Five Kinds of Hedge, and Which One to Cut
The common advice about speaking honestly is to strip out the softeners. Stop saying "I think". Stop saying "maybe". Say the thing.
Linguistically this is backwards, and it produces a specific failure: people who have removed their hedges are not being more truthful, they are being less accurate about how much they actually know.
Honest speech needs more markers, not fewer. It just needs the right ones in the right places, and almost nobody has been taught which is which.
Some languages make you declare your source
English lets you say "it rained last night" whether you watched it or someone mentioned it. Many languages do not offer that convenience.
Quechua — spoken across the Peruvian highlands, including the regions where Chuncho cacao grows — is one of the well-studied cases. It carries a set of suffixes usually described as evidential markers: broadly, one form for something you witnessed directly, another for something you were told, and another for something you are inferring. Speakers attach them routinely, and leaving one off is not neutral. It is noticeable.
The relevant point for anyone trying to speak plainly is not that one language is better. It is that the distinction between what I saw, what I was told and what I concluded is real enough that entire grammars are built around it — and English speakers collapse all three into the same flat sentence dozens of times a day without noticing.
Most conversations that go wrong over "honesty" go wrong exactly there.
Five kinds of marker, and what each one is really doing
The linguistic literature on hedging goes back to work by George Lakoff in the early 1970s, and the useful move is to stop treating "hedge" as one thing. There are at least five jobs being done, and they are not interchangeable.
1. Evidentials — where it came from
I saw · she told me · apparently · I read · going by the numbers · I'm inferring this
These report your source. They are the single most undersupplied category in ordinary English conversation, and the one that resolves the most disputes. "Apparently the order was late" and "I saw the order arrive late" are different claims that most people utter identically.
2. Confidence modals — how sure you are
must · probably · might · almost certainly · I'd be surprised if
These grade certainty independently of source. You can be highly confident from an indirect source, or genuinely unsure about something you watched. Keeping these separate from category 1 is most of the skill.
3. Approximators — the content is fuzzy
about · roughly · a few · sort of · in the region of
These hedge the thing itself, not you. "Roughly forty" is not a weaker claim than "forty", it is a more accurate one when you do not have the exact figure. Removing approximators to sound decisive is how false precision enters a conversation.
4. Commitment shields — how far you'll stand behind it
I think · it seems to me · my read is · I could be wrong but
These are the ones people are told to delete. They mark the distance between you and the claim, which is real information. Deleting them does not increase honesty; it silently upgrades a tentative view into a firm one.
5. Face-softeners — managing the relationship, not the truth
no offence but · just · I hate to say this · does that make sense?
This is the only category that genuinely obscures. Face-softeners do not describe your evidence or your certainty. They manage the other person's reaction, and they are what make a hard message land as mush.
The rule that falls out of the taxonomy
Cut category five. Keep one to four. Add category one deliberately.
Nearly all bad advice about speaking directly comes from confusing four with five. "I think we have a problem with the schedule" is not hedged mush — it is an accurately marked claim. "I hate to bring this up, and it's probably just me, but does that make sense?" is mush, and it is mush because of the softeners, not because of the "probably".
Fifteen minutes
This is a rehearsal, not a written exercise. Say the sentences out loud; the whole point is what your mouth does under pressure.
- Minutes 0–4. Make the cup and name the thing. One sentence, out loud, saying what you have been avoiding saying. Do not edit it yet. Whisk while you say it — occupied hands make the first attempt easier.
- Minutes 4–7. Strip category five. Say it again with every face-softener removed. No "just", no "no offence", no "does that make sense". It will feel blunter than you intend. Notice that it is not actually ruder — it is only less padded.
- Minutes 7–11. Add the evidential. Say it a third time, beginning with where it came from. I saw · I was told · I'm inferring this from. This is usually the version that surprises you, because it exposes how much of a firm position rests on something you were told once.
- Minutes 11–14. Grade your confidence honestly. Fourth pass, with the right modal. If it is "probably", say probably. Overclaiming and underclaiming are both misreports.
- Minute 14. Keep the fourth version. That is the sentence. Say it once more so it is in your mouth rather than on a page.
On dose and on the drink
10–15 g is right for a practice like this; there is no argument for a ceremonial dose to rehearse four sentences. Around 65 °C, so the cup lasts the fifteen minutes rather than going cold in the middle of pass three.
The one place the drink earns its keep is step one, where saying the difficult sentence aloud to an empty room is genuinely uncomfortable and having something to do with your hands makes it happen. Our Rose Vanilla blend is the gentlest of the range, which suits a practice you may be slightly dreading.
What this will not fix
Precision is not the same as courage. You can mark a sentence perfectly and still not say it to the person who needs to hear it, and no amount of vocabulary work substitutes for that.
What the taxonomy does fix is the version of the problem where you did speak up, you were accurate, and it landed as either aggression or as nothing at all. That failure is almost always a marker problem, and marker problems are learnable.
