Schemas vs Mental Models: Does the difference make a difference?

Why the distinction between “schemas” and “mental models” tell us more more about our language than our minds.

“For a difference to be a difference, it must make a difference.”

Gregory Bateson

Gregory Bateson’s aphorism is both elegant and inconvenient. Elegant, because it captures in a single phrase what it means for an idea to have explanatory power; inconvenient, because it forces us to ask whether the distinctions we love to make actually change anything. Bateson, a British anthropologist and systems theorist, coined the phrase in Steps to an Ecology of Mind (1972) while defining information as “a difference that makes a difference.” In other words, a variation only counts as meaningful if it produces a change in the system that perceives it. A distinction that alters nothing is not knowledge but noise.

Education is especially prone to mistaking verbal distinctions for meaningful ones. We coin new terms, slice familiar concepts into finer categories, and congratulate ourselves on theoretical precision. But unless those distinctions alter what teachers do or how students learn, they amount to little more than semantics.

The current enthusiasm for separating schemas from mental models is a case in point. On paper, the difference looks tidy: schemas are automatic, implicit frameworks for organising experience; mental models are explicit, deliberate representations of how things work. But does this conceptual distinction make a practical difference in the classroom? Does it meaningfully change how we plan, teach, or assess? Or is it, in Bateson’s sense, a difference that makes no difference at all?

This tendency to multiply distinctions that make no real difference is closely related to what psychologists call the jingle–jangle fallacies. The jingle side occurs when different constructs are treated as identical simply because they share a name; the jangle side happens when identical or near-identical constructs are treated as different because they’ve been given new labels. Education suffers more from the latter. Each time a familiar idea is renamed or slightly reframed, it acquires the sheen of novelty and the promise of innovation.

What begins as an attempt at conceptual clarity soon becomes an act of rebranding. “Growth mindset” sounds more progressive than “self-efficacy”; “adaptive teaching” feels fresher than “differentiation.” “Knowledge organisers,” “learning maps,” and “concept maps” are all close cousins, yet each arrives trumpeted as a new discovery. The jangle fallacy flourishes in such conditions because the field prizes novelty and suffers from short institutional memory. But without changes in practice or outcomes, new terminology merely repackages old ideas, giving us the illusion of movement while we stand still.

The argument for a difference

In a recent post, Paul Kirschner draws a clear line between schemas and mental models. Schemas, he says, are automatic and implicit — the mind’s way of organising experience into tidy, retrievable frameworks. Mental models, by contrast, are explicit and deliberate. They are the representations we consciously manipulate when reasoning, planning, or predicting.

Kirschner’s distinction has intuitive appeal. Schemas allow us to recognise patterns without thinking; mental models allow us to think about patterns deliberately. When a skilled driver approaches a roundabout, the schema guides the automatic acts of signalling, clutch, and mirror-checking, while the mental model governs the reasoning: if that car accelerates, I’ll need to yield.

From an instructional design perspective, the distinction seems powerful. For novices, we must first build schemas, automatic recognition and fluent recall of patterns. Only once those are solid should we move to mental models, enabling reasoning about systems. In that sense, schema-building and model-building represent successive stages in learning: first fluency, then understanding.

The trouble is, we have no direct access to long-term memory. No one can inspect the contents of LTM or observe its architecture. All we can do is construct explanatory models that allow us to speculate about how our minds work and why we behave as we do. Terms like schema and mental model are not discoveries of mental furniture but metaphors for unseen processes, useful fictions that help us organise our thinking about cognition. Their value lies not in what they reveal about the brain but in how they guide the design of teaching.

