Blog / Why Traditional Forced-Choice Rankings Create Ipsative Profiles—and When Thurstonian IRT Changes That

Why Traditional Forced-Choice Rankings Create Ipsative Profiles—and When Thurstonian IRT Changes That

Separate the response format, the traditional scoring rule, and a modern comparative-judgment model before interpreting a forced-choice profile.

October 8, 2026 · 6 min read · By Cepingku Editorial Team

A forced-choice item asks a respondent to compare statements in one block: perhaps rank them from most to least like me, or choose the most and least like me. That is a response format, not a guarantee about how a score should be interpreted. A problem arises with some traditional scoring rules. They distribute a fixed number of points among different dimensions in every block. Add the blocks and every respondent receives the same total. Brown and Maydeu-Olivares call these ipsative data in their 2012 Psychological Methods paper: traditional scoring of multidimensional forced-choice questionnaires can give all individuals a common total test score.

It would therefore be inaccurate to say that every forced-choice questionnaire is unreliable or that no forced-choice result can ever compare people. Ask three separate questions instead. What comparison does the response format require? Which scoring rule turns that response into a score? Does that rule model comparative judgement appropriately? Brown and Maydeu-Olivares argue that the ipsative problems come from inadequate traditional scoring, and that suitable Thurstonian IRT modelling can produce scores free from those ipsative problems.

A fictional three-dimension ranking has a fixed total

This is an invented teaching block, not a DISC item or this site’s scoring method. Three statements represent dimensions A, B, and C. Every respondent must rank all three. Traditional scoring awards 2 points to first, 1 to second, and 0 to third. Whatever the ranking, 2 + 1 + 0 = 3. On PDF pages 7–8, Brown and Maydeu-Olivares explain this inverse-rank scoring: a fixed number of points is distributed in each block, so the total across blocks is the same for every respondent.

Fictional rankingTraditional scores and total
Person A: A first, B second, C thirdA=2, B=1, C=0; total=3.
Person B: C first, B second, A thirdA=0, B=1, C=2; total=3.

Person A’s profile can describe a relative ordering within that block: A above B above C. Person B’s describes C above B above A. Both totals equal three, but that does not mean they have the same absolute standing on all three dimensions. Nor does A’s score of two establish that Person A is absolutely higher on A than Person B. Under this constraint, awarding more points to one dimension necessarily awards fewer elsewhere, even when people’s absolute trait levels could differ greatly.

A constant total changes the question a profile can answer

With a fixed total, a profile cannot show that every dimension is high or that every dimension is low. A high point first means that the dimension was ranked ahead of the others in its block; it is not an independent measure of an absolute level. The paper defines ipsative data as scores whose sum is constant for any individual, and explains that this restricts cross-person interpretation of absolute, or normative, trait standing.

Within-person comparison can still be a limited description. If Person A repeatedly ranks A before B and C in comparable blocks, a careful statement is that A was relatively more often preferred in those comparisons. It does not follow that A is absolutely high, or that Person A must exceed Person B on A. A cross-person claim needs scoring and validation evidence for the particular instrument that is not constrained by a constant total.

Thurstonian IRT changes the model and scoring

Thurstonian IRT is not simply a new set of numbers substituted for 2, 1, and 0. It models the comparative-judgement response process and estimates latent trait locations using item parameters. The paper’s abstract and PDF page 7 describe how appropriate Thurstonian IRT modelling can overcome the ipsative problems of traditional scoring and permit comparison between individuals. Whether it succeeds for a specific questionnaire still depends on the item design, model fit, parameter estimation, and validation. The word IRT alone is not proof that any questionnaire is sound.

The research cannot be transferred to this site. This article does not state whether the DISC Test uses forced-choice items, traditional scoring, or Thurstonian IRT. The paper also supplies no reliability, validity, norms, or basis for cross-person comparison for this assessment.

Read a behaviour-dimension report within its documented scope

First confirm the actual format, version, scoring explanation, and stated scope of a behaviour-dimension report. If it uses traditional ranking scores, the careful question is, ‘Which dimensions are relatively higher within this person’s profile?’ To compare two people, establish that the particular tool has an applicable cross-person scoring model and validation evidence. The DISC Test can prompt reflection on responses to behaviour descriptions. Do not treat this fictional 2/1/0 block, an external IRT paper, or forced-choice theory as evidence that this site uses a particular format, scoring method, validity standard, or cross-person comparison.

Want to see where you land?

Take the DISC Test and get your results in minutes.

Take the DISC Test →

← Back to the blog