Checking Quotes, Statistics and Screenshots

Earlier lessons in this course covered how to size up a source (lesson 02) and how to trace an image (lesson 03). This lesson covers the other things that fly around a community group chat: a sharp sentence under someone's photo, a statistic that sounds alarming, a chart that makes a small change look dramatic, and a screenshot of a post or text message that nobody can find anymore. All four are easy to make, and all four feel solid because they come as a finished little package. The fix is to open the package and look for the original.

This course is educational. Every method here is about checking what others show you. Nothing here is a guide to making deceptive material.

Quotes: find the first place it was said

A quote can be wrong in four distinct ways, and the way matters because the correction differs.

The SIFT method calls its last move "trace claims, quotes, and media back to the original context" (Mike Caulfield's wording). For a quote, that means finding the first moment it was spoken or printed and reading it in place.

1. Copy a distinctive run of four to six words from the quote, put it in quotation marks and search it with the speaker's name. Distinctive means not "we must work together." Pick the odd phrase.
2. Look at the earliest results by date, not the top results. Quotes get copied thousands of times and the copies rank above the origin.
3. For a public official, go to the primary record: meeting minutes and video from the board or council, a transcript posted by the agency, a press release, a court or legislative record, or the full recording of an interview. Many local boards post agendas, minutes and meeting video on their own websites.
4. Read the whole passage, a few paragraphs on each side. Check whether the speaker was quoting an opponent, describing a view they reject, or answering a hypothetical.
5. If the quote is attributed to someone long dead, expect that it is misattributed. Look for a citation to a book, page and edition. A line with no source older than the internet is a warning.
6. If you find nothing, say so plainly: "I could not find this in any record." That is a finding, and it is different from "this is false."

Quote cards have a few common tells. There is no date, no place and no event. The type is large and the attribution is a name and a photograph, with no link. The person is described by a title that was true at some other time. The wording is modern for a person from another century. The sentence flatters one group and insults another so neatly that it reads like a slogan, which is what it was designed to be. None of these proves a quote is fake. Each one is a reason to look for the origin before sharing.

Video clips and audio quotes

With a clip, find the whole recording. A 20 second excerpt of a three hour meeting can be a fair summary or a distortion. Look at what was said right before and right after, and who spoke next. If the clip appears with captions, compare the captions to the audio, because captions are text someone added. Lesson 06 covers cloned voices and altered video. Lesson 07 covers content credentials, which can sometimes show where a recording came from.

Statistics: five questions before you believe a number

Most misleading statistics are not invented. They are true numbers placed so that you draw the wrong conclusion. Five questions handle most of them.

1. Counted what? "Reported incidents," "arrests," "complaints" and "confirmed cases" measure different things. A rise in reports may reflect more reporting.
2. Out of how many? This is the denominator. A count without it is half a number.
3. Compared with what? Last year, the state average, a similar town, before and after a rule change? A comparison picked after the fact can be chosen to look dramatic.
4. Over what period? One bad month or one odd year can fall outside the longer pattern. A starting point just before a spike or after a dip changes the story.
5. From whom? Who collected it, how, and for what purpose? An agency's data file, a peer-reviewed study, a trade group's survey and an anonymous poster's figure deserve different levels of trust. Ask what the data source is called and whether you can open it.

Counts versus rates

When towns, schools or counties are compared, divide by population (or by the right base, such as students or households). That gives a rate. Rates are usually quoted per 1,000 or per 100,000 so the numbers are easy to read.

Percentages of small numbers

"Up 50 percent" can be 2 incidents rising to 3. Whenever you see a percent change, find the two raw numbers it came from. A change from 2 to 3 is 1 divided by 2, which is 50 percent, and also one extra event. A change from 2,000 to 3,000 is also 50 percent and means 1,000 more. The headline is identical and the meaning is not.

Also separate percent from percentage points. If a school's participation rate goes from 40 percent to 50 percent, that is 10 percentage points and a 25 percent increase (10 divided by 40). Both statements are correct, and writers sometimes use whichever sounds bigger.

Base rates and tests that are "99 percent accurate"

When a claim says a test, a detector or a screening method is 99 percent accurate, ask how common the thing being tested for is. The answer changes what a positive result means.

Charts that exaggerate

The most common chart trick is the truncated axis. A bar chart should have an axis that starts at zero, because the eye reads the length of a bar as the value. If the axis starts at 95, a bar for 98 is three times longer than a bar for 96 even though the numbers differ by about 2 percent. Line charts can reasonably zoom in, since the eye reads change along the line, but a zoomed line still needs labeled axes so a reader can see the scale.

Other chart habits to watch: a second vertical axis that lets two unrelated lines appear to move together; a time window that begins at a peak or a dip; a pie chart whose slices add up to more than 100 percent; a map that shades whole counties by a count rather than a rate, so big, empty counties dominate the picture; and a chart with no source line. The last one matters most. If a chart gives no source, no date and no definition of the measure, treat it as an illustration, not data.

Correlation is another trap. Two lines that rise together may both follow a third factor, such as population growth, or may match by coincidence. A claim that one thing caused another needs more than a matching line, and honest sources say what else was considered.

Finding the source of the data

Work back from the chart or number to the dataset: the table title, the agency, the year and the exact definition. Go to the agency site directly rather than trusting a reposted figure. Check whether the number you were given matches the table at the source and whether the table's footnotes change the meaning (for example, "provisional" or "excludes cases pending"). If an organization is citing a study, find the study's own abstract and look at who funded it and how many people or events it covered. Then check whether newer data exist, since old figures travel for years.

