Communicating uncertainty: what we know, what we don't, and the numbers
State what is known, what is not and what is being done, give numbers as absolute figures with their source, and pair probability words with numbers.
- Start here: the essentials
- 1. The statement: known, not known, being done
- 2. Saying it: ranges, steady voice and no hedging words
- 3. Probability words and the numbers behind them
- 4. Numbers: absolute figures, denominators and time frames
- 5. What would change your conclusion, and updating
- 6. See it
- 7. Practice
- 8. Your interview
- 9. Check your work
- 10. Knowledge check and scenario
Start here Essentials
At minimum, write what you know, what you do not know yet and what is being done to find out, and put each number you plan to say next to its source. Everything marked Go deeper can wait.
- Write what you know, what you do not know yet, and what is being done to find out (the three-part statement).
- Replace each vague word, such as "small" or "likely," with a number or range and where it came from (probability words).
- Turn any "doubled" or "50 percent higher" into before and after numbers with the risk converter (absolute figures).
- Write what would change your conclusion (what would change it).
- Say the statement aloud in a steady voice and time it (practice).
Minimum time: about 30 minutes. Done when: the uncertainty sheet has its three parts filled, and every number in the numbers table has its source and its absolute form.
1. The statement: known, not known, being done Understand
Uncertainty is the gap between what a person knows and what would settle a question. It shows up in almost every interview about science, risk, numbers or a developing event, and the interviewee has to decide how to say it. The guidance from public-health agencies and from research on communication converges on one structure. CDC's emergency communication manual puts it in one sentence: state what you know, what you do not know and what you are doing to find out more, and do not speculate (CDC, CERC: Messages and Audiences). To speculate is to offer a guess about something you cannot yet support with facts. Convention
The World Health Organization's guideline on risk communication in public health emergencies makes the same point as a graded recommendation: authorities should include explicit information about uncertainties in risks, events and interventions, and say what is known and not known at a given time. WHO rates this a strong recommendation resting on moderate-quality evidence, and it reports that a systematic review found acknowledging uncertainty among the factors that predict higher trust (WHO, communicating risk in public health emergencies). The same guideline warns that people's, and experts', understanding of likelihood information is prone to error, and that media emphasis on rapidly changing, conflicting information increases public uncertainty and can reduce trust.
This manual statement has three core parts and three that follow from them. The sheet is the uncertainty statement.
| Part | What it holds | Lena's version |
|---|---|---|
| What we know | Facts you can support, each with its source | The bloom covers about a third of the lake by satellite image. The beach is closed. In the past five summers, 2 of the 14 blooms on this lake tested above the state's advisory level for toxins. |
| What we don't know yet | The question that is open, said plainly | Whether toxins in this bloom are above the advisory level. |
| What is being done to find out | The process, described simply | The state lab is testing samples. |
| When we expect to know more | A time you control or someone else announces | The results are due in about two days. |
| What people can do now | Actions from the authority responsible, not your own advice | Follow the beach closure and the state health department's guidance. |
| What would change this | The result or fact that would change what you say | A lab result above or below the advisory level. |
Three notes on the table. First, CDC's model wording for acknowledging uncertainty is simple: the situation is developing, we do not yet have all the facts, and based on what we know, we expect X. The guidance adds that you say a process is in place to learn more and describe it simply, and that you "don't over-reassure," meaning you keep calm concern rather than promising an outcome (CDC, CERC: Spokesperson). Second, "what people can do now" belongs to the authority responsible; this manual does not give medical, safety or risk advice, and an interviewee who is not that authority says whom to follow. Third, CDC accepts "I don't know" when it is followed by a commitment to find the answer, which is what the "being done" and "when" parts supply.
What is uncertain, and why
Researchers who reviewed the field separate three things that can be uncertain: facts (is this the case), numbers (how many, how big) and science, meaning hypotheses or models (van der Bles and colleagues, Royal Society Open Science, 2019; full article via DOI). They also separate direct uncertainty, about the thing itself, from indirect uncertainty, about the quality of the evidence. Their advice to communicators includes identifying what you are uncertain about and why: natural variation, measurement, limited knowledge or disagreement among experts. Say which one applies. "The lab has not finished the tests" is limited knowledge. "Blooms differ from year to year" is variation. Different causes need different sentences, and a listener who hears the cause can tell whether the answer is coming or not.
