Journalism
A path through The Default Library
========================================
How news is found, checked, written and paid for, and why a free press is worth defending.


Step 1. Reporting & Interviewing: How to Find Out What Happened
---------------------------------------------------------------
Why: Finding sources, asking questions and getting it right.
A news story is only as good as the people and papers behind it. Reporters learn to ask who actually knows, to say in advance what can be printed, to let a silence do some work, and to get the exact words and the exact facts right.

A water main bursts under a street you have never visited. By lunchtime a school is closed, a shop is flooded, and three different people have told you three different reasons. The water company says the pipe was in good order. A man on the corner says it was leaking for days. Someone in a parents' group chat says the council cut the repair budget.
  Nothing in that paragraph is yet a story. It is a pile of claims, each from someone with a reason to say it. Reporting is the work of turning the pile into something a stranger can rely on: finding who actually saw or knows, getting them to talk, being exact about what they said, and being honest about what is still unknown.
  Every journalism programme starts here, because everything else (writing, verifying, ethics, investigation) is built on the notebook. This piece is about the notebook.

  
    The course in one box
    
      Who takes itJournalism and communication majors, almost always in the first year, usually called Reporting, Newswriting, or Reporting and Writing. Schools in the American accreditation tradition (ACEJMC), UNESCO's model curricula and the great majority of programmes worldwide treat it as the entry point. Many communications, politics and law students take a short version.
      What it is forFinding information that is not already public, from people who may not want to give it, and recording it exactly enough that you can stand behind it.
      Comes afterNothing in this major: it is the entry point.
      Leads toNews Writing & Storytelling (35/93), Verification & Fact-Checking (35/94), Journalism Ethics (35/95) and, in year 4, Investigative Journalism (35/98).
      What this piece gives youYou will not be able to land a source in a hostile newsroom after this. You will know how to tell a source from an echo, what "off the record" does and does not protect, why one kind of question gets stories and another gets nods, and what a quote owes the person who said it. Easier first reads: The Art of Asking Great Questions (22/06), Plain Writing (26/10) and Documentary and the Problem of Truth (10/07). Every person, town and document in the tools below is invented.
    
  

  Start with what you need to know, not what you think
  Beginners start with a conclusion ("the council cut the budget") and go looking for people who agree. Experienced reporters start with a list of questions and a list of kinds of people who could answer them: who saw it, which records exist, who is affected, who is independent, and who has something to gain. Then they ask which of those they do not yet have.
  The useful split is between firsthand and secondhand knowledge. A witness saw the water. A record was made at the time. A neighbour who heard about it from the witness is a lead, not a source: valuable for finding the witness, weak as the thing you print. And every source has an interest. The water company's press office is not lying by default, but it is the one party that most wants one version to stand.
  
    Interactive · Who do you call? (invented town)
    The Harbour Road main in the invented town of Marlow Creek has burst. Tap sources to call them. The board tracks which of five needs you have covered.
    
    
    All names and details are invented. The point is the coverage, not the story.
  
  Without JavaScript: a good first round covers an eyewitness, a record made before the event, someone affected, an independent expert, and the party with a stake. A press release alone covers only the version its author wants known.
  The rule of thumb that falls out is simple. For each fact you plan to print, you should be able to name who knows it firsthand, or which document shows it. If the honest answer is "several people said so", keep going.

  Sources: who actually saw it, and who is repeating it
  Many newsrooms ask for two independent sources before printing a serious claim. The word that matters is independent. Six items that all say the same thing are not six confirmations if five of them copied the first. This is how a wrong fact becomes "well-known": an early report is repeated, then cited by the repeaters, and the reader sees a crowd where there is one voice.
  The test is to trace each claim upstream. Where did this outlet get it? Where did that one get it? Stop when you reach someone who saw it, measured it, or holds the record, or when the trail ends at "someone said". Then count the origins, not the items.
  
    Interactive · Count the origins
    The claim: "The Harbour Road pipe was about 80 years old." Tick the items you would count as confirmations.
    
    
    Invented items. Where an item repeats another, the chain is shown in the readout.
  
  Without JavaScript: in the tool, six items repeat the claim, but five of them trace back to the water company's statement; only the retired engineer's own examination is a second origin. Six items, two origins.
  Real cases are murkier. Two sources can share a hidden upstream, such as two officials briefed by the same person. A document can be sound and still be someone's version. A source can be firsthand about one thing and hearsay about the next sentence. Good reporters ask a sentence at a time: "Did you see that yourself?"

  On the record, off the record: say the terms first
  People talk to reporters on conditions. Newsrooms use a small vocabulary for those conditions, and the first rule is that the terms must be agreed before the information, not after. The words are used as follows by many English-language newsrooms; they are conventions, not law, and outlets differ, so a careful reporter says out loud what they mean.
  
    TermUsually means
    On the recordEverything can be printed, with the person's name.
    On backgroundIt can be printed, but not with the person's name; an agreed description such as "a senior official" is used.
    On deep backgroundIt can inform the story, but is not attributed to the person or to any identifiable description.
    Off the recordIt is not to be printed. Whether the reporter may use it to ask others is the most disputed point; many reporters insist on agreeing that explicitly.
  
  
    Interactive · Can I print it?
    Agreed terms
    
    What you want to do
    
    
    A teaching summary of common conventions; your newsroom's rules and the agreement you actually made decide.
  
  Without JavaScript: on the record allows everything; background allows an agreed description but not the name; deep background allows the information without any attribution; off the record allows none of those, and following up with others depends on what you agreed.
  Three practical points. A source cannot normally turn a statement "off the record" after saying it; many reporters honour a request as a courtesy, but the convention is that terms come first. Ask yourself why the person wants anonymity, and what you can prove without them (the ethics of that choice are in Journalism Ethics, 35/95). And a promise is a promise: in Cohen v. Cowles Media Co. (1991) the US Supreme Court held that the First Amendment did not stop a source suing two newspapers that had broken a promise to keep his name out, because the law of promises applies to everyone, including the press.

  Questions: the open one, the follow-up and the silence
  A closed question has a short answer: "Was the pipe old?" "Yes." An open question asks for a story: "Walk me through that morning." A leading question carries its own answer: "The council cut your budget, didn't it?" It gets agreement or a defence, and either way you have learned about your assumption, not the world. Reporters use all three, but they know which one they are using.
  The craft is in what happens next. The follow-up picks up a detail the person just offered and asks for the specific: "You said it was flagged. What happened to the flag?" The silence is the one tool beginners fear. After an answer, many interviewees fill a pause with the thing they were weighing whether to say. And the last question is almost always the same: "Who else should I talk to, and what should I have asked?"
  
    Interactive · Interview the supervisor
    You have five questions with Dana Okafor, an invented city water supervisor. There are five facts to find. Choose carefully: some questions need an earlier answer, and an accusation changes the mood.
    
    
    Start again
  
  Without JavaScript: in the tool, an open question about the morning surfaces the stuck valve; an open question about what she knew surfaces an old survey flag; the follow-up on that flag surfaces the budget decision; a pause after an answer gets one more admission; the closing question gets a lead. The two closed or leading questions get almost nothing.
  Questions also carry power. Interviewing a grieving relative, a child, or a frightened witness calls for a different manner from questioning a minister, and a reporter who treats them the same fails at one of the two. Pressing a powerful person who is dodging is part of the job; pressing a vulnerable person for a better quote is not.

  Notes, recordings and the exact words
  A quotation is a promise to the reader: these are the words the person used. That promise is easy to break by accident. Memory rounds off sentences, notes drop the small words that carry the meaning ("I would not say it failed" becomes "it failed"), and tidy-up edits change what a speaker meant. This is why reporters use a notebook for key phrases (some learn a shorthand system) and, where the source agrees, an audio recording as a check.
  Recording has legal rules that vary by country, and in the United States by state. US federal law requires the consent of only one party to a conversation. The Reporters Committee for Freedom of the Press lists about eleven states with mainly all-party consent requirements (including California, Florida and Pennsylvania) and several with partial rules. Other countries differ again, so the safest professional habit is to ask, on the recording, "Is it all right if I record this?" That is both honest and the best protection. This piece describes practice and isn't legal advice.
  The second danger is the clean edit that changes the sense. Cutting words out of the middle of a quote, or keeping only the half that fits the story, can reverse it.
  
    Interactive · Which version is fair?
    An invented programme director said this, recorded:
    "I wouldn't say the programme failed. I'd say it failed the families who couldn't get to the sign-up desk, and that's on us."
    Three published versions. Pick the one you would run.
    
    
  
  Without JavaScript: the version printing only "the programme failed" reverses what she said. The full line is the safest. A trimmed version that keeps her second sentence is defensible only if the story is clear she rejected the broader word.

  Accuracy: the dull part that is the whole job
  Accuracy is mostly small things done every time. Names are spelled by asking the person, not guessing: a reporter asks for the spelling even of "Smith". Titles, ages, dates and numbers are checked against a document or confirmed again. Figures are written with their units and their base: a number without "out of what?" is incomplete. And claims you cannot confirm are labelled as the claim of a named person, not stated in your own voice (News Writing & Storytelling, 35/93, and Verification & Fact-Checking, 35/94, go further).
  Practices differ on one question. Should a reporter read a quote or a passage back to the source? Many newsrooms say no to full drafts, because it invites rewriting and gives sources control; many allow checking a specific fact or a technical term; some science and magazine desks go further. There is a real trade-off between independence and error, and a reporter should know the house rule. Whatever the rule, errors do happen. How to correct them is in Journalism Ethics (35/95).
  The honest habit that sits under all of it: write down what you do not know. A reporter's notebook has a column for "unconfirmed" and a list of calls still to make, and the story is not ready until the list is shorter than the pile.

  What a graduate just sees
  
    Every claim has an origin; count the origins, not the repetitions.
    Firsthand beats secondhand, a record made at the time beats a memory, and every source has an interest.
    Terms come before information: on the record, background, deep background and off the record mean different things, so say which.
    An open question gets a story, a closed one a nod, a leading one your own assumption back.
    The follow-up and the silence produce more than the prepared list.
    A quote is a promise about someone's words; trimming is where promises break.
    Ask the last question every time: who else, and what should I have asked?
  

  Where it shows up next
  
    News Writing & Storytelling (35/93): the notebook becomes a lede, an attribution and an order of paragraphs.
    Verification & Fact-Checking (35/94): the same instinct, for photos, videos and viral claims.
    Journalism Ethics (35/95): anonymity, harm and what you owe a source.
    Investigative Journalism (35/98): the same craft over months, with documents and public records.
    In the world: HR interviews, user research, police and medical history-taking, oral history, and any job that starts with "find out what happened".
  

