Five-point scales and their traps
Not at all, rather not, neither, rather, entirely: this format has dominated measurement since 1932. It carries three well-identified biases.
A format from 1932
The agreement scale was proposed by Rensis Likert in his 1932 thesis. Its dominance rests on three practical qualities: it is quick to fill in, it codes into numbers, and it lends itself to the usual statistical treatments without discussion.
Yet it rests on a questionable assumption: that the intervals between degrees are equivalent, in other words that the gap between rather not and neither is worth the gap between rather and entirely. Nothing guarantees it, and the treatments applied proceed as if.
Acquiescence
Some respondents agree more than they disagree, independently of the item's content. The phenomenon is stable, more marked in certain populations, and it mechanically inflates scores on dimensions whose items are all phrased in the same direction.
The remedy is simple and not always applied: phrasing some items in the opposite direction, so that agreeing means sometimes one pole, sometimes the other. This publisher's emotional profile test proceeds in that way, with an explicit distribution by dimension.
Response style
Some people readily use the ends of the scale, others stay near the middle. This preference is stable within an individual and varies strongly from one population to another, which makes comparisons between countries delicate.
It produces an unpleasant artefact: a person with an extreme style will appear more marked on every dimension, including those on which they do not stand out. Corrections exist; they require having the full set of answers and accepting that scores be transformed.
The reference group
The most insidious bias has to do not with the format but with the meaning of words. Declaring oneself rather calm presupposes a comparison with those around one, and those people are not the same for everyone. Two equally composed people, one surrounded by reserved people, the other by demonstrative ones, will not tick the same box.
This mechanism explains counter-intuitive international comparison results, where the populations reputed to be the most disciplined declare themselves the least disciplined. Everyone compares themselves with their neighbours, not with humanity.
What this imposes on the reader
A score is never an absolute quantity. It is a position relative to a sample, obtained by a format that carries its own effects, and accompanied by a margin of error that is rarely displayed.
The practical consequence fits in one sentence: a gap of a few points between two dimensions means nothing. An honest report comments on clear positions, leaves minimal gaps unsaid, and says on which sample it is pronouncing.
Key points
The agreement scale assumes equivalent intervals, which nothing guarantees.
Acquiescence is corrected by phrasing some items in the opposite direction.
The reference group effect makes comparisons between populations unreliable.
This article describes documented mechanisms and disagreements. It recommends no position, no party and no vote, and commits to no reading other than that of the work it cites.
And you, where do you stand?
The Civix test situates your convictions on six dimensions and matches you with one of the twelve profiles, without ever telling you what to think.
Start my test