The same article makes a point that matters for interviews: keep the size of an effect separate from how sure you are of it. Its example is processed meat, which it describes as a carcinogen with high certainty but a low effect. A sentence such as "we are very sure this has a small effect" says two different things in one clause, and a reporter may quote only one. Say the size and the certainty as separate statements.
What audiences want
The National Cancer Institute's workbook on communicating data observes that lay audiences often want definitive, prescriptive information, using the example of a mechanic: nobody wants to hear that there is a 30 percent chance it is the alternator. Its advice for uncertainty is to address it directly: be honest that the science is tentative, explain why a definitive answer is not yet available, and say what the uncertainty means for people (NCI, Making Data Talk). So the statement is not an apology for what you lack. It tells the listener what you can be definite about, what you cannot, and what happens next. Convention
This chapter teaches how to state uncertainty. It does not judge whether a study, estimate or test is sound, and it gives no medical, safety or risk advice. The author of the estimate, an expert in the field or the agency responsible settles those questions. The research on how audiences respond to stated uncertainty is mixed and comes from experiments on particular topics; no study here tells you how a particular reporter, outlet or audience will respond.
2. Saying it: ranges, steady voice and no hedging words
There are two ways to signal that a number is uncertain. You can give a range, such as "between 3.7 and 4.1 percent," or you can use a word, such as "somewhat," "roughly" or "an estimated." The experiments that compared them are the best evidence this manual found on what to do. In a set of five experiments with 5,780 participants, including a field experiment on the BBC News website, van der Bles and colleagues tested a numeric range and a verbal hedge on facts about unemployment, tigers in India, global temperature rise and migration (van der Bles and others, PNAS, 2020). The numeric range barely changed trust in the source; the verbal hedge ("some uncertainty around this estimate, it could be somewhat higher or lower") lowered it more. People also did not notice uncertainty when it was implied only by words like "estimated" or "about." The authors' summary is that communicating uncertainty numerically had only a minor effect on trust.
This manual takes one rule from this: give the number as a range or with its margin, and do not rely on a hedging word alone. Instead of "about 4 percent, give or take," say "between 3.7 and 4.1 percent." Instead of "roughly a third of the lake," say "about a third, from the satellite image," and name the source of the "about." The range tells the listener how uncertain; the source tells them why.
The studies have limits that this manual repeats because they matter. The experiments covered uncertainty about past and present facts, not forecasts. The BBC sample was self-selected. The effect on trust was explained by whether people perceived the uncertainty. Other studies, covered in Go deeper, found that emphasizing uncertainty can erode trust in some contexts, and that overstated certainty is punished when conflicting information arrives. A National Academies review of the research agenda says avoiding uncertainty "may be a mistake," because leaving it out can convey false certainty that undermines trust if the information is later revised, and also says evidence that transparency increases trust is limited (National Academies, Communicating Science Effectively). So this manual states what each study found and what it does not settle, and the default is modest: say the range, in a steady voice.
The steady voice
Van der Bles's team describes communicating uncertainty confidently and unapologetically, which they call "muscular uncertainty." In practice that means saying "we do not know yet" in the same tone as "the beach is closed." CDC notes that microphones and nerves make people talk faster and advises slowing down and pausing between sentences. Say the open question as plainly as the known fact, with no rising pitch and no apology. Preference
Do not over-reassure, and do not speculate
The two ways to go wrong are opposite. Over-reassurance says "there is nothing to worry about" before the result is in. Speculation says "it is probably the fertilizer" without evidence. CDC's spokesperson chapter advises against both: "don't over-reassure," and don't speculate, assume or make premature promises that may have to be reversed later. A sentence that avoids both has a fact, a gap and a next step: "The beach is closed while we wait for the lab. We don't yet know whether the toxins are above the advisory level. The results are due in two days."
Nine ways to express uncertainty
The van der Bles review lists nine ways to express direct uncertainty, in decreasing order of precision. The numbered list is theirs, in this manual's words.
- A full probability distribution.
- A summary of a distribution.
- A rounded figure, a range or an order of magnitude.
- A predefined category.
- A qualifying verbal statement, such as "around x" or "probably x."
- A list of possibilities.
- Humility: mentioning that you could be wrong, or that unknown unknowns exist.
- No mention of uncertainty.
- An explicit denial that uncertainty exists.
Their advice is to choose an expression that matches your actual precision and to test it with your audience. For an interview, items 3 and 4 are the usual choices: a range, or a category with a number. The article recommends the FDA's guide on communicating risks and benefits for testing.
