationships give rise to ‘emergent powers’ which ‘cannot be reduced to those of their constituents’ (Archer, 1995: 43–51). For all the efforts of positivists and ethnomethodologists, it is difficult to stick with the belief that ‘army is just the plural of soldier and all statements about the army can be reduced to statements about the particular soldiers comprising it’ (Jarvie, 1959, quoted in Bhaskar, 2011: 70).

Structures, on the other hand, cannot be regarded as fixed. They interact to generate complex historical change. Consider, for example, the powerful combination of freedom to buy and sell land (enclosures), mercantile imperial expansion and industrial technologies which led to the powerful industrial capitalism shaping society and culture in 19th-century Western Europe. This, in turn, led to neo-imperialism, computer technologies, globalisation, finance capital, and so on, but not in a linear process; it resulted in a transformation rather than abolition of capitalism. All these forces have deeply affected modern education, creating the circumstances for economic and political formations (significantly neo-liberalism) to shape new forms of school governance, datafication, increasing pressures on childhood, an economistic curriculum – but never absolutely determining what will result:

People … are capable of resisting, repudiating, suspending or circumventing structural and cultural tendencies, in ways which are unpredictable because of their creative powers as human beings. In other words, the exercise of socio-cultural powers is dependent inter alia upon their reception and realization by people: their effect is not direct but mediated, for there are no other ways in which it could be exercised. (Archer, 1995: 195)

Social powers do not have to be physical structures (for example, bureaucracies, division of labour) but can also be cultural norms of expected behaviour (for example, new public management, school choice) or systems of meaning (discourses). Systems of meaning can condition behaviour as much as physical structures do, and hegemonic discourses play an important role in maintaining unequal structures. They are not, however, deterministic.

This inevitably raises concerns about the role and limitations of mathematical representations in research. Whilst statistics enables us to map the world and see a stretch of reality at a glance, it is inevitably selective, concealing certain features: it intrinsically tends towards a loss of complexity. This is why we should heed Rose’s (2005: 9) warning, speaking as a biologist: ‘Not everything is capable of being captured in a mathematical formula. Some properties of living systems are not quantifiable, and attempts to put numbers on them produce only mystification’. The complex dynamics of emergence and stratification of powers in open systems present multiple challenges to social science, making diverse variants of reductionism a constant threat, as scientific methodologies are applied which do not measure up to the ontological complexity.

Finally, of course, emergence is central to the field of education; indeed, emergence is precisely what education is about. Consequently, it is rare that children are equally talented in all areas, which is somewhat inconvenient for advocates of unitary intelligence. It means that children rarely progress in smooth linear ways, which radically undermines the mechanics of test-based accountability systems. Further implications of critical realism for education will be discussed later.

Reductionism in the education field: pedagogy and the learner

The history of 20th-century pedagogical theory can be written as a struggle against reductionism, a struggle which was itself seminal in its two leading figures, Dewey and Vygotsky. Reich et al. (2016) summarise Dewey’s contribution as a ‘constant struggle against behaviourism and instrumental reductionism … insistence on context’, with behaviourism guilty of forgetting ‘emotional, social and cognitive aspects of experience’ (1002). Bringing the argument up to date, these authors speculate on Dewey’s reaction to contemporary versions of reduction – ‘for example, in neuroscience where the mind is reduced to the brain’ (1003).

