MindOS Learning Manual: Spatial-Contiguity State | If Related Information Is Too Far Apart, the Learner Spends Effort Searching Instead of Learning

MindOS · Spatial-Contiguity State · Related Elements → Physical Separation → Search/Integration Cost → Integrate → Explain Relation → Reduce Layout Support → Reconstruct → Transfer → Return

Wait, What? A Learner Can Understand the Diagram and Understand the Text—and Still Lose the Idea Because the Two Are Too Far Apart

A diagram is on the left. Its explanation is three paragraphs below. The learner looks down, reads one sentence, looks back up, searches for the matching part, looks down again, forgets which phrase they were holding, and starts over.

Nothing is factually wrong.

The problem is spatial.

When mutually dependent information is physically separated, the learner may have to spend working-memory capacity on searching, holding and mentally integrating pieces that could have been aligned directly. This is the territory of spatial contiguity and split attention.

Quick Answer

Owned learner job: when related visual and verbal information must be repeatedly searched and mentally joined, place corresponding elements close enough to reduce unnecessary integration cost, then verify that the learner can still reconstruct the relationship when the supportive layout changes.

The RFE is not “put everything next to everything.” It is:

Remove avoidable spatial search only where two elements must be mentally combined to understand the same idea.

The Owned Boundary: This Is Not Signaling State

Signaling State changes the prominence or organisation of relevant information using cues such as arrows, colour or headings.

Spatial contiguity changes the physical relationship between corresponding information. A label can be perfectly clear but still costly if it sits far from the structure it describes.

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

Working Memory Load owns the broader problem of coordinating too many unstable elements.

Spatial contiguity is one specific design lever when part of that load comes from maintaining one representation while visually searching for its corresponding partner elsewhere.

The Owned Boundary: This Is Not Representation State

Representation State asks whether the learner can understand and move among different forms of the same idea.

Spatial contiguity assumes the representations are individually interpretable but unnecessarily separated while the learner must integrate them.

What Split Attention Looks Like

a diagram on one page and its labels on another;
a graph at the top of a page and an explanation far below;
a geometry figure and a separate legend that must be repeatedly decoded;
a worked equation with explanatory notes in a distant margin;
a table whose units or definitions are separated from the relevant columns;
a digital interface requiring repeated scrolling between evidence and explanation.

Not every separation is harmful. The critical condition is that the learner must mentally integrate the two sources to perform the learning job.

Why Integration Can Help

When related information is spatially integrated, the learner can often devote fewer resources to searching and more to understanding the relation itself.

This can reduce:

visual search;
temporary holding of text while locating a diagram part;
re-reading caused by losing the match;
errors caused by connecting the wrong label to the wrong element;
extraneous working-memory demand unrelated to the actual concept.

The useful mechanism is not proximity by itself. It is proximity of information that must be integrated.

Observable Learner Signatures

The learner repeatedly scans between two regions before every explanation.
They understand each source separately but make matching errors between them.
Performance improves immediately when labels are moved next to the relevant parts.
They lose their place while scrolling between text and diagram.
They can explain the concept after manually annotating corresponding elements.
A learner with strong prior knowledge is less affected because they no longer need the separated explanation.
An integrated format helps on complex material but adds little on a simple, already familiar representation.

These signs are not proof. Weak reading comprehension, unfamiliar notation, poor visual design, attention drift and conceptual misunderstanding can create the same outward behaviour.

Discrimination Test 1: Keep Content Constant, Change Only Layout

Present the same words and picture once in a separated format and once with corresponding elements integrated.

If the integrated version improves explanation without adding information, spatial search is a plausible contributor.

Discrimination Test 2: Does the Learner Need Both Sources?

If the diagram is already self-explanatory or the text is redundant, moving them together may not help. Contiguity matters most when the sources are mutually dependent.

Discrimination Test 3: Search Cost or Concept Gap?

Point directly to the relevant pair. If the learner still cannot explain the relation, the weak link is not merely spatial search.

Discrimination Test 4: Integrated Support or Permanent Dependence?

After learning with integrated material, change the layout. Can the learner reconstruct which elements correspond?

If not, the layout has become a cue that still needs fading.

The MindOS Spatial-Contiguity Protocol

Step 1 — Name the Required Integration

Which two or more elements must the learner combine to understand the idea?

Step 2 — Observe the Search Route

Where does the learner look first, how many times do they scan back and forth, and where do matching errors occur?

Step 3 — Integrate Corresponding Elements

Move the short label beside the diagram part, the explanation beside the equation step, or the unit beside the table column. Do not duplicate entire paragraphs unnecessarily.

Step 4 — Require the Learner to State the Relation

The learner should explain what the two elements contribute together.

Step 5 — Reduce the Support

Move to fewer labels, shorter annotations or a less integrated but still manageable format.