What the evidence says

Unfortunately, empirical support for this neat division is thin. There is strong evidence for schema theory itself. Decades of research confirm that knowledge is stored and retrieved through interconnected frameworks. Schema formation reduces cognitive load by allowing information to be chunked, meaning we can process complex material without overloading working memory.¹

Likewise, there is robust evidence for mental model theory in reasoning and decision-making research. Studies in psychology, human–computer interaction, and spatial cognition show that people construct internal models of systems and use them to predict outcomes.²

What we lack, however, is evidence that schema-use and model-building are truly distinct processes, or that instruction targeting one yields measurably different results from instruction targeting the other. Cognitive scientists such as Busselle (2017) note that the terms are often used interchangeably and that mental models may simply be complex schemas viewed at a higher level of abstraction. Experimental studies that deliberately manipulate the two, “schema-based” versus “model-based” instruction, are virtually non-existent in educational settings.

Essentially, both constructs are well supported, but the line dividing them is more theoretical convenience than empirical boundary. There is a huge conceptual overlap between schema theory and mental model theory andhe fact that the two are often used interchangeably maybe suggests that mental models can best be understood as complex or elaborated schemas.

Why the distinction might still matter

Despite this, Kirschner’s distinction is not without practical value. It can act as a heuristic for thinking about what kind of knowledge we are trying to build. In teaching writing, for example, we might begin by developing schemas of sentence patterns: repeated, guided practice that builds automatic recognition of clause boundaries, conjunctions, and syntactic balance. Only once these patterns are fluent do we shift toward mental models of how sentences create tone, rhythm, or emphasis.

Distinguishing between schemas and mental models may helps by making explicit the different cognitive demands placed on students at different stages of learning. If everything is treated as ‘understanding,’ we risk asking novices to reason before they have anything stable to reason with. By separating schema-building from model-building, teachers can see that fluency and understanding are not simultaneous but sequential. Schema-building tasks aim to automate recognition and recall, so that working memory is freed for higher-order reasoning. Mental model tasks, by contrast, require that freed capacity to be used for deliberate manipulation: planning, predicting, evaluating.

For early instruction, then, the focus is on automaticity: imitating and internalising patterns until they no longer consume attention. For advanced instruction, the focus is on reasoning: manipulating those patterns to achieve deliberate effects. The schema gives the writer fluency; the mental model gives the writer control.

This progression echoes a broader truth about expertise. The novice struggles to recall and apply isolated facts; the expert operates through rich networks of knowledge that can be reasoned with, adapted, and applied flexibly. Whether we label this as moving from schema to model or from shallow to deep understanding is less important than recognising the transition itself.

Why it might not

The danger lies in mistaking a useful metaphor for a hard distinction. The line between automatic and deliberate knowledge is porous. Many schemas are consciously shaped before they become automatic; many mental models operate unconsciously once well rehearsed. A child learning multiplication facts, for instance, begins with deliberate reasoning, repeated counting or grouping, before those procedures condense into automatic schemas. Conversely, an expert driver’s understanding of how gears transmit torque is rarely articulated, yet it guides behaviour through an implicit mental model. The border between the two is not fixed but fluid, shifting as practice deepens and familiarity grows.

If teachers become fixated on terminology rather than practice, the concept risks joining the long list of edu-jargon whose meaning evaporates the moment it hits the staffroom. Teachers have lived through plenty of such semantic migrations: differentiation that became a paperwork exercise; metacognition that turned into a checklist; formative assessment that collapsed into endless data collection. “Building schemas” and “developing mental models” could easily meet the same fate if they are repeated as slogans rather than used to inform instructional design.

Moreover, since no one has yet demonstrated clear differential effects, there’s little justification for claiming that “mental models are not schemas” as if this were a settled scientific fact. Doing so risks over-claiming: constructing a pedagogical edifice on what remains, at best, a conceptual nuance. Without empirical evidence showing that different instructional approaches are required, the distinction functions as a metaphorical lens, not a law of cognition. Treating it as doctrine confuses description with explanation and distracts attention from what actually matters — designing sequences of instruction that move students from practice to understanding, whatever theoretical label we choose to attach.

