Vocabulary is the collection of words and established expressions available in a language, a subject, or a person’s understanding and use. One important thing vocabulary helps us do is make a distinction available beyond the person who first noticed it. We can explain something to another person, give it a shared name, and record enough information for someone else to understand it later.
This article develops an eduKate teaching framework: One mind → many minds → future minds. We call these the Single State, the Multiple State, and the Durable Record State. The last is what we may be tempted to call a “Permanent State”, but permanence is not guaranteed. A surviving record still needs to be accessible and interpretable. These are useful perspectives on communication and preservation, not literal phases of matter or a validated classification of brain states.
Here is the test that makes the framework concrete: can you help someone understand a light bulb without saying “light bulb”? Yes. But identifying the object, understanding its purpose, explaining its mechanism and learning to manufacture it are different achievements. The test reveals where meaning lives, what a name contributes, and what must survive when the original speaker is no longer present. This is a companion to Vocabulary | Ideas Into Sounds Into Stone.
The three states · The light-bulb test · Abstraction, representation and tokenisation · The future-reader test · Try the lesson
1. Single State, Multiple State and the wish for a Permanent State
Imagine noticing that one particular part of a desk lamp is the part that produces light. At first, this particular observation is yours. You may picture the part, remember where it fits, or anticipate what it does without currently recalling its name. In this teaching framework, that is the Single State: one person’s present formulation of an idea. It does not mean that nobody else in the world has independently had the same idea.
You then explain your observation. Another person distinguishes the same part from the shade, stand and switch. Now there is a Multiple State: relevant understanding is available to more than one person. Success does not require identical memories or an exact copy of your mental experience. It requires enough agreement for the task, which we can check.
Finally, you write a description beside a drawing. A later reader who never met you uses the record to identify the intended part. The record has helped your explanation cross a gap in time. That is the Durable Record State. “Future minds” describes its purpose; it does not promise that the record will last forever or that every future reader will understand it.
| State in this framework | What becomes possible? | What would count as evidence? |
|---|---|---|
| Single: one mind | A person forms or notices a distinction. | The person can identify, compare or demonstrate the intended distinction. |
| Multiple: more than one mind | Other people can work with a related distinction. | They can explain it, distinguish alternatives or use it appropriately. |
| Durable record: later minds | A representation remains available beyond the original encounter. | A later reader can recover the relevant meaning without the original speaker’s help. |
When we call an individual a “cell” in a wider social system, cell is a metaphor for a person. It does not mean that an idea resides inside one biological cell. For a general reader, “one mind” avoids that unnecessary ambiguity.
2. This strengthens the original idea; it does not replace the definition
Ideas Into Sounds Into Stone draws attention to expression and its medium. The new framework asks a different question: who can recover the meaning, and when? A spoken explanation, a signed conversation, a written paragraph and a labelled drawing can take different routes towards that goal. We should not make sound a compulsory stage through which every idea must pass.
Vocabulary is important, but a word list alone does not create this journey. Grammar, context, shared conventions, examples, attention, feedback and recording practices also matter. The stronger claim is therefore: vocabulary, used within a communication system, helps make selected distinctions addressable, shareable, revisable and recoverable.
There is also a difference between lacking one name and lacking access to language. A person who cannot recall a particular word may still understand its referent. A 2024 Nature perspective by Fedorenko, Piantadosi and Gibson argues from evidence separating language and thought that language is not a prerequisite for all complex thinking. That is not a reason to dismiss language’s contribution. Lupyan, Rakison and McClelland’s category-learning experiments found that redundant verbal labels could facilitate learning novel categories. Neither source establishes our three-state framework as a scientific law.
3. Explain a light bulb without saying its name
For this exercise, choose a picture of an ordinary desk lamp with a replaceable bulb. Use pictures rather than asking children to handle electrical fittings. The target is one familiar example, not a technical definition covering every lighting product.
First attempt: give a broad property
“It makes light.” That could describe a candle, a screen or several other things. The statement may be true of the target, but it does not yet distinguish it. A correct property is not necessarily a sufficient explanation.
Second attempt: add a relevant relationship
“It uses electricity to produce light.” This narrows the possibilities but does not identify the intended part of the pictured lamp. The listener may still point to the entire lamp. We have made progress, not completed the task.
Third attempt: establish the boundary
“In this picture, I mean the replaceable part fitted into the lamp that produces light when supplied with electricity—not the shade, switch or stand.” The explanation now combines function, position and a boundary. It gives the listener something more specific to identify.
Notice what happened. We did not use the target name, but we used other vocabulary: replaceable, part, fitted, electricity and not. The experiment shows that meaning can be communicated without one particular label. It does not show that the explanation is independent of language.
