In 1968, Robert Rosenthal and Lenore Jacobson published Pygmalion in the Classroom, a study that told teachers certain randomly selected students were about to show unusual intellectual growth. The headline finding was that these students subsequently gained more IQ points than controls. The interpretation was that teacher expectations causally produce student outcomes: that what a teacher believes about a child shapes what that child becomes. The study became one of the most cited in all of educational psychology and has been the foundation of half a century of teacher training, management consultancy, and popular accounts of the power of positive expectations. From the business world to parenting advice to school improvement frameworks, the Pygmalion effect has functioned as received wisdom: believe in people, and they will rise to your belief.
This article argues that the Pygmalion effect, in the form in which it has achieved popular cultural and policy authority, does not exist. What the original study actually demonstrated was methodologically compromised, empirically fragile, and ontologically misconceived. The subsequent literature has consistently shown that teacher expectations predict student outcomes primarily because they are accurate perceptions of a distributed mediational reality, not because they are autonomous causal forces that manufacture the outcomes they anticipate. The popular version of the Pygmalion effect inverts this relationship: it treats accurate social perception as irrational bias and proposes that correcting that bias, by raising expectations, will change outcomes. This proposal is wrong about the causal structure, wrong about the nature of expectations, and wrong about where the determinants of educational achievement actually lie.
The argument draws on Living Value Theory and on the analytical framework developed in the companion article Unearned Confidence: Why Psychology Mass-produces Compelling But Misleading Claims. That article identified three systematic ontological errors in the psychological paradigms it examined: the container ontology that locates causal variables inside individuals rather than in distributed mediational configurations; the failure to distinguish between recursivity levels, particularly the confusion of L1 seamless coordination with L3 articulable cognitive states; and the failure to distinguish between nonrecursive, selfrecursive, and interrecursive processes, resulting in the consistent misapplication of methods designed for the first to phenomena that belong to the third. The Pygmalion study commits all three errors with particular clarity, and its cultural authority rests on the same ideological convenience that has sustained the marshmallow test, self-efficacy theory, and growth mindset: it locates the source of unequal educational outcomes inside individual minds, where they can be addressed by training and attitude adjustment, rather than in the structural and material conditions that actually produce those inequalities.
I. What Rosenthal and Jacobson Actually Did
The original study selected students at a San Francisco elementary school using a standard IQ test, which Rosenthal and Jacobson relabelled the Harvard Test of Inflected Acquisition, a fictional instrument designed to sound authoritative to teachers. They then randomly selected approximately twenty percent of students in each classroom and told teachers that these students had been identified by the test as likely to show unusual intellectual growth in the coming year. The selection was random. There was nothing actually distinctive about these children. The experiment then measured IQ gains across the academic year to see whether the falsely identified bloomers outperformed their peers.
The reported findings were striking. In grades one and two, the randomly selected bloomers showed substantially greater IQ gains than controls. The interpretation was that teacher expectations had caused these gains: that believing certain children were about to flourish, teachers had treated them differently in ways that produced genuine intellectual development. Rosenthal and Jacobson's account of the mechanism was vague, involving differential attention, warmer interactions, more challenging material, and more encouraging feedback, but the causal claim was stated with confidence. Teacher expectations are self-fulfilling prophecies. What teachers believe about children shapes what children become.
The methodological problems were identified almost immediately and with unusual severity. Robert Thorndike's 1968 review in the American Educational Research Journal documented that the IQ test used, the TOGA, was psychometrically inappropriate for young children, and that the average pretest score for one first-grade class fell in a range typically classified as intellectual disability, an impossible result for a normal classroom functioning at a normal level. This meant the apparent IQ gains in grades one and two, where all the effects were concentrated, almost certainly reflected regression to the mean: scores that had been aberrantly low at baseline returned to more accurate levels on retest, and this statistical artefact appeared as intellectual growth attributable to teacher expectations. Thorndike's verdict was unsparing: when the clock strikes thirteen, doubt is cast not only on the last stroke but on all that have gone before. Richard Snow's 1969 review made converging points about the statistical handling of floor and ceiling scores. The effect sizes reported also varied wildly across publications, with some reporting a fifteen-point IQ gain and others a three-and-a-half-point gain for the same data, a discrepancy that no legitimate measurement should produce.