  If you want to go deeper
  
    📘 Textbook. The Missouri Group (Brian S. Brooks, George Kennedy, Daryl R. Moen, Don Ranly, with later co-authors), News Reporting and Writing (Bedford/St. Martin's, 13th edition 2020). A standard US text, strong on sourcing, interviewing and accuracy.
    📘 Gentler. William Zinsser, On Writing Well has a chapter on interviews; Janet Malcolm, The Journalist and the Murderer (1990) is a famous, provocative essay on what reporters owe their subjects.
    🌐 Free course. The Knight Center for Journalism in the Americas runs online courses at journalismcourses.org; some are free, others carry a fee, so check each one. The Ethical Journalism Network publishes free reporting guidelines.
  

  
    The takeaway
    Reporting is the discipline of knowing how you know. Find the people who saw it, count origins instead of repetitions, agree the terms before the information, ask open questions and then listen, and treat every quotation as a promise about someone's actual words.
    Try this: next time you hear a surprising claim, ask "how does the person who said it know?" and trace it back two steps.
  

  
    Sources & Further Reading
    Sources. US recording law: Reporters Committee for Freedom of the Press, Reporter's Recording Guide (federal one-party consent; about eleven mainly all-party states). Cohen v. Cowles Media Co., 501 U.S. 663 (1991). On-the-record terms described as common newsroom conventions; they vary by outlet. The scenario, town, people and documents are invented. Go deeper.
    
      📘 The Missouri Group, News Reporting and Writing
      📘 Janet Malcolm, The Journalist and the Murderer
      🌐 Knight Center for Journalism in the Americas


Step 2. News Writing & Storytelling: Put the Thing That Matters First
---------------------------------------------------------------------
Why: The lede, the structure and the story.
Readers give a news story a sentence or two to prove itself, and editors cut from the bottom. News writers learn to build for that: a lede that earns the second sentence, an order that survives being trimmed, a promise about why you should care, and a clear line between what is known and who says so.

Open any news site and look at what is visible before you tap: a headline, perhaps one line under it. In that space somebody decided what mattered most. If they chose well, you know the outline of what happened without reading further, and if you do read on, each paragraph adds detail in descending order of importance. If they chose badly, you scroll past.
  News writing is the craft of making that choice quickly, under deadline, about facts someone else may be disputing. It is not "plain writing with a dateline". It is its own form, shaped by how and where people read, and a graduate of the course can look at a news story and see its skeleton.

  
    The course in one box
    
      Who takes itJournalism and communication majors, usually in year 1, often paired with the reporting course. Public relations, politics and law students take a version because the same structure drives press releases, briefings and wire copy.
      What it is forTurning a notebook of facts and quotes into copy that tells a reader what happened, in an order that serves them, with every claim tied to a source.
      Comes afterReporting & Interviewing (35/92) and Narrative (35/29), which covers time and point of view in fiction and nonfiction.
      Leads toJournalism Ethics (35/95), Media Law & Press Freedom (35/96), Data Journalism (35/97) and Investigative Journalism (35/98).
      What this piece gives youYou will not be a news editor after this. You will be able to read a story as a structure, write a first paragraph that does its job, and see why "said" is the most useful word in news. Easier first reads: Plain Writing (26/10) and Narrative: Who Tells It, and When (35/29). Every story, name and figure in the tools below is invented.
    
  

  The lede: one sentence that earns the second
  The first paragraph of a news story is the lede (journalists spell it that way to avoid confusion with the metal "lead" in old printing rooms). In a hard-news story it answers the questions a reader would shout first: who did what, when and where, and, if possible, why it matters. The usual advice is one sentence of about 25 to 35 words. Longer, and the reader loses the verb; shorter, and it often leaves out the part that matters.
  Two common faults. The topic lede says what a story is about instead of what happened ("There was a meeting at City Hall"). The stuffed lede crams every fact into one breathless sentence. Between them is the one that picks the single most important fact, states it plainly, and leaves the rest to paragraph two.
  
    Interactive · Which lede for which job?
    The invented facts: a city council voted 5–4 on Tuesday to close the Eastside library branch on 30 June, saving about $410,000 a year; about 1,900 people use it. Pick a lede, then a job.
    
    
    
    Word counts are computed from the text. The “answers” tags are my own reading of each sentence.
  
  Without JavaScript: for a breaking-news alert the summary lede (about 22 words) is right. The anecdotal lede suits a feature and needs a nut graf soon after. The topic lede and the question lede tell the reader nothing they could act on.

  The inverted pyramid, and why it survived
  The classic form is the inverted pyramid: the most important facts at the top, then the next most important, down to background and detail. The shape does three jobs. A reader can stop at any paragraph and still have the gist. An editor can cut from the bottom to fit a space without rewriting. And the lede can be sent as a bulletin or alert on its own.
  Where it came from is debated. A popular account credits the telegraph: lines were costly and unreliable, so reporters sent the main facts first in case the wire failed. Studies of nineteenth-century American papers, reported by Errico and others, suggest the form spread decades after the telegraph, so other pressures mattered too, such as a push towards factual, event-based reporting. Horst Pöttker's 2003 article in Journalism Studies, "News and its communicative quality: the inverted pyramid—when and why did it appear?", is the standard place to read the argument. And the form is not universal in history: the Associated Press told the Lincoln assassination in April 1865 in a chronological report, while the New York Herald that day put the main facts in its first paragraph (Chip Scanlan's essay discusses both).
  
    Interactive · Trim the story from the bottom
    The same eight facts, in three orders. Slide to choose how many paragraphs a reader (or an editor with a space problem) keeps.
    
    Paragraphs kept
      
    
    
    The eight facts and their importance scores (0–10) are invented for illustration; your own scoring would differ, which is the editor’s judgment.
  
  Without JavaScript: in the pyramid order, keeping just the first paragraph keeps the decision; keeping three keeps over half of the story’s weight. The chronological order, which starts with the building’s history, keeps under a quarter at three paragraphs and not the decision.

  Beyond the pyramid: other shapes, and when to use them
  The pyramid is for a story whose point is a thing that happened, and where the reader’s first question is “what?”. Many pieces are better in another shape:
  
    ShapeHow it worksGood for
    Inverted pyramidDecision first, detail descending.Breaking news, alerts, wire copy, short reports.
    HourglassPyramid first, then a turn (“Here is how it happened”) and a chronological second half.Crime, accidents, disasters, where both the result and the sequence matter.
    Anecdotal (the “Wall Street Journal” formula)A person or scene, then a nut graf stating the point, then evidence, then a return to the opening.Features and analysis; stories about trends.
    NarrativeTold in time, with scenes and a build; the point may be saved.Long-form, investigations, profiles. See Narrative (35/29).
    List or Q&AChunks a reader can jump between.Explainers, guides, interviews, live blogs.
  
  The trade is clear. The pyramid is efficient but can feel mechanical and rewards the loudest fact over the most meaningful one. Narrative holds a reader and teaches through scenes but costs the reader’s time and is easy to cut badly. The skill is picking the shape the story and the reader need, and then not pretending it is another.

  The nut graf: tell me why I am still reading
  A story that opens with a scene or a person has to tell the reader, soon, what it is about and why it matters now. That paragraph is the nut graf (“nutshell paragraph”). It states the point, gives the news peg (why this story today), and often the scale of the issue. Without it, an anecdotal story is a charming scene that goes nowhere, and readers leave.
  How soon is “soon”? There is no law. A common rule of thumb is by the third to fifth paragraph, or within about the first hundred words. The tool measures it in words and in reader time, so you can feel the cost of waiting.
  
    Interactive · How long does the reader wait?
    Position of the nut graf among 6 paragraphs
      
    
    
    Paragraph lengths are invented; reading time uses 238 words per minute, the average adult silent-reading figure used across this library. A rule of thumb, not a law.
  
  Without JavaScript: put the nut graf third and the reader reads about 80 words before knowing why; put it sixth and it is over 200 words, nearly a minute of reading with no promise.

  Attribution: who says so
  Every claim in a news story is either something the reporter can establish (a vote recorded in the minutes, a date in a document, an event seen first-hand) or somebody’s statement. The first can be written flatly. The second needs the speaker: “the finance office estimates”, “said Councillor Patel”. Attribution tells the reader how much to trust a line and who to blame if it is wrong. It also protects the reporter from putting an opinion or an unproven claim in the paper’s own voice.
  The standard verb is said, and the reason is that it adds nothing. “Claimed” hints the speaker is lying. “Admitted” hints guilt. “Insisted” hints stubbornness. “Argued” fits a case being made; “noted” and “pointed out” suggest the thing is true. Each is sometimes right, and each carries a verdict.
  
    VerbWhat the reader hearsUse
    saidNeutral.The default.
    claimed“Probably not true.”Rarely; only if you are saying it is disputed.
    admitted“Confessing a fault.”Only when the speaker did confess.
    argued“Making a case.”An opinion or position.
    noted / pointed out“A fact the reporter endorses.”Only if you verified it.
  
  
    Interactive · Does this need a name on it?
    For each invented sentence, decide whether it needs attribution.
    
    
  
  Without JavaScript: the vote, the opening date and Patel’s vote are record facts that can be stated flatly. The savings figure, “the busiest branch” and “a mistake” need a speaker.
  Anonymous attribution (“a senior official said”) is a last resort with its own rules, covered in Reporting & Interviewing (35/92) and Journalism Ethics (35/95).

  Writing for the ear, the screen and the thumb
  Audio and video. A listener cannot re-read. Broadcast style therefore uses short sentences, one idea each; puts attribution before the claim (“The finance office says the closure will save…”), because by the time you hear a surprising claim you want to know who is making it; rounds numbers (“about four hundred thousand dollars”); and avoids packing a sentence with subordinate clauses. It is written to be spoken, so reporters read it aloud. A common newsroom rule of thumb is that a presenter reads about two and a half words a second, so a 20-second radio read is around 50 words and a 90-second television package perhaps 200. Pace varies by presenter and station.
  Screen and phone. A mobile reader scrolls fast and arrives from a link or a push alert. The headline and first sentence carry the load; paragraphs are short; the key fact is not buried below an ad or a tall image. Alerts and push notifications are a miniature lede. Search and social add a requirement: the headline must make sense out of context, away from your site.
  
    Interactive · Same facts, three outlets
    
    
    
    Words in a read-out (calculator)
      
    
    Invented text. Calculator uses 150 words a minute (2.5 a second) as the working rule of thumb; real reading speeds vary.
  
