MindOS Learning Manual: Pretraining State | Learn the Parts Before the System Starts Moving

MindOS · Pretraining State · Identify Parts → Learn Key Characteristics → Start Dynamic Explanation → Build Relations → Remove Labels → Reconstruct System → Transfer → Return

Wait, What? Sometimes the Explanation Is Too Early

A learner watches a clear explanation of how a complex system works. Every sentence is accurate. The animation is smooth. The teacher is not speaking too quickly.

Yet the learner loses the explanation.

The problem may not be the explanation itself. The learner may be trying to do two difficult jobs at once: identify unfamiliar parts and understand how those parts interact over time.

If the names, locations or basic states of the components are still unstable, each new causal step competes with basic orientation. The learner is not only asking “What happens next?” They are also asking “Which thing is that?”

Pretraining State separates those jobs.

Quick Answer

Pretraining is a learner-support operation in which the names and essential characteristics of the main components are learned before a complex explanation, process or system is presented.

Owned learner job: stabilise the minimum component model needed so that the learner can spend the main lesson building relationships rather than repeatedly rediscovering what each part is.

The RFE is not “make the lesson easier.” It is:

Move identification work out of the way only when doing so helps the learner build, retain and later operate the whole system independently.

The Owned Boundary: This Is Not Prior-Knowledge Activation

Prior-Knowledge Activation State asks what the learner already knows and whether that knowledge will help or mislead the next learning episode.

Pretraining can teach information the learner does not yet know. It deliberately creates a small prerequisite representation before the main instruction begins.

The distinction is important:

Activation: retrieve existing prerequisite knowledge.
Pretraining: install missing component knowledge before the main relational explanation.

The Owned Boundary: This Is Not Working-Memory Load State

Working Memory Load owns the broader question of whether too many elements must be held and coordinated at once.

Pretraining is one specific intervention for one specific load pattern: the learner is trying to construct component models and a relational system model simultaneously.

The Owned Boundary: This Is Not Segmenting

Segmenting slows or divides the main explanation into manageable pieces. Pretraining changes what the learner knows before the explanation begins.

A learner may need one, both or neither. If the components are familiar but the process is too fast, segmentation may be the better lever. If the process is well paced but every component is unfamiliar, pretraining becomes more plausible.

What Counts as a Component Model?

Pretraining should teach only what the learner needs in order to follow the later relationships. That usually means some combination of:

the name of a part;
where it is located;
what basic state or role it can have;
how to recognise it in the representation;
the minimum vocabulary needed to distinguish it from neighbouring parts.

Pretraining is not an excuse to front-load the entire chapter. If the “pretraining” becomes a full lecture, the learner has simply received the difficult lesson in a different order.

Observable Learner Signatures

The learner repeatedly stops a process explanation to ask what a labelled part is.
They understand individual sentences but cannot maintain the causal chain because component identity keeps dropping out.
When a diagram is shown, they spend most of their effort locating parts rather than explaining interactions.
A process becomes much easier after a short component-identification exercise.
The learner knows the main idea verbally but cannot map terms to the diagram or representation.
Once labels are removed, the learner can no longer tell which component is acting.
The learner confuses two components with similar names, causing the entire later mechanism to collapse.

These signs do not prove that pretraining is needed. The main explanation may simply be poor, the pacing may be too fast, the representation may be ambiguous, or prerequisite concepts may be deeper than component names.

Discrimination Test 1: Can the Learner Name the Parts Without Explaining the System?

Before teaching the mechanism, show the static representation and ask the learner to identify the main components and one defining characteristic of each.

If the learner can already do this fluently, pretraining may add little. If component identification itself is effortful, the intervention has a plausible target.

Discrimination Test 2: Is the Missing Knowledge a Name—or a Concept?

Knowing that a structure is called a valve is different from understanding pressure, one-way flow or control.

If the learner cannot understand the component’s function without deeper prerequisite knowledge, naming alone will not solve the problem. Route to concept repair rather than pretending a label is understanding.

Discrimination Test 3: Pretraining or Slower Main Instruction?

Try the same explanation with learner-controlled pauses. If performance recovers, pacing may be the dominant weak link. If the learner still spends each pause reconstructing what the parts are, pretraining remains a better candidate.

The MindOS Pretraining Protocol

Step 1 — Define the Main System Job

What must the learner eventually explain or operate? Keep the final performance visible so the pretraining does not expand without limit.

Step 2 — Select Only Load-Bearing Components

Choose the smallest set of names and characteristics needed to follow the mechanism. Decorative vocabulary belongs later.

Step 3 — Build Fast Recognition

Point, name, identify, match and distinguish. The learner should not need a long verbal search to recognise each component.

Step 4 — Add One Functional Characteristic

For each component, add only the characteristic needed for the main explanation: opens/closes, carries signal, changes pressure, stores energy, receives input, performs transformation.

Step 5 — Start the Main Dynamic Explanation

Now build the causal or procedural relationships. The learner should be spending more effort on why one state leads to another and less effort on identifying the actors.

Step 6 — Remove Labels

Pretraining has an expiry condition. Hide the names and ask the learner to identify the components while explaining the process.

Step 7 — Reconstruct the Whole System

Close the model. Ask for the components, their roles and the relationships among them. The final object is the system, not the flashcard set.

Step 8 — Change the Representation

Use a different diagram, wording or surface example. If the learner recognises components only in the original picture, the training has become representation-dependent.

Worked Example: Science

Before explaining the cardiac cycle, a learner may need stable identification of atria, ventricles, valves and major vessels. If every causal statement is interrupted by “Which valve is that?”, the mechanism is competing with orientation.