The most rigorous meta-analysis of the subsequent literature, conducted by Stephen Raudenbush in 1984 and covering eighteen experimental studies that followed the Rosenthal and Jacobson design, found an average effect size of approximately d equals 0.11, which is small. But the crucial finding within the meta-analysis was not the average effect size. It was the relationship between effect size and the teacher's prior knowledge of the students. When teachers had known their students for more than two weeks before receiving the manipulated expectation information, the effect approached zero. When teachers received the manipulation at the start of the school year, before they had developed their own observations of their students, the effect was larger. The implication is precise and damaging to the Pygmalion narrative: artificially induced expectations operate only in the absence of real knowledge. Once teachers have genuine observational data about a student, accurate perception overwrites the induced expectation. The self-fulfilling prophecy is not a robust feature of teaching. It is a laboratory artefact produced by informational deprivation.
II. Expectations as Accurate Perception
The most important body of subsequent research on teacher expectations has been conducted by Lee Jussim, whose systematic examination of the question spans from 1989 to the present. Jussim's central finding, developed across multiple studies and synthesised in a 2005 review with Kent Harber covering thirty-five years of empirical research, is that the primary reason teacher expectations predict student outcomes is accuracy, not self-fulfilling prophecy. Teachers form expectations that are largely correct predictions, grounded in everything they observe: prior achievement, classroom engagement, the quality of preparation students bring to class, signals about home support and parental involvement, material readiness, and the full range of factors that actually predict academic outcomes.
This finding is deeply inconvenient for the Pygmalion narrative, and it has been correspondingly underemphasised in the popular and policy reception of the expectation literature. If teacher expectations are largely accurate perceptions of a distributed causal reality, then the finding that high expectations correlate with better outcomes does not mean that high expectations cause better outcomes. It means that the conditions that produce better outcomes also tend to produce higher expectations in teachers who are observing those conditions accurately. The causal arrow runs not from expectation to outcome but from the distributed mediational configuration of the student's life, their material resources, their relational support, their embodied history of educational engagement, to both the expectation and the outcome simultaneously. Teacher expectation is a symptom of the causal reality, not its driver.
The policy implication of this reframing is radical. The Pygmalion narrative says: raise teacher expectations, and outcomes will improve. The accuracy account says: teacher expectations are already largely calibrated to reality, and outcomes will improve when the reality changes, not when the expectation is adjusted. A teacher who expects less from a student whose home environment is chaotic, whose nutrition is unreliable, whose school has inadequate resources, and whose peer culture does not reward academic engagement, is not exhibiting a cognitive bias that training can correct. They are making an accurate assessment of the mediational configuration that determines what this student is currently capable of achieving in this institutional context. To call this expectation a self-fulfilling prophecy is to pathologise accurate pattern recognition and to propose, as a remedy, that teachers adopt beliefs about their students that are less well calibrated to the actual conditions those students face.
Wang, Rubie-Davies, and Meissel's 2018 review found that forty percent of studies on teacher expectations did not control for baseline achievement, meaning that apparent expectancy effects in those studies may simply reflect the correlation between prior ability and subsequent performance. Jussim and Harber's review found that self-fulfilling prophecy effects in the classroom are typically small and do not accumulate greatly across teachers or over time. The popular version of the Pygmalion effect, in which a teacher's belief about a child functions as a powerful, durable, and general causal force on that child's intellectual development, is not supported by the empirical record of the subsequent literature. What is supported is a small, conditional effect that operates primarily when teachers lack real knowledge of their students, and that is dwarfed in magnitude by the accuracy of teachers' perceptions of the actual conditions that determine educational outcomes.
III. The Container Ontology and the Misidentification of Cause
The Pygmalion study's foundational error, read through the LVT framework, is its application of the container ontology to a constitutively interrecursive phenomenon. The classroom is not a collection of individual containers, each with internal properties that independently produce behaviour. It is a coordinative field in which what the teacher does depends on what the students do, which depends on what the teacher does, all of it embedded in a wider mediational configuration involving the students' home environments, material resources, peer cultures, family histories, institutional legacies, and the full range of embodied, relational, spatial, material, and symbolic conditions within which learning either occurs or fails to occur.
