What is a crisis to one person is a Tuesday to another.
What counts as a crisis for one person is an ordinary Tuesday for someone else. That is not a complaint about people being dramatic or stoic. It is a problem with the words we use to measure them.
Where this ends up: any word meaning more-than-normal needs a normal, and almost nobody says whose. Inside one person those words work fine, because your own reference point holds still. Across people they measure nothing, and two of the questions on my own shared sheet break that rule.
Significant distress. Markedly impaired. Excessive. Difficulty functioning. Poor sleep. Heavy workload. Every one of those means more than normal, and not one of them says whose normal, or measured against what.
Hypothesis A comparative word without an anchor is not a measurement. It is a question handed to the reader, who answers it privately, using a reference point nobody asked about and nobody recorded.
Why this is not pedantry
Someone whose baseline has always been high distress may not register significant distress at all. It is not significant to them; it is what things are like. Someone whose baseline is calm registers a small change as alarming and reports it. Put both answers in the same column and the column is measuring three things at once: what happened, where the person sits normally, and how harshly they judge themselves.
The same problem sits in a workplace. Ask ten people whether their workload was heavy this week and you have not measured workload. You have measured the distance between this week and each person’s private idea of a normal week, which is not a quantity anyone can compare.
Where the words actually do their damage
Suggestion Something interesting shows up when you look at where trained people agree and where they do not.
In a study of clinicians diagnosing real patients across 28 centres and 13 countries, agreement on some conditions was high, with figures around .87 for schizophrenia and .84 for bipolar I on a scale where 1 is perfect (Reed et al., 2018). Agreement in other studies, particularly ones where clinicians read written case descriptions rather than meeting anyone, comes out much lower.
The pattern that makes sense of this: people agree about clear cases and disagree at the boundary, and the boundary is exactly where the relative words are doing the work. Nobody needs the word "marked" to identify an obvious case. It only matters when the question is whether this counts, and that is precisely the situation where two people with different anchors will answer differently.
So low agreement is not evidence that trained people cannot tell. It is evidence about where the vagueness lives.
The version of this in my own work
Observation I write self-observation sheets, and I have a rule that only questions worded identically for everyone can ever be compared between people. Everything else belongs to one person and is never pooled.
Two questions on my own shared list break that rule. Ate properly. Enough? Both are comparative words with no anchor, and I marked them as comparable across people.
They are good questions. Inside one person they work, because your own reference point stays roughly stable across a month, so your answer on day 3 and your answer on day 27 mean something next to each other. Across people they measure nothing, because a demanding person answers "sort of" to a day someone else would call "yes".
I already had the fix and did not apply it. On the sheet I wrote for a nurse, the question is ate a real meal: no, snacked, yes. Anchored, countable, and it travels. I solved it for her and left the vague version on the shared page.
What to do about it
Hypothesis Two moves, and the first is nearly free.
Anchor the question or stop pooling the answer. Either give the word something to hold onto, or accept that the answer belongs to one person and never put it in a column with anyone else’s. Both are fine. Doing neither is what produces confident nonsense.
Prefer the countable version where one exists. Hours slept beats slept well. Went outside beats got some fresh air. Ate a real meal beats ate properly. You lose some richness and you gain the ability to compare, and if you want the richness it belongs in the written answer rather than the checkbox.
Question The one I have not resolved. Some of the most important things a person could tell you have no countable version. Whether the day felt like theirs. Whether it was enough. Those are the questions people actually want to answer, and they are exactly the ones that cannot be pooled. I do not think the answer is to stop asking them. I think it is to stop pretending the answers are comparable, and to be much clearer about which of the two kinds of question you are asking.
Where this stops. The argument here is reasoning about measurement, not a research finding, and it is labelled that way. The clinician agreement figures are real and come from one large study of real patients; my explanation of the pattern is my own reading and could be wrong.
What I am not saying. That the words are useless or that the judgements behind them are arbitrary. A trained person using a vague word about a case they have seen is doing something quite different from a form asking a stranger to rate their week. The problem is not the word. It is the column it gets put in afterwards.