eduKateSG Learning Node Series · 0053

How Learning Progressions Work | Map the Intermediate States Between Novice and Expert

A learner is rarely simply “does not know” on Monday and “knows” on Friday. Between those endpoints are intermediate states: partial models, useful but limited strategies, misconceptions that work in some cases, vocabulary without structure, procedures without selection, and explanations that are almost right but fail at one critical relationship.

Learning progressions are an attempt to map those intermediate states. They describe increasingly sophisticated ways of thinking or performing that learners may move through when instruction and experience support the movement. The map is not a timetable and it is not a claim that every child must walk one identical path. It is a model of likely developmental structure that helps teachers ask a more useful question than “Have we taught this chapter?”

The control question is: what is the learner able to see, explain and do now, what more powerful state should come next, and what evidence would show that the transition has actually happened?

The 50-Second Read

• A learning progression maps increasingly sophisticated states of understanding or performance.
• It sits between a curriculum sequence and an individual learner diagnosis: the curriculum says what should be learned; the progression helps describe how understanding may become more powerful.
• The intermediate states matter because teaching can only act on the state that actually exists.
• Good progressions include concepts, practices, representations, common misconceptions and evidence of transition.
• They are hypotheses informed by research and evidence, not immutable laws of development.
• Assessment becomes more useful when it locates a learner on a progression instead of producing only a score.
• The next instructional move should be small enough to be learnable and important enough to change what becomes possible next.
• Progress should eventually survive reduced support, changed contexts, time delay and authentic performance.

Canonical Owner Boundary

This page owns the intermediate-state map. How Curriculum Works owns the broader selection and sequencing of what a system intends learners to study. How Formative Assessment Works owns the evidence-to-action loop during instruction. How Mastery Learning Works owns progression decisions when later learning depends on stable prerequisites. A learning progression connects these owners by describing the meaningful states through which capability can become more sophisticated.

1. Why Two Endpoints Are Not Enough

Imagine a Primary Mathematics goal: understand fractions as numbers. One endpoint is a learner who treats a fraction as two unrelated whole numbers. The other is a learner who can place fractions on a number line, compare unlike fractions, generate equivalents, reason about magnitude and use fractions flexibly in later ratio and algebraic work.

The educational work happens between those endpoints. A child may first understand halves and quarters through sharing, then recognise part-whole notation, then coordinate numerator and denominator, then understand equal partitions, then connect fractions to measures, then recognise equivalence, then reason multiplicatively. If instruction treats all errors as one category called “weak fractions,” it loses the location of the actual transition.

This is why a progression is more useful than a checklist. A checklist can say which outcomes are missing. A progression can suggest which missing outcome is structurally next.

2. Progression Is Not the Same as Pace

A fast learner and a slow learner can occupy the same conceptual state for different lengths of time. A learner can also move quickly through one region and stall in another. Progression describes structure; pace describes time.

This distinction protects teachers from a common error: assuming that because six weeks have passed, six weeks of conceptual development must have happened. Calendars advance automatically. Learner states do not.

3. A Learning Progression Is a Model, Not a Railway Track

The National Research Council described learning progressions as successively more sophisticated ways of thinking about a topic that may follow one another over extended periods of learning. It also emphasised that these pathways depend on suitable instruction. Later National Academies work on science assessment treated progressions as descriptions of developing understanding between beginning states and more mature disciplinary expectations. See Taking Science to School and Developing Assessments for the Next Generation Science Standards.

The important word is model. The progression is a disciplined hypothesis about development. Real learners may enter from different starting points, skip an apparent state, combine two states, regress temporarily, or display one level in a familiar context and another in an unfamiliar one.

A useful progression therefore behaves more like a topographic map than a railway timetable. It shows terrain, slopes, bottlenecks and plausible routes. It does not guarantee that every traveller walks the same path at the same speed.

4. The Five Things a Strong Progression Should Make Visible

First: the target capability. What does mature performance actually require? “Understand algebra” is too vague. A useful target names relationships, representations, operations, selection and reasoning.

Second: intermediate states. What less sophisticated but intelligible forms can precede the target? These should describe thinking, not merely scores.

