Deep Layer: Comparison Requires Commensurability

The deepest comparison rule is simple: compare like with like.

Two objects do not become comparable merely because they appear in the same question. They need a shared dimension, compatible units, an aligned time frame, a common baseline or a clearly explained reason why the difference in context still permits a meaningful relationship.

This property is called commensurability: the states can be placed on a common frame.

Aisha cannot directly compare 3 minutes with 150 seconds until she puts them in one time unit. Clara cannot compare a national policy’s ten-year effect with a local pilot’s first-month outcome as though the time and scale were identical. Ben cannot compare one writer’s tone with another writer’s argument structure and call it one difference. Ryan cannot compare two methods by number of written lines if one method solves a more general problem than the other.

The comparison operation begins by building the common frame.

The Common-Denominator Question

What single question can I ask of both A and B?

If the answer is “How much did each cost?”, cost is the dimension. If it is “How quickly did each respond?”, speed is the dimension. If it is “How does each writer present authority?”, representation of authority is the dimension.

If no single question applies to both pieces of evidence, the comparison may be mismatched.

Categories, Orders and Quantities

Comparisons can operate at different levels.

• Categorical: same category or different category?
• Ordinal: more, less, stronger, weaker, earlier, later?
• Quantitative: how much more or less?
• Structural: do the parts or relationships operate in the same way?
• Functional: do they perform the same job or produce different effects?

A learner should not claim quantitative precision when only an ordinal relationship is supported. If one source feels “more critical,” the passage may support that ordering without allowing a numerical statement about how much more critical.

Similarly, if one strategy costs $2 million and another costs $1 million, the quantitative relationship is explicit: A costs twice as much as B. Use the strongest comparison the evidence genuinely supports.

Absolute, Relative and Normalised Comparisons

The same numbers can support different relationships depending on baseline.

Suppose A rises from 10 to 20 and B rises from 100 to 110.

• Absolute increase: both rise by 10.
• Relative increase: A rises 100%, B rises 10%.
• Final size: B remains much larger.

Which comparison matters depends on the task.

This is why good quantitative comparison identifies the denominator behind a percentage. “A improved more” can be true under relative change and false under absolute change.

Do not hide the baseline.

Normalisation Makes Unequal Scales Comparable

Sometimes raw totals are unfair because the objects differ in size.

A large school may record more absences than a small school simply because it has more students. Absences per 100 students may be a more useful comparison. A country may emit more carbon in total but less per person. A business may earn more profit in dollars but a lower profit margin.

Normalisation means adjusting for a relevant scale so the comparison answers the real question.

But normalisation itself is a modelling choice. Per-capita measures answer different questions from totals. The learner should not normalise automatically; choose the measure that fits the intended dimension.

Baseline Comparisons

A baseline provides a common reference.

• before versus after;
• treatment versus control;
• actual versus target;
• current year versus reference year;
• method A and B versus the same benchmark;
• two texts compared against the same question or theme.

If A improved by 10 points and B improved by 8, that comparison becomes more meaningful when starting conditions are also known. If A began much lower, the same absolute change may carry different significance.

Baselines prevent context from disappearing.

Indexing: Set One State to 100

At advanced levels, different series can be compared by indexing them to a common base, such as setting the starting value to 100.

This can reveal relative growth even when original magnitudes differ greatly.

The idea is useful conceptually even where formal index numbers are not in the syllabus: sometimes the fairest comparison asks, “How did each change relative to where it began?” rather than “Which ended with the larger raw value?”

Comparison Across Time Requires Aligned Windows

A one-year growth rate and a ten-year cumulative change are not the same type of quantity.

Before comparing time-based outcomes, ask:

• same start date?
• same end date?
• same interval length?
• same stage of the process?
• same seasonal conditions?
• same frequency of measurement?

If the windows differ, either align them or state the limitation explicitly.

Trend Comparison: Level, Direction, Rate and Shape

Comparing two trends requires more than asking which line is higher.

• Level: which has larger values?
• Direction: rising, falling, stable?
• Rate: which changes faster?
• Acceleration: is change speeding up or slowing down?
• Turning points: when does direction change?
• Volatility: which fluctuates more?
• Thresholds: does behaviour change after a point?
• Convergence/divergence: are the series moving closer together or further apart?

A and B can share direction while differing in rate. They can share final value while having different paths. They can have similar averages but different variability.

Whole-pattern comparison is richer than endpoint comparison.

Slope Comparison

When graphs use compatible scales, slope can encode rate of change.