Vygotsky first made his mark with a direct challenge to Pavlovian behaviourism (known at the time as ‘reflexology’) in a speech at the Second All-Russian Congress of Psychoneurologists in 1924. The clearest statement of his position appeared the following year in ‘Consciousness as a problem in the psychology of behavior’ (Vygotsky, 1925), an exemplary critique of the reductionism of this crude materialism which ignored consciousness and was over-reliant on the stimulus–response reflex. Within several pages, Vygotsky has challenged the ‘reflexologists’ for:

Failing to distinguish between human and animal behaviour;

Following a precise imminent critique, Vygotsky moves on to look at humanity from other perspectives. He points to three key advantages of human learning: ‘historical experience, social experience, and doubled experience’:

All our life, our labour and behaviour draw broadly on the experience of former generations, which is not transmitted at birth from father to son. We may provisionally designate this as a historical experience … If I know the Sahara and Mars although I have never travelled outside my country and have never looked into a telescope, obviously the origin of this experience is due to the experience of other people who have travelled to the Sahara and have looked into a telescope … Let us call this the social component of our behaviour. (Vygotsky, 1925)

As Blunden deduces, reductionist psychology not only results in very limited understanding; it debases our humanity by disempowering us:

The aim of controlling human behavior answers to the needs of capitalist, prison guard, interrogator, marketer, politician and bureaucrat, but an emancipatory psychology aims to free people from manipulation so that they can have voluntary control over their own behavior. (Blunden, 2012: 127)

Reductionist misunderstandings of poverty-related underachievement

There are many reductionist pitfalls in theorising the relationship between poverty and schooling. An adequate understanding requires bringing together different forms of knowledge (sociology and urban geography, curriculum and pedagogy, organisational theory, etc.) and drawing on appropriate disciplinary paradigms for examining a stratified, conflictual and open situation. For example, sociological analysis might range from statistical correlations to youth and classroom ethnography.

Recent years have seen some outstanding research on inequality (for example, Dorling, 2014; Pickett and Wilkinson, 2010), revealing acute polarisation of income and ownership. The insights are important, but this kind of research does not identify the underlying causes. For this, a systemic understanding of the dynamics of finance capital and neo-liberal globalisation is needed, including the mechanisms of exploitation and accumulation by dispossession (see Harvey, 2005, 2017).

The same problem arises with Standing’s (2011) widely accepted notion of ‘precariat’. Leaning towards the deficit ‘underclass’ theories (Murray, 1996) which underpin derogatory ‘poverty porn’ accounts of working-class life, this concept misses the point that wage workers (Marx’s original definition of proletariat) are inherently precarious, given that their position in the economic system is defined by not owning the means of production and therefore depending for work on the small minority who do. This intrinsic precarity is exaggerated by the economic powers of globalised capital, and the diminished power of workers in societies defined by neo-liberalism. Precarity is not simply the condition of a distinct social group of unskilled, unemployed or marginally employed, but increasingly affects large sections of the population, including many employees with good qualifications and in white-collar or professional work.

This understanding, however, requires a theoretical understanding of class which does not rely on surface appearances such as lifestyle (as with Savage, 2015). Even in Marx’s time, ‘the largest group of wage labourers was not the industrial working class but domestic servants, most of whom were female. The working class, then, is not always male, brawny and handy with a sledgehammer’ (Eagleton, 2011: 171). An understanding of the educational impact of poverty requires stratification. The concept of class is rooted in economics, and it is exploitation and dispossession rather than cultural reproduction that generates poverty – people are poor because of exploitative work conditions, not the bad habits of underclass parents. At the same time, we have to recognise the interactions between economics, culture and psychology, without allowing culture or psychology to float free (Wrigley, 2013b; Smyth and Wrigley, 2013). To recognise the major influence of parents’ qualifications on children’s attainment (Sammons et al., 2008) is not to deny economics, but to recognise the cultural means by which economic advantage transmits between generations (Bourdieu and Passeron, 1990).

Reductionist understandings of the link between social position and educational achievement take many forms, but crucially omit key layers and causal forces. For example, blaming teachers for low attainment fails to take seriously the economic, cultural and psychological pressures. Conversely, theories of genetically inherited intelligence represent a reduction from psychological to biological levels of explanation. The ‘lack of aspiration’ argument, and the demand for greater personal ‘resilience’, shifts the burden to individualistic psychological explanations, forgetting that it is difficult to sustain high aspirations without corresponding opportunities. The Conservative reframing of social justice as ‘social mobility’ within an assumed meritocracy (Littler, 2018) is a denial of structure and inflation of individual potentiality.

