Do schools kill creativity?
The case for knowledge before novelty and why constraint enables original thought
“Do schools kill creativity?” asked Ken Robinson in his wildly popular 2006 TED Talk. He tells the story of a little girl sitting at the back of the classroom drawing. Her teacher asks what she is doing. “I’m drawing a picture of God,” the girl replies. The teacher protests that nobody knows what God looks like. “They will in a minute,” the girl says. Da dum tish!
Children, Robinson argues, are naturally creative. They are willing to take risks and offer bold answers. Schooling, by contrast, teaches caution. We learn that our job is to remember the correct answer rather than invent something new. By the time we reach adulthood, he suggests, that fearless imaginative impulse has been steadily educated out of us. Schools are bleak, Victorian institutions run on a factory model which exist to mould chldren into standardised shapes and knock off all their interesting edges.
Robinson claimed education systems fail to help students discover their natural talents and argued that schools should place creativity on the same footing as literacy. His solutions were less maths and more dancing, greater encouragement of divergent thinking, and more personalised pathways that allow students to find the activities in which their abilities and passions meet.
It sounds appealing, doesn’t it? Tens of millions of viewers have been seduced by Robinson’s easy charm and the elegant simplicity of his diagnosis. But if the solution is so obvious, why don’t schools simply do what he suggests?
Can creativity really be taught? And if it can, is the process as straight forward as Sir Ken suggested? Before answering those questions, we need to untangle a few awkward problems. The first is definitional.
What exactly do we mean by creativity?
The American Psychological Association defines creativity as “the generation of ideas that are new and useful in a particular situation.”1 this echoes Robison’s own definition: “The process of having original ideas that have value.” Both these definitions hinge on the conccept of usefulness or value. Novelty or originality alone are not enough. An idea may be novel yet trivial, eccentric or simply wrong. Usefulness implies a standard. A mathematical conjecture must withstand proof; a scientific hypothesis must survive experiment; a poem must work within, or against, recognised forms and expectations. Utility is always judged from within a domain. Taking utility seriously, forces us to acknowledge that creativity cannot float free of area of knowledge it occupies.
Mihaly Csikszentmihalyi offers a useful corrective to individualistic accounts of creativity. In his systems model, creativity does not reside solely within a person. It emerges from the interaction between three elements: the individual, the domain and the field.2 The domain consists of the symbolic rules and procedures of a discipline. The field consists of those gatekeepers who judge whether a contribution counts. A work is only creative if the field recognises it as a valuable contribution to the domain. On this view, creativity is not merely psychological but social. It is not enough to generate novelty; novelty must be accepted within a tradition.
It may also help to distinguish between different levels of creativity. Researchers, James Kaufman and Ronald Beghetto propose a Four C model:
Mini-c — personal insight and interpretation.
Little-c — everyday problem solving and expression.
Pro-C — professional-level expertise.
Big-C — eminent, field-changing achievement.
Schools primarily operate at the level of mini-c and little-c. Confusion arises when classroom practice is judged by standards appropriate to Nobel laureates. Clarifying scale prevents inflated expectations and misplaced disappointment.
Can creativity be measured?