Third: common misconceptions and partial models. Errors often have structure. A student who believes heavier objects fall faster is not empty of knowledge; she has a causal model that can generate predictions. Instruction needs to meet the model, not an imaginary blank slate.

Fourth: evidence markers. What could a learner say, draw, solve, classify, predict, construct or explain that would justify placing them in one state rather than another?

Fifth: transition moves. What experience, comparison, representation, explanation, practice or feedback is likely to help the learner reorganise the current model into the next one?

5. Mathematics: From Procedure to Structure

Consider solving linear equations. A shallow sequence might be: easy equations, harder equations, exam questions. A learning progression asks what changes in the learner’s internal model.

1. Arithmetic unknowns are solved by guessing or reversing a single operation.
2. The equal sign is understood as balance rather than “the answer comes next.”
3. Equivalent transformations are recognised as preserving equality.
4. Inverse operations are applied systematically to isolate an unknown.
5. Multi-step equations are coordinated without losing sign structure.
6. Equations are recognised inside unfamiliar representations and word problems.
7. The learner can select, justify and verify an algebraic route independently.

Those are not merely seven worksheets. They are seven changes in what the learner notices and controls.

6. English: From Retrieval to Interpretation

In comprehension, a learner may move from finding explicitly repeated words, to matching paraphrases, to locating evidence, to connecting evidence across sentences, to inferring unstated relationships, to evaluating how language and structure produce an effect.

A teacher who knows the progression does not respond to every weak inference answer with “read more carefully.” The teacher asks which relationship the learner cannot yet construct. Is the vocabulary blocking the sentence? Is pronoun reference unstable? Is the learner failing to distinguish evidence from interpretation? Is the inference plausible but unsupported?

The progression changes feedback from generic advice into state-specific repair.

7. Science: Models Become More Powerful

Science is especially suited to progression thinking because learners often hold explanatory models that become progressively more mechanistic. A student may know that plants “need sunlight,” later understand photosynthesis as a process producing glucose, later coordinate matter and energy, later connect cellular processes and ecosystem flows, and later reason about evidence, model limitations and competing explanations.

The National Academies’ work on science learning progressions is useful here because it distinguishes mere accumulation of facts from increasing sophistication in explanation and practice. A mature model does not only contain more pieces. It organises them differently.

8. CivDJ Cross-Domain Comparison: Learning Progression as Engineering Maturity

An engineering prototype does not jump from idea to certified final system. It may pass through proof of concept, functional prototype, integrated prototype, representative-environment testing and production readiness. Each state answers different questions and supports different next decisions.

Learning works similarly. “Can perform with a model visible” is not the same state as “can retrieve independently.” “Can execute a familiar method” is not the same as “can select it among competitors.” “Can solve untimed” is not the same as “can sustain the method under examination conditions.”

The engineering lesson is simple: never certify a later state using evidence that only proves an earlier one.

9. The Rainbolt Missing-Node Scan: Look for the Unnamed Middle

A strong observer notices what everyone else passes over. In learning, the missing node is often an unnamed intermediate capability.

A student can differentiate when told “use the chain rule” but fails in mixed questions. The missing node is not differentiation execution. It is method recognition.

A student can define diffusion but cannot explain perfume spreading across a room. The missing node is not the definition. It is transfer from term to mechanism in context.

A student can write a good paragraph from a sentence frame but collapses when the frame disappears. The missing node is internal paragraph architecture.

The missing-node scan asks: what capability would have to exist between the performance I can see and the performance I want?

10. Assessment Should Locate, Not Merely Rank

A score compresses information. Seventy per cent can be useful, but it does not tell the teacher whether the learner has a misconception, a retrieval problem, a representation problem, a method-selection problem or a checking problem.

Progression-aligned assessment asks items that discriminate among states. Instead of only asking whether the student can produce the final answer, it may ask the learner to explain, classify, compare, predict, represent or transfer.

This makes assessment operational. The result points toward the next instructional move.