The trust studies side by side
| Study | What was tested | What it found | Limits |
|---|---|---|---|
| van der Bles and others, PNAS 2020 | A numeric range or a verbal hedge on facts and numbers | A range had a minor effect on trust; a verbal hedge lowered trust more; "estimated" or "about" alone was not perceived as uncertainty | Past and present facts, not forecasts; the BBC sample was self-selected |
| Kreps and Kriner, Science Advances 2020 (DOI) | Five survey experiments with more than 6,000 US adults on COVID-19 projections | Emphasizing uncertainty in projections can erode public trust in some contexts; downplaying it can raise support in the short term, but reversals may weaken that effect or reduce trust in science | Projections, not facts; contested political context |
| Batteux and others, BMJ Open 2022 (DOI) | A "certain" or "uncertain" government announcement about COVID-19 vaccines, then conflicting information | The "certain" group lost more vaccination intention and more trust afterward than the "uncertain" group; the authors conclude that communicating uncertainties can protect people from the effect of later conflicting information | 328 UK adults online; this manual read the abstract only |
| National Academies, 2017 | A research agenda that reviews the literature | Avoiding uncertainty may convey false certainty; evidence that transparency increases trust is limited; communicating uncertainty can sometimes reduce perceived authority | A call for research, not a test |
| Altenmüller and others, Public Understanding of Science 2021 (DOI) | Whether researchers who admitted errors and promised reform were seen as credible | Self-criticism with intent to reform raised perceived trustworthiness and credibility | About researchers and their own work, not crisis spokespeople |
The studies do not give one answer, and this manual does not pretend they do. They differ in topic, in whether the uncertainty is about a fact or a forecast, and in the political setting. What the table supports is our modest default: a stated range in a steady voice, a "when" for the open question, and an update when the facts change.
3. Probability words and the numbers behind them
A verbal probability is a word or phrase such as "likely," "unlikely" or "a small chance" that stands in for a number. People read these words differently. The National Academies review reports that people disagree widely about what verbal probability terms mean, that attempts to legislate translations, such as the Intergovernmental Panel on Climate Change's, have not succeeded, and that putting numbers alongside the words gives more accurate interpretation than either alone. The FDA's evidence-based guide to communicating risks says the same from another direction: words alone, such as "a low chance," are ineffective, raise risk perception and are read differently by different people; low numeracy is not a reason to avoid numbers (FDA, Communicating Risks and Benefits). Convention
Expert systems that define these words do not agree with each other. The table shows three words from two published scales. The probability words page has the full tables.
| Word | IPCC AR5 guidance note | US Intelligence Community Directive 203 |
|---|---|---|
| Likely | 66 to 100 percent | 55 to 80 percent |
| Very likely | 90 to 100 percent | 80 to 95 percent |
| Unlikely | 0 to 33 percent | 20 to 45 percent |
The IPCC's ranges (IPCC AR5 uncertainty guidance note) are nested and overlapping. Intelligence Community Directive 203 (ICD 203, analytic standards) sets bins that do not overlap. Both were written for specialists reading a report, and the IPCC note itself says it was not subjected to formal IPCC review. The point for an interview is not which is right. It is that the same everyday word has different numbers in different expert systems, so a listener who knows one system will hear something different from a speaker who uses another. Convention
This manual rule: pair the word with a number, or say only the number. "That is unlikely" becomes "2 of the last 14 blooms on this lake, about 1 in 7." The word can follow the number if you want it, as "that is a minority of blooms," but the number comes first, so the listener has something to hold.
One number, two words. Dr. Lena Ostrowski (fictional) says in a prep call, "I'd call a toxic bloom unlikely." Her figure is from her own data (fictional): 2 of the 14 blooms in five summers tested above the state's advisory level, which is about 14 percent. On the IPCC scale a probability in the range 0 to 33 percent is "unlikely"; on ICD 203, 5 to 20 percent is "very unlikely" or "highly improbable." The same data fits two different words, and a listener who knows the second scale would hear a stronger statement than Lena means.
Known: 2 of the 14 blooms tested above the advisory level in the past five summers; the toxin results for this bloom are due in 48 hours.
Assumed: Lena's rate of about 1 in 7 applies to this bloom only if it resembles past blooms on the lake. That is an assumption, and she says "based on past blooms" rather than "this bloom has a 1 in 7 chance."
Open: whether toxins are above the advisory level in this bloom, which the state lab's results settle; the state health department issues any advisory, not Lena. If the results come back, her past-bloom number stops mattering for this bloom. The number is what she says now; the lab result replaces it.
