Chapter map

1. 1. Objects
2. 2. Distances
3. 3. Squared matrix
4. 4. Centering matrix
5. 5. Row means
6. 6. Grand mean
7. 7. First Gram entry
8. 8. Full Gram matrix
9. 9. Center check
10. 10. Eigenvalues
11. 11. Leading coordinate
12. 12. Reconstruction
13. 13. Translation
14. 14. Negative eigenvalue
15. 15. Dimension choice
16. 16. Claim

Visit the How Mathematics Works hub for the wider series and its connections to education, reasoning, science, computing, design and careers.

---

1. Objects

This is where modelling choices begin to affect evidence. Use points A=0,B=1,C=3.

A concrete situation

Three objects create three distances.

What the mathematics does not license

Object order must match rows and columns. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Record IDs. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

2. Distances

The central mathematical idea is concrete. Compute AB=1,BC=2,AC=3.

A concrete situation

The symmetric distance matrix has zero diagonal.

What the mathematics does not license

Use one declared metric. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Record D. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

3. Squared matrix

Here the geometry becomes operational. Square every entry.

A concrete situation

D²=[[0,1,9],[1,0,4],[9,4,0]].

What the mathematics does not license

Square values, not matrix-multiply D. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Record D². Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

4. Centering matrix

This step deserves a careful pause. Form J=I-11ᵀ/3.

A concrete situation

Diagonal entries are 2/3; off-diagonals -1/3.

What the mathematics does not license

Use n=3 exactly. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Record J. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

5. Row means

The useful question is not only how to calculate, but what the calculation assumes. Average D² rows.

A concrete situation

Means are 10/3,5/3,13/3.

What the mathematics does not license

Keep fractions to avoid drift. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Record means. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

6. Grand mean

A small example makes the mechanism visible. Average all nine D² entries.

A concrete situation

Grand mean=28/9.

What the mathematics does not license

Include symmetric entries and diagonal consistently. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Record mean. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

7. First Gram entry

This is where modelling choices begin to affect evidence. Compute -.5(0-10/3-10/3+28/9).

A concrete situation

B11=16/9.

What the mathematics does not license

Follow double-centering signs. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Show step. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

8. Full Gram matrix

The central mathematical idea is concrete. Apply the rule to all cells.

A concrete situation

B=[[16,4,-20],[4,1,-5],[-20,-5,25]]/9.

What the mathematics does not license

Matrix must be symmetric. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Record B. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

9. Center check

Here the geometry becomes operational. Sum each row and column.

A concrete situation

Every sum is zero.

What the mathematics does not license

Failure signals a centering error. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Verify. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

10. Eigenvalues

This step deserves a careful pause. Eigendecompose B.

A concrete situation

Eigenvalues are 14/3,0,0 up to roundoff.

What the mathematics does not license

Sort descending. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Record spectrum. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

11. Leading coordinate

The useful question is not only how to calculate, but what the calculation assumes. Scale the leading eigenvector by sqrt(14/3).

A concrete situation

Coordinates equal [-4/3,-1/3,5/3] up to reflection.

What the mathematics does not license

Eigenvector signs are arbitrary. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Record coordinates. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

12. Reconstruction

A small example makes the mechanism visible. Calculate coordinate distances.

A concrete situation

Distances return 1,2,3.

What the mathematics does not license

Compare every pair. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Audit. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

13. Translation

This is where modelling choices begin to affect evidence. Add ten to all reconstructed coordinates.

A concrete situation

Distances remain unchanged.

What the mathematics does not license

MDS centres one representative. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Explain invariance. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

14. Negative eigenvalue

The central mathematical idea is concrete. Perturb a distance inconsistently.

A concrete situation

B can become indefinite.

What the mathematics does not license

Do not silently discard without reporting. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Inspect spectrum. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

15. Dimension choice

Here the geometry becomes operational. Retain one positive coordinate.

A concrete situation

The example is exactly one-dimensional.

What the mathematics does not license

Real data need reconstruction and task checks. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Plot error. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

16. Claim

This step deserves a careful pause. Name metric, centering and spectrum.

A concrete situation

One dimension exactly reconstructs the teaching distances.

What the mathematics does not license

Do not infer latent truth. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Bound conclusion. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

A seven-day practice lab

1. Day 1: define the decision, observation unit, X variables and Y or reference condition.
2. Day 2: reproduce the smallest worked example and check every dimension, unit and sign.
3. Day 3: build a transparent baseline and write what it assumes.
4. Day 4: fit the main method inside a training-only pipeline; record every parameter.
5. Day 5: challenge one assumption with a deliberately difficult dataset or operating regime.
6. Day 6: compare validation metrics with the practical workload created by errors.
7. Day 7: present one supported conclusion, one unresolved uncertainty and one safe next experiment.

Questions for a student, parent or mentor

• Which part of the model is a mathematical definition, and which part is a choice?
• What small calculation would catch a sign, scale or dimension error?
• What evidence comes from training, validation and the untouched test?
• What benign condition could look like a fault or prediction failure?
• How would the answer change for another response, asset, school project or time period?
• What should a human review before acting on the score?