11. The Same Learner Can Occupy Different States at Once

A Secondary student may understand quadratic structure conceptually but still make frequent algebraic manipulation errors. She may explain a Science mechanism accurately in speech but struggle to express it in exam language. He may retrieve vocabulary in a flashcard context but fail to recognise the same word in a dense passage.

Learning is multidimensional. Progressions should therefore avoid pretending that one number can completely describe the learner. Useful dimensions can include conceptual understanding, representation, retrieval, selection, execution, explanation, transfer and independence.

12. Progressions and the First Weak Link

The first weak link is the earliest unstable dependency that meaningfully constrains downstream performance. A progression helps find it because the teacher can test the chain in order.

If the learner cannot compare fractions, do not immediately assign difficult ratio word problems. If the learner cannot identify what a pronoun refers to, advanced inference questions may be noisy evidence. If the learner cannot distinguish independent and dependent variables, laboratory evaluation questions may fail for reasons that have little to do with evaluation.

Progression thinking therefore makes diagnosis cheaper. It tells us where to probe before adding more volume.

13. Progressions and Mastery Are Related but Different

A progression describes states. Mastery learning decides when evidence is strong enough to progress across a dependency boundary. The two are easy to confuse.

You can have a learning progression without enforcing strict mastery gates. You can also run mastery cycles with a coarse progression. The strongest design uses both: a sufficiently detailed state map and sensible thresholds for high-leverage prerequisites.

14. A Practical Teacher Workflow

1. Name the mature capability. Describe what successful independent performance looks like.
2. Decompose the capability. Identify conceptual, representational, procedural and strategic components.
3. Observe typical partial states. Use student work, questions, think-alouds and common errors.
4. Order only what has a defensible dependency. Do not force every difference into a sequence.
5. Write discriminating checks. Each check should help distinguish neighbouring states.
6. Choose the smallest high-leverage transition. Teach what changes the learner’s model, not merely what fills the next worksheet.
7. Retest under reduced support. A transition is more convincing when the scaffold can shrink.
8. Test transfer. Change wording, context, representation or problem surface.
9. Update the map. If real learners repeatedly move differently from the model, revise the model.

15. Failure Mode: Turning the Progression Into a Label

“She is a Level 2 learner” can become as unhelpful as “she is weak.” The state should describe current evidence in a domain, not become the child’s identity.

Repair: phrase placement as temporary and specific. “In equivalent-fraction reasoning, current work shows State 2 evidence; here is the next transition we are teaching.”

16. Failure Mode: Assuming One Universal Route

Some learning has strong dependencies. Other learning allows multiple routes. A progression becomes dangerous when a useful pattern is mistaken for a biological law.

Repair: distinguish necessary prerequisites from common pathways. Let evidence decide whether the map predicts real performance.

17. Failure Mode: Confusing Task Difficulty With Conceptual Sophistication

A longer calculation can be harder while requiring no deeper model. Conversely, a short conceptual question can reveal a major shift in understanding.

Repair: describe what is changing in the learner’s reasoning, not only how many steps the task contains.

18. Failure Mode: Making Progressions So Detailed They Cannot Be Used

A map with hundreds of tiny states can become a second curriculum bureaucracy. Teachers stop using it because locating the learner costs more than the decision is worth.

Repair: keep only distinctions that change instruction, assessment or progression decisions.

19. Evidence and Limits

Learning progressions have a substantial research base in science and mathematics education, but the quality of a progression depends on the construct, the evidence behind the proposed sequence and the instructional contexts in which it was developed. The National Academies has repeatedly cautioned that some progressions are more empirically established than others. A progression should therefore be treated as a testable educational model rather than a fixed developmental truth.

That caution improves the idea rather than weakening it. The teacher is not required to worship the map. The teacher uses the map to make better predictions, compares those predictions with student evidence, and revises when reality disagrees.

20. The Return Path

Return to the learner between novice and expert.

Without a progression, the middle can look like a collection of mistakes.

With a progression, the same evidence becomes a map of what is already built, what remains unstable and what more powerful form of understanding is close enough to teach next.

Learning progressions work when they turn the invisible middle of learning into a usable map—detailed enough to locate the learner, modest enough to remain revisable, and practical enough to tell a teacher what transition is worth building next.