Fourth attempt: check the listener rather than admire the sentence
Ask the listener to point to the intended part. Then ask, “Why not the switch?” and “Why not the whole lamp?” If the answers reveal a misunderstanding, adjust the description or add a drawing. Communication is the task, not producing a sentence that merely sounds complete.
This use of feedback is consistent with Clark and Wilkes-Gibbs’s research on referring as a collaborative process. Their communication task examined how pairs worked together to establish references to complex figures. Here, our own light-bulb exercise uses that general principle: a receiver’s response helps establish whether the intended reference has been understood.
Fifth attempt: introduce the compact label
Now say, “That part is called a light bulb.” Once the name is connected to the intended concept, the next instruction can be shorter. “Compare these two light bulbs” may replace a much longer description. The shorter expression works because some explanatory work has already been done, or because the listener already had the relevant knowledge.
The word has not swallowed the complete object, its physics and its manufacturing history. It is a conventional handle through which a listener can access relevant knowledge. A person unfamiliar with the term may still need the longer explanation. For the related writing principle, see lexical economy.
4. Abstraction, representation and tokenisation do different jobs
We can now separate three questions that otherwise become tangled. They interact, but they are not interchangeable names for the same operation.
Abstraction: which features or relationships matter here?
Compare several pictured bulbs. For an introductory identification task, we may set aside colour, brand and decorative shape and focus on their role as electrical light sources. For a different task, such as comparing appearances, shape may become important again. Abstraction is not simply removing detail until nothing remains; it is preserving structure relevant to the question.
An instructive nonexample is the lamp’s switch. It participates in controlling the light, but it is not the light-producing part we selected. The contrast helps the learner distinguish a cause or control from the target component. The textbook chapter on abstraction develops this movement between examples, differences and shared relationships.
Representation: how will we make the idea available?
A description, a drawing, a photograph, a label or a diagram can represent something about the selected part. Each makes some information easier to recover and leaves other information out. A photograph may help recognition. An annotated diagram may clarify the part’s relationship to the rest of the lamp. Neither automatically explains every mechanism.
Representation is therefore not identical to naming. A drawing can be a representation without displaying the English name. A spoken explanation is also a representation, even when it uses many words rather than one established label.
Tokenisation, in eduKate’s broad teaching usage: what counts as one workable unit?
In a furniture inventory, the entire desk lamp may be one item. In our picture exercise, we distinguish the bulb from the stand, shade and switch. In a more detailed analysis, we could examine structures within the selected component. The unit changes with the task.
eduKate’s What Is a Token? uses “tokenisation” in this broad, explicitly qualified sense: choosing or recognising manageable units. Naming the unit is a further possibility, not the definition of the whole process. A person can attend to a component before knowing its established name.
In one line: abstraction selects relevant structure; representation presents it; broad cognitive tokenisation selects workable units. These are analytical distinctions for the lesson, not a claim that every brain runs three discrete steps in a fixed order. Choosing a representation can reveal a poor boundary; comparing examples can change the abstraction.
5. Do not confuse three different uses of “token”
Our broad teaching usage needs a visible boundary. In the philosophical type–token distinction, a token is a particular instance, such as one written occurrence of a word; a type is the repeatable kind. Writing “lamp” twice produces two written word tokens of one word type. That distinction is not itself a theory of how a child forms concepts.
In natural-language processing, tokenisation is an engineered procedure for dividing text into units. Hugging Face’s documentation explains subword approaches whose segmentation depends on the algorithm and vocabulary. The expression “light bulb” has two written words, but an AI token count cannot be inferred simply from its conceptual unity or word count. None of these uses establishes that the brain contains a literal computer tokenizer.
6. What exactly has been transferred?
Suppose the listener identifies the intended bulb correctly. What have we demonstrated? We have evidence of successful identification under the conditions of this task. We have not demonstrated that the listener can explain the detailed physics, choose a safe replacement, or manufacture the object. Those are separate questions requiring additional knowledge and appropriate checks.
This protects the framework from an attractive exaggeration: “I said the word, so I transferred the entire idea.” Instead, ask what level of understanding the task requires. Can the listener recognise the object? Describe its purpose? Distinguish it from a nearby alternative? Explain a relevant relationship? Apply the distinction to an unfamiliar example?
A useful way to describe the encounter is my understanding → an expression → your interpretation → feedback. The expression gives you evidence from which to construct an interpretation; it does not let me inspect or duplicate your private experience. We judge success by relevant responses, not by pretending that two minds have become identical.