Whatever our definition of creativity, how can we measure it? If we can’t measure it, how would we know if we were getting better at teaching it? In the 1960s, J.P. Guilford came up with an interesting proxy for measuring creativity: ‘divergent thinking’, the idea that there are multiple solutions to any problem. He devised a number of tests to quantify divergent thinking including the 1967 Alternative Uses Test which asks participants to suggest as many alternative uses as they are able for everyday objects including toothpicks, bricks and, most famously, paper clips.3
In 1968, George Land conducted a research study to test the creativity of 1,600 children ranging in age from three to five years old. This was the same creativity test he devised for NASA to help select innovative engineers and scientists, sometimes referred to as ‘the paperclip test’ in which participants have to come up with as many different uses for a paper clip as they can think of. Most people are able to come up with 10 to 15 uses. People who are good at divergent thinking come up with around 200.4
Unfortunately — at least according to some interpretations — our capacity for divergent thinking deteriorates with age. Land’s longitudinal study found that 98 per cent of them were at genius level in divergent thinking at age five. Five years later, when they were aged 8 to 10 years, those at genius level had dropped to 30%. By the time they were 15, the average score was just 12%. Adults taking the same test tend to score around 2%. Creativity guru, Ken Robinson argued that the main intervention these children have had is a conveyor belt education that tells them, “There’s only one answer. It’s at the back. And don’t look. That’s called cheating.”5
Despite the fact that a great many of the most creative people who’ve ever lived – including Robinson – have been products of the school system, the real problem with Land’s analysis is that simply giving us percentages tells us nothing about the quality of the answers the children provided. It’s possible that adults think of fewer uses for a paperclip because they have the ability to filter out nonsense responses. Children naturally possess a relatively limited set of schema, and thus their ability to be creative (to combine schema from different domains to generate interesting variations) is also limited. What probably impresses us as adults is the uninhibited way children make productions based on their limited knowledge of the world. Princesses are painted, doctors and nurses are role-played, cardboard robots are constructed without embarrassment, and these unselected productions are usually (and quite naturally) rewarded by praise and attention.
However, as we get older we begin to identify more with our peers and, even if our parents still encourage us, our creativity starts to undergo more internal selection before production. Over time we come to understand that the first thoughts that pop into our heads are rarely true expressions of genius, and the nature of our creativity starts to expand and change. As we learn more the schema available to us expand, and we become more discerning about what represents quality.
This is precisely the effect of good schooling. Consider two classrooms. In one, students are told to “write a creative story” with minimal guidance. Many default to familiar tropes drawn from a narrow store of experience. Plots drift, characters are flat and stories buckle under the weight of pointless dialogue. In another, students have read widely within a genre. They understand narrative structure, archetype, pacing and point of view. When they write, they are not merely ‘“expressing themselves”; they’re manipulating conventions. The difference is not innate creativity, it’s knowledge.
Recent research increasingly supports the view that creativity depends on the structure of knowledge in long-term memory. Kenett et al’s semantic network analysis suggests that creative thinking often involves navigating complex networks of ideas and making connections between distant concepts. People who perform well on creative tasks tend to have richer and more flexible conceptual networks, allowing them to move more easily between related ideas. Without a sufficiently rich store of concepts, there is little available to recombine.
It is also worth distinguishing between fluency and quality. Divergent thinking tests typically reward the number of responses produced, sometimes alongside measures of flexibility or statistical originality. They do not, and arguably cannot, assess depth, coherence or value in any robust sense. Generating 200 uses for a paperclip may demonstrate fluency, but it tells us little about whether those uses are feasible, elegant or worth pursuing. The leap from quantity to creativity is far from secure.
John Baer argues that much of what is supposed about divergent thinking may be wrong. Divergent thinking, as with any other form of creativity, is not a stable trait that can be transferred between domains of knowledge. We are all capable of being creative in one area without being able to be creative in others, but the strategies a student might use to be creative on the sports field, or when writing an essay, will not necessarily make them creative in mathematics or science, and vice versa. Although it seems intuitive that linking unusual uses for a paperclip must be similar to thinking of unusual recipes for soufflé, simply lumping things together and giving them the same name doesn’t actually make them the same.6
This was demonstrated by Ellis Paul Torrance, the leading developer of creativity tests. He created two different versions of a test to measure figural (thinking creatively with pictures) and verbal (thinking creatively with words) divergent thinking, and gave the same tests to different groups of people.7 There was almost no correlation between scores on these two tests of divergent thinking indicating that the two sets of divergent thinking skills were completely unrelated.
If figural and verbal divergent thinking barely correlate, then the idea of a single, transferable creative capacity begins to look doubtful. A programme that claims to teach “creative thinking skills” in the abstract rests on a category error. What transfers is not a generic spark but structured knowledge and strategies embedded within particular forms of thinking.