Notice what she says instead: "In the past five summers, 2 of the 14 blooms on this lake tested above the state's advisory level for toxins. Whether this one does, the state lab will tell us in two days." The number comes with its denominator, its time frame and the source of both.
The IPCC and ICD 203 scales compared
The two scales differ in structure as well as in numbers. The IPCC's likelihood terms nest: "very likely" (90 to 100 percent) sits inside "likely" (66 to 100 percent), so the ranges overlap and each term covers everything above its lower bound. ICD 203 uses seven bins that do not overlap and that stop short of 0 and 100 percent, each with two permitted wordings, and says not to mix rows. Both separate likelihood from confidence. The IPCC's confidence scale has five qualifiers, from very low to very high, based on the amount and agreement of the evidence, and the note says this is distinct from statistical confidence. ICD 203 says not to put a confidence level and a likelihood in the same sentence. The probability words page shows the full tables. This manual did not check whether a later IPCC assessment changed the AR5 terms, and it flags the ICD 203 host for re-checking because the directive's address now resolves to an archive site.
The reason to learn the scales is to recognize the words when someone else uses them, not to adopt them. Budescu and colleagues' study of 25 samples in 24 countries and 17 languages, as van der Bles's review cites it, found that people read the IPCC terms "regressively": they underestimate high probabilities and overestimate low ones.
4. Numbers: absolute figures, denominators and time frames
A number in an interview has four jobs: it says how many, out of how many, over what time, and from where. The FDA's guide to communicating risks lists recommendations for quantitative information. The ones that matter most for an interview are these (FDA, Communicating Risks and Benefits): give numbers; provide absolute risks, not just relative risks; keep denominators constant; keep time frames constant; use whole numbers; give both the positive and the negative frame. Convention
- Absolute figures. An absolute risk is the actual chance, such as 1 in 7,000. A relative risk compares two risks, such as "doubled." Relative framing makes a change look larger than absolute framing, according to the FDA chapter, for lay people and medical students alike.
- Constant denominators. The denominator is the "out of how many." The FDA's example pairs 1 in 10,000 with 337 in 10,000, one denominator for both. The National Cancer Institute observes that people may think 1 in 200 is a greater risk than 1 in 25.
- Constant time frames. Say "per year" or "over five summers" for each figure you compare, not one for each.
- Whole numbers. Whole numbers beat decimals, so use a larger denominator for a small risk.
- Both frames. If 2 of 14 blooms tested above the level, 12 of 14 did not. The FDA's example is that 60 percent impotent also means 40 percent not. Say both when the frame could change how the figure is heard.
- Few numbers, no false precision. The NCI workbook advises three or fewer numbers when you want to persuade or motivate, familiar metrics, no unnecessary precision ("about 1 in 4" rather than 23 percent) and rounding decimals (9.6 becomes 10).
In October 1995 the UK Committee on Safety of Medicines warned that third-generation contraceptive pills doubled the risk of blood clots, "that is, by 100%." The absolute numbers, as reported by Gigerenzer and colleagues in 2007, were about 1 in 7,000 users of the second-generation pills and about 2 in 7,000 users of the third-generation pills, an absolute increase of 1 in 7,000 (Gigerenzer and others, Psychological Science in the Public Interest, 2007). The paper reports that the warning was followed by an estimated 13,000 additional abortions in England and Wales the next year. "Doubled" was accurate. It left out the baseline. The risk converter starts with these numbers: baseline 1 in 7,000, change plus 100 percent, new risk 2 in 7,000, absolute difference 1 more in 7,000.
The converter does arithmetic on the numbers you give it. It does not check them, and it does not tell you what a risk means for any person. It will not tell you whether your source is sound.
The sentence a reporter can paraphrase
Numbers rarely survive as quotes. NPR's guide to choosing soundbites says that producers write basic facts into narration instead of using them as bites, and that sweeping statements without facts undermine credibility (NPR Training, soundbite essentials). So give the number inside a plain sentence a reporter can lift into their own words: "In five summers, 2 of 14 blooms on this lake tested above the state's advisory level." Then say what you want people to take from it. Your feelings and plain judgments get aired; your figures tend to be paraphrased, so make them easy to paraphrase correctly. Give the source and the date in the same sentence, and write the exact figure down on the fact sheet to read from.