  Without JavaScript: the broadcast version uses three short sentences and puts who-says-so first; the print version is one 28-word sentence. At 2.5 words a second, 50 words is 20 seconds.
  None of these is the “wrong” version. A good newsroom writes the same story three ways because the same facts reach different people in different conditions.

  What a graduate just sees
  
    Every news story has a skeleton: a lede, a nut graf if it needs one, and an order that survives being cut.
    A lede states the news, not the topic; if you cannot, you do not yet know the story.
    The pyramid is for a reader who may stop early. Narrative is for a reader who will stay.
    If a story opens with a person, find the paragraph that says why that person matters today.
    “Said” is neutral; every other verb is a small verdict.
    Facts the reporter can establish are stated flatly; claims and opinions have a speaker.
    Writing for the ear means short sentences and attribution first; read it aloud.
  

  Where it shows up next
  
    Journalism Ethics (35/95): the line between a fact and a claim is also a line between reporting and endorsing.
    Media Law & Press Freedom (35/96): attribution and the distinction between fact and opinion are central to defamation law in many countries.
    Data Journalism (35/97): the lede on a number story is a comparison, not a figure.
    Narrative (35/29): how the same events are told in different orders.
    In the world: press releases, executive summaries, incident reports, emails that begin with the decision, and the first screen of any report.
  

  If you want to go deeper
  
    📘 Textbook. The Missouri Group, News Reporting and Writing (Bedford/St. Martin's, 13th edition 2020), a standard US text with chapters on ledes, structure, attribution and broadcast and online writing. Many courses also use the AP Stylebook.
    📘 Gentler. William Zinsser, On Writing Well (HarperCollins, many editions), and Roy Peter Clark, Writing Tools: 50 Essential Strategies for Every Writer (Little, Brown, 2006).
    🌐 Free course. The Knight Center’s journalismcourses.org lists online journalism courses; some are free and some carry a fee. Poynter’s website also has free writing guidance.
  

  
    The takeaway
    News writing puts the point first, builds an order that survives being cut, promises early why a story matters, ties every claim to someone, and then rewrites for whoever will hear it or thumb through it.
    Try this: take a news story you just read and write its lede in 25 words without looking back. Then check whether the story’s own lede did better.
  

  
    Sources & Further Reading
    Sources. Inverted-pyramid history: the Wikipedia entry on the inverted pyramid (telegraph account; Errico and others on spread decades later; the New York Herald of 15 April 1865 and the Associated Press chronological account, via Chip Scanlan’s essay); Pöttker, Journalism Studies, 2003. 238 words per minute: the library’s standard reading-speed figure. Broadcast pacing is a newsroom rule of thumb. All stories, names and figures in the tools are invented. Go deeper.
    
      📘 The Missouri Group, News Reporting and Writing
      📘 Roy Peter Clark, Writing Tools
      🌐 Knight Center for Journalism in the Americas


Step 3. Narrative: Who Tells It, and When
-----------------------------------------
Why: How stories work: voice, time and point of view.
A story is what happened. A narrative is how it is told. Almost everything a storyteller does, from the first sentence to the last twist, is a choice about order, speed and who gets to speak.

Two friends tell you about the same night. One starts with the car breaking down. The other starts with the moment they were laughing on the roadside, then goes back to explain why. Same night, same events, and yet you feel different things, and you trust one friend more than the other. Nothing in the events changed. The telling did.
  That gap between what happened and how it is told is the first thing a literature student learns to see, and it never stops being useful. It explains why a mystery can hide a fact for 300 pages, why a book narrated by a murderer reads differently from one narrated by a detective, and why a story told by a woman to save her life became one of the most influential books ever made.

  
    The course in one box
    
      Who takes itEnglish, literature and comparative-literature majors, usually in year 1 or 2; creative-writing students meet it as craft. Many programmes fold it into an "introduction to literary study" or to the novel, and the field of study has its own name, narratology.
      What it is forTaking a story apart into its choices (order, speed, voice, character) so you can see why it works on you.
      Comes afterNothing; it sits beside Close Reading & Poetry (35/28) in year 1.
      Leads toEpic, Myth & the Ancient Literatures (35/30), Drama & Tragedy (35/31), The Novel (35/32) and Literary Theory & Criticism (35/34).
      What this piece gives youYou will not be able to write a seminar paper on Genette. You will see five dials a storyteller turns, you will turn them yourself, and you will meet stories from Baghdad, Florence, Tokyo and Shanghai. Easier first reads: The Shape of Every Story (10/03) and How to Watch a Film Like a Director (10/04).
    
  

  Story and telling: the same events in two orders
  The Russian Formalist critics of the 1910s and 20s gave this split two names: fabula, the events in the order they happened, and syuzhet, the events in the order the text gives them. The French critic Gérard Genette later refined it into three layers: the story (events), the narrative (the text on the page) and the narrating (the act of telling, with a someone doing it). Aristotle, in his Poetics (4th century BCE), had already said that the arrangement of events is the heart of a story, more important than the characters.
  Below are six events from an invented story. Pick a telling order and watch what changes. The readout counts how many pairs of events are told in a different order from the one in which they happened: zero means a plain chronological telling.
  
    Interactive · Reorder the telling
    
    
    
    Invented events, numbered by when they happen. Out-of-order pairs are counted by the computer (a measure called inversions); 15 is the most possible for six events.
  
  The same events, told in the third order, become a mystery: we learn the mother is dead and a letter has been found before we learn why the sisters stopped speaking. Told in the fourth, the story begins at its end and the whole telling becomes an answer to "how did she get here?" A beginning in medias res, "in the middle of things," is how Homer opens the Iliad and Odyssey. Nobody has told it any other way since.

  Plot and time: speed is a choice
  Everyone knows the English novelist E. M. Forster's example from Aspects of the Novel (1927). "The king died and then the queen died" is a bare story. "The king died, and then the queen died of grief" is a plot, because now one event causes another. Plot is story plus why. It is what makes a reader keep turning pages: not "and then" but "because".
  The second dial is time itself. A narrative can summarise years in a sentence ("She waited thirty winters") or stretch five seconds across a page. Genette called the ratio of text to time the duration or speed. Slow means the reader is meant to pay attention. Fast means something is background. Count the words per story-minute for the same walk to a door.
  
    Interactive · Three speeds
    
    
    
    Three versions of the same scene, written for this piece. Words are counted by the page; story time is stated for each version.
  
  Notice that the fastest version holds almost no feeling and the slowest holds almost no plot. Most of what a story says about what matters is carried by where it slows down.

  Who tells it: person, point of view and the unreliable narrator
  Someone is always telling. Sometimes the teller is a character in the story (first person, "I"), sometimes a voice outside it (third person, "she"), and rarely a voice that addresses the reader as a character (second person, "you"). But the person is only half of it. The other half is the point of view, or in Genette's word the focalisation: whose eyes and mind does the reader see through? A third-person narrator can be locked inside one head (limited), can move among all of them (omniscient), or can see only what a camera would (objective).
  The same moment, five ways. For each, the table says what the narrator is able to tell us.
  
    Interactive · Same scene, five narrators
    
    
    
    Scene written for this piece. The hidden fact: Ben has already torn up his ticket and plans to stay.
  
  A first-person narrator can lie, or be wrong, or simply not know. That is an unreliable narrator, and it works by making the reader see more than the teller does. In the first version above, Ana says Ben does not care. A reader who notices that he is speaking to the floor will suspect she is wrong. Famous cases worldwide: Lu Xun's "Diary of a Madman" (1918), a Chinese story told by a man who believes his neighbours are cannibals and whose madness may be the sanest part; and Machado de Assis's Posthumous Memoirs of Brás Cubas (1881), a Brazilian novel narrated by a man who is already dead and does not mind admitting he was a bit of a fool. Both are in the public domain.

  Character: made of choices, not adjectives
  Aristotle again: character is revealed by choice under pressure. A narrator who says "Dora was greedy" has told you a fact. A narrator who shows her sliding the last three pieces of cake onto her own plate and covering them with her hand has made you conclude it, and conclusions feel truer than claims. This is the rule behind the writers' advice "show, don't tell". Try it.
  
    Interactive · Which one shows?
    
    
    Four pairs. Which sentence shows the trait through an action?
    Sentences written for this piece.
  
  The advice is a tendency, not a law. Telling is fast and honest, and a story that only showed would take forever. Forster, in the same book, called characters flat if they are built around one idea or quality and round if they can surprise us convincingly. A flat character is not a bad one; Dickens built a world of them. What matters is that the reader gets what the story needs.

  Frame tales: stories inside stories
  One of the oldest tricks is the frame: an outer story whose only job is to hold other stories. The Sanskrit Panchatantra (a collection of animal fables, composed in Sanskrit in India, with roots two thousand years old or more) uses it. So does Boccaccio's Decameron, written about 1348 to 1353: ten young people, seven women and three men, leave plague-struck Florence for a villa and tell one another one hundred tales. So do Chaucer's Canterbury Tales (c. 1387 to 1400), told by pilgrims on the road.
  The greatest frame tale is the Thousand and One Nights. Its frame: King Shahryar, betrayed once, marries a new wife each day and has her killed at dawn. The vizier's daughter Shahrazad (also spelled Scheherazade) volunteers, tells a story that breaks off at its most exciting point, and is spared another night. This repeats for 1,001 nights. The earliest known fragment of the book, a few pages from Syria, dates from the early 9th century; the oldest substantial Syrian manuscript was copied some time between about 1450 and 1593. The frame itself probably descends from a lost Persian book, the Hezar Afsan ("A Thousand Tales"), mentioned by the 10th-century writer Ibn al-Nadim. The collection grew over centuries with stories from India, Persia, Baghdad and Cairo.
  Two details are worth knowing. The European versions began with Antoine Galland's French translation (12 volumes, 1704 to 1717), which added "Aladdin" and "Ali Baba," tales that appear in no Arabic manuscript of the time. Galland said he heard them from a Syrian storyteller, Hanna Diyab, during Diyab's visit to Paris in 1709. And the stories nest. "The Fisherman and the Jinni" is told by Shahrazad, and contains a story the fisherman tells the jinni, which contains tales told by a king and a vizier. Tap through the layers.
  
    Interactive · How deep are we?
    
    Press a tale to see who is telling it to whom, and how many layers down it sits.
    Nesting of "The Fisherman and the Jinni" as summarised in standard reference accounts. Depth counts layers below the outer frame (Shahrazad telling Shahryar).
  
