Variation Theory

Variation Theory of Learning

This skill applies the variation theory of learning, developed by Ference Marton and colleagues, to structure content for optimal learning.

Reference Files

This skill includes detailed references in references/:

Core Theoretical Principles

The Object of Learning

The object of learning is what is to be learned. Understanding develops when learners discern the critical aspects of the object of learning.

Marton's central principle: "to learn something, the learner must discern what is to be learned. Discerning the object of learning amounts to discerning its critical aspects."

Variation and Invariance

The necessary condition for discernment: learners must experience variation in a dimension corresponding to that aspect, against the background of invariance in other aspects.

Key insight: "When some aspect of a phenomenon varies while another aspect remains invariant, the varying aspect will be discerned."

Critical Aspects as Focus of Variation

FUNDAMENTAL PRINCIPLE: Variation must occur in the critical aspects of the object of learning. Arbitrary variation does not lead to learning.

Identifying Critical Aspects First

Before designing any pattern of variation:

1. Define the object of learning - What should learners understand/do?
2. Identify the critical aspects - Which features must be discerned?
3. Design variation IN those aspects - Create patterns varying the critical dimensions

Dimensions vs Values

The principle "one thing at a time" applies to dimensions (aspects), NOT to values within a dimension.

• Dimension (aspect): A category/feature type (e.g., "punctuation marks", "file mode")
• Values/features: Specific instances (e.g., punctuation marks like period, question mark, exclamation mark; or file modes like r, w, a)

The principle:

• Vary ONE dimension at a time - to separate aspects from each other
• Contrast MULTIPLE values together - within that dimension

Research evidence: Teaching punctuation marks separately: 15% improvement. Teaching all three together with contrast: 63% improvement.

Common Mistake: Varying Non-Critical Aspects

Anti-pattern: Creating variation in aspects irrelevant to the learning objective.

Example: Teaching "why files need open/close":

• Wrong: Vary filename or content (not critical)
• Right: Vary what happens when close() is/isn't called (critical aspect is resource management)

Applying Contrast in Documentation Examples

Variation theory is not limited to classroom teaching — it applies whenever a reader must discern differences between alternatives. Documentation, READMEs, and literate programs routinely show multiple ways to achieve the same goal. The contrast pattern ensures the reader's attention falls on the method (the critical aspect) rather than on incidental details.

Principle: Keep the Problem Invariant, Vary the Method

When showing alternative approaches to the same task, use the same concrete names, filenames, and identifiers across all examples. If the names change between examples, the reader must determine whether the name change is meaningful — this distracts from the actual difference (the method).

Where the variation analysis goes: notes vs. body

Decide with one question first: is the \enquote{what varies / what stays invariant} analysis part of the object of learning for this document's reader, or commentary on how the document pursues that object of learning? Getting this backwards is a common mistake — do not reach for \ltnote reflexively.

• The document teaches a subject (a tutorial the reader works through, a lecture, README, literate program, crypto notes). The reader's object of learning is the subject, not the pedagogy. The variation analysis is meta-commentary about how the material is built, so it goes in \ltnote{...} (via the didactic-notes skill). Keep the visible prose focused on the example itself and the feature the reader should use. This is the common case.

• The teaching design is the subject (a research paper, a methods or pedagogy writeup, an article analysing how to teach against a misconception — or a literate source whose tangled output is the learner-facing tutorial, so the woven prose documents that tutorial for educators/maintainers). Here the reader's object of learning is the teaching design itself, and the variation analysis is the content — the very finding or rationale the document reports — so it belongs in the body prose. Putting it in \ltnote hides the document's own argument. \ltnote is then reserved for reasoning about the document's own construction: why a section is ordered this way, why this example was chosen, where a topic is placed. In the literate-tutorial case the layers form a chain — the tangled tutorial teaches the student, the woven prose teaches the teacher the tutorial's design, and \ltnote teaches the future author how the prose targets the teacher's learning.

This is the same scope test the didactic-notes skill states: if the note would still make sense to a reader who only cares how the thing being described works, it is body text. When the thing being described is the teaching design, the variation analysis is body text.

