Personality & Individual Differences · Simplified
One Moment, or the Whole Fortnight?
You catch two people at one moment each. One of them is having a bad hour.
Step 1 of 2
Two people, one reading each
Ada and Bo are two fictional people in a study that asks them, at random moments, how they are feeling right now on a scale of 0 to 100. Here is the first reading from each of them, taken at the same moment on the same morning.
Ada, Tuesday 09:20
67
how Ada rates the momentBo, Tuesday 09:20
47
how Bo rates the moment
Which of them has the higher typical level, the one you would put on their record as what they are generally like?
Step 2 of 2
Fifty-six moments each
Every reading from both of them, four a day for fourteen days. Move the slider to average more of those moments together, starting from the one you were shown.
Key terms
- State
- Where somebody is at one moment.
- Trait
- What somebody is generally like, across many moments. It takes more than one reading to estimate and more than an average to describe.
- Within-person variability
- How much one person moves about from moment to moment. It is reasonably stable for a given person, so it is a characteristic of them rather than noise to be cleaned away.
Each dot is one reading. The moments you are averaging are filled in; the rest are hollow. The horizontal line across each person's band is the average of the moments you have included so far.
What this shows
The average answers one question and hides another
Key idea: A single reading is a state. It tells you where somebody was at that moment, and at 09:20 on Tuesday Ada really was twenty points above Bo. It is a poor estimate of what either of them is generally like, and it is a worse estimate for Bo than for Ada, because Bo moves about more. That is why the gap between the two averages collapses as you add moments: it was never a difference between the people, only a difference between two mornings. But notice what averaging does not remove. However many moments you include, Ada's readings stay bunched and Bo's stay scattered, and that difference is the most reliable thing on the chart. Two people can share an average and have quite different lives behind it, so a questionnaire that asks "in general, how do you feel?" and records one number per person throws that difference away.
Two invented people, with readings generated from a fixed random seed rather than observed, so the same fortnight appears every time this page is opened. Ada and Bo were built to have the same typical level and different variability, which is the comparison the activity exists to make; real pairs are rarely so tidy. Two other things this cannot show. A single reading also contains measurement error, which does shrink with averaging, and in one reading it cannot be told apart from genuine fluctuation. And a fortnight is not a lifetime: a level that is stable across fourteen days can still move across fourteen months, so nothing here says a typical level is fixed.
The longer version adds two more people, controls for variability, measurement error and life events, a re-seedable dataset and two challenges. It is at State versus Trait Tracker in the main collection.