  The nesting does what a frame is for. Every tale is told to do something: to save a life, to change a mind, to delay a death. Telling is shown as an action with stakes, which is what makes "narrative" a thing people do, not only a thing on the page.

  The short story: a whole world in one sitting
  The short story as a form grew up in the 19th century, with magazines that wanted something finished in one sitting. In 1842 Edgar Allan Poe, reviewing Hawthorne's Twice-Told Tales, argued that a short tale should aim at a single "unity of effect," every sentence working toward one impression. Around the world, short fiction took its own shapes: Gogol and Chekhov in Russia, Maupassant in France, Tagore in Bengal, Lu Xun in China, Quiroga in Uruguay, Akutagawa in Japan.
  Akutagawa's "In a Grove" (1922) is a good test of this whole piece, because it turns the narrator dial into the plot. A samurai is found dead. Seven people give testimony in turn: a woodcutter, a priest, a policeman's informant, an old woman, a bandit who confesses, the wife who says she did it, and the dead man himself, speaking through a medium. They do not agree. The story never says who is right, and the reader learns that the way people tell a thing is part of what happened. (Akira Kurosawa's film Rashomon of 1950 takes its structure from it.) In the next century, Borges, Mahfouz, Alice Munro and many others made the short story a place for thinking about how any story gets told.

  What a graduate just sees
  
    There are always two timelines: the order things happened and the order they are told. Ask why they differ.
    Plot is not "and then" but "because". Cause is what makes a story feel like a story.
    Where a narrative slows down is where it is saying "this matters".
    Someone is always telling, and the teller is a choice: who sees, who speaks, what they do not know.
    Whenever a narrator tells you what to feel, check whether the events agree.
    A character is a pattern of choices, and showing is often (not always) stronger than telling.
    A frame is a story about why stories get told, and what they are for.
    The same facts told by three witnesses are three stories.
  

  Where it shows up next
  
    Epic, Myth & the Ancient Literatures (35/30): the epics begin in the middle, nest tales, and are told by bards who change them each time.
    The Novel (35/32): the novel is the form where voice, time and point of view get their fullest play, from Murasaki to Cervantes to Austen.
    Literary Theory & Criticism (35/34): narratology is one of the schools, and structuralism asks whether all stories share a grammar.
    In the world: films and TV (flashbacks, voiceovers), news reporting, podcasts, game design, advertising and the way your friend tells a story at dinner.
  

  If you want to go deeper
  
    📘 Textbook. H. Porter Abbott, The Cambridge Introduction to Narrative (Cambridge University Press; 3rd edition, 2021). A standard university text that covers everything above in plain prose.
    📘 Textbook. Gérard Genette, Narrative Discourse: An Essay in Method (Cornell University Press, English translation 1980). The source of the time, speed and focalisation vocabulary. Harder, and worth it.
    📘 Gentler. E. M. Forster, Aspects of the Novel (1927), short, chatty and still the best place to meet flat and round characters; and James Wood, How Fiction Works (2008).
    🌐 Free course. Open Yale Courses, ENGL 300 Introduction to Theory of Literature (Paul Fry), which includes lectures on Russian Formalism and on Semiotics and Structuralism, where the story/telling split comes from.
  

  
    The takeaway
    Every story has two clocks: the events, and the telling. A storyteller chooses the order, the speed and the voice, and each choice changes what the reader feels and trusts. Learn to see the choices and every story, from Baghdad to Tokyo, becomes a machine you can open.
    Try this: this week, retell something that happened to you in a different order, or from another person's point of view. Notice what you now have to leave out, and what it makes you understand.
  

  
    Sources & Further Reading
    Sources. Fabula/syuzhet and Genette's story, narrative and narrating are standard narratology (Abbott; Genette, Discours du récit, 1972, English 1980). Forster, Aspects of the Novel (1927): plot versus story and flat versus round characters; the example is paraphrased, not quoted. Nights facts (9th-century Syrian fragment, Galland manuscript c. 1450 to 1593, Galland's 1704 to 1717 translation, Hanna Diyab in 1709, Hezar Afsan): Wikipedia, "One Thousand and One Nights", checked against its cited scholarship. Nested tales of "The Fisherman and the Jinni": Wikipedia, same title. Decameron (ten tellers, one hundred tales, c. 1348 to 1353): Wikipedia. Akutagawa, "In a Grove" (Shincho, January 1922; seven testimonies): Wikipedia. Poe on Hawthorne (1842) and the single effect: widely taught, from Poe's review in Graham's Magazine. Lu Xun 1918; Machado de Assis 1881; both public domain. Scenes and sentences in the interactives were written for this piece. Counts in the interactives were computed by the page and checked in a browser.
    
      🌐 Open Yale Courses, ENGL 300 Introduction to Theory of Literature
      📘 H. Porter Abbott, The Cambridge Introduction to Narrative
      📘 E. M. Forster, Aspects of the Novel
      📘 Gérard Genette, Narrative Discourse


Step 4. Verification & Fact-Checking: How to Know What You Know
---------------------------------------------------------------
Why: How to verify a claim, a photo or a video.
A photo arrives with a caption, a video with a date, a quote with a name. Verification is the trained habit of asking where each came from, who else can confirm it, and how sure you are, in steps, out loud.

.opt{display:flex;gap:10px;align-items:flex-start;padding:8px 6px;border-bottom:1px solid var(--line);font-family:-apple-system,sans-serif;font-size:.92rem}
  .opt input{margin-top:.25em;min-width:22px;min-height:22px}
  .opt label{flex:1;min-width:0}
  .lvl{font-family:-apple-system,sans-serif;font-weight:700;font-size:1.05rem}
  .l0{color:var(--hot)}.l1{color:var(--accent2)}.l2{color:var(--cold)}.l3{color:var(--good)}
  .chain{font-family:-apple-system,sans-serif;font-size:.88rem}

  
  
  
  

  At 6:40 on a Tuesday morning a photograph reaches a newsroom: a brown river carrying away a footbridge, a crowd on the bank, a caption saying this is the town of Marlow, today. It has been shared four thousand times. It is vivid, it fits the weather forecast, and a well-meaning editor could publish it in two minutes.
  A verifier asks different questions. Who took it? Is this the first time it appeared, or the fortieth? Does the bridge exist? Does the light fit the time of day? Does anyone with no stake in the story say the same? None of this needs special software. It needs a method and a willingness to say "not yet" in public. Journalism's own textbook calls verification the essence of the craft: what separates it from rumour, entertainment and propaganda.
  Every photo, source, document and town in this piece is invented, so the methods can be practised without touching a real person. This describes professional practice and isn't a guide to targeting anyone: the ethics section is part of the method, not an extra.

  
    The course in one box
    
      Who takes itJournalism students, usually in year 2 or 3 after a reporting course; the same skills are taught at fact-checking organisations, in human-rights groups and by libraries and schools. Journalism programmes teach it as part of reporting and, increasingly, as a stand-alone course on open-source investigation.
      What it is forDeciding what you can say is true, how sure you are, and how to show your working, for claims, images, video and AI-made media.
      Comes afterReporting & Interviewing (35/92).
      Leads toData Journalism (35/97), Investigative Journalism (35/98), and it feeds Journalism Ethics (35/95) and Media Law & Press Freedom (35/96).
      What this piece gives youYou will not become a forensic analyst. You will know the six moves a verifier makes, why "this detector says fake" is a weak answer, and where the ethical lines are. Easier first reads: How to Spot Bullshit (17/03), Propaganda, Misinformation and How to Resist It (18/09) and Critical Thinking and Epistemic Hygiene (22/02). This goes further into method.
    
  

  A claim is a bundle: count the independent sources
  "The Harbor Bridge reopens Friday" feels like one fact. A verifier splits it: who says so, how would they know, and who else knows independently? The last question catches the commonest failure in news, circular reporting: five outlets repeating one unnamed official look like five confirmations and are one.
  The rule of thumb in reporting is two independent sources, or one document you can check. "Independent" is the hard word: it means each source has its own route to the fact, not that each has its own logo. A report that cites another report adds nothing but reach.
  Fact-checking grew into its own institution out of this. Dedicated outlets now operate on every inhabited continent: for example Snopes (US, 1994), Chequeado (Argentina, 2010) and Africa Check (2012, with offices in South Africa, Nigeria, Kenya and Senegal). Many publish under the International Fact-Checking Network's code of principles, run from the Poynter Institute since 2015, which asks for non-partisanship, transparent sources and methods, and open corrections. Fact-checkers check claims; the newsroom's verification desk checks material. Same discipline, different input.
  
    Interactive · How many confirmations is that, really?
    Invented claim: "The Harbor Bridge reopens Friday." Tap each item you would count as a separate confirmation.
    
    
    
    Invented outlets and officials. "Traces to" shows where each item got the claim.
  
  Without JavaScript: five items repeat the claim, but they trace back to only two origins, an unnamed official and a city press release. That is two confirmations at most, and the unnamed official is a source you still cannot check.

  Confidence comes in steps, not a switch
  Professional verifiers rarely say "true" or "false" about a piece of material on first look. They say how far up a ladder it has climbed. Different organisations use different words, but the rungs are similar: unverified, plausible, probably genuine, verified. Each rung is earned by evidence that a stranger could re-check.
  The classic checklist for a piece of user-generated content has four tests: provenance (is this the original, and who made it?), source (can I reach the person, and does their history fit?), date (when was it made?), location (where?). Add independent confirmation and you have the working ladder below. Saying the rung out loud ("we have the location and the date, but could not reach the photographer") is more honest than a headline that hides the gap.
  The Bellingcat investigators' work on the shooting down of Malaysia Airlines flight MH17 over eastern Ukraine (17 July 2014) is the best-known case of this ladder in public: videos of a missile launcher on a convoy were placed on a map by matching buildings, road signs and a petrol station, and their dates by social-media posts. A Dutch court later convicted three men in absentia (17 November 2022) in a criminal case that relied on a wide body of evidence, of which open-source material was one part.
  
    Interactive · Climb the confidence ladder
    Invented photo of a flooded footbridge in "Marlow". Tick what you have established so far.
    
    
    A teaching ladder, not an industry standard: real desks set their own thresholds, and weights here are simple rules (below).
  
  Without JavaScript: a rule of thumb. Original file + location + date + independent confirmation = verified. Three of those = probably genuine. Two = plausible. Fewer, or a screenshot of unknown origin = unverified, so do not state it as fact.