Much of the statistical data generated to track the link between poverty and education is short on theory and, whilst acknowledging that correlation is not causality, the texts are full of casual conclusions that ‘X causes Y’ (Gorard and See, 2013: 22). Further, important distinctions which would allow the complex realities to speak through the data are often missing, as with the overuse of free school meal (FSM) entitlement as a proxy indicator of poverty. The simple binary FSM/not FSM does not reflect the intense multiple deprivation experienced by young people in deindustrialised areas or run-down coastal towns. The absence of FSM entitlement covers everything from extreme affluence and privilege to being seriously poor but unable to meet the official FSM criteria. The FSM category does not reflect unhealthy housing conditions, low levels of parental qualifications or occupations, or the fatalism and disillusionment which result from growing up in a deindustrialised region. Even complex multiple regression analysis can only deal with averages and suggest tendencies, and does not cope well with tracing paths of influence. The mathematics of tendencies and effects, which is at basis additive/subtractive, cannot reflect the possibility that a combination of influences, including adverse childhood experiences, can create a tipping point.

A non-reductionist attempt to ‘close the gap’ must look beyond as well as within the school. This would surely involve preventing child poverty, free activity schemes in school holidays, looking at how teachers understand the lives of the communities they serve, and so on. To theorise this requires bringing together different subdisciplines of education and sociology at a macro, meso and micro level. It also requires considering such organisational/pedagogical factors as segregation by ‘ability’, the impact of high-stakes accountability on teacher recruitment, or the need to enrich young people’s experience.

Reductionist theories of school development

School evaluation and development are understood in many countries through paradigms of School Effectiveness and School Improvement. Recurrent keywords such as ‘accountability’, ‘effectiveness’ and ‘leadership’ work ideologically by emphasising technical rationalism, eclipsing questions of political or moral purpose.

School Effectiveness is a quantitative paradigm which calculates the relative impact of a school on student learning and then attempts to identify what has brought that about. Endemic to this is a methodological reductionism (for example, mechanistic causality, including the assumption of linear cause-and-effect relationships) and contextual reductionism (a failure to examine environmental influences when tracing causal relationships internal to a school).

Methodological reductionism is exemplified by assumptions of simple linear pathways of influence as inputs lead to higher test scores, whereas, in reality, key factors can operate recursively, events operate as catalysts, and forces contradict each other or fail to actualise due to adverse context – indeed, all the complex possibilities understood by Critical Realism (see earlier). Contextual reductionism includes the tendency of neo-liberal accountability systems to treat schools as isolated entities, disconnected from pupils’ lifeworlds beyond school, youth culture, commercial influences or indeed the effects of the national education system.

The harnessing of School Improvement theory to School Effectiveness reinforced teleological reductionism by privileging attainment data as the only important aim of education. Keywords such as ‘vision’, ‘mission’ and ‘values’ are not only hollowed-out terms; they are a kind of spiritual or vitalist add-on which disguises the values reductionism of the School Effectiveness paradigm. The previous emphasis on willing, thoughtful participation, whilst surviving as rhetoric, was replaced by top-down power relationships of a system defined by surveillance and control (Wrigley, 2004; Gunter, 2001; Hargreaves, 1994). There is no concession that some innovations are ill-conceived, that professionals have a right and duty to question them, or that some changes need to be resisted.

Causation is poorly conceptualised by School Effectiveness’s list of ‘key characteristics’, which reify potentially significant activities, concerns and interaction, and are in any case too vague and ambiguous to be helpful. What, for example, does ‘a focus on teaching and learning’ mean? It could be anything from transmission of Hirschian lists of facts (Hirsch, 1987) to place-based discovery methods. An emphasis on assessment might mean anything from authentic assessment or assessment for learning to high-stakes testing. To see all of this as beneficial forgets the collateral damage which high-stakes accountability entails.