Then there’s the very real risk of conceptual confusion. Essentially, schemas and mental models must coexist within the mind. Any attempt to isolate one from the other distorts the interdependence between automaticity and reasoning. To state categorically that one is this and the other is that is not only inaccurate but actively misleading. It invites teachers to picture schemas as static, monolithic structures — fixed mental filing cabinets into which knowledge is deposited — while imagining mental models as temporary, ad hoc constructs that collapse after use. Both notions are misconceptions born of the very distinction they seek to clarify. In reality, schemas are dynamic, continually reorganised through experience, and mental models often persist, evolving as they integrate with broader knowledge networks. To overplay their separation is to misrepresent cognition itself, and in doing so, risk narrowing rather than enriching teachers’ understanding of how learning unfolds.

A pragmatic conclusion

The distinction between schemas and mental models is useful only insofar as it clarifies intention. When planning a lesson, ask: am I helping students recognise patterns or reason about them? Am I trying to build fluency or understanding? If the distinction sharpens your design, use it. If it adds nothing, discard it.

If there is a useful difference at all, it lies not in the underlying psychology but in the language of practice. “Schema” and “mental model” are not rival explanations of cognition so much as two ways of talking about what we do with knowledge. Teachers can maybe use both terms to mark a change in emphasis rather than a change in kind.

When planning curriculum and early instruction, schema might be the more helpful term, foregrounding, as it does, structure, accumulation, and organisation. Talking about “building schemas” reminds teachers that knowledge must be stable and retrievable before it can be flexibly applied. It justifies repetition, modelling, and tightly scaffolded practice, the kinds of teaching sometimes dismissed as mechanical but which are essential to long-term retention. Using the language of schema helps to legitimise the deliberate cultivation of fluency, and to explain why overloading novices with open-ended reasoning tasks is counterproductive.

By contrast, mental model can serve as a productive metaphor when the goal shifts from consolidation to application. It invites teachers and students to think of knowledge as something that can be run, tested, and revised: a simulation that allows us to predict, explain, or imagine outcomes. Framing a lesson around “building mental models” could legitimise exploratory and reflective activity once sufficient knowledge exists to support it. It suits discussion, hypothesis-testing, or creative recombination: the phase of learning where we begin to reason about how things work rather than simply knowing that they do.

In this sense, the two terms operate like zoom levels on the same map. Schema keeps the focus on the terrain, the detailed knowledge that must be mastered; whereas mental model pulls back to show the routes between places, how that knowledge interacts to form systems of understanding. Used together, they offer teachers a vocabulary for sequencing: from map-making to map-reading, from accumulation to integration. The distinction, then, is linguistic rather than cognitive and is, at best, a way of varying our talk to keep sight of both the ground beneath our feet and the landscape of ideas that lies beyond.

For now, the available evidence seems to suggest that schemas and mental models are two faces of the same process: the gradual organisation of knowledge into structures that first automate performance and then enable flexible reasoning. They differ in emphasis rather than kind. Suggesting that anyone is wrong for using or preferring one term over another is both unnecessary and unhelpful. Anyone dogmatically asserting that using one term instead of the other is incorrect is obviously wrong.

So, does the difference make any difference? Perhaps only this: good teaching moves students from doing without thinking to thinking about what they do. Whether we call that progression schema-building or model-building is, in the end, largely a matter of taste.

1. See Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285; Chi, M. T. H., Glaser, R., & Rees, E. (1982). Expertise in problem solving. In R. J. Sternberg (Ed.), Advances in the psychology of human intelligence (Vol. 1, pp. 7–75). Hillsdale, NJ: Lawrence Erlbaum; Anderson, J. R. (1983). The Architecture of Cognition. Cambridge, MA: Harvard University Press.

2. See Johnson-Laird, P. N. (1983). Mental Models: Towards a Cognitive Science of Language, Inference, and Consciousness. Cambridge University Press; Gentner, D., & Stevens, A. L. (Eds.). (1983). Mental Models. Hillsdale, NJ: Lawrence Erlbaum; Norman, D. A. (1983). Some Observations on Mental Models. In D. Gentner & A. L. Stevens (Eds.), Mental Models (pp. 7–14); Johnson-Laird, P. N. (2010). Mental Models and Human Reasoning. Proceedings of the National Academy of Sciences, 107(43), 18243–18250.