  Images and video: find the first appearance
  The commonest fake is not a forgery. It is a real picture with the wrong caption: an old flood, another country, a film set. So the first question about any image is "where did this first appear?"
  Reverse image search (tools such as Google Lens, TinEye and Bing Visual Search, which differ in coverage) takes a picture, finds copies and sorts them by date. The earliest copy, not the most popular one, is your lead. If the "today" photo turns up three years ago, you have your answer, and a lesson about caption laundering.
  For video, grab still frames and search those, then look for cuts, music added later, watermarks from other platforms and a dubbed voice that does not match the lips. Metadata (the camera, time and sometimes place stored inside a file) is a clue and nothing more: most platforms strip it on upload, and it is easy to edit, so a verifier treats it as a lead to be checked against something else. Then geolocation: compare what is visible (a ridge line, a shop sign, the shape of a bridge) with satellite imagery and street-level photos. It is detective work on landmarks, never on people.
  
    Interactive · Which copy is the first?
    The footbridge photo is captioned "Marlow, today, 12 March". Tap a sighting to inspect it.
    
    
    Invented sightings and dates. Real search results are messier: crops, filters and re-uploads hide the earliest copy.
  
  Without JavaScript: the earliest sighting is a stock-photo library entry dated 4 June 2019, about six years before the claimed event, and it names a different river. The photo is genuine; the caption is false.

  Shadows, sun and the clock: placing a photo in time
  Light is a free witness. A vertical object of height h casts a shadow of length L, and the sun's elevation above the horizon satisfies tan(elevation) = h / L. Measure the shadow-to-height ratio in a photo and you know the sun's elevation, which depends on three things: latitude, date and time of day. This is chronolocation: if you trust the place, the shadow constrains the time; if you trust the time, it constrains the place.
  sin(elevation) = sin(lat)·sin(dec) + cos(lat)·cos(dec)·cos(H)dec is the sun's declination on the date (about ±23.4° over the year); H is the hour angle, 15° per hour from solar noon
  Two cautions. Solar time is not clock time: time zones, daylight saving and longitude shift the clock by up to hours. And the same elevation occurs twice a day, once in the morning and once in the afternoon, so shadow direction (west in the morning, east in the afternoon) breaks the tie. A single photo gives a range of possibilities, not a stamp.
  
    Interactive · Sun-angle chronolocation
    Invented photo: a lamp post 5 m tall casts a shadow 7 m long (ratio 1.4). Pick a place and a day; the page finds the solar times that fit.
    Latitude (north +, south −)
    Day of year
    Solar time of day
    
    
    Computed with the standard declination approximation (accurate to roughly a degree or two). The photo is invented.
  
  Without JavaScript: at latitude 40° N around 21 March, a shadow 1.4 times the object's height happens at about 9:16 solar time in the morning and 14:44 in the afternoon.

  AI-generated media: why "spot the fake" is a losing game
  Synthetic images, audio and video are now cheap. Early fakes had tells: extra fingers, garbled signs, glassy eyes. But those tells are artefacts of a generation of models, and they fade as the models improve. The harm is real: in March 2022 a fake video of Ukraine's president telling soldiers to surrender circulated before it was debunked, and in early 2024 a Hong Kong finance worker at the engineering firm Arup was reported to have been tricked into transferring about US$25 million after a video call with deepfaked colleagues.
  The honest position among verifiers is that pixel-level detection cannot carry the weight. Meta's 2020 Deepfake Detection Challenge, with more than two thousand participants, found that the best model reached only about 65% accuracy on a held-out test set designed to resemble the real world; detectors also tend to get worse on kinds of fakes they were not trained on. So newsrooms lean on what has always worked: source, context, corroboration. Where was the person? Do the other cameras agree? Who benefits from the clip? Does the official channel confirm?
  A second approach is provenance: instead of proving a file fake, prove where a real one came from. The C2PA standard (the Coalition for Content Provenance and Authenticity, founded in February 2021 by Adobe, Arm, the BBC, Intel, Microsoft and Truepic) attaches signed "Content Credentials" showing the capture device and edits. It tells you about origin, not truth, and the information is lost if a file is re-processed. Treat it as a helpful label.
  Detectors also have a base-rate problem, which is the arithmetic below. Even a good test produces mostly false alarms when real fakes are rare.
  
    Interactive · A detector flags an image. Is it fake?
    Share of images that really are fake
    Detector catches this share of fakes
    Detector clears this share of real images
    
    
    Illustrative numbers, not any real detector's. Each square is 10 images out of 1,000; red squares are flagged. See Statistics (34/05) for the Bayes arithmetic.
  
  Without JavaScript: with 2% fakes and a detector that is 90% sensitive and 90% specific, of 1,000 images about 18 fakes are caught and about 98 real images are wrongly flagged, so only about 16% of flagged images are fake.

  OSINT ethics: you could, but should you?
  Open-source intelligence means using public material to establish facts. The same skills (matching a window to a satellite image) can verify a massacre or stalk a neighbour. Professional codes (Berkeley Protocols for digital open-source investigations, first published by the UN human rights office and UC Berkeley in 2020, are one) make the difference explicit.
  
    Verify events, not private individuals. Placing a conflict video is accountability; working out where an ordinary person lives from their photos is not journalism.
    Minimise harm. Blur faces, plates and house numbers of bystanders and victims, unless naming them is the public-interest point.
    No getting round locks. No guessing passwords, bypassing paywalls or logging into accounts. Public means public.
    Mind your own footprint. Don't make contact from a personal account, and don't expose a source by what you ask.
    Show your method. Publish how you did it so others can check you, with the personal details left out.
    Beware the crowd. After the Boston Marathon bombing in April 2013, online sleuths wrongly named innocent people, including a missing student whose family was then harassed; he had in fact died weeks earlier. Crowd-sourced "investigation" without editors is how that happens.
  
  
    Interactive · Publish, blur or hold?
    Four invented situations. Choose what a careful desk would do.
    
    Pick an answer for each.
  
  Without JavaScript: the rule is to verify events and not private people, to minimise harm, to stay within what is public, and to hold anything that cannot be confirmed.

  What a graduate just sees
  
    Five outlets quoting one official are one source with five logos.
    The first appearance of a picture matters more than its popularity.
    A caption can be false while the picture is real, and that is the common case.
    Shadows give a range of times and places, not a single answer, and solar time is not clock time.
    "The detector says fake" is weak; where it came from and who confirms it is strong.
    Say how sure you are, in words a reader can check, and correct in the open.
    Verify events, not private people.
  

  Where it shows up next
  
    Data Journalism (35/97): the same questions asked of a spreadsheet: where did it come from, what does it count?
    Investigative Journalism (35/98): a leak is a document that must be verified before anything is built on it.
    Journalism Ethics (35/95) and Media Law & Press Freedom (35/96): corrections, harm and the cost of being wrong in print.
    In the world: human-rights groups documenting abuses, insurers checking claims, courts weighing digital evidence, and anyone deciding whether to forward a video.
  

  If you want to go deeper
  
    📘 Textbook. Bill Kovach and Tom Rosenstiel, The Elements of Journalism (Crown; several editions), whose chapter on "the discipline of verification" is the standard statement. Craig Silverman (ed.), Verification Handbook (European Journalism Centre, free online), is the practical companion.
    📘 Gentler. Claire Wardle and Hossein Derakhshan, Information Disorder (Council of Europe, 2017, free), a clear map of mis-, dis- and mal-information.
    🌐 Free course. First Draft's training material (its archive now sits with the Information Futures Lab at Brown University) and Bellingcat's free guides; the Knight Center for Journalism in the Americas runs free online courses (including in Spanish and Portuguese) on verification.
  

  
    The takeaway
    Verification is a method: split the claim, trace each source to its origin, find the first appearance, test place and time against the world, and report your confidence by rung. Tools, and now AI detectors, help but cannot replace source and context, and the work stops at the line between verifying events and exposing people.
    Try this: before you forward the next dramatic photo, find its earliest appearance with a reverse image search and see whether the caption survives.
  

  
    Sources & Further Reading
    Sources. Kovach and Rosenstiel, The Elements of Journalism; Poynter's International Fact-Checking Network code of principles; fact-checker founding years from the organisations' own pages; Bellingcat's MH17 reporting and the Hague District Court judgment of 17 November 2022; Meta's Deepfake Detection Challenge results (2020); the C2PA's own account of its founding (22 February 2021); the Berkeley Protocol on Digital Open Source Investigations (UN OHCHR and UC Berkeley, 2020); contemporary reporting on the March 2022 Zelensky deepfake and on the 2024 Arup fraud; sun geometry from the standard solar-position equations. Solar numbers in this page were computed, not copied.
    
      📘 Silverman (ed.), Verification Handbook
      🌐 Bellingcat
      🌐 IFCN Code of Principles


Step 5. Journalism Ethics: Four Duties That Pull Against Each Other
-------------------------------------------------------------------
Why: Truth, harm, independence and accountability.
Tell the truth, avoid needless harm, answer to no one you cover, and own up when you are wrong. Every journalist agrees. The craft is in the cases where two of the four collide, and in an argument, still open, about whether "neutral" is even the right goal.

A reporter has a document that proves a hospital hid errors. It came from a nurse who is frightened. It also names a patient who died, and the patient's daughter has not been told. Publishing serves the public. It could also expose the nurse and wound the daughter. A rule cannot settle this; a way of thinking can.
  Journalism ethics is not mainly a list of prohibitions. It is a set of duties that sometimes pull against each other and a habit of weighing them in the open. A graduate of the course learns the codes, but more usefully learns to ask the questions the codes were written to prompt, and to write down the answer.

  
    The course in one box
    
      Who takes itJournalism majors, usually in year 2 (a short version appears in year 1 reporting courses). Accrediting bodies such as the US ACEJMC list ethics among the core professional values, and UNESCO's model curricula do the same. Media, politics and communications students often take it too.
      What it is forMaking and defending decisions about truth, harm, independence and accountability, before and after publication.
      Comes afterReporting & Interviewing (35/92).
      Leads toMedia Law & Press Freedom (35/96), Investigative Journalism (35/98) and The Business & Future of News (35/99).
      What this piece gives youYou will not be a standards editor after this. You will know what the main codes promise, how to weigh public interest against harm, what makes anonymity defensible, and how the argument about objectivity is actually structured. Easier first reads: Propaganda, Misinformation & How to Resist It (18/09), Free Speech (18/13) and Ethical Discernment & Moral Courage (22/10). The cases in the tools are invented; this is not legal advice.
    
