Personality & Individual Differences · Simplified
Same Score, Different People
Sam and Priya both come out at 66 for conscientiousness. Nobody who knows them would call them alike.
Step 1 of 3
First, a judgement
Two fictional people complete the same personality questionnaire. Their broad conscientiousness scores come out equal. Watched over several weeks they behave differently, in ways everyone around them recognises. What is the most likely explanation?
Step 2 of 3
Underneath the broad score
Conscientiousness is usually described as having two narrower aspects: getting things done, and keeping things in order. Sam and Priya are close to opposites on them.
Key terms
- Domain and facet
- The broad trait, and the narrower components it is made of. A domain score is an average over its facets.
- Bandwidth and fidelity
- Broad measures predict broad outcomes; narrow measures predict narrow ones. Choosing a level is a decision about what you want to predict, not about which is more real.
- Item mix
- Which items a particular questionnaire happens to contain. Two questionnaires for the same domain can draw on its facets in quite different proportions.
| Measured at | Sam | Priya | Difference |
|---|
Here is why that matters. Four things they were observed doing, with what you would expect of each person if you used their two facet scores, and what you would expect if you used only the broad score.
| Observed behaviour | Facets say Sam | Facets say Priya | Broad score says, for both |
|---|
Step 3 of 3
Change the questionnaire, not the people
Sam and Priya do not change from here on. The only thing that changes is how many of the questionnaire's items are about getting things done rather than keeping things in order.
Each line is one person's broad conscientiousness score as the questionnaire changes. The upright marker is the questionnaire you have built. Where the lines cross, the two people score the same.
What this shows
A broad score is an average, and averages hide things
Key idea: Sam and Priya scored the same because a domain score averages over its facets, and they are high and low on opposite ones. The broad score is not wrong. It predicts broad things well, and if you want to know who will generally be dependable it is a reasonable thing to ask for. What it cannot do is tell those two people apart, and the four observations show exactly where it fails: the facet prediction separates them on every one, and the broad prediction gives the identical number to both. The second half is less comfortable. Their scores were only equal because the questionnaire happened to draw half its items from each facet. Shift that balance and the same two unchanged people swap places. So the sentence "Sam and Priya are equally conscientious" is not purely a fact about Sam and Priya. It is partly a fact about which questionnaire someone reached for.
Two invented people and four invented observations, with tendencies produced by a formula rather than measured. Real facet structures are richer than two aspects per domain, the aspects themselves correlate rather than trading off cleanly, and real behaviour has error in it that this model does not. Two cautions about the argument rather than the numbers. Narrower is not automatically better: facets predict narrow outcomes better and broad scores predict broad ones better, so the level to measure at follows from what you want to predict. And nothing here says either person is more conscientious in any sense that exists apart from a measure. The point is the opposite, that the comparison depends on a measurement choice somebody made.
The longer version adds three domains to investigate, a matching task in which you assign each observation to a person from behaviour alone, and a studio where you build your own pair of profiles. It is at Facet-Level Detective in the main collection.