But visual steepness is meaningful only when axes and scales are aligned. A graph with a compressed vertical axis can look flatter than another with identical data plotted on a different scale.

Ryan checks axis units and scale before comparing gradients visually.

Where numerical gradients are available, compare the actual rates rather than page geometry.

Variability Comparison

Two groups can have the same mean and different spread.

Two methods can have similar average output but one is much more variable.

If consistency matters, compare range, standard deviation or another taught measure of variability—not just central tendency.

Again, use only statistical tools appropriate to the course. The deeper principle is that “same average” does not mean “same distribution.”

Distribution Comparison

Advanced comparisons may consider shape, centre, spread and outliers.

A distribution can have a higher median but more extreme low values. Another can have a lower mean but tighter spread. The best summary depends on the question.

Students should avoid reducing a full distribution to one statistic when the task requires broader comparison.

Compare Mechanisms, Not Only Outcomes

Two systems can produce similar outcomes through different mechanisms.

Two policies may both reduce congestion: one by increasing public transport capacity, another by reducing car demand. Two study methods may both improve scores: one through retrieval strengthening, another through better organisation. Two texts may both criticise authority: one through irony, another through direct accusation.

Outcome comparison answers “what happened?” Mechanism comparison answers “how did each get there?”

Advanced answers often need both.

Compare Inputs, Processes and Outputs

A useful systems frame is:

• inputs;
• process;
• outputs;
• feedback;
• constraints.

For two experiments, inputs may differ while the measurement process is the same. For two policies, the target group may be similar while mechanisms differ. For two algorithms, inputs may be identical while process efficiency differs.

This framework helps comparison avoid random fact matching.

Fair Comparison Requires Controlled Conditions

If two methods are tested under different conditions, outcome differences may not come from the methods alone.

Aisha asks whether the comparison is fair:

• same sample?
• same equipment?
• same duration?
• same environmental conditions?
• same measurement?
• same starting state?

If not, qualify the comparison. Differences can be observed without claiming the tested factor caused them.

Confounding Can Make Comparisons Misleading

A confounder is a factor linked to both the groups and the outcome that can distort a simple comparison.

If students choosing Method A were already stronger before the intervention, higher final scores do not by themselves show Method A caused the difference.

If two cities have different transport outcomes but also radically different population density, density may matter to the comparison.

Comparison can be accurate descriptively and weak causally. State the relationship at the strength the evidence supports.

Selection Effects

Groups may differ because of how members entered them.

Students who volunteer for extra tuition may be more motivated. Customers choosing a premium service may have higher incomes. Regions adopting a policy early may differ institutionally from regions adopting it later.

If selection differs, a simple outcome comparison may mix treatment difference with group difference.

This does not make the comparison useless. It changes what can be inferred.

Contextual Comparison

Some objects should not be forced into identical context.

A historical policy in wartime and one in peacetime can still be compared, but context becomes part of the relationship. A tropical city and an arid city can be compared on water management, but climate differences must be built into the analysis. Two literary texts from different genres can be compared, but genre conventions may shape their language choices.

Fair comparison does not mean pretending context is identical. It means making relevant context visible.

Counterfactual Comparison

Sometimes the most important comparison is between what happened and what might have happened otherwise.

This counterfactual comparison can support causal evaluation:

• treatment versus control;
• policy versus no policy;
• actual result versus baseline trend;
• method with feature versus method without feature.

Examinations may not use the word “counterfactual,” but many comparative questions implicitly ask what difference the factor made.

Text Comparison: Match Evidence Function

In English, two quotations can be thematically related and still perform different functions.

Ben asks:

• Do both quotations express the same attitude?
• Do they use similar or different methods?
• Do they appear at similar structural positions?
• Do they address the same object?
• Do they produce similar effects?

A fair language comparison aligns function before technique labels.

Semantic-Field Comparison

Two texts may use different vocabulary fields to frame the same subject.

Writer A may use economic vocabulary—“investment,” “return,” “cost”—while Writer B uses moral vocabulary—“duty,” “fairness,” “responsibility.”

The comparison is not merely that the words differ. It is that each writer constructs a different lens through which the issue is understood.

Register Comparison

Register can be compared through formality, technicality, intimacy, authority and audience relationship.

Ben might write:

Writer A uses technical, institutional language that creates distance and authority, whereas Writer B uses conversational first-person language that creates immediacy.

The dimension is not simply “word choice.” It is writer–audience relationship through register.

Structural Comparison in Texts

Two texts may share themes but organise them differently.