A short pretraining phase can establish location and one basic role for each structure. Then the learner follows pressure changes, valve state and blood movement through the cycle.

The return test removes the labels and changes the diagram orientation. The learner must still reconstruct the flow and explain why the valves change state.

Worked Example: Mathematics

Before a learner studies a complex transformation of functions, the notation itself may be unstable: input, output, domain, range, translation vector, reflection axis.

Pretraining can stabilise the minimal vocabulary and representations. But the intervention earns its place only if the learner subsequently uses those components to explain and predict transformations without needing the vocabulary sheet.

Worked Example: English

Before teaching a multi-step argument-analysis routine, a learner may need to distinguish claim, evidence, assumption and counterclaim. If those roles are not stable, a sophisticated argument map becomes a moving blur.

Pretrain the four roles with short examples. Then analyse a complete passage. Later remove the labels and require the learner to identify each role inside unfamiliar prose.

How Do We Know?

The classic multimedia-learning account proposes that pretraining helps when learners must otherwise build representations of the components and the causal system at the same time. Mayer, Mathias and Wetzell tested this logic in 2002 using systems such as a car braking system and bicycle tyre pump. Learners who received component pretraining performed better on transfer in all three experiments and on retention in two.

Later handbook syntheses reported substantial support for the pretraining principle in complex multimedia lessons. However, the evidence must not be frozen at those early summaries. A 2025 meta-analysis of Richard Mayer’s multimedia-learning corpus found strong variation across design principles and boundary conditions. The average estimate for pretraining in that corpus was positive but not statistically significant (g = 0.28).

That newer result strengthens the MindOS boundary rather than destroying the state. Pretraining should be used when the learner actually faces component-identification load inside complex material—not as a universal decoration added to every lesson.

Evidence Boundary

Pretraining is best supported for complex instructional situations in which essential component knowledge is needed to follow a system or process.
The effect is not uniformly positive across all studies and contexts.
The 2025 corpus meta-analysis found the average pretraining estimate non-significant, so boundary conditions matter.
Learning component names does not prove understanding of the relationships among them.
Pretraining cannot repair a false or misleading main explanation.
Excessive pretraining can increase study time and introduce information that the learner does not need.
A learner who already knows the components may gain little and may experience the support as redundant.

AI Boundary: Do Not Let the Glossary Become the Lesson

An AI tutor can generate component lists instantly. That convenience creates a risk: the learner receives twenty definitions before knowing which five matter.

A safer AI-assisted sequence is:

name the final system the learner must explain;
ask AI for the minimum load-bearing component set;
verify the definitions against an appropriate source;
learn and identify those components;
close the glossary;
study the full mechanism;
reconstruct without AI;
test on a changed representation.

The tool should reduce orientation cost and then disappear. If the learner cannot operate without the generated labels, independence has not been reached.

Staged Practice

Labelled recognition: identify key components with names visible.
Unlabelled recognition: identify the same components without names.
Characteristic retrieval: state the one functional property needed for the main system.
Dynamic explanation: follow how components interact.
Whole-system reconstruction: explain the sequence without the model.
Changed representation: identify and use components in a new diagram or wording.
Mixed task: decide when component identification is required and when it is irrelevant.

Scaffold Fade

Begin with labelled components and short definitions. Next remove the definitions, then remove the labels, then remove the original diagram. Eventually the learner should meet a fresh system representation and orient themselves without a pretraining card.

The mature learner can also decide when pretraining is unnecessary. Independence includes knowing when not to build a scaffold.

Examination Craft

Examinations rarely announce, “Here are the components you should pretrain.” They present the system and expect orientation plus reasoning.

Therefore final practice must remove the component sheet. The learner should be able to recognise parts from standard notation, diagram conventions and contextual clues, then use them in the explanation or solution.

Transfer Test

Give a new system with partly familiar architecture but different labels or surface features. Ask the learner to identify which components need to be stabilised before they attempt the full explanation.

Transfer is demonstrated when the learner can choose and create an appropriate pretraining layer themselves rather than waiting for a teacher to supply one.

Delayed Return Test

Several days later, present the system without the original component sheet. Ask the learner to name the critical parts, explain their roles and reconstruct the dynamic relation. If the labels survive but the mechanism does not, pretraining became the endpoint instead of the doorway.

Parent and Tutor Teaching Guide

When a child gets lost in a complicated explanation, do not immediately repeat the whole explanation more slowly. First ask:

“Can you point to each important part?”
“What is this called?”
“What one thing does it do?”
“Which two parts are you mixing up?”
“Now that the parts are stable, can you tell me what changes first?”
“Let us hide the labels. Can you still run the system?”

This is diagnosis before prescription. If component knowledge was not the weak link, stop pretraining and route elsewhere.

MindOS Direction Graph

Complex system → component check → unfamiliar critical parts? → minimal pretraining → unlabelled recognition → dynamic explanation → whole-system reconstruction → changed representation → delayed return → independent decision about future pretraining.

If the learner already knows the components but loses the sequence, route to Working Memory Load or segmentation/scaffold work. If the prerequisite knowledge already exists but is not activated, use Prior-Knowledge Activation State. If the learner can follow the full model but not work independently, use Worked Example, Completion Problem or scaffold-fading routes. If the mechanism fails on a new surface, use Transfer State.

MindOS rule: learn the parts first only when doing so frees the learner to understand the relationships. The scaffold has succeeded when the learner can run the whole system after the labels disappear.