Layers of Learning

Layers of Learning is a build order for learning a topic. Layers of Learning sits inside the Deep Processing dimension. The order runs from the inside out: first the few relations that hold the topic together, then the ideas those relations connect, then the facts that fill in the ideas. Sources present those three kinds of piece mixed together, so each new fact has to be placed with no structure already there to receive it. Building the layers in this order gives every later piece a place before it arrives, which makes it cheaper to place.

Core takeaways

• A topic has three layers. The logic layer is the frame of the topic: the major relations, mechanisms, and organizing principles. The concept layer is the key ideas that populate that logic. The detail layer is the facts, numbers, exceptions, and sub-concepts that fill those ideas.
• Textbooks and lectures mix the layers together. Following a source start to finish inherits the mix, so relevance swings from piece to piece and difficulty stays high.
• Build the logic first, then the concepts, then the details. Each piece then arrives at a place that is already visible, and it is processed cheaply.
• When many facts will not stick, the frame is thin. The repair is to work on the logic and concept layers. Repeating the facts does not repair the frame.
• The frame is cheap to fix only while it is still small. A confidently wrong frame costs more than no frame, because everything gets attached to it.
• The first pass is finished when the few main relations can be said out loud with the source closed. Difficulty should fall with each pass after that.

The layers

Knowledge sits in layers. Which layer a piece belongs to is decided by the question that produces it. The size of the idea does not decide it.

The logic layer is the major relations, mechanisms, and organizing principles of a topic: why things connect, and in which direction. This layer is also called the backbone. The backbone is not the first heading in a chapter. It is the handful of points that carry the most connections to everything else. Recalling from those heavily connected points brings back the overall structure of the topic, with the gaps at the edges. Recalling the same number of peripheral points brings back individual facts with nothing in the middle to connect them. Two questions build this layer: why is this important, and how does it relate to everything else. The first question cannot be answered without doing the second. That is why the pair produces relations and not a list of headings.

The concept layer is the key ideas that populate that logic. Its questions are what the thing is, and how it happens. The logic layer answers why and in which direction. The concept layer answers what and how.

The detail layer is the specific facts, numbers, exceptions, and sub-concepts that fill those ideas. Some details are encoded early, because they are the concrete anchor that makes an abstract concept holdable. A concept given only in impressions cannot be pictured. One specific, checkable fact makes it picturable. For example, a painting described as bright, vibrant, and an outdoor scene cannot be pictured until someone says how wide it is. The test for a detail is whether knowing it helps the concept it is attached to. If it does, it is encoded now. If it does not, it waits.

The leftovers are arbitrary details. They are pieces that do not yet help anything get understood. They are learned because the assessment asks for them. They are still learned. They are learned last, and they are learned by repetition. As the structure grows, they stop being arbitrary.

The logic layer gives the topic its frame. The concept layer gives it its content. The detail layer gives it its resolution.

Why sources mix the layers

A resource does not mark these layers. Textbooks and lectures mix logic, concepts, and details together. A heading is sometimes a logical grouping, sometimes a concept, and sometimes a detail that has been given its own heading.

The grouping a resource offers as its logic may be replaced once the learner judges what is important. Importance-Based Chunking is the tool for overruling the resource’s grouping. It groups the pieces of each layer by why they matter instead of by the heading they sat under.

The layers have to be extracted from the source and rebuilt. That work is encoding. The learner does not invent the curriculum of the topic. The learner rebuilds that curriculum in a usable order. Following the source start to finish inherits the mix. Relevance then swings from piece to piece with no relation to what the learner already holds, and difficulty stays high no matter how much has been covered.

When a piece arrives and the place for it is already visible, it is processed cheaply. When relevance swings, effort goes first to deciding whether the piece matters, and only then to what it means. That effort stays high and never settles. Cognitive Load & What Mental Effort Is Trying to Cue covers what that effort is a signal of.

How relevance changes as the structure grows

The layers are not fixed. How relevant a detail is depends on whether the structure that would give it relevance already exists. Logic generates relevance for concepts. Concepts generate relevance for details. The difficulty of each pass falls as the network becomes robust enough to hold new pieces.

A large number of facts that will not stick means the frame is thin. A hard subject is not the cause.

When a fact will not stick, the move is to repair the logic and concept layers. Repeating the fact does not repair them. A topic with a very large number of details still has only a handful of main relations. Working on those relations is what makes the details attach.

Those layers are layers of a Schema. A schema is the network a new piece attaches to, and it is the thing that generates the piece’s relevance.

Building the frame first

Repairing the frame starts with finding the logic. The first pass skims the source for organizing principles: the main relations, the mechanisms, and the causal chains. This is the expensive pass, because it takes evaluation and not just reading. That evaluation is the work of Higher-Order Learning, and it is what produces the frame. While the map holds only a few relations, rearranging it is cheap. Once everything is on the map, correcting it costs more than building it did. A topic’s basic logic rarely changes much, so the time spent settling it is not spent again.

The first pass names only the few largest relations and how they connect. Concepts go onto the frame next, and each concept now has a location. Details go last, and far fewer of them arrive with nowhere to attach. Skipping to details before the logic is in place encodes information that has no location to land in.