Another major keyword is ‘culture’ – potentially a very powerful concept because it bridges between the material and the spiritual. Culture is matter which means something, as when an artist reshapes material to convey a meaning. This complex concept has been reduced from a powerful tool for examining patterns of life and their inherent value to a means of domestication and control. This is exemplified in Deal and Peterson’s (1999) argument that the head teacher’s role is ‘to manage culture’.

An alternative, non-reductive understanding of school culture would recognise the inequalities of power and the dynamic interrelation of school world and lifeworld, enabling a conversation about how some schools are able to moderate the impact of disadvantage. This might involve, for example:

examining the cultural messages of classrooms which are dominated by the teacher’s voice, closed questions and rituals of transmission of superior wisdom;

An empowerment culture which enables marginalised young people to develop as learners and members of society might involve, for example, teamwork in planning social-constructive and cooperative learning; a curriculum which connects with students’ lives while building towards powerful academic knowledge; a school ethos which enables learners to find a voice; and a counterculture to the despair of an impoverished or oppressed community.

Using culture in this sense would avoid reductionism. It does not neglect power, difference, privilege, barriers of communication, differences of perspective or the links between practices and (explicit or tacit) theories. It remains open to contradictions and tensions within the school, and between the school and the outside world (see Alvesson, 2002).

The naively positive developmental models of change used in mainstream School Improvement are reductionist simplifications. Schools require a better understanding of open systems, and could even learn from the science of living things, in order to reference openness, scale, stratification, complexity, systemic pressures and human purpose. One alternative would be to emulate (but not just copy) ecological models:

The evolution of living systems is self-organising and dynamic (i.e. not simply the result of causes, but of a long and accident-prone learning history on the system–environment border).

Increasingly, school development is seen in terms of ‘leadership', which, despite variants such as ‘distributed leadership’, invariably defaults to individualised top-down surveillance and control. A rare alternative is Wilkinson and Kemmis’s (2015) variation on actor-network theory (Latour, 2005), which recognises the agency of non-humans but without losing the sense of agency due to participants’ interpretations and intentions. Following Schatzki (2002), its focus is on how ‘practice architectures’ bring together ‘sayings, doings and relatings’ which are ‘not split off from one another’ but ‘hang together’ in the overall project of a practice (Kemmis et al., 2012: 35). These fit within wider arrangements – cultural-discursive, material-economic and social-political preconditions that make practices possible. Thus, for example, what is said within the school is shaped by the discourses of the wider society and its views of childhood and education. Such a model avoids the default individualism and apolitical neglect of structural power inequality, and the marginal recognition of material structural powers which characterise the majority of ‘leadership’ thinking.

The reductionist assumptions of ‘evidence-based teaching’

The methodology of ‘evidence-based teaching’ (Wrigley, 2018) – RCTs aggregated through meta-analyses, then further amalgamated and compared through what we might call ‘meta-meta-analyses’ or ‘mega-synthesis’ (Hattie, 2009 or the Education Endowment Foundation’s ‘Teaching and Learning Toolkit') – is exemplary of reductionism in its many forms. There have been powerful challenges to the statistical assumptions (Simpson, 2017, 2018; Lilley, 2018), but here I will briefly highlight a few arguments which connect with issues of reductionism (for more detailed explanation and illustration, see Wrigley, 2018).