Rosenthal and Jacobson extracted one element from this coordinative field, teacher expectation, and treated it as the causal driver of the whole system. This is exactly the container ontology that the Unearned Confidence analysis identified as the foundational error of the psychological paradigms it examined: locating the cause of distributed outcomes inside an individual cognitive state, then designing an intervention targeted at that state, then interpreting the finding that the state correlates with outcomes as evidence that the state causes the outcomes. In the Pygmalion case, the container is the teacher's mind. The cognitive state is the expectation. The causal claim is that manipulating the expectation will alter the outcome. And the fundamental error is the same: the outcome is constituted by a mediational configuration that the expectation reflects but does not produce.
A full multimediational analysis of the classroom reveals just how much the container ontology suppresses. In the embodiment domain, the student's physical state, their nutrition, sleep, health, sensory development, and the accumulated bodily history of their educational engagement, all shape what learning is possible in any given classroom session regardless of what the teacher believes about them. A child who arrives at school hungry, sleep-deprived, or managing chronic health conditions is not going to achieve better outcomes because their teacher has been told to expect intellectual growth. Their embodiment stakes are unmet, and no expectation can substitute for the material and relational conditions that would meet them. In the being-with domain, the relational texture of the classroom, the quality of the teacher-student relationship built over time, the peer dynamics that either support or undermine academic engagement, the family relationships that carry or fail to carry the child's educational development into the home, all determine the interrecursive ecology within which expectation operates. The teacher's expectation is one element of this ecology, and not the most powerful one. In the dwelling domain, the stability of the child's home environment, the availability of quiet space for homework, the spatial and temporal organisation of a household that either supports or disrupts educational routine, all shape outcomes through channels that a teacher's belief cannot reach. In the multimateriality domain, the resources of the school, the quality of its equipment and infrastructure, the class sizes that determine how much individual attention any student can receive, the material conditions of the student's home, all constitute the material infrastructure of educational achievement in ways that are entirely independent of any teacher's cognitive states. In the multisymbolisation domain, the cultural frameworks through which academic achievement is understood and valued in the student's community, the symbolic stakes attached to educational performance in their social world, the institutional categories through which their abilities have been named and assessed in their prior educational history, all shape what learning means and what it can achieve in ways that a single teacher's expectation, delivered to the teacher by a researcher in September, cannot override.
What the Pygmalion experiment actually did was manipulate one symbolic input, a declaration about a child's potential, into a specific being-with configuration, the teacher-student relationship, while holding all other mediational conditions constant. It then measured one output, IQ test performance, and attributed the variance in that output to the symbolic input. The measurement was contaminated by the psychometric problems Thorndike identified. The effect disappeared when teachers had real observational data. The mechanism was never specified. And the causal inference from manipulated expectation to outcome variance ignored the entire mediational configuration within which both expectation and outcome were embedded. This is not a study that established a general causal principle about the power of belief. It is a study that produced a methodologically compromised finding in a specific population under specific informational conditions, and that finding was then elevated into a cultural myth about the transformative power of positive thinking.
IV. Accurate Perception Pathologised as Bias
The Pygmalion narrative commits a second ontological error that is specific to the expectation literature and does not appear in quite the same form in the other cases examined in the Unearned Confidence analysis. It pathologises accurate perception. A teacher who expects less from a student from a deprived socioeconomic background is, according to the Pygmalion framework, exhibiting a bias: a cognitive distortion that produces a self-fulfilling prophecy of underachievement. The remedy is to raise the expectation, to believe more in the student's potential, to override the perception with an aspiration.
This framing is not merely empirically wrong. It is a category error about what teacher expectations are and how they are formed. Teacher expectations are not free-floating cognitive attitudes that generate their objects independently of the world. They are perceptions formed through the full interrecursive ecology of the classroom, through accumulated observation of how this student performs on this kind of task, engages with this kind of instruction, responds to this kind of challenge, arrives at school in this kind of state, and is supported by this kind of home environment and peer culture. They are, as Jussim and Harber's review established, largely accurate predictions: they track the distributed mediational reality that actually determines student outcomes, and they do so because teachers are embedded in that reality as participants and observers, not because they are biased interpreters of an independent causal force.