Step 6 — Change the Layout

Use a new textbook, rotated diagram or different placement. The learner must locate correspondences rather than memorise coordinates.

Step 7 — Reconstruct Independently

Give the learner an unlabelled representation and ask them to add the necessary annotations themselves.

Step 8 — Delay and Return

Several days later, present the relation in a new spatial arrangement and test whether understanding survives.

Worked Example: Science

A plant-transport diagram labels xylem and phloem in a legend below the figure. A novice repeatedly scans between legend and stem cross-section.

Place short labels directly beside the structures during early learning. Ask the learner to explain function and position. Later remove labels and present a different cross-section. The learner must identify the tissues from structure and function rather than from remembered label location.

Worked Example: Mathematics

A geometry solution lists reasons in a separate column far from the marked diagram. The learner repeatedly loses which reason belongs to which angle relation.

Temporarily place brief reasons beside the relevant marks. Once the learner understands the correspondence, return to a conventional exam layout and require them to generate the reasons independently.

Worked Example: English

A model paragraph is shown on one page and a commentary key on another. A novice spends more time matching sentence numbers than understanding how evidence becomes inference.

Integrate short commentary beside the sentences for the first examples, then remove it and ask the learner to annotate a new paragraph.

How Do We Know?

Schroeder and Cenkci’s 2018 meta-analysis examined 58 independent comparisons involving 2,426 participants and found an overall benefit of spatially integrated designs of about g = 0.63. Earlier meta-analytic work by Ginns also found strong benefits of reducing spatial and temporal split attention, particularly for complex learning material.

However, newer synthesis adds an important boundary. Cromley and Chen’s 2025 meta-analysis of Richard Mayer’s multimedia-learning corpus found a small, non-significant average contiguity effect within that particular corpus (g about 0.13), while noting that broader meta-analyses outside Mayer’s own studies report substantially larger effects. The authors suggest media type may help explain the discrepancy: static text-plus-diagram materials often show stronger contiguity effects than animation-heavy contexts.

That disagreement is useful. It means spatial integration should be diagnosed from the actual search-and-integration problem rather than applied as a universal decoration rule.

Evidence Boundary

Spatial integration is strongly supported in several meta-analyses, but effect sizes vary across research corpora and media types.
Contiguity matters most when the learner must integrate mutually dependent sources.
Simple or familiar material may benefit less.
Integration can create clutter if too much text is embedded directly in a figure.
Spatial proximity does not repair an incorrect or incomprehensible explanation.
Supported performance in an integrated layout is not proof that the learner can handle conventional separated formats later.

Common Misconceptions

“Everything should be next to everything.” Only mutually dependent information needs integration.
“If the page looks cleaner, it must teach better.” Aesthetic cleanliness and cognitive integration are different goals.
“Integrated diagrams remove the need to learn labels.” The layout is a scaffold, not the final knowledge state.
“Split attention means the learner has poor attention.” It describes a design problem, not a clinical or personality diagnosis.

AI and Technology Boundary

AI can reformat notes, place annotations beside figures and create integrated study sheets. That can reduce search cost, but the technology may also perform the correspondence mapping for the learner.

Ask: who identified which text belongs with which element?

Early support can be useful. Then give the learner an unannotated figure and require them to place the labels and explanations independently.

Staged Practice and Scaffold Fade

fully integrated expert layout;
learner explains each correspondence;
fewer embedded labels;
short legend plus nearby markers;
conventional separated format;
new layout;
learner creates their own integrated annotation;
learner reconstructs without layout support.

Transfer Test

Give the learner a fresh diagram whose explanatory text is separated. Ask them to identify which elements must be mentally integrated and to annotate only those relationships. Transfer is present when they can diagnose split attention themselves rather than needing a redesigned page.

Delayed Independent Return

Several days later, change the layout completely. The learner should still identify the correspondence, explain the relationship and reconstruct missing labels without returning to the original integrated sheet.

Examination Implication

Examination papers may use legends, separate data tables or unfamiliar figure layouts. Final preparation should therefore move beyond beautifully integrated notes and include conventional exam-style representations where the learner must perform the mapping independently.

Parent and Tutor Teaching Guide

“Which two things are you trying to connect?”
“Are you spending more effort searching than thinking?”
“Would moving this label beside the diagram help?”
“What relationship does the new layout reveal?”
“Now I will move the information apart again. Can you still connect it?”
“Can you annotate a fresh diagram yourself?”

MindOS Direction Graph

Mutually dependent sources → separated in space → repeated search/matching errors → integrate corresponding elements → explain relation → reduce layout support → changed arrangement → independent reconstruction → delayed return.

If the learner simply does not know what matters, use Relevance-Filtering or Signaling. If the diagram itself is not understood, use Representation State. If too many elements remain unstable even after integration, route to Working Memory Load.

MindOS rule: when two pieces of information must become one idea, do not waste the learner’s working memory making them hunt across the page first.