Dana's one number. Dana Ruiz (fictional) will say that the Eastside Tool Library has 260 registered members, from her sign-up sheet. She is not sure the figure is current. On the numbers table: claim, "registered members"; number as she will say it, "260"; range, blank; absolute form, "260 people"; per how many, none; time frame, "as of the last count"; source, "my sign-up sheet." On her statement: she knows 410 tools and 260 members at her last count; she does not know whether new people signed up this week; she is counting the sheet before the call at 10 a.m. Thursday; and what would change it is a new sign-up.
Known: 410 tools and 260 registered members at her last count, three years of Saturday mornings.
Assumed: that the count has not changed since; that is Dana's assumption.
Open: whether 260 is current, which Dana settles by counting the sign-up sheet before the call (Your main points). The sheet is optional on this path because she has one number and one source.
A correct format does not make a number correct. This manual teaches how to state numbers; it does not check yours. Check each against your records, the organization's records or the published source. Where a number comes from an estimate or study, its author or the agency responsible settles whether it is sound, and where a number bears on someone's health or safety, ask that agency before you say it.
Frequencies or percentages
The FDA guide's chapters do not fully agree. Chapter 7 on quantitative information reports that results comparing frequencies with percentages have been equivocal. Chapter 18, on news coverage, quotes Gigerenzer on relative risks without base rates, and recommends frequency statements, absolute risks and natural frequencies. The NCI workbook advises "about 1 in 4" rather than 23 percent for an audience that does not need the precision. Our default is a count out of a constant denominator ("2 of 14"), which both chapters accept, and a percentage if your audience uses percentages. This manual found no source that settles the choice for every audience, and the FDA guide recommends testing the wording with coworkers, friends and family.
Fact boxes
A fact box sets the main benefits and harms side by side in a small table, built from the best available evidence, for people without a statistical background. The Harding Center for Risk Literacy at the University of Potsdam, which develops them, says several studies show they work; its page does not cite effect sizes (Harding Center, fact boxes). If your interview is about a decision with benefits and harms, a fact box is a handout you can offer to the reporter after the interview.
5. What would change your conclusion, and updating
The last two parts of the statement look forward. The first is what would change your conclusion. ICD 203 asks analysts to name the causes of uncertainty and, as appropriate, to identify indicators that would alter the levels of uncertainty, which is a written-down version of saying what would change your mind. For an interview, one sentence is enough: "If the lab finds toxins above the state's advisory level, the state health department will say what that means for the lake." This does two things. It tells the listener that your current answer is not fixed, and it shows the audience what to watch for. Convention
The second is updating when the facts change. CDC's messages chapter advises updating quickly if facts change and coordinating the change with other organizations, and its spokesperson chapter advises being regretful, not defensive, when acknowledging a mistake. The WHO review lists avoiding rapid changes and conflicting information across agencies among the factors that predict higher trust. In the one study this manual found on admitting error, researchers who criticized their own work and said they would reform were seen as more trustworthy than researchers who did not; it concerned researchers, not spokespeople (Altenmüller and colleagues). When your number changes, say the old number, the new number and why: "Yesterday I said about a third; the new satellite image shows about half."
An update is also a news event. A reporter who ran your first number may ask what changed. Have the reason ready, and send the corrected figure by the same route that carried the first one. After the interview covers corrections.
Say it aloud
Put the whole statement into spoken sentences and time it with the answer timer. A first statement of three or four sentences fits the short answers other chapters ask for. If it runs long, cut the "when" and the "what people can do," which the reporter can usually ask for, and keep the known, the unknown and the next step. Then say it slowly, with a pause between sentences, the way CDC suggests.
6. See it See it
Lena's statement for the 6 p.m. hit. All details are invented. Lena Ostrowski, a hydrologist at Kettle River State University (fictional), is booked for a live remote interview on Lakes 7 News (fictional) about a blue-green algae bloom that has closed the beach at Pine Hollow Lake (fictional). She fills the statement sheet:
- What we know: the bloom covers about a third of the lake by satellite image; the beach is closed; in the past five summers, 2 of the 14 blooms on this lake tested above the state's advisory level for toxins.
- What we don't know yet: whether toxins in this bloom are above the advisory level.
- What is being done: the state lab is testing samples.
- When we expect to know more: in about two days, when the lab reports.
- What people can do now: follow the beach closure and the state health department's guidance.
- What would change this: a lab result above or below the advisory level.
Her numbers table has one row: claim, "blooms that tested above the advisory level"; number as she will say it, "2 of the 14"; absolute form, "2 in 14, about 1 in 7"; per how many, 14; time frame, "past five summers"; source, "my own records of this lake." She says the statement aloud and times it: 63 words, about 25 seconds at 150 words a minute, so she plans to give the first four parts and wait for the producer's next question for the rest.