  

  Codes: what journalists promise, and who holds them to it
  Almost every journalists' organisation has a code. They are mostly voluntary: no state licenses journalists in most democracies, and the codes work through professional pressure, press councils and the reader's right to point at a broken promise. A few of the main ones:
  
    CodeCore of it
    SPJ Code of Ethics (US)Four principles: Seek Truth and Report It; Minimize Harm; Act Independently; Be Accountable and Transparent. Revised in 2014, and again in 2026: SPJ members approved a revised code at the 2026 convention in Columbus (the board endorsed it on 25 September; 93% voted in favour, reported 5 October 2026), keeping the four headings and adding material on AI, verification and more. Check spj.org for the final text.
    IFJ Global Charter of Ethics for JournalistsAdopted on 12 June 2019 at the International Federation of Journalists' 30th World Congress in Tunis. Sixteen articles. It builds on the IFJ's 1954 Declaration of Principles on the Conduct of Journalists (adopted at its congress in Bordeaux). The IFJ was founded in Paris in 1926.
    Ethical Journalism NetworkFive core principles: truth and accuracy; independence; fairness and impartiality; humanity; accountability. A charity that works across countries.
    German PressekodexAdopted in 1973 by the German Press Council. Its first article puts respect for the truth and protection of human dignity first among the press's duties.
  
  Two cautions. First, the same family resemblance across countries hides different emphases: German practice puts dignity first; US practice leans on independence and the public's right to know. Second, "ethics" can be turned into a weapon. In some countries, governments have set up official bodies or "codes" that critics say are used to punish reporters who displease them. A code that a profession writes for itself and a code that a state writes for the press are not the same thing, whatever they are called.

  Seek truth: when the reporter is the problem
  The first duty is accuracy, and most accuracy failures are honest mistakes (see Reporting & Interviewing, 35/92). The scandals that shaped the profession were different: reporters who made things up.
  In September 1980 The Washington Post published "Jimmy's World" by Janet Cooke, about an eight-year-old heroin addict. It won a Pulitzer Prize for Feature Writing on 13 April 1981. Within days, doubts about her biography led to a confession; she resigned and the prize went to another writer. Editors found no evidence she had interviewed such a child. In May 2003, The New York Times published a front-page account of more than 7,000 words (7,239, by one count) of how its reporter Jayson Blair had plagiarised from other outlets and invented quotations and reporting trips. The paper called it "a low point" in its history.
  What do such cases teach? That fabrication tends to exploit trust, speed and a rising star; that editors need to ask routine questions of everyone (where was that? who is she? may I see your notes?); and that the remedy is public. Truth also includes fairness: when a story accuses someone, giving them a real chance to respond before publication is a basic duty in most codes, and it often improves the story.

  Minimize harm: weigh, do not flinch
  Journalism hurts people; it is meant to. A story that exposes a corrupt official does harm to the official. The ethical question is about needless harm and about harm to people who did not choose publicity: crime victims, bereaved families, children, ordinary people caught up in news. Codes ask for compassion, for particular care with minors and the grieving, and for a test: is the detail necessary to the public-interest point of the story?
  One area has specific evidence-based guidance. The World Health Organization's Preventing suicide: a resource for media professionals, update 2023 advises that how suicide is reported can affect vulnerable readers, and encourages stories about recovery and where to find help. If you or someone you know is struggling, findahelpline.com lists services by country.
  
    Interactive · A harm and public-interest weigh-in (invented case)
    A story about a fire could name a 17-year-old witness who was identified in a social post. Set the facts and see which questions the case raises. It is a prompt for an editors' discussion, not a verdict.
    Public importance of naming her
    Possible harm to her if named
    
      Her name is needed for the story
      She is a public figure acting in role
      She and a parent consented, knowing the risks
    
    
  
  Without JavaScript: if the name is not needed, leave it out. If the person is a public figure acting in role and the public importance is at least the harm, naming is the ordinary case. If a minor has consented, check for pressure and ask a senior editor.

  Anonymity: a promise with conditions
  Granting anonymity is a trade. It buys information a source would not give otherwise, and costs the reader the ability to judge the source. Most newsrooms therefore set conditions: the information must be important; there is no other way to get it; the editor knows who the source is; the reporter has weighed the source's motive; and the story describes the source as precisely as the promise allows ("a county official with direct knowledge").
  The most famous anonymous source in American journalism, "Deep Throat", helped the Post reporters Bob Woodward and Carl Bernstein on the Watergate story in the early 1970s. Woodward had promised never to reveal the source while he was alive. In 2005 Mark Felt, a former FBI official, revealed himself in Vanity Fair, and Woodward confirmed it. The promise held for three decades; it is also a reminder that the cost of such a promise is real.
  A promise is not just a courtesy. In the United States the Supreme Court held in Cohen v. Cowles Media Co. (1991) that a source whose name was published despite a promise could sue under ordinary promise law. Rules protecting journalists' sources differ widely between countries (see Media Law & Press Freedom, 35/96).
  
    Interactive · Is anonymity defensible?
    Tick what is true of the (invented) source and the claim.
    
    
    A checklist drawn from common newsroom policies; each outlet's rules differ.
  
  Without JavaScript: a newsroom typically wants all of these: the information is important, there is no other route, the source is at real risk, the motive is understood, an editor knows who it is, and the story is as specific as the promise allows.

  Independence: the conflicts readers cannot see
  "Act independently" means a journalist's loyalty is to the reader's right to know, not to a source, an advertiser, an owner, a party, or a friend. Conflicts come in three kinds: actual (your spouse runs the company you cover), potential (you take a gift that might sway you), and apparent (nothing is wrong but it looks wrong). Most codes say that appearance matters too, because trust is the only thing a newsroom sells.
  The usual remedies are to decline (free tickets, gifts, paid trips), to disclose (to editors and, where it matters, to readers) and to step aside (hand the story to a colleague). Commercial content raises a related duty: advertising that looks like news must be labelled so readers know who paid.
  
    Interactive · Fine, disclose, or step aside?
    
    
    Invented cases. Real policies differ between outlets and countries; the point is the reasoning.
  
  Without JavaScript: a spouse working for the subject means step aside; free tickets from a promoter mean decline or pay your own way; a broad fund holding the company is usually fine; a former colleague now doing PR for a source, and an owner with business before the council the paper covers, mean disclose.

  Corrections: how to be wrong in public
  Every newsroom publishes errors. What readers judge is the response. A good correction is prompt, prominent (visible where the error was seen), specific (says what was wrong and what is right), and unexcused (no spin). Online, the best practice is to fix the text and add a note stating the change; quietly editing without notice damages trust. Many outlets have a corrections page, a standards editor or a readers' editor ("ombudsman" in older usage), and press councils in some countries will hear complaints.
  
    Interactive · Which correction is best?
    An (invented) story said the Eastside library would close on July 30. It will close on June 30. Choose a correction.
    
    
  
  Without JavaScript: the best correction says what the article said, what is right, and that the article has been updated. A vague “contained errors” note, or a silent change, is worse.

  Objectivity and moral clarity: a debate told fairly
  "Objectivity" in news means different things, and the debate gets confused unless you separate them. In the United States it has been the dominant ideal since the 1830s (a common explanation is that cheap mass newspapers needed readers of every party); an Associated Press chief in Washington told Congress in 1856, "My business is to communicate facts." Walter Lippmann is usually credited with the modern idea, urging scientific discipline in news-gathering after the excesses of the yellow press. Much of the press outside that tradition, in Europe and Latin America for example, has long been openly partisan, and readers know which paper is which.
  Then came the critique, from several directions at once. Edward Herman and Noam Chomsky argued that "neutral" reporting leans on official sources and so favours the powerful. David Mindich argued that detached coverage of 1890s lynchings, treating the alleged "crimes" of victims alongside the killings, normalised the violence. Brent Cunningham (2003) argued that objectivity can excuse lazy reporting. Jay Rosen, borrowing Thomas Nagel's phrase, called the pose of impartial authority a "view from nowhere". And in recent years, debates in newsrooms have used phrases like "moral clarity" for the view that some facts, such as racist violence or an attack on an election, should not be reported with equal doubt on both sides.
  Defenders of the older ideal answer in turn. Kovach and Rosenstiel, in The Elements of Journalism, argue that the objective part is the method (verification, openness about sources), not the journalist, who is expected to have views. Others fear that reporters who announce a side will lose the trust of half their audience, and that a press that is mainly advocacy cannot be corrected by its readers. Philosopher Judith Lichtenberg argued that critics who say objectivity is impossible yet condemn journalists for failing at it contradict themselves. None of this is settled; research on audience trust finds different things in different countries and subjects.
  Notice what kind of disagreement it is. The empirical questions (does neutral framing in fact favour officials? does open advocacy cost trust?) can be studied. The normative ones (what is journalism for?) cannot. And there is an overlap that nearly all sides accept: false balance, giving a fringe claim equal space against overwhelming evidence, is a failure, which defenders of balance call a failure of weighing and critics call a failure of balance itself.
  
    Interactive · Five positions, each with its best case
    
    
  
  Without JavaScript: the five positions are method-objectivity (Kovach and Rosenstiel), detached neutrality, transparency (Rosen), attachment (Martin Bell, 1998) and open advocacy or a “view from somewhere” (Lewis Raven Wallace, 2019).

  What a graduate just sees
  
    Ethics is weighing duties in the open, with the reasoning written down; it is not a rulebook.
    Harm is not the test; needless harm to someone who did not choose publicity is.
    Anonymity is a trade: it must buy something important and cost the reader as little as possible.
    A conflict that looks bad is a conflict even if nothing wrong happened.
    A good correction names the error, gives the right fact, and does not hide.
    "Objective" can mean method, voice or stance; separate them before arguing.
    Codes written by a profession and "ethics rules" written by a state are different things.
  

  Where it shows up next
  
    Media Law & Press Freedom (35/96): where ethics ends and defamation, privacy and source-protection law begin.
    Investigative Journalism (35/98): the hardest anonymity, harm and independence calls come in investigations.
    The Business & Future of News (35/99): sponsorship, platforms and AI put old conflicts in new places.
    In the world: medicine, law and engineering have codes too; the same logic of weighing duties applies, and so does the habit of publishing your reasoning.
  

  If you want to go deeper
  
    📘 Textbook. Bill Kovach and Tom Rosenstiel, The Elements of Journalism (Crown; first published 2001, with later revised editions), and Stephen J. A. Ward, Global Journalism Ethics (McGill-Queen's University Press, 2010), which argues for ethics across borders.
    📘 Gentler. Janet Malcolm, The Journalist and the Murderer (1990), a short, provocative book on what a reporter owes a subject.
    🌐 Free resources. The Ethical Journalism Network publishes guidelines, and the IFJ publishes its Global Charter in full.
  