Screenshots: a picture of a post is not the post

A screenshot is an image. Editing text inside an image takes little skill, and fake message templates are common. Even an unedited screenshot can mislead: it can be cropped to leave out the reply that changes the meaning, it can show a post that was later retracted, or it can come from a parody account. So the rule is simple: use a screenshot to find the original, not as the original.

1. Note what the screenshot claims: who posted, on which platform, roughly when, and what it said. Write down the account name or handle exactly.
2. Go to that account on the platform itself and look for the post. If it exists, read it in place with its replies and edit history if one is shown.
3. If the post is gone, look for a dated archived copy of the account page or the post address (see below), or for news coverage that quoted it at the time.
4. Check the account. Is the handle exactly right, or one character off? Did the account exist at the date shown? Is it verified or is it a parody or fan account? Lesson 02 covers checking an organization or an account's history.
5. For a screenshot of a text or direct message, you can not check a private conversation from outside. Ask the people involved, or look for corroboration: an official statement, a record, other recipients.
6. Look at the image itself. Mismatched fonts or spacing, a status bar clock that disagrees with the message time, a battery or signal icon that looks wrong for the phone, and cut-off edges are reasons to be careful. Their absence is not proof. A skilled forgery shows none of them.

Archived pages

Archive services save a dated copy of a web page. The Internet Archive's Wayback Machine is the best known; it lets you look up a web address to see saved versions by date, and it provides a "save page now" function for pages you want preserved. Some pages can not be archived, because of how they are built or because the site blocks it, and archives of logged-in social media content are often incomplete. Features and rules change, so look at the service's own help page.

Use archives for three jobs. First, to see what a page said on an earlier date, such as a county notice before it was edited. Second, to keep a record of a claim you are checking before it gets deleted. Third, to share a stable link with the date visible, so others can look at the same thing. An archive proves that the page said something at a time. It does not make the content true.

Idaho and election-related claims

Claims about local government and elections travel as quotes, numbers and screenshots, so these checks apply there. For election procedure, use official sources. VoteIdaho (voteidaho.gov) is the state's voter information site, with tools to check a voter record, find a polling place and look up your county clerk, and an election calendar. The county clerk's office runs elections locally. Lesson 09 goes through that in detail, including what Idaho law says about false statements and manipulated media in election communications.

Who does what

Questions and answers

How long should I spend checking a quote?

Two minutes usually settles it. Search a distinctive phrase, look at the earliest dated results, and look for a primary record. If a quote is going to drive a decision or be repeated to many people, spend longer and find the full context.

The quote is real but the meaning seems different in the post. Is it false?

It is misleading. Say so precisely: the words are accurate, and the surrounding passage shows the speaker meant something else. That is more useful, and more believable, than calling it fake.

Where do I find the numbers behind a statistic?

Search the agency, the table or report title and the year. Local figures often sit on a county or city website, a state agency data page or a federal statistics portal. If the post names no source, treat the figure as unverified and say so.

What is the quick way to tell whether a percentage is meaningful?

Find the two raw numbers it came from. A 50 percent jump from 2 to 3 is one event. Then check how many years of data there are, since small counts swing from year to year.

Is a chart with a cut-off axis always dishonest?

Not always. Line charts of things such as temperature often zoom in, and a labeled zoom is fine. The problem is a bar chart, where length is the message, or a zoom that hides the scale from the reader.

Can I trust a screenshot from someone I know?

Trust the person for honesty and still check the content. They may have received it from someone else. Ask where they got it and look for the original post.

Why not just use the date in the screenshot?

Dates and times in an image can be typed or edited. A screenshot also shows only the clock of the phone, not when a message was sent. Use an archive, a platform's own timestamp or an independent record.

Do archive services work for everything?

No. Some pages block archiving, many login-only pages do not save, and dynamic pages can come out incomplete. Save more than one copy when it matters, and also keep your own notes: the address, the time you looked and a description of what you saw.

Hard questions

If I can not verify a quote, should I say it is false?

No. Say what you did: "I searched meeting records and video for these words and found nothing." If the speaker or their office denies it, quote that. A flat claim of falsehood from a failed search can itself be wrong, and it hands the original poster an easy rebuttal.

Statistics can be bent in both directions. How do I avoid bending them myself?

Apply the five questions to figures you agree with as strictly as to those you dislike. Show the denominator, the period and the source in your own posts. If a correction later shows your number was off, say so openly. Lesson 10 covers corrections.

A screenshot shows something an official said, and the official's office will not confirm. What then?

Without confirmation or a record, it stays unverified. A local reporter can ask for records through Idaho's public records law (Idaho Code Title 74, Chapter 1), which covers public agencies' written records, including many messages sent on public business. A private resident can not compel that, so the honest position is "I saw a screenshot I could not confirm."

Is it a crime to share a doctored screenshot or fake quote?

Not in general. Speech that is false is often still protected by the First Amendment, and Idaho has no broad statute against sharing a false quote. Some specific laws apply: Idaho Code 67-6628A allows a candidate to seek relief over deceptive synthetic audio or video in an electioneering communication, and general laws on fraud, impersonation or defamation may apply to particular facts. Lesson 09 explains the election-specific rules and their limits.