Firstly, the call for education to imitate medicine fails to recognise the specifics of particular fields and sciences, and loses the complexity and specificity of situations. Its tacit – and mistaken – assumptions are that medical research consists entirely of drugs trials; that the clinical judgement of the doctor has become marginal; that there is an easy transfer from findings of drugs trials to individual diagnosis and treatment; that double-blind trials are possible in education; and that theory is unimportant. Fundamental assumptions of drugs trials are impossible in teaching, including double-blinding and ‘placebo’. What exactly should your ‘control group’ be doing – a deliberate avoidance of the behaviour under investigation or ‘business as usual’? Imagine the simple case of an RCT about the benefit of more open questions. Should the control group experience only closed questions or should the teacher simply be asked to do as they normally do and not think too much about the type of question? As Pawson (2006: 51) states: ‘This is not the world in repose. This is no vacuum … Control groups or control areas are in fact kept very busy’. With regard to theory, Pawson reminds us that:

Medical treatments … are the embodiment of years of theory-testing. They are already scientific inquiry incarnate before the first Phase III RCT is even designed. By this stage, medical science knows pretty well how a treatment works and it entrusts to the RCT a slightly different question about how well it works in a particular manifestation. (Pawson, 2006: 47)

This reflects the empiricist assumption that the closed situation of experiments sufficiently explains what happens in real-world open systems. As Critical Realists explain, experiments are not a reflection of reality, and the data does not simply speak for itself. That is only the start of the problem. I will highlight some further simplifications here:

Data analysis frequently relies on the reductionist assumption of fairly regular and reliable linear progress. This is no longer tenable in accountability data (see analysis from Education Datalab, 2015).

As the higher levels of meta-analysis and mega-synthesis are reached, complexity is progressively buried. Feinstein (1995: 76) argued that: ‘Important inconsistencies are ignored and buried in the statistical agglomeration’. Gene Glass, who originated the idea of ‘meta-analysis’, has subsequently issued this sharp warning about heterogeneity:

Our biggest challenge is to tame the wild variation in our findings not by decreeing this or that set of standard protocols but by describing and accounting for the variability in our findings. The result of a meta-analysis should never be an average; it should be a graph. (quoted in Robinson, 2004: 29; my emphasis)

Final remarks

This article has examined multiple forms of reductionism, including the deep shortcomings of an empiricist philosophy of natural and social science. This is not to undermine ontological realism or the need to empirically verify or exemplify ideas, but to insist on stronger foundations and methodologies. It has argued for the need to attend to openness, stratification and emergence.

The reductionist problems highlighted here have deep historical roots. Because modern science began with Galilean and Newtonian mechanics, its principles are often assumed to apply across the board, including life sciences and indeed social fields. This can be seen in Bennett’s (2013) populist diatribe against ‘non-scientific’ research in education, and the accusation that research which does not emulate laboratory experiments is quackery.

Rose (2005: 9) has referred to the ‘physics envy’ of many biologists. We are faced with something like ‘engineering envy’ among many educational statisticians and policymakers desperate for efficiency and certainty. This accompanies an increasing faith that neuroscience and behavioural genetics will open the window to understanding cognition.

Biesta (2010: 496) presents the case that, in education, we are dealing with open, recursive and semiotic systems – open because they interact with their environment and do not operate deterministically; recursive because they ‘feed back into themselves, so that the behaviour of the system is the result of a combination of external factors and internal dynamics’; and semiotic because they ‘do not operate through physical force but through the exchange of meaning’. Even this is too straightforward. In formal education, we are actually dealing with multiple nested and interlocking systems (the individual learner, groups, classrooms, the accountability machine) and various forms of interaction and interface (teacher–pupil, teachers–management, school–parents). Learners are pulled between school world and lifeworld, which is particularly challenging if there is a distance between school and community cultures. Even a solo act of learning involves interaction between subject and object, mediated by various instruments and cultural tools, in a dance involving a backwards-and-forwards motion between abstract concepts and sensory experience. And this is not even to begin considering social, economic and cultural power differences or the politics of the curriculum.

Education is fundamentally about emergence, which operates at the various interfaces mentioned above. The quest for certainty and efficiency, though probably inevitable for formal schooling (Biesta, 2015), is unattainable and undesirable if taken too far. Education (Bildung, pedagogy) turns into its opposite – a machine for the transmission of official dogma, received opinion and hegemonic beliefs.