  
    The takeaway
    Journalism ethics is four duties (truth, minimizing harm, independence, accountability) that sometimes collide. Codes help by naming them; judgment decides, in the open. Whether neutrality or moral clarity serves truth better is still argued, and the useful move is to say which question you are asking.
    Try this: next time a news story names a private person, ask whether the story would still work without the name.
  

  
    Sources & Further Reading
    Sources. SPJ four principles and the 2026 revision: Society of Professional Journalists and Editor & Publisher (5 October 2026). IFJ Global Charter: ifj.org (adopted 12 June 2019, Tunis; sixteen articles; 1954 Declaration). Ethical Journalism Network principles: ethicaljournalismnetwork.org. German Pressekodex: German Press Council (1973). Cooke and Blair: Wikipedia entries, which cite the Post and the Times. Felt and Deep Throat: Wikipedia, "Mark Felt". Cohen v. Cowles Media Co., 501 U.S. 663 (1991). WHO, Preventing suicide: a resource for media professionals, update 2023. Objectivity history and critics: Wikipedia, "Objectivity (journalism)"; Kovach and Rosenstiel, The Elements of Journalism. Cases in the tools are invented. Go deeper.
    
      📘 Kovach and Rosenstiel, The Elements of Journalism
      📘 Stephen J. A. Ward, Global Journalism Ethics
      🌐 Ethical Journalism Network


Step 6. Probability & Statistics: Reasoning with Uncertainty
------------------------------------------------------------
Why: The statistics behind every poll, chart and claim.
Nothing you measure is exact, and no part comes out the same twice. Statistics is the discipline of knowing how much to trust a number.

A machinist turns a hundred steel shafts to "10 millimetres". Measure them with a good micrometer and no two are the same: 9.981, 10.012, 10.004. Nothing is wrong. That scatter is a fact of every process, every material and every sensor.
  Now stack five of those parts end to end in a gearbox. Will the stack still fit the housing? A pollster calls a thousand people and says 52% will vote yes, give or take three. A drug trial says "significant". A forecast says 30% chance of rain. Every one of these is the same question: given scatter, how much should I trust this number?
  A graduate of probability and statistics doesn't see a single number any more. They see the shape of the scatter behind it. This piece builds that sight with eight things you can run: dice, tolerances, polls, a drug-trial trap, a belief that updates, and a gearbox stack.

  
    The course in one box
    Who takes it: almost every engineering and science student, usually in the second year; computer scientists meet it again before machine learning.
    What it's for: deciding what to believe, and how far to trust a measurement, when outcomes are uncertain.
    Comes after: Calculus I (34/01), which helps with continuous distributions, though the ideas here need little more than arithmetic.
    Leads to: Machine Learning (34/32) needs it directly. It also shows up in Materials Science & Manufacturing (34/13: process control and defects), Systems Engineering (34/22: reliability) and Computer Security (34/33: randomness and guessing).
    Honest line: you won't be able to design a clinical trial after this. You will see why "significant" doesn't mean "important", why a bigger sample beats a cleverer guess, and why most real quantities look like a bell curve. For easier first reads, see the library's Probability: How to Not Fool Yourself (12/03), Statistics for Reading the World (12/05) and Bayes' Theorem (12/17); this piece goes further into the engineering side.
  

  Random variables: a shape with a centre and a spread
  A random variable is a number that comes out of a process you can't predict one run at a time. The key move is to stop asking "what will it be?" and ask "what is the shape of what it could be?" Roll two dice and add them: 7 is six times as likely as 2 or 12, so the shape is a tent. Roll many times and the tent appears.

  
    Interactive · Roll dice, grow a shape
    
      Dice added together
      Roll 10 timesRoll 1,000 timesClear
      
      
    
    Without JavaScript: two dice add to 7 with probability 6/36 and to 2 or 12 with probability 1/36 each. The mean total is 7 and the standard deviation is about 2.42.
  
  Two numbers summarise a shape. The mean is its balance point: 7 for two dice. The standard deviation (the "spread") says how far values typically stray from the mean: about 2.4 for two dice, which means totals usually land between roughly 5 and 9. The variance is the standard deviation squared, and it is the one that adds up neatly when you combine independent things.
  standard deviation = √( average of (value − mean)² )
  Now the shape that matters most. Measured quantities (shaft diameters, heights, sensor noise) tend to form the normal curve, the bell. Its great convenience is a fixed rule: about 68.3% of values fall within one standard deviation of the mean, 95.4% within two, and 99.7% within three. Engineers use it every day. Say the shaft's spec is 10.00 mm ± 0.05 mm. Slide the process around and see what share of parts get thrown away.

  
    Interactive · How many parts fail the spec?
    
      Average diameter
      Spread (standard deviation)
      
      
    
    Without JavaScript: a centred process with standard deviation 0.025 mm (the spec limits are exactly ±2 standard deviations) scraps about 4.55% of parts. Cut the spread to 0.01 mm and about 0.00006% are scrapped. Shift the average 0.02 mm off-centre with spread 0.02 mm and about 6.7% are scrapped.
  
  Two lessons graduates keep. First, spread, not average, decides quality: a process centred perfectly but wide still scraps parts. Second, being off-centre costs a lot even when the spread is fine. That is why factories track both.

  The central limit theorem: why the bell keeps appearing
  Why should so many things be bell-shaped? The answer is one of the most useful theorems in mathematics. Take almost any random process, however lopsided, and average a few dozen independent draws. The averages pile up in a bell, centred on the true mean, with spread shrinking like 1/√n. Early forms are due to de Moivre in 1733 (for coin flips) and Laplace, whose general version dates from about 1810. Pick a lopsided source and drag the sample size.

  
    Interactive · Average anything, get a bell
    
      Flat (0 to 1)
      Very lopsided
      Coin (0 or 1)
    
    
      Draws averaged together: n
      
      
    
    Without JavaScript: for the lopsided source (mean 1, standard deviation 1) the average of n = 30 draws has standard deviation 1/√30 = 0.183, and the 3,000 averages form a close-to-normal bell around 1.
  
  At n = 1 you see the raw source. By n = 30 even the lopsided one is a near-bell. The prediction (the curve) uses only two numbers from the source, its mean and its standard deviation. The shape of the source is forgotten. That is why a measurement made of many small independent errors is normal, and why so much of statistics can pretend things are bell-shaped.
  standard error of an average = σ / √n
  In words: the average of n draws wobbles less than one draw, by the square root of n. To halve the wobble you need four times the data, not twice. It is the quietest expensive fact in all of experimental science.

  Sampling and confidence intervals: how sure is a poll?
  You can't ask everyone, so you ask n people and treat the share saying yes as an estimate. How far off might it be? The standard error from the last section gives the answer. For a share p the wobble is √(p(1−p)/n), and a 95% confidence interval is the estimate plus and minus about two of those. For n = 1,000 and a share near half, that is ±3.1 percentage points: the familiar "margin of error".
  The wording is subtle, and the interactive makes it concrete. "95% confidence" is a statement about the method, not about one interval: if you ran the procedure again and again, about 95 intervals in 100 would capture the true value. Here the truth is secretly 52%. Run 100 polls.

  
    Interactive · 100 polls, one truth
    
      People asked in each poll
      Run 100 polls
      
      
    
    Without JavaScript: with 1,000 people per poll the margin of error is about 3.1 points; run 100 polls and about 95 of the intervals cover the true value. With 4,000 people the margin halves to about 1.5 points.
  
  Watch two things. Roughly five in a hundred intervals miss the truth even when everything is done correctly. And quadrupling the sample from 1,000 to 4,000 only halves the interval. A poll of a country and a poll of a town need the same sample size for the same precision, as long as the sample is random. The real danger is not the margin of error; it is a sample that isn't random at all, which no formula fixes.

  P-values and Bayes: two ways to weigh evidence
  A hypothesis test asks whether data are surprising if nothing is going on. The p-value is the probability, assuming nothing is going on, of seeing a result at least as extreme as yours. Flip a coin 100 times and get 60 heads: for a fair coin the chance of a split at least that lopsided in either direction is about 5.7%. It is not the probability that the coin is fair, and it is not the probability that the result was "due to chance". The American Statistical Association issued a rare formal statement on this in 2016, with six principles. Two of them: p-values do not measure the probability that a hypothesis is true, and a p-value does not measure the size or importance of an effect.
  The classic trap is testing many things. At the usual cut-off of 0.05, a test of a true "nothing" comes out "significant" one time in 20 by luck. Test twenty unrelated things and the chance that at least one lights up is 1 − 0.9520, about 64%. Try it: each press tests twenty effects that are, by construction, nothing.

  
    Interactive · Test twenty things that do nothing
    
      Run 20 testsReset the tally
      
      
    
    Without JavaScript: twenty tests of pure noise at the 0.05 level give at least one false "discovery" about 64% of the time, and one in 20 tests on average.
  
  If you report only the one that lit up, you have a headline and no finding. This is part of why many published results fail to repeat: in the Open Science Collaboration's 2015 attempt to replicate 100 psychology studies, 97% of the originals had statistically significant results but only 36% of the replications did.
  Bayes' rule answers a different question: given the data, how should my belief move? Start with a prior belief, multiply by how well each possibility explains the data, and renormalise. Say you test parts and want the defect rate. Know nothing at first (any rate from 0% to 100% equally likely) and add parts one at a time.

  
    Interactive · Learn a defect rate from parts
    
      Add a good partAdd 10 good partsAdd a defective partReset
      
      
    
    Without JavaScript: after 20 good parts and no defects, a flat starting belief gives a most-plausible defect rate near 0%, an average of 4.5%, and a 90% credible range of roughly a quarter of a percent to 13%. Twenty clean parts cannot rule out a defect rate near one in eight.
  
  Frequentists and Bayesians disagree about the philosophy: one treats probability as long-run frequency and never assigns it to a hypothesis, the other as degrees of belief that data update. In practice both are used, often on the same project, and on big data sets they usually agree. The Bayesian version has one practical advantage here: after twenty clean parts it honestly says "we still can't rule out one in eight".

  Correlation, regression and the trap of causation
  Correlation (called r) measures how tightly two quantities move together along a straight line, from −1 to +1. Regression fits the best line through the cloud (by least squares, as in Linear Algebra, 34/04). Both are summaries, and any summary can hide the story. In 1973 the statistician Francis Anscombe made four data sets with, to two decimals, the same mean, the same spread, the same r (0.82) and the same line (y = 3 + 0.5x). Look at each.