A method-specific audit before deployment

Classical multidimensional scaling reconstructs centred coordinates from a matrix of pairwise Euclidean distances. Square the distances, double-centre the matrix with J=I-11ᵀ/n, form B=-1/2 JD²J, and eigendecompose B. Positive eigenvalues and their eigenvectors provide coordinates whose Euclidean distances approximate the input.

Place three points on a line at 0,1,3. Their squared-distance matrix is [[0,1,9],[1,0,4],[9,4,0]]. Double centring gives B=[[16/9,4/9,-20/9],[4/9,1/9,-5/9],[-20/9,-5/9,25/9]]. Every row and column sums to zero, as a centred Gram matrix should.

B has one positive eigenvalue 14/3 and two zero eigenvalues apart from numerical roundoff. Multiplying the leading unit eigenvector by sqrt(14/3) gives coordinates equivalent, up to reflection, to [-4/3,-1/3,5/3]. Their pairwise distances are again 1,2,3, so one dimension reconstructs the example exactly.

Negative eigenvalues reveal that a supplied dissimilarity matrix is not exactly Euclidean under the stated treatment, but they do not diagnose the cause. Noise, rounding, non-Euclidean metrics, missing-data procedures and transformations can all matter. Report distance definition, centring, eigenvalues, retained dimensions, negative-eigenvalue handling and reconstruction checks.

---

Frequently asked questions

What is classical multidimensional scaling?

A spectral method that reconstructs centred Euclidean coordinates from pairwise distances through double centring and eigendecomposition.

Why square the distances?

Squared Euclidean distances have a linear relationship with the centred Gram matrix.

What do positive eigenvalues provide?

Coordinate axes and squared magnitudes for the Euclidean representation.

What do negative eigenvalues mean?

The supplied dissimilarities are not exactly Euclidean under the stated treatment, though the cause requires investigation.

Is the coordinate orientation unique?

No. Translation is removed and rotations or reflections preserve distances.

What must be reported?

Objects, distance definition, transformations, centring, eigenvalues, dimension rule, negative-eigenvalue treatment and reconstruction error.

---

Primary source and fact-check note

Warren S. Torgerson, “Multidimensional Scaling: I. Theory and Method,” Psychometrika 17(4), 401–419 (1952), DOI 10.1007/BF02288916. Definitions and core claims in this article were checked against this primary publication. Calculations shown here were independently recomputed. Examples are educational illustrations; they are not reported industrial performance results.

Useful next reading

• Kruskal Stress-1
• Shepard diagrams
• Isomap residual variance
• How AI reduces dimensions with PCA
• Vector spaces in AI

Keep the options open. Mathematics education supports clearer questions, transferable problem-solving skills and informed choices across science, engineering, computing, AI, statistics, design and many other careers. It does not guarantee intelligence, income, admission or employment; those outcomes depend on far more than one subject or model.

Extended laboratory: configuration evidence and responsible use

Distance audit

Verify symmetry, zero diagonal, units, missing entries and whether the dissimilarity is intended to be Euclidean.

Centring audit

Recalculate J and B, then check symmetry and zero row and column sums before eigendecomposition.

Spectrum audit

Publish positive, near-zero and negative eigenvalues with a numerical tolerance and retained-dimension rule.

Reconstruction audit

Rebuild pair distances from coordinates and localise residuals instead of reporting eigenvalues alone.

Mechanism workshop: calculate, challenge and transfer

Manual reconstruction

Rebuild the example with identifiers and record distance matrix, squared distances, centring matrix, Gram matrix, eigenvalues, coordinates and reconstruction residuals. Explain every transformation before software.

One-entry experiment

Change one distance, coordinate, loading or row while holding the comparison rule fixed. Recalculate every affected matrix cell and score.

Scaling experiment

Apply uniform and unequal scaling. State which changes are removed by the method and which alter the evidence.

Sign or reflection experiment

Reverse an axis or mirror a configuration. Apply the declared invariance policy and explain the scientific consequence.

Correspondence experiment

Swap two object, landmark or variable labels. Show why arithmetic can remain tidy while the comparison becomes meaningless.

Stability experiment

Repeat samples, starts, rotations, factor matches or preprocessing choices and report a distribution.

Parent explanation

Explain why two maps can look different yet match after alignment, or look similar while their distance or loading structure differs.

Student transfer

Apply the mechanism to school maps, sensor tables, questionnaire factors or image embeddings using real identifiers.

Limit statement

Write one supported claim, one overclaim and one repaired claim naming preprocessing, invariances and uncertainty.

Next measurement

Choose a complementary local, global, predictive or substantive check and explain its separate question.