Common case — the document teaches a subject

Bad --- student-facing pedagogical meta-commentary:

These two examples form a deliberate contrast.

\begin{description}
\item[What varies] Whether [[show_url]] is set to [[False]].
\item[What stays invariant] The helper, question, and environment.
\end{description}

Good --- instructor-facing pedagogical annotation:

\ltnote{%
  \textbf{Variation pattern}: Contrast
  \textbf{What varies}: Whether [[show_url]] is set to [[False]].
  \textbf{What stays invariant}: The helper, question, and environment.
}

Carve-out — the document is about the teaching design

In a research paper whose subject is the variation-theory designs themselves, state the analysis in the body and keep \ltnote for authoring decisions only.

Good --- the analysis is body prose (it is the finding):

We keep the program invariant and vary only the truth value of the condition,
so the reader discerns that a false \mintinline{python}{if} does not terminate
the program and the \mintinline{python}{else} does not also run.

Good --- \ltnote carries a decision about the paper, not the design:

\ltnote{Placed in the types chapter rather than functions/variables because
students first meet it as a property of the string type.}

Example: CLI Documentation with Two Delivery Methods

A tool offers two ways to share a log: push/clone via Git, or export/play via a bundle file. The object of learning is "how to share a log"; the critical aspect is "which delivery method to use."

BAD — names vary alongside the method, obscuring what actually changed:

# Method 1: push to remote
learnlog set-remote git@gitlab.kth.se:dbosk/demo-log.git
learnlog push

# Method 2: export as bundle
learnlog export -o lecture01.bundle
learnlog play lecture01.bundle

The reader sees demo-log vs lecture01 and wonders: does the name matter? Is a bundle different from a repository? The irrelevant variation in the name competes with the relevant variation in the method.

GOOD — the name is invariant, only the method varies:

# Method 1: push to remote
learnlog set-remote git@gitlab.kth.se:dbosk/lecture01.git
learnlog push
...
learnlog clone git@gitlab.kth.se:dbosk/lecture01.git
learnlog play

# Method 2: export as bundle
learnlog export -o lecture01.bundle
...
learnlog play lecture01.bundle

Now lecture01 is invariant across both examples. The only thing that changes is the delivery mechanism (push/clone vs export/play), which is exactly the critical aspect the reader should discern.

When to Apply This

Activate this pattern whenever you encounter:

• Multiple code examples showing alternative approaches
• CLI usage sections with different flags or subcommands
• API examples with different authentication methods
• Configuration examples with different backends
• Any documentation where "you can also do X instead of Y"

Self-test: If two examples differ in more ways than the one dimension you intend to contrast, eliminate the incidental differences.

Mentipy for Live Variation

Mentipy can operationalize variation patterns in live teaching. Use a short sequence of mc, scale, open_text, or word_cloud questions to vary one critical aspect at a time while keeping the object of learning invariant. This is especially useful for diagnostic contrast before an explanation and for checking whether students now discern the critical aspect afterward.

The Three Patterns of Variation

1. Contrast

Purpose: Help learners recognize that an aspect exists by experiencing what it is versus what it is not.

How it works: Present examples that differ in one critical aspect while keeping all other factors constant.

Example: To understand "height," show two objects identical in all respects except height.

Note: Contrast achieves separation—the critical aspect becomes discernible through experiencing variation.

2. Generalization

Purpose: Help learners recognize that a pattern or principle holds across different contexts.

How it works: Present the same critical value in varied appearances. Keep the critical aspect invariant while varying other (non-critical) aspects.

Example: Show the same geometric principle applied to triangles, rectangles, circles.

3. Fusion

Purpose: Enable learners to experience multiple critical aspects simultaneously as an integrated whole.

How it works: Vary several critical aspects at once so learners must attend to their simultaneous interrelationships.

Example: In understanding circuits, vary resistance and voltage simultaneously.

Pedagogical Sequence

Research suggests using patterns in this order:

1. Contrast - Vary the critical aspect while keeping other aspects invariant. This separates (makes discernible) the critical aspect.
2. Generalization - Keep the critical value invariant while varying other aspects. Shows the pattern holds across contexts.
3. Fusion - Vary multiple critical aspects simultaneously. Enables learners to experience interrelationships.