To call this accurate perception a bias and propose that correcting it through training will improve outcomes is to make exactly the error that the LVT framework identifies as the redistribution of recursivity: the transfer of the authority to define what experience counts as, away from the person embedded in the coordinative field and toward an institutional actor whose authority derives from their symbolic position. The teacher, embedded in the actual interrecursive ecology of the classroom and observing the actual mediational conditions that determine what this student can achieve in this context, is told that their perception is a bias because it deviates from the normative expectation of expectation-blindness. The institutional authority of the psychological framework then overrides the teacher's situated knowledge with an ideological prescription: believe more, and the outcomes will change.
The practical consequences of this error are significant. Teacher training programmes built on the Pygmalion narrative direct resources toward raising expectations rather than toward changing the conditions that expectations accurately reflect. School improvement frameworks built on the expectation literature target teachers' cognitive states rather than the material, relational, and institutional conditions that the teachers are correctly perceiving as obstacles to their students' achievement. And students from deprived backgrounds continue to underachieve not because their teachers believe in them insufficiently but because the mediational configurations that actually produce educational achievement, the stable homes, the adequate nutrition, the well-resourced schools, the peer cultures that reward academic engagement, the family relationships that carry learning into the home, are not being addressed by any intervention targeted at the contents of a teacher's mind.
V. The Measurement Constitutes the Phenomenon
The third structural error in the Pygmalion study is its treatment of the experimental manipulation as a neutral measurement of an existing state rather than as a constitutive social act that reorganises the interrecursive ecology it claims to be measuring. Telling a teacher that a specific child has been identified by a scientific instrument as a late bloomer about to show unusual intellectual growth is not a neutral experimental procedure that activates a pre-existing causal mechanism called teacher expectation. It is a social act with specific interrecursive consequences: it creates a new social fact in the relationship between this teacher and this child, one that carries the authority of scientific credentialing and that reorganises what both parties expect from their interactions.
The teacher who has been told a child is a late bloomer now inhabits a different interrecursive ecology with that child. The declaration does not simply raise a number on the teacher's internal expectation scale. It creates a new relational configuration in which the teacher is watching for signs of the growth they have been told to expect, interpreting ambiguous performance signals through the frame of anticipated bloom, and perhaps providing marginally different attention, encouragement, or challenge in ways that accumulate over the academic year. The child, even without knowing what has been told to the teacher, inhabits a classroom environment that has been subtly altered by the declaration, because the teacher's altered behaviour, however small, is part of the interrecursive ecology within which the child is developing.
This is the sankalpa mechanism identified in the self-efficacy analysis, operating at the level of a third-party declaration rather than a self-declaration. The researcher's statement to the teacher constitutes a new social fact that reorganises the interrecursive field. What is then measured as an expectancy effect is partly the consequence of this constitutive intervention, not a readout of a stable pre-existing mechanism called teacher expectation. The experiment cannot separate the constitutive effect of the declaration from the causal effect of the expectation it purported to manipulate, because the manipulation and the declaration were the same act. There was no pre-existing expectation that was being raised or lowered. There was a social event that created a new configuration, and the study measured some consequences of that configuration while calling them evidence of a general causal mechanism.
Raudenbush's finding that the effect disappears when teachers have prior knowledge of their students is, on this account, precisely what should be expected. When the teacher has real observational data about the child, the constitutive force of the researcher's declaration is overridden by the teacher's accumulated interrecursive history with that child. The new social fact created by the declaration cannot successfully reorganise an interrecursive ecology that is already densely constituted by direct observation and relational history. The effect is not, as the Pygmalion narrative implies, that accurate knowledge of students prevents teachers from holding high expectations. The effect is that real relational knowledge overwrites an artificially constituted social fact, because interrecursive relationships constituted through sustained direct engagement are more durable than those constituted through a single declaration by a researcher with no ongoing relationship to either party.