A begins with a personal anecdote before broadening to policy; B begins with statistics before narrowing to an individual case. A delays its judgement; B states it immediately.

Comparison can explain how ordering changes the reader’s route through the argument.

Source Comparison: One Enquiry, Two Pieces of Evidence

Historical or social-science source comparison should begin with the common enquiry.

If the enquiry is public reaction, compare what each source indicates about public reaction. If it is government intention, compare evidence about intention. Do not compare one source’s origin with another source’s content and treat that as one relationship.

Provenance can explain why messages differ after the content comparison has been established.

Compare Causal Factors Without Premature Ranking

When comparing causes, first identify their roles:

• structural versus triggering;
• long-term versus short-term;
• direct versus indirect;
• necessary versus contributory;
• broad versus local;
• independent versus interacting.

This is still comparison.

Only if the question asks “which was more significant?” should the answer move into explicit evaluation and weighting.

Compare Assumptions

Two models can differ less in equations than in assumptions.

One model assumes constant growth; another allows saturation. One assumes independence; another models dependence. One text assumes the reader values efficiency; another assumes fairness matters most.

Comparing assumptions can explain why models or arguments reach different conclusions from similar evidence.

Compare Error Profiles

Methods can have similar average performance and different failure modes.

Method A may be fast but occasionally produce catastrophic errors. Method B may be slower but consistently close. A measuring instrument may be precise but biased; another may be unbiased on average but noisy.

Comparing error profile can be more informative than comparing average outcome alone.

Compare Robustness

A robust method performs reasonably across changing conditions.

A fragile method performs very well only under narrow assumptions.

When relevant, compare how A and B respond to noise, uncertain inputs, different populations or changed parameters.

This dimension often appears in advanced evaluation but can be compared neutrally before ranking.

Compare Efficiency Correctly

Efficiency is output relative to resource use, not simply speed.

Two algorithms can be compared by time and memory. Two study methods can be compared by learning gained per unit time. Two policies can be compared by outcome per dollar.

If one method is faster but wastes more material, “more efficient” depends on which resource the question cares about.

Worked Comparison 1: Science Data

Experiment A reaction time: 60 s, 58 s, 62 s. Experiment B: 45 s, 70 s, 50 s.

Average performance may be compared, but variability matters too.

A shows tightly clustered results around 60 s. B contains a wider spread. If the question asks about consistency, A is more consistent even if B sometimes records a faster time.

The dimension controls which evidence matters.

Worked Comparison 2: Science Processes

Compare diffusion and osmosis at the level appropriate to the syllabus.

Shared dimension: movement driven by concentration differences.

Difference dimension: osmosis specifically concerns water movement across a partially permeable membrane under the taught definition, whereas diffusion is broader particle movement from higher to lower concentration.

The comparison aligns mechanism and conditions rather than presenting two separate definitions.

Worked Comparison 3: Mathematics Functions

Compare y = 2x + 1 and y = x².

On rate of change, the linear function has constant gradient 2, while the quadratic’s gradient changes with x.

On graph shape, the first is a straight line and the second a parabola.

On long-run growth for positive large x, the quadratic eventually increases faster.

Each paragraph stays on one dimension.

Worked Comparison 4: Mathematics Methods

Compare solving a quadratic by factorisation and formula.

Generality: the formula works for any quadratic in the appropriate standard form, while simple factorisation depends on recognisable factors.

Speed: factorisation may be faster when factors are obvious.

Error profile: the formula introduces substitution and sign risks; factorisation introduces factor-selection risks.

These are comparisons. Choosing the “better” method for a particular question would be evaluation.

Worked Comparison 5: English Viewpoints

Writer A argues that remote work improves autonomy. Writer B focuses on isolation.

Common dimension: effect of remote work on employees.

Relationship: A foregrounds control and flexibility, while B foregrounds loss of social connection.

Evidence should then pair quotations or examples that support those contrasting frames.

Worked Comparison 6: English Tone

Passage A uses modal verbs such as “may” and “could,” creating a cautious tone. Passage B uses “must” and “will,” creating greater certainty and urgency.

The dimension is certainty. The evidence is matched grammatical choice. The relationship is more cautious versus more forceful.

Worked Comparison 7: Historical Sources

Source A says the policy restored order; Source B says it suppressed dissent.

Common dimension: interpretation of the policy’s effect.

Relationship: both recognise strong state action, but A frames it as stabilisation while B frames the same action as repression.

Provenance may later explain why the frames differ.

Worked Comparison 8: Geography Places

City A and City B both face flood risk.