This is self-regulated scaffolding: the topic is taken apart and rebuilt in the order that makes each step manageable. The work that introduced scaffolding described someone else fading the support. The self-directed version is a later extension.

Order control is the mechanism that makes this possible while working through sources. The next piece is chosen by what it can already attach to. The chapter order does not choose it.

Where the model fails and how to tell

The first pass returns nothing that can be pointed at. It costs more than reading and produces no visible notes. That is why it is the pass that gets skipped.

Some material has no backbone to find. In a vocabulary list, or a procedure with no causal structure, the pieces barely interact. The investment in a first pass does not return, and learning the leftovers by repetition is the whole job.

A confidently wrong frame is more expensive than no frame, because everything gets attached to it. The limit is the same as before: the frame is cheap to fix only while it is still small.

Layering does not remove the memorization tier. It shrinks it. Under a deadline, the leftover details still have to be learned.

No published model matches this exact cut. What outside research supports is narrower: frames and elements learned before relations reduce load and improve later learning, and expertise is organized by principles rather than surface features.

There are signs that the layers are being built. A detail that would not stick before the frame was in place is retained on first contact after it. A fact that still will not stick is a signal that the frame needs repair. The first pass ends when the few main relations can be said out loud with the source closed. If they cannot, that pass is not finished, however much was read. Difficulty falls from one pass to the next. If every pass costs the same, the layers are not being built, and the session is still following the source’s order.

Where the model is used

Layers of Learning is a model. It is not a method. The model says what to build first, and what is still cheap to fix. It does not give permission to skip the outer layers. Those layers still have to be learned.

Prestudy answers when. Prestudy is a short survey that finds the frame before the main study session. That survey spreads the cost of the big picture across sessions. Layering answers what: which layer the session is building. Prestudy builds the logic layer, and the main study sessions fill in the concepts and details. Prestudy can also be a light pass over the concepts.

Shoot is where the layers get built against real sources, question by question. Order control is the mechanism there: moving between resources for the most relevant next piece instead of following one source start to finish.

The Bear Hunter System is the four-pass loop that walks the layers in order.

How to practice this

1. Make a first pass over the source that looks only for the main relations, mechanisms, and causal chains. Notice that this pass produces no visible notes and costs more than reading.
2. Close the source and say the few main relations out loud. If you cannot, the first pass is not finished, however much you have read.
3. Put the key ideas onto that frame next, asking what each thing is and how it happens. Notice that each idea now has a location.
4. Add details last, and for each one ask whether knowing it helps the idea it is attached to. Learn the detail now if it helps; leave it for last and learn it by repetition if it does not. Notice that far fewer details arrive with nowhere to attach.
5. When a fact will not stick, go back to the logic and concept layers instead of repeating the fact. Notice whether the fact is retained the first time you meet it after the repair.
6. Compare the effort of each pass with the pass before it. Difficulty should fall from one pass to the next. If every pass costs the same, the session is still following the source’s order.

Related pages

• Deep Processing: the parent dimension this model sits inside
• Prestudy: first-pass implementation at topic level, a short survey that finds the frame before the main event
• Shoot: where the layers get built against real sources, question by question
• Schema: the structure the layers are layers of
• Higher-Order Learning: evaluating and relating, which is the work the logic pass is made of

3 more related pages

• Cognitive Load & What Mental Effort Is Trying to Cue: what the expensive, never-settling feeling under a mixed source order is trying to cue
• Importance-Based Chunking: how the pieces of each layer get grouped, and the tool for overruling a resource’s own grouping
• Bear Hunter System: the four-pass loop that walks the layers in order

Sources

• Ausubel, D. P. (1960). The use of advance organizers in the learning and retention of meaningful verbal material. Journal of Educational Psychology. A frame in place before the material arrives improves what is learned after.
• Mayer, R. E. Pre-training principle. Names and characteristics of the main elements, learned first, improve later learning of how those elements interact.
• Pollock, E., Chandler, P., & Sweller, J. (2002). Assimilating complex information. Learning and Instruction, 12. Isolated elements before their relations lowers load and improves what is learned after.
• Chi, M. T. H., Feltovich, P. J., & Glaser, R. (1981). Categorization and representation of physics problems by experts and novices. Cognitive Science, 5(2). Expertise is organized by principles rather than surface features.
• Sweller, J., van Merriënboer, J. J. G., & Paas, F. (1998/2019). Cognitive architecture and instructional design. Mixed, highly interactive input keeps load high.
• Bransford, J. D., & Johnson, M. K. (1972). Contextual prerequisites for understanding. Journal of Verbal Learning and Verbal Behavior, 11. Material that is arbitrary without a frame becomes comprehensible with one.
• Wood, D., Bruner, J. S., & Ross, G. (1976). The role of tutoring in problem solving. Journal of Child Psychology and Psychiatry. Scaffolding as other-regulated fading support, the bound on the self-directed reading.
• Holton, D., & Clarke, D. (2006). Scaffolding and metacognition. International Journal of Mathematical Education in Science and Technology. The self-directed extension.