Evidence notebook: reconstruct every mathematical step

Evidence notebook 1: Objects

Notebook step 1 isolates objects. Use points A=0,B=1,C=3. The reproducible result is: Three objects create three distances. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Object order must match rows and columns. Record IDs. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 2: Distances

Notebook step 2 isolates distances. Compute AB=1,BC=2,AC=3. The reproducible result is: The symmetric distance matrix has zero diagonal. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Use one declared metric. Record D. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 3: Squared matrix

Notebook step 3 isolates squared matrix. Square every entry. The reproducible result is: D²=[[0,1,9],[1,0,4],[9,4,0]]. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Square values, not matrix-multiply D. Record D². Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 4: Centering matrix

Notebook step 4 isolates centering matrix. Form J=I-11ᵀ/3. The reproducible result is: Diagonal entries are 2/3; off-diagonals -1/3. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Use n=3 exactly. Record J. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 5: Row means

Notebook step 5 isolates row means. Average D² rows. The reproducible result is: Means are 10/3,5/3,13/3. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Keep fractions to avoid drift. Record means. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 6: Grand mean

Notebook step 6 isolates grand mean. Average all nine D² entries. The reproducible result is: Grand mean=28/9. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Include symmetric entries and diagonal consistently. Record mean. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 7: First Gram entry

Notebook step 7 isolates first gram entry. Compute -.5(0-10/3-10/3+28/9). The reproducible result is: B11=16/9. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Follow double-centering signs. Show step. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 8: Full Gram matrix

Notebook step 8 isolates full gram matrix. Apply the rule to all cells. The reproducible result is: B=[[16,4,-20],[4,1,-5],[-20,-5,25]]/9. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Matrix must be symmetric. Record B. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 9: Center check

Notebook step 9 isolates center check. Sum each row and column. The reproducible result is: Every sum is zero. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Failure signals a centering error. Verify. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 10: Eigenvalues

Notebook step 10 isolates eigenvalues. Eigendecompose B. The reproducible result is: Eigenvalues are 14/3,0,0 up to roundoff. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Sort descending. Record spectrum. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 11: Leading coordinate

Notebook step 11 isolates leading coordinate. Scale the leading eigenvector by sqrt(14/3). The reproducible result is: Coordinates equal [-4/3,-1/3,5/3] up to reflection. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Eigenvector signs are arbitrary. Record coordinates. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 12: Reconstruction

Notebook step 12 isolates reconstruction. Calculate coordinate distances. The reproducible result is: Distances return 1,2,3. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Compare every pair. Audit. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 13: Translation

Notebook step 13 isolates translation. Add ten to all reconstructed coordinates. The reproducible result is: Distances remain unchanged. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: MDS centres one representative. Explain invariance. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 14: Negative eigenvalue

Notebook step 14 isolates negative eigenvalue. Perturb a distance inconsistently. The reproducible result is: B can become indefinite. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Do not silently discard without reporting. Inspect spectrum. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 15: Dimension choice

Notebook step 15 isolates dimension choice. Retain one positive coordinate. The reproducible result is: The example is exactly one-dimensional. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Real data need reconstruction and task checks. Plot error. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Evidence notebook 16: Claim

Notebook step 16 isolates claim. Name metric, centering and spectrum. The reproducible result is: One dimension exactly reconstructs the teaching distances. The evidence concerns recovering Euclidean coordinate structure from pairwise distances. Keep object, landmark, variable or row identities beside every value. The safeguard is: Do not infer latent truth. Bound conclusion. Change one input while holding centring, scaling, matching and formula conventions fixed. Recalculate the affected distance, Gram entry, eigenvalue, transform, residual, dot product, trace or coefficient. Explain whether the movement came from data, preprocessing, correspondence or model flexibility. If another reader cannot reconstruct the value, it is not ready for a claim. Finish with the narrow conclusion and the extra evidence still required.

Failure cases and responsible recovery

Non-Euclidean distances

A negative eigenvalue appears. Check metric construction, missing data and rounding before choosing a correction or approximation. Recalculate a small counterexample, repeat under another defensible setting and state the extra evidence needed to decide.

Squared-matrix mistake

D² means elementwise squared distances in this derivation, not D multiplied by D. A small notation error changes every later result. Recalculate a small counterexample, repeat under another defensible setting and state the extra evidence needed to decide.

Hidden order mismatch

Row labels and columns refer to different object orders. Symmetry or reconstruction checks can expose the join error. Recalculate a small counterexample, repeat under another defensible setting and state the extra evidence needed to decide.

Dimension cherry-pick

A convenient two-dimensional display ignores material eigenvalues. Publish the spectrum and reconstruction trade-off. Recalculate a small counterexample, repeat under another defensible setting and state the extra evidence needed to decide.

Negative eigenvalues discarded

Dropping them produces a Euclidean approximation, not proof the original dissimilarities were Euclidean. Recalculate a small counterexample, repeat under another defensible setting and state the extra evidence needed to decide.

Axis storytelling

Eigenvector orientation can rotate or reflect without changing distances. Interpret stable relational structure, not arbitrary left-right placement. Recalculate a small counterexample, repeat under another defensible setting and state the extra evidence needed to decide.

Out-of-sample use

Classical MDS coordinates the supplied objects. Define and validate an extension before placing new cases. Recalculate a small counterexample, repeat under another defensible setting and state the extra evidence needed to decide.

Responsible conclusion

State reconstruction quality under a declared metric and stop before claims of causal or psychological truth. Recalculate a small counterexample, repeat under another defensible setting and state the extra evidence needed to decide.