Hazard mechanism differs: A’s risk is dominated by river overflow; B’s by intense urban surface runoff.

Management therefore differs: A prioritises river defences and floodplain management; B prioritises drainage, storage and permeability.

Comparison links hazard mechanism to response rather than merely listing two city profiles.

Worked Comparison 9: Economics Policies

Compare a tax and a subsidy designed to influence behaviour.

Mechanism: a tax raises private cost of the discouraged activity, while a subsidy lowers private cost of the encouraged alternative.

Fiscal direction: tax can generate government revenue; subsidy requires expenditure.

Distribution and behavioural response can then be compared separately.

Worked Comparison 10: Business Options

Option A: open an owned outlet. Option B: use a franchise partner.

Capital: A requires more direct investment than B.

Control: A gives greater operational control.

Speed: B may expand faster using partner capital and local capacity.

Comparison keeps the options aligned. Evaluation would decide which dimension matters most for the case.

The Comparison Matrix

For training, use rows as dimensions and columns as A, B and relationship.

Example rows:

• purpose;
• input;
• mechanism;
• output;
• time;
• cost;
• risk;
• evidence quality.

The learner should not fill every row automatically. Select rows relevant to the question. The matrix is a way to keep dimensions aligned before prose.

The Comparison Compression Protocol

For short-answer comparison:

dimension → A state → B state → relationship.

Example:

A has a steeper gradient than B, so A’s value increases more rapidly over the interval.

One sentence can carry the complete comparison.

The Long Comparison Protocol

1. Define the comparison object and scope.
2. Choose Dimension 1; compare A and B directly.
3. Choose Dimension 2; compare A and B directly.
4. Choose Dimension 3 if marks justify it.
5. Explain context differences that qualify fairness.
6. If asked, synthesise the overall pattern or move into evaluation.

The Comparison Under Pressure Protocol

When time is short, protect the relationship words and matched evidence.

Do not write a beautiful description of A while planning to “get to B later.”

Start comparatively:

Both…, but A…, whereas B…

This ensures the command is answered even if the clock tightens.

Thirty Comparison Prompts for Training

1. Compare two numerical data sets.
2. Compare two graph trends.
3. Compare two experimental methods.
4. Compare diffusion and osmosis.
5. Compare two energy-transfer processes.
6. Compare two mathematical solution methods.
7. Compare linear and quadratic functions.
8. Compare exact and approximate methods.
9. Compare two writers’ viewpoints.
10. Compare two writers’ tones.
11. Compare structural choices in two texts.
12. Compare formal and informal register.
13. Compare two historical sources’ messages.
14. Compare two sources’ purposes.
15. Compare two causal factors by role.
16. Compare two cities’ flood risks.
17. Compare two development strategies.
18. Compare two economic policies.
19. Compare two market structures.
20. Compare two business-growth options.
21. Compare two study methods.
22. Compare two samples’ variability.
23. Compare total and per-capita measures.
24. Compare short-run and long-run effects.
25. Compare a model inside and outside its observed range.
26. Compare two explanations’ assumptions.
27. Compare two algorithms’ time and memory demands.
28. Compare two sources against one enquiry.
29. Compare two interventions’ mechanisms without evaluating them.
30. Compare two exam answers for completeness and directness.

The Seven-Day Comparison Repair

• Day 1: common dimensions and AB sentences.
• Day 2: similarities, differences and direction.
• Day 3: absolute, relative and normalised comparisons.
• Day 4: trend, distribution and mechanism comparison.
• Day 5: text/source comparison.
• Day 6: mixed cross-subject timed comparisons.
• Day 7: final-quarter performance and independent audit.

The Twelve-Week Comparison Arc

• Weeks 1–2: one dimension, direct language.
• Weeks 3–4: multiple dimensions and matched evidence.
• Weeks 5–6: quantitative and trend comparison.
• Weeks 7–8: mechanisms, context and fair-comparison limits.
• Weeks 9–10: cross-subject transfer and compare-plus commands.
• Weeks 11–12: full-paper speed, fatigue and independence.

Why Comparison Matters Beyond Examinations

Comparison is how differences become interpretable.

Scientists compare treatment and control. Engineers compare designs. Doctors compare outcomes. Consumers compare products. Investors compare returns and risk. Governments compare policies and countries. Writers compare texts and arguments.

The discipline is always the same: align the dimension before trusting the difference.

Ben eventually stops thinking of comparison as a connective-word exercise.

He sees it as alignment.

A and B enter the same frame. Only then does the difference mean something.