VI. The Political Economy of the Pygmalion Myth
The Pygmalion effect achieved cultural authority for the same reasons that the marshmallow test, self-efficacy, and growth mindset achieved cultural authority: it produced a finding that is ideologically convenient for governance frameworks that prefer individual-level explanations of structural inequalities. The claim that teacher expectations are powerful causal forces on student outcomes locates the source of educational inequality in the cognitive states of teachers, where it can be addressed by training, attitude adjustment, and professional development programmes. This is vastly cheaper and more politically acceptable than addressing the structural conditions that Jussim's accuracy account reveals as the actual determinants: poverty, housing instability, inadequate school funding, under-resourced communities, and the full range of material and relational conditions that produce the mediational configurations within which some children thrive and others do not.
The Pygmalion narrative also has a specific moral appeal that compounds its ideological utility. It presents raising expectations as an act of respect and justice toward marginalised students. The teacher who holds high expectations for all students regardless of background is, in this framing, affirming their students' potential against a society that underestimates them. The teacher who holds lower expectations for students from deprived backgrounds is, in this framing, perpetuating inequality through a self-fulfilling prophecy of failure. This moral framing is powerful and emotionally resonant. It is also, on the accuracy account, largely misdirected. The teacher who accurately perceives that a student from a chaotic home environment, attending an under-resourced school, in a peer culture that penalises academic effort, is unlikely to achieve the same outcomes as a student from a stable, well-resourced environment with strong institutional and family support, is not perpetuating inequality. They are recognising it. The inequality is produced by the mediational conditions, not by the perception of those conditions.
The moral and political energy generated by the Pygmalion narrative has been consistently directed toward changing teachers' minds rather than changing the material conditions that teachers' minds are accurately reflecting. School improvement programmes have spent decades training teachers to hold high expectations for all students, to avoid deficit thinking, to see potential rather than limitation, to refuse the self-fulfilling prophecy of low expectation. These programmes are not without value: there are genuine cases of irrational bias, particularly racial and ethnic bias, in which teacher expectations deviate from accurate assessment of actual performance and potential, and addressing these deviations matters. But the overall orientation of expectation-based educational reform has been to target individual cognitive states while the structural conditions that those states accurately reflect remain unaddressed. The result, predictably, is that educational inequality persists, and each generation of expectation-based intervention is succeeded by the next, because the causal machinery has never been engaged.
The parallel with the marshmallow test is exact. Just as that study directed policy attention toward training children's self-control rather than addressing the material instability and institutional unreliability that the test was actually measuring, the Pygmalion study has directed policy attention toward raising teachers' expectations rather than addressing the poverty, under-resourcing, and structural disadvantage that those expectations are accurately perceiving. In both cases, the intervention is targeted at an individual cognitive state that is downstream of a distributed mediational configuration, and in both cases the failure of the intervention to produce the promised outcomes is attributed to insufficient implementation, insufficient commitment, or some other individual-level failure, rather than to the fundamental misdirection of the causal theory.
VII. What the Pygmalion Literature Actually Shows
The preceding critique should not be taken as a denial that expectation effects exist under any conditions. The Jussim and Harber review does not make this claim, and this article does not either. There is a genuine and non-trivial phenomenon somewhere in the expectation literature, and honesty requires acknowledging it precisely rather than dismissing it entirely.
Expectancy effects do occur under specific conditions: when teachers have minimal prior knowledge of a student, when differential treatment is pronounced and sustained, and in very young children whose self-concepts and performance patterns are still forming. In these conditions, a teacher's belief about a student can become part of the interrecursive ecology of their development in ways that have small but real consequences for outcomes. The mechanism is interrecursive, not cognitive: the teacher's altered behaviour, however modest, changes the relational and symbolic environment within which the child is developing, and this changed environment produces marginally different outcomes over time. This is a genuine interrecursive effect, and it is interesting precisely because it illustrates how the being-with mediation can carry consequences into the embodiment and multisymbolisation domains through sustained relational exposure.
But the magnitude of this effect, d equals 0.11 in the most rigorous meta-analysis, under the most favourable conditions of informational deprivation, is far smaller than the popular version of the Pygmalion effect implies. And the effect is dwarfed by the accuracy component: teacher expectations predict student outcomes primarily because they are accurate perceptions of the distributed mediational reality that actually produces those outcomes. The causal contribution of the expectation itself, over and above its accuracy, is modest, conditional, and concentrated in circumstances that are quite specific and that become less relevant as teachers develop real knowledge of their students.