Known: the closure, the satellite estimate, the number of blooms, the lab timetable (each from her records or the lab).
Assumed: that past blooms predict this one; she labels it "based on past blooms."
Open: whether toxins are above the advisory level in this bloom (the state lab's results settle it, and the state health department issues any advisory); whether she may speak for the university or only from her expertise (the university's media relations office settles it, Before you say yes); what "likely" would mean if she used it (she gives the number, 2 in 14, instead).
Free sources that show the method:
- FDA, Communicating Risks and Benefits: An Evidence-Based User's Guide: Chapter 7 on numbers is the source of the recommendations above.
- National Cancer Institute, Making Data Talk: a workbook on numbers for public audiences, including a crisis chapter.
- Harding Center for Risk Literacy, fact boxes: the benefit and harm table format.
- WHO, communicating risk in public health emergencies: the graded recommendations on trust and uncertainty.
7. Practice Practice
Minimum: the three core parts of the statement filled, one number with its source and absolute form in the numbers table, and the statement said aloud once. Time: about 30 to 40 minutes. Materials: the statement sheet and the numbers table (essential if your setup says there are numbers, risks or facts not yet known, or you are unsure; optional otherwise), the risk converter, the probability words page, the answer timer and a phone to record.
- Write the three core parts. What you know, what you do not know yet and what is being done. One or two sentences each.
- List every number you plan to say in the numbers table. For each: the number, the absolute form, the denominator, the time frame and the source. Leave a cell blank if you do not have it, and mark who has it.
- Replace each vague word. Find "small," "likely," "a lot," "significant," "rare." Write the number or range and its source. If you have no number, write the reason, such as "not yet measured," and say that.
- Convert each relative change. For any "doubled," "up 50 percent" or "twice as likely," enter the baseline in the converter and write the before and after counts.
- Write what would change your conclusion, and the other three parts if they apply.
- Say the statement aloud, at a steady pace, and time it. Record it once and listen for hedging words and rising pitch.
- Write one sentence a reporter could paraphrase for each number, with its source in the same sentence.
The smallest version has one number and one source, as in Dana's example, and a statement in three sentences.
Take a statistic from a recent story in your field that says "increased" or "doubled" and rewrite it as before and after counts with the converter. Then pick two probability words you tend to use and write which numbers you mean by them. Compare with the two scales on the probability words page.
8. Your interview Your interview
9. Check your work Check your work
Criteria
- Does the statement say what is known, what is not known yet and what is being done, without speculation?
- Does every number have its source, its denominator and its time frame?
- Is every relative change ("doubled," "up 50 percent") also stated as before and after numbers?
- Does each probability word have a number beside it, or is it replaced by the number?
- Does the statement say what would change the conclusion?
- Can you say it aloud at a steady pace, without hedging words, in the time you planned?
Worked example
Lena's sheet (fictional) meets the first criterion with her three core parts and meets the second for her one number: 2 of 14, five summers, her own records. It has no relative change, so the third criterion does not apply. It replaces "unlikely" with the number, which meets the fourth. It states a lab result as what would change the conclusion. Her time is about 25 seconds at 150 words a minute. The conclusion it supports is limited: her past-bloom rate of about 1 in 7 is a base rate, and it applies to this bloom only under the assumption that the bloom resembles earlier ones. The lab results settle the actual question, and the state health department announces any advisory.
Common mistakes
- A hedging word in place of a range. "Roughly," "around," "about" with no range. Add the range and its source.
- "Doubled" with no baseline. Convert it to before and after counts.
- Changing denominators. "1 in 7,000 versus 2 in 7,000" is easier to compare than "1 in 7,000 versus 1 in 3,500."
- Probability words with no number. "Likely" means 66 to 100 percent in one scale and 55 to 80 in another. Say the number.
- Speculating to fill the gap. "It's probably the fertilizer." Say what is not known and what is being done to find out.
- Over-reassuring. "There's nothing to worry about" before the result is in.
- Many numbers in one answer. Pick the one or two a listener needs.
What is still unresolved
This chapter cannot settle whether an estimate or study is sound, what a risk means for any person, or how a particular reporter or audience reacts to stated uncertainty. The author of the estimate, an expert in the field or the agency responsible settles the first two. If your interview involves bad news or a personal story, Bad news and personal stories covers a first statement that includes these parts.