Important: Within each pattern, contrast multiple values together. "One at a time" applies to dimensions/aspects, not values.

Why contrast must come first (verified in Marton, Necessary Conditions of Learning, 2014, pp. 45–47 — cite these pages when writing this claim):

• Generalization before contrast is induction, and induction cannot open a dimension the learner has never discerned: "if you cannot see the greenness of any of three green, though otherwise different, things, you cannot see what they have in common, either" (p. 46).
• Of two teachers juxtaposing sets of objects to make greenness discernible, "[o]ne is betting on induction, the other on contrast. The latter wins" (p. 47).
• Discernment comes from "the simultaneous experience of things that differ in color but are the same otherwise" (p. 46) — so juxtapose the contrasting instances side by side (in one figure, one code listing pair, one slide). Instances presented apart must be juxtaposed in the learner's memory instead: a cognitive-load argument for contrast-first, side-by-side design.

Temporal Sequencing: Examples Before Generalizations

CRITICAL: Examples must precede generalizations. Students need concrete instances creating necessary variation before abstract principles become meaningful.

Why: Variation must be experienced before invariants can be discerned. When you state a general principle first, students have no variation pattern to map it onto.

Anti-pattern:

% BAD: Generalization before examples
Filer behövs för persistens, datautbyte, och skalbarhet.

\begin{example}[Spara spelets progress]
  ...
\end{example}

Good pattern:

% GOOD: Examples create variation, then generalize
\begin{example}[Spara spelets progress]
  Ett spel behöver komma ihåg spelarens poäng...
\end{example}

\begin{example}[Dela data mellan program]
  Ett program genererar data som ett annat använder...
\end{example}

\begin{remark}[Varför filer behövs]
  Filer behövs för persistens, datautbyte, och skalbarhet.
\end{remark}

See references/common-violations.md for detailed violation types and fixes.

Common Generalization Violations

See references/common-violations.md for detailed examples.

Violation types:

1. Generic/placeholder code before concrete examples
2. Block/remark environments before examples
3. Incomplete skeletons before complete solutions
4. Explanatory principles before demonstrating examples

Fix pattern for each violation:

1. Remove the generalization from its current position
2. Ensure 2-3 concrete examples exist creating necessary variation
3. Add back the generalization AFTER examples
4. Verify students can now discern the pattern from the variation

Side-by-Side Contrast

See references/latex-examples.md for details.

For Beamer presentations, use mode splits:

\begin{frame}
  \mode<presentation>{%
    \textbytext{%
      \begin{definition}[Primärminne]
        Flyktigt minne med snabb åtkomst.
      \end{definition}
    }{%
      \begin{definition}[Sekundärminne]
        Oflyktigt minne, långsammare.
      \end{definition}
    }
  }
  \mode<article>{%
    \textbytext*{%
      ...
    }{%
      ...
    }
  }
\end{frame}

Generalizations in Semantic Environments

Capture generalizations in semantic environments AFTER examples:

Language Consistency

CRITICAL: When documenting variation theory in notes (e.g., \ltnote), match the document's instructional language.

Swedish terminology:

• "Variation Pattern" → "Variationsmönster"
• "Contrast" → "Kontrast"
• "What varies" → "Vad som varierar"
• "What remains invariant" → "Vad som hålls invariant"
• "Critical aspects" → "Kritiska aspekter"

When Applying This Skill

1. Identify the object of learning (what should be understood)
2. Determine the critical aspects (what must be discerned)
3. Structure content using the three patterns to create necessary conditions
4. Remember: "there is no discernment without variation"

Connection to Try-First-Tell-Later

The try-first-tell-later skill complements variation theory: use try-first prompts to diagnose which critical aspects students can already discern, then design variation patterns to teach aspects they cannot yet see.

Key References

• Marton, F. (2015). Necessary Conditions of Learning. Routledge. (Primary reference)
• Marton, F., & Booth, S. (1997). Learning and Awareness. Lawrence Erlbaum.
• Marton, F., & Pang, M. F. (2006). On Some Necessary Conditions of Learning. Journal of the Learning Sciences, 15(2), 193-220.
• Marton, F., & Tsui, A. (2004). Classroom discourse and the space of learning. Lawrence Erlbaum.