It seems that divergent thinking is many completely different sets of skills. Unfortunately, many school districts, and even some researchers who use Torrance’s tests, frequently forget this and act as if these tests were measuring some readily transferable, domain-general skill. The result has been confusion about what creativity is and the likelihood that many research findings that should have been rejected have been accepted uncritically.8
Although divergent thinking tests produce easily quantifiable data, what does that data actually tell us? How could teachers really know if students had become more creative as a result of their efforts? Attempts to measure creativity suffer from a lack of validity and reliability. Creativity covers many different forms and fields. A test which assesses a proxy like counting uses of a paper clip suffers from a lack of validity. How, say, might Mozart or Van Gogh have performed on this test? If they were unable to think of many uses would that make their music or art less creative? If creativity is not, as seems likely, a global trait, the tests will also suffer from poor reliability as it would be unlikely that people would get the same score in different situations.9
The most ambitious recent attempt to measure creativity comes from OECD’s PISA 2022 Results: Creative Minds, Creative Schools, which introduced an international assessment of ‘creative thinking.’ Students were asked to generate, evaluate and improve ideas across tasks involving writing, visual expression, scientific reasoning and social problem solving. For example, one task asked students to generate ideas for improving an everyday object. Students were shown a familiar item such as a school bag and asked to suggest ways it could be redesigned to solve particular problems. They were then prompted to explain which of their ideas would work best and why. Other tasks asked students to write a short story based on an unusual prompt, design a visual poster communicating a message, or suggest solutions to a social problem. Responses were scored according to criteria such as originality, diversity of ideas and the extent to which students evaluate and refine their suggestions.
But the very act of measuring creativity reveals the difficulty. Large-scale assessments require clear definitions, tightly specified tasks and standardised scoring criteria. Creativity, however, prevails in the opposite conditions. The kinds of originality that matter in science, literature or engineering usually emerge slowly through sustained work within a discipline, not through short bursts of idea generation under exam conditions. Compressing creativity into timed tasks inevitably turns it into a proxy for something else: fluency, quick thinking or the ability to produce plausible answers on demand.
In a recent study Mark Connolly and colleagues have pointed out the philosophical and methodological problems this creates. Attempts to quantify creativity often assume that it is a stable psychological trait that can be captured through standardised tasks and compared across individuals and contexts. Yet creativity probably doesn’t behave at all like this, depending as it does on domain knowledge, disciplinary conventions and social recognition. What counts as a ‘creative insight’ in mathematics is unlikely to resemble creativity in dance or politics. Once these contextual differences are stripped away, what remains may be measurable, but it is not at all obvious that what’s being measured is not what most people would understand by ‘creativity.’ The more closely we look, the more elusive the concept becomes. The very tools designed to capture creativity seem to obscure the phenomenon they seek to illuminate.
On top of this, research on transfer in cognitive psychology adds further weight to the domain-specific view. Skills learned in one context rarely transfer wholesale to another without shared structure and extensive practice. The ability to solve problems in chess does not automatically generalise to mathematics; musical training does not guarantee improved logical reasoning.10 If even well-defined cognitive skills struggle to transfer, it seems optimistic to suppose that a generic “creative thinking” programme will do so. What transfers is not a free-floating capacity but knowledge organised around deep structure.
Intelligence vs creativity
One way around the problem would be to measure a more global trait, which is highly correlated with creativity. There’s been much debate in the psychological literature about whether intelligence and creativity are part of the same process (the conjoint hypothesis) or represent distinct mental processes (the disjoint hypothesis). Some researchers believe that creativity is the outcome of the same cognitive processes as intelligence, and that it is only labelled when the outcome of a cognitive process happens to produce something novel.11 The most plausible hypothesis is that below a certain level of cognitive ability, creative performance is constrained. Once that level is reached, further increases in IQ explain progressively less variance. At that point, domain knowledge, persistence, openness to experience, risk tolerance and motivation become more predictive of creative achievement. In other words, intelligence appears necessary but not sufficient. It sets the stage; what happens on that stage depends on knowledge, personality and context.
Empirical work lends support to the threshold hypothesis. Studies using breakpoint analysis suggest that intelligence predicts creative potential up to a certain level, after which the relationship weakens. Beyond that point, factors such as openness to experience, persistence and domain engagement become more decisive. This does not collapse creativity into intelligence, but neither does it detach the two entirely. Cognitive ability sets limits; within those limits, personality, motivation and knowledge shape what is achieved.12