The more important question that the expectation literature points toward is not how to raise expectations but what the actual mediational conditions are that produce the outcomes teachers are accurately perceiving, and how those conditions can be changed. This is the question that the Jussim accuracy account makes unavoidable: if expectations are correct predictions of a causal reality, then changing the predictions without changing the reality will not change the outcomes. The reality in question is the full multimediational configuration of a child's educational life: their embodied health and development, the relational support they receive at home and at school, the spatial and material conditions within which learning occurs, the symbolic frameworks through which academic achievement is understood and valued in their community. Changing this reality requires not training teachers to believe differently but investing in the material, relational, and institutional conditions that make educational achievement genuinely possible for children whose current mediational configurations do not support it.
VIII. The Wider Pattern
The Pygmalion effect fits the pattern identified in the Unearned Confidence analysis with sufficient precision to warrant its inclusion as a fifth paradigm case alongside the marshmallow test, self-efficacy, the illusion of explanatory depth, and growth mindset. Like all of them, it begins from the container ontology, locating the cause of distributed outcomes inside an individual cognitive state. Like all of them, it fails to distinguish between the recursivity levels at which the phenomenon operates: accurate L1 and L2 coordination between teacher and student, built through sustained interrecursive engagement, is collapsed with the L3 and L4 cognitive state called expectation, and the two are treated as equivalent when they are structurally distinct. Like all of them, it applies an experimental methodology designed for nonrecursive natural science to a constitutively interrecursive phenomenon, using a controlled manipulation to produce a specific interrecursive configuration and then misidentifying that configuration's properties as evidence of a general cognitive mechanism. And like all of them, it achieves cultural authority because it is ideologically convenient: it locates the source of structural inequality in individual cognitive states, where it can be addressed by cheap, scalable interventions that do not require engaging with the structural conditions that the cognitive states are accurately reflecting.
There is one dimension in which the Pygmalion case adds something to the pattern that the other cases do not make as visible: the pathologisation of accuracy. The marshmallow test pathologises eating the marshmallow, treating an accurate response to institutional unreliability as evidence of individual cognitive deficit. The Pygmalion study pathologises accurate perception, treating a teacher's correct assessment of a distributed mediational reality as a bias that produces the very outcomes it predicts. In both cases, what the individual is doing is epistemically reasonable: they are responding accurately to the actual conditions of the situation they are in. And in both cases, the psychological framework reinterprets this accurate response as a cognitive failure and proposes an intervention that targets the response rather than the conditions that make the response reasonable.
This is the deepest form of conceptual harm that the LVT framework identifies: not simply measuring the wrong thing, but pathologising the correct response to a structural reality and thereby redirecting attention and resources away from that reality and toward the individual who is accurately perceiving it. The child who eats the marshmallow because adult promises cannot be trusted is not failing at self-control. The teacher who holds calibrated expectations based on accurate observation of mediational conditions is not perpetuating inequality through bias. Both are doing something epistemically correct in response to conditions that are structurally produced. The psychological frameworks that reinterpret their accurate responses as individual failures are not merely wrong about the causal structure. They are performing a specific ideological function: making structural conditions invisible by rendering individual responses to those conditions as the problem to be solved.
Psychology will continue to produce findings of this kind for as long as it begins from an ontology that cannot see the irreducible multimediality of human coordination, cannot distinguish between the levels at which coordination operates, and cannot differentiate between the recursive domains within which different phenomena belong. The Pygmalion study is not an outlier in this regard. It is a paradigm case of what the discipline produces when these ontological blind spots are combined with the ideological pressures of a research economy that rewards individually targeted, scalable, and structurally convenient findings. Acknowledging this clearly is the first step toward a different relationship with psychological claims: one that asks, for any finding, not just whether it replicates, but whether the thing it claims to measure exists in the form it describes, whether the causal mechanism it proposes is coherent with the ontological character of the phenomenon, and whose interests are served by the particular form the finding takes.