7. The future-reader test: preserving marks is not enough
Now remove the original speaker. Put the description and drawing on a card and give it to a third person. The new reader cannot ask what you were pointing at yesterday. Does the card still identify the object? Does it explain what “part” refers to? Does the drawing show the lamp as well as the selected component?
Compare two records: “That one makes it work” and “In this drawing, the labelled component produces light when supplied with electricity; it is distinct from the switch that controls the circuit.” The second is more self-contained for this task. The first relies on a shared scene that may no longer be available.
Carving the first sentence into stone would not repair its missing reference. Durability of the surface and recoverability of the meaning are different problems. A useful record needs an interpretable representation, sufficient context, and a way to find and access it. These are practical conditions, not a guarantee against every future misunderstanding.
For digital records, preservation also involves ongoing care. The Library of Congress’s personal-archiving guidance recommends descriptive organisation, copies in different locations and periodic checks that files remain readable. “Saved once” is not the same as “preserved indefinitely”. The textbook develops this distinction in preserving an idea and making a record for a future reader.
8. The three states are not a compulsory staircase
The simple arrow is memorable, but two independent questions give us a more accurate model: how widely is the meaning understood, and how recoverable is its record over time?
A private notebook can record an idea before anybody else understands it. A spoken explanation can be shared without being written down. An unread inscription can retain marks without having a current reader who understands them. A recorded conversation can survive without first being converted into writing. These counterexamples show why distribution and durability must not be collapsed into one variable.
Nor does making a record make a claim true. A mistake can be copied accurately. A durable record can preserve both a useful explanation and an error. For knowledge to improve, later readers also need ways to question, compare and correct what they receive.
We also do not need a speculative story about “cavemen grunting” to explain the model. Imagine two people who do not yet share a name for an object. They may point, draw or demonstrate. That is a teaching scenario, not a reconstruction of language’s origin. In particular, improvised gestures must not be confused with complete signed languages. NIDCD describes American Sign Language as a complete natural language with its own grammar, not a primitive substitute for speech.
9. A shared lesson: understand, explain, name, record, recover
Use pictures of a lamp and its parts, paper and a pencil. Explain that the task is about making meaning clear, not winning a guessing game. Do not show the target name at the start. The following is an illustrative teaching activity, not a standardised assessment or a diagnostic test.
- Choose the unit. Privately select the light-producing component in the picture. Decide what the listener must distinguish it from.
- Explain without the target name. Give a description or drawing. Let the listener ask questions. Record which details actually helped.
- Check the concept. Ask the listener to identify the target and reject one plausible alternative with a reason. A lucky guess is not the same as an explained distinction.
- Introduce the name. Connect “light bulb” to the established example and its relevant properties. Ask the learner to use it in a new sentence.
- Create a record. Write a short explanation with a labelled drawing, a useful example and a nonexample. Remove vague references such as “that thing”.
- Test recovery. Give the record to another reader without extra spoken help. Ask them to explain what it means. Revise any part that depended on information left in the first speaker’s head.
Try the same pattern with a genuinely abstract idea such as a promise. Present one situation containing a commitment, another containing only a hope, and a third containing a prediction. Ask what relationship distinguishes the cases. Then introduce or refine the label. The point is not to prove that every concept has one perfectly sharp definition. It is to make the intended distinction discussable and test it against alternatives.
Keep separate notes on recognition, explanation, word recall and use in a new context. A difficulty on one task should lead to a better question, not a sweeping judgement about the learner’s intelligence. Return to the textbook’s chapter on locating the break in understanding for that distinction.
10. What this adds to the eduKate vocabulary ecosystem
The ingredients are not new to the site. Vocabulary OS: Tokens and Compressions already uses the lightbulb as an example of a compact expression linked to richer knowledge. What Is a Token? already distinguishes units, representations and abstractions. The 48-chapter textbook already connects meaning, expression, abstraction and records.
This companion puts those ingredients into one testable teaching sequence. Remove the name to reveal the explanatory work. Introduce the name to show what compact reference adds. Remove the speaker to test whether the record can carry enough meaning forward. The contribution is this integrated explanation and exercise—not a claim to have discovered the origin of thought or a new law of cognition.
A practical evaluation would compare learners’ explanations and performance on new examples, not just whether they can repeat these three labels. If the framework makes distinctions clearer and exposes missing context, it has served its teaching purpose. That usefulness remains something to test rather than assume.
The label is not the concept. The signal is not the understanding. The record is not the reader. Vocabulary becomes powerful when these connections work: a person can make a meaningful distinction, another person can reconstruct enough of it to act or reason, and a later person can recover it without depending entirely on the first speaker’s presence.
Read the full vocabulary textbook · Explore the Vocabulary Hub · Practise meaning and precision