Always, Never, Everyone: What Extreme Wording Actually Signals (2026)

It never works. Everybody on my team complains about it. I spend my entire day in this tool.

These are the quotes that end up on the slide, because they are vivid and they sound decisive. They are also, taken literally, almost always false -- and the speakers know it. The extremity is not a claim about frequency. It is a claim about how much resistance the speaker expects to meet.

Anita Pomerantz named this pattern in a 1986 paper in Human Studies: the extreme case formulation, an expression that names a maximum case of the matter at hand. All the time, everybody, no one. Once you can see them, you will find them clustered in exactly one place in your transcripts -- around complaints and justifications -- and you will stop reading them as measurements.

What an extreme case formulation is

The vocabulary

Derek Edwards, in a 2000 paper in Research on Language and Social Interaction, lists the working set: all, none, most, every, least, absolutely, completely, and their relatives. The function is to defend or justify a description or an assessment, particularly where the speaker anticipates a challenge.

That last clause is the whole finding. An ECF is deployed against an imagined objection. Its presence in a transcript is evidence that the participant expected to be doubted.

Pomerantz's three uses

Her account identifies three jobs these formulations do:

• Asserting the strongest case in anticipation of a non-sympathetic hearing. The speaker braces for disagreement and overshoots to leave room to be argued down.
• Proposing the cause of a phenomenon. Extremity is used to make a causal claim sound inevitable rather than contingent.
• Speaking for the rightness or wrongness of a practice. The formulation carries a moral position, not a count.

To legitimize a complaint and portray a situation as worth complaining about, a speaker may describe the offence or the suffering in extreme terms. None of these three uses is a frequency report. All three are routinely coded as one.

The inversion: the softened claim is the stronger one

Here is where analytic instinct gets it backwards.

Extreme claims are factually brittle

Edwards describes ECFs as factually brittle, for a precise reason: a single counterexample refutes them. It never works is defeated the moment anyone produces one occasion where it worked. The claim is maximally vivid and minimally defensible.

Speakers can manage that brittleness by softening -- mostly, almost, nearly always, hardly anyone. A softened formulation is a logically weaker claim, and it is for that very reason more robust: there is no cheap counterexample that kills it.

So rank the two on the dimension that matters for your report. The punchy absolute quote is the fragile one. The hedged version is the one that survives scrutiny. An analyst who reaches for it never works over it almost never works has systematically selected the least defensible sentence in the transcript, because it was the most quotable.

What you can actually bank

The practical consequence is a reversal of the usual weighting. When a participant softens a claim unprompted, treat that as a signal of care rather than of weakness -- they have made a statement you can defend. When a participant produces an unsoftened extreme, treat it as a strong signal about stance and investment and a weak signal about magnitude.

Edwards also observes that ECFs are overwhelmingly left unsoftened and overwhelmingly go unchallenged by recipients, which is itself evidence that everyone in the conversation is treating them as rhetoric rather than as arithmetic. Everyone, that is, except the analysis pipeline downstream.

What to do when you hear one

Probe for the incident, not the number

Do not argue with an ECF and do not accept it. Convert it. The move that works is to ask for a specific occasion: when did that last happen? or tell me about the most recent time.

This is productive for two reasons. It replaces an unfalsifiable formulation with a datable event you can count and compare across participants. And it does so without challenging the participant, because you are treating their claim as true and asking for detail -- which is the only way to probe a defensive formulation without triggering more defence.

Do not tally extremes as frequencies

If your analysis counts always and never as literal quantifiers, extreme wording will distort severity rankings in a specific direction: issues that participants feel defensive about will outrank issues that are merely frequent. The fix is to record the ECF as a marker of stance, and to source magnitude from a question designed to measure it.

There is a second finding hiding in the pattern, and it is often the better one. A cluster of extreme formulations around a topic means participants expect not to be believed about it. That usually means they have raised it before and been dismissed. That is a genuine, reportable insight about the relationship -- just not the one you were about to write down.

Where this stops and the hedging article starts

Three boundaries.

This article is about the participant's wording inside an answer. The analyst's own hedging, when writing a finding they cannot fully support, is a different discipline covered in the graded hedge.

It is also the opposite case from conversational implicature, which covers what an answer conveys by saying too little. An ECF says too much. Both are places where the literal text and the intended meaning diverge, and they diverge in opposite directions.

Finally, an ECF produced because your question invited it is a question-design problem; see avoiding leading questions.

How Koji handles this

Extreme formulations are best dealt with during the interview, while the participant is still available to be asked when did that last happen? -- which is exactly the moment a survey tool has already lost.

• AI follow-ups convert rhetoric into incidents. Koji's AI interviewer probes vague and absolute answers automatically, asking for the specific recent occasion behind a sweeping claim. A traditional form -- SurveyMonkey, Typeform, Qualtrics -- records it never works and stops there. Koji comes back with the follow-up a good moderator would have asked, on every interview, without scheduling anyone.
• Structured questions measure magnitude properly. Never source a frequency from an open-ended answer if the frequency matters. Koji supports six question types -- open_ended, scale, single_choice, multiple_choice, ranking, and yes_no. Use open_ended for the story and the stance, and a scale or single_choice question for how often it actually happens. The structured questions guide covers combining them in a single study.
• Counting stays honest. Koji reports themes with the underlying evidence attached, so a theme that looks severe can be traced back to whether it was severe or merely emphatic. How you count themes changes the ranking more than most teams expect, which is the subject of counting themes by mention or by participant.
• Quotes keep their context. Because every quote in a Koji report links to its passage, a reviewer can see whether an extreme statement was softened two sentences later -- the softening that gets stripped when a quote is lifted.

Frequently asked questions

What is an extreme case formulation?

It is a description that names a maximum case -- always, never, everybody, nobody, completely. Pomerantz identified it in 1986. Its function is to defend or legitimize a claim the speaker expects to be challenged, which makes it evidence about anticipated resistance rather than about frequency.

Does never really mean never?

Almost never, and the speaker does not intend it to. Extreme formulations are treated as rhetoric by everyone in the conversation: recipients rarely challenge them, and speakers rarely soften them. Treat one as a strong statement of stance and a weak statement of magnitude.

Should I quote the extreme version or the softened one?

The softened one, if you want a finding that survives review. An unsoftened extreme is factually brittle -- one counterexample defeats it -- while almost never has no cheap refutation. The most quotable sentence in a transcript is frequently the least defensible one in it.

How do I probe an extreme claim without arguing?

Ask for the most recent instance: when did that last happen? You are accepting the claim and requesting detail, which avoids triggering further defensiveness, and you get back a datable event you can actually count. Koji's AI interviewer runs this probe automatically.

What does a cluster of extreme wording tell me?

That participants expect not to be believed about that topic, which usually means they have raised it before and been dismissed. That is a real finding about the customer relationship, and it is separate from whatever the literal complaint was about.

How does Koji stop this from skewing my results?

By probing absolute answers during the interview instead of leaving them for the analyst, and by giving you structured question types for anything where magnitude matters. Koji also keeps every quote linked to its surrounding passage, so a stripped-out extreme can be checked against what the participant said next.

Related Resources

• Structured questions in AI interviews -- the six question types, and why frequency should never come from an open-ended answer.
• What participants mean but never say -- the opposite failure, where the answer says too little.
• The graded hedge -- hedging on your side of the report rather than the participant's.
• Probing questions in user interviews -- turning a sweeping claim into a specific incident.
• Counting themes by mention or by participant -- how a counting choice reorders your severity ranking.
• Avoiding leading questions -- making sure the extremity did not come from your own wording.