What is hedging rate?
By Noam Shemla · Measurements
Hedging rate is the percentage of your words belonging to a set of twelve softening constructions: I think, I guess, I'm not sure, maybe, probably, perhaps, kind of, sort of, somewhat, it seems, it might, could be.
It is the simplest of our measurements to describe and the easiest to misread, because the same phrase can be an honest statement of uncertainty or a nervous habit, and a word list cannot tell which.
The short version Hedging is measured exactly like filler rate (matched against a list, reported as a percentage of words) and costs three clarity points per percentage point, half what filler costs. Four of the twelve phrases are also on the filler list, so they are charged twice.
One qualification to the summary above: the three-point charge applies to transcripts from live sessions. On a recorded practice take, your hedges are counted and shown on the report, but the clarity formula is given your filler rate only. Both paths are described below.
Matching the hedges
Pattern matching over the transcript
Measured
- Hedge count
- Hedging rate as a percentage of words
Inferred, not measured
- Nothing directly, but hedging is what the Executive Presence coaching track is told to prioritise
The patterns run over the whole transcript, ignoring case, and each one matches only whole words. Every match adds one to the count, whether it is a single word like maybe or a three-word phrase like I'm not sure. The count is then divided by the number of words in the transcript.
A worked example. Take a 200-word answer in a live session that contains I think twice, maybe once and sort of once, and no other fillers. That is four hedges, or 2% of the words, which costs six clarity points on its own. But I think and sort of are also fillers, so the same three phrases add 1.5% to the filler rate and cost another nine points there. The answer loses fifteen points from four phrases.
The double charge
I think, I guess, kind of and sort of appear on both the filler list and the hedging list.
The nine-point figure applies where both rates are charged, which is the session path. On a recorded practice take, those four phrases cost six points as filler and nothing further.
Two paths, two charges
On a live session, the transcript is split by speaker and each speaker gets a filler rate and a hedging rate. Both go into the clarity formula: 100, minus six per percentage point of filler, minus three per percentage point of hedging.
On a recorded practice take, the report counts your hedges and labels them. With no hedges it says None; one is Minimal; two or three is Occasional; four or more is Frequent. The clarity formula on this path receives the filler rate and not the hedging rate, so a hedge that is not also a filler costs nothing arithmetically. The coaching model still sees the count, and may well comment on it.
Neither path is wrong. They differ, and your clarity score from a session and from a practice take are not built identically.
When a hedge is a problem
Hedging is the clearest case in this collection of a measurement that must not be read as a rule.
Softening a claim is not a defect. It is how English marks uncertainty, and a speaker who says I think about something they think (as opposed to something they know) is being accurate. Removing every hedge tends to produce overclaiming, and an audience that catches you overclaiming once discounts everything after.
The case for measuring it is narrower. Hedges cluster where a speaker is uncomfortable, and they cluster far more densely in some speakers' delivery than the content warrants. A rate well above your own baseline, on material you know well, is a signal about nerves rather than about epistemics. This collection draws out that distinction at length in powerless speech.
Research on courtroom testimony shows why the dense version matters. When testimony carried a cluster of hedges, hesitations and similar markers, which the authors called "powerless" speech, listeners rated the witness as less credible than when the same content was delivered without them.1 The effect came from the cluster, not from any single I think.
It matches strings The instrument cannot tell a warranted hedge from a nervous one. It matches strings. A researcher correctly qualifying a finding and a speaker apologising for having an opinion produce identical measurements, and the coaching that follows treats them the same way.
Where it is used most
The Executive Presence coaching track instructs the model to focus on hedging and directness, and to flag phrases like I think, maybe and kind of explicitly. If you are on that track, this parameter drives a disproportionate share of your feedback. That is by design, and worth knowing when you read it.
Hedging at a glance
- 12 patterns matched. English only.
- 3 pts clarity cost. Per percentage point of words.
- 4 charged twice. Phrases that are also on the filler list.
The hedging measurement, in full.
What the hedge count leaves out
It does not know whether the hedge was true.
It does not catch hedging that is not lexical. Trailing off, an upward inflection at the end of a statement, and a question asked instead of a claim made are all hedges an audience hears and this measurement does not.
It has no baseline. Three percent might be ordinary for you and high for someone else; the score treats them identically.
And the twelve-phrase list is short. I suppose and a bit soften a claim just as well and are not on it.
Hedging also has uses. Holmes found hedges doing real interactive work (softening, inviting, managing disagreement) rather than merely signalling doubt2, and Hyland's study of research writing showed them to be a required feature of careful claims3. Our rate counts them; it cannot tell which kind you were doing.
Reading your own rate
Compare your rate on material you know well with material you are unsure of. If the two are similar, the hedges are a habit rather than a judgement.
Then look at where they fall. A hedge on your main claim weakens the point the talk exists to make. A hedge on a side remark about something you have not checked is often the honest choice, and worth keeping.
References
- Erickson, B., Lind, E. A., Johnson, B. C., & O’Barr, W. M. (1978). Speech style and impression formation in a court setting: the effects of “powerful” and “powerless” speech. Journal of Experimental Social Psychology, 14(3), 266–279. https://doi.org/10.1016/0022-1031(78)90015-x
- Holmes, J. (1990). Hedges and boosters in women’s and men’s speech. Language & Communication, 10(3), 185–205. https://doi.org/10.1016/0271-5309(90)90002-s
- Hyland, K. (1998). Hedging in Scientific Research Articles. John Benjamins. https://openlibrary.org/works/OL85756W