  
    Interactive · Four data sets, one set of numbers
    
      Set 1Set 2Set 3Set 4
    
    
      
      
    
    Without JavaScript: all four Anscombe sets have mean x = 9, mean y = 7.50, correlation 0.816 and line y = 3.00 + 0.50x. Set 1 is a noisy line, set 2 a curve, set 3 a line with one outlier, and set 4 a single far point that creates the whole slope.
  
  The numbers agree and the stories do not: one noisy line, one curve, one line with a rogue point, one vertical stack held up by a single point. Always plot first.
  The other trap is the famous one. Ice-cream sales and drownings rise together, because summer heat drives both. A confounder (a hidden third factor) makes two things move together without either causing the other. The remedy graduates know is randomised assignment: if a coin decides who gets the treatment, nothing else can be systematically linked to it. Where you can't randomise, causal claims need much more than a line through a cloud.

  Engineering: tolerance stacks, reliability and Monte Carlo
  Five parts, each 10 mm ± 0.1 mm, line up in an assembly. What is the worst the total can be out? Worst-case stacking adds the tolerances: ±0.5 mm. It is safe and expensive, because it assumes every part is at its extreme in the same direction at once. If the parts vary independently, variances add, so the standard deviations add in quadrature: root-sum-square, √5 × 0.1 = ±0.224 mm. For five parts that is under half the worst-case figure.
  When the formulas get hard, engineers use Monte Carlo: let the computer build the assembly ten thousand times with random parts and count how often it fails. Here each part's tolerance is treated as three standard deviations. Choose how many parts are stacked.

  
    Interactive · Monte Carlo tolerance stack
    
      Parts in the stack
      Build 10,000 assemblies again
      
      
    
    Without JavaScript: five parts at ±0.1 mm (three standard deviations) give a worst-case limit of ±0.5 mm, a root-sum-square limit of ±0.224 mm, and in simulation only about 0.3% of assemblies exceed the root-sum-square limit; none come close to 0.5 mm.
  
  The honest caveat is that root-sum-square assumes the parts vary independently and are centred. A tool wearing in the same direction on every part breaks that assumption, and so does one supplier's batch. Engineers disagree about when to trust statistical stacking over worst-case; safety-critical designs often insist on worst-case.
  Reliability follows the same arithmetic. A system that needs all of its parts is a series system, and probabilities multiply: ten components that each work 99% of the time give 0.9910 ≈ 90.4% for the whole. Add a backup in parallel and failure needs both to fail: two units that each work 90% of the time give 1 − 0.1² = 99%. The two rules drive why aircraft duplicate systems and why complicated machines need very reliable components.

  What a graduate just sees
  
    Every number has a spread behind it; the first question is "how scattered?", not "what is it?".
    Spread, not the average, decides how many parts are scrapped, and centring matters too.
    Averages of many independent things are bell-shaped, whatever the source looked like.
    Halving the uncertainty costs four times the data.
    A 95% interval is a claim about a method: it misses one time in twenty by design.
    A p-value is "how surprising is this if nothing is going on", not "how likely is it true".
    Test twenty things and one will look significant; ask how many were tried.
    Plot the data; correlation is not causation, and only randomising breaks a confounder.
    Independent errors add in quadrature; series reliabilities multiply.
  

  Where it shows up next
  
    Machine Learning (34/32): training is fitting a model to noisy data; ideas of overfitting, error bars and Bayes are this course made bigger.
    Materials Science & Manufacturing (34/13): process capability and scrap rates are the normal curve against a spec limit.
    Mechanics of Materials (34/11) and Machine Design (34/19): safety factors exist because loads and strengths are distributions, not numbers.
    Systems Engineering (34/22): reliability budgets multiply series probabilities and add redundancy to push them back up.
    Computer Security (34/33) and Algorithms (34/24): good random numbers, randomised algorithms, and the odds of guessing a secret.
  

  If you want to go deeper
  
    📘 Douglas Montgomery and George Runger, Applied Statistics and Probability for Engineers (Wiley; 7th edition). The standard engineering course text, with plenty of tolerance, quality and reliability problems.
    📘 Joseph Blitzstein and Jessica Hwang, Introduction to Probability (CRC Press; 2nd edition, 2019). Best for deep intuition about probability itself, from Harvard's Stat 110.
    📖 David Spiegelhalter, The Art of Statistics (2019): the most readable route for learning alone, built around real cases and honest about uncertainty.
    🌐 MIT OpenCourseWare 18.05, Introduction to Probability and Statistics (Spring 2022): free notes and problem sets, with Bayesian and classical statistics side by side.
  

  
    The takeaway
    Nothing you measure is exact, so every number is a shape. Summarise a shape by its centre and its spread, trust averages more as the square root of the sample, read a confidence interval as a promise about a method, treat a p-value as a measure of surprise, and plot before you believe a correlation.
    Try this: next time you see a poll or a "significant" headline, find the sample size and ask how many things were tested. Then, in the dice widget above, add four dice and watch the tent turn into a bell.
  

  
    Sources & Further Reading
    Sources. The normal-curve rule (68.3%, 95.4%, 99.7%), the central limit theorem, confidence intervals, the binomial test, the Beta posterior, root-sum-square stacking and series/parallel reliability are standard textbook results (Montgomery and Runger; MIT 18.05); every number shown was recomputed in code. Wasserstein and Lazar, "The ASA's Statement on p-Values: Context, Process, and Purpose", The American Statistician (2016). Open Science Collaboration, "Estimating the reproducibility of psychological science", Science (2015): 97% of originals significant against 36% of replications. Anscombe, "Graphs in Statistical Analysis", The American Statistician (1973). Early forms of the central limit theorem: de Moivre (1733) and Laplace (c. 1810), as surveyed in recent histories such as Ulyanov, "From classical to modern central limit theorems" (arXiv 2405.19828).
    Go deeper. The books and course above; Wasserstein and Lazar's statement is free and short (doi.org/10.1080/00031305.2016.1154108).


Step 7. Media Law & Press Freedom: Who Can Be Sued for What They Print
----------------------------------------------------------------------
Why: Defamation, privacy, sources and press freedom around the world.
Every country must decide how much a published mistake may cost the person who published it. The answers run from almost nothing to prison, and they decide which stories get written at all.

A local reporter learns that the owner of a care home has been cited for safety failures. She has a regulator's report, two former staff who will talk, and one detail she cannot confirm. Whether she can publish, and what it might cost her if she is wrong about that one detail, depends almost entirely on where she is standing.
  In one country the owner would have to prove that she knew the story was false, or did not care. In another the owner need only show serious harm, and the reporter must prove it true. In a third, the owner can file a criminal complaint and the reporter can face prison. All three say they protect both expression and reputation. They weigh the two differently.
  Not legal advice. This piece describes how several legal systems work, in simplified form. Laws differ by country and change, and the tools below are illustrations, not tests that tell anyone their rights. For a real question, ask a media lawyer where you are.

  
    The course in one box
    
      Who takes itJournalism majors, usually in year 3, and law students as an elective. The US accrediting council for journalism schools lists the law of speech and press among the competencies it expects, and schools elsewhere teach the same material under names like "law of the press" or "media regulation".
      What it is forKnowing what you may publish, what you must be able to prove, how to protect a source, and why the answers change at a border.
      Comes afterLegal Systems of the World (35/01) for the common-law and civil-law families, and Journalism Ethics (35/95).
      Leads toInvestigative Journalism (35/98), where every claim is checked against what you can prove. See also Constitutional Law (35/03) and European Union Law (35/11).
      What this piece gives youYou will not be able to clear a story for publication. You will see the few rules that shape the choice: who must prove what, why rich claimants can win without winning, how sources are protected, and how press freedom is measured. Easier first reads: Free Speech: The Arguments, the History, and Where Every Democracy Draws the Line (18/13) and Rights, Constitutions & the Rule of Law (18/04).
    
  

  Defamation: protecting a name without silencing criticism
  Defamation is harm to someone's reputation by a false statement of fact. Every legal system asks: when someone is hurt by something published about them, who has to prove what? There are four levers. Must the claimant prove the statement is false, or must the publisher prove it is true? Does it matter that the claimant is a public official? Does the publisher's state of mind matter: knew, was careless, or took care? And is the remedy money, or prison?
  Opinion is treated differently from fact, and truth is a defence in most civil systems. After that the systems part. The tool below lets you set three facts about one invented story and see how a simplified version of each regime responds.
  
    Interactive · One story, four rule-sets
    Who is suing?
      A public officialA private person
    Is the key statement true?
      FalseTrue
    What did the publisher do?
      Knew, or did not careHonest but sloppyChecked well; public interest
    
    Stylised rules for teaching. Real cases turn on facts, forum and procedure. "Inter-American benchmark" is the Commission's 2000 declaration of principles, a standard rather than one country's law. India's column is simplified criminal law.
  
  Without JavaScript: for a carelessly published false statement about an official, the US claim fails, England and Wales turns on the public-interest defence, the Inter-American benchmark allows no sanction, and India's criminal law can still apply.

  The American rule: breathing space for error
  In 1960 a full-page advertisement in The New York Times described the treatment of civil rights protesters in Alabama. It contained small errors, including how many times Martin Luther King Jr. had been arrested. L. B. Sullivan, a Montgomery city commissioner who was not even named, sued. An Alabama jury awarded him $500,000. In New York Times Co. v. Sullivan (376 U.S. 254, decided 9 March 1964) the Supreme Court unanimously threw the judgment out.
  A public official can win only by showing the statement was false and that it was made "with knowledge that it was false or with reckless disregard of whether it was false or not."
  That is the actual malice rule. It is not about spite but about what the publisher knew. The Court's reasoning: open debate will include some mistakes, and punishing every one would make publishers fall silent about the powerful. Debate needs "breathing space". The rule was later extended to public figures (Curtis Publishing Co. v. Butts, 1967), while for private people Gertz v. Robert Welch, Inc. (1974) lets states set a lower fault standard, such as negligence, but not impose liability without fault.
  The rule has critics from several directions. Critics say it leaves people little recourse against spreading falsehoods, and Justice Clarence Thomas has written that the Court should reconsider it. Defenders reply that anything looser would be used by the powerful against reporters. The test below uses four invented stories. Decide whether a court could find actual malice.
  
    Interactive · Predict, then reveal
    
    Could be actual maliceNot actual malice
    Pick one, then read the result.
    Next story
    All four stories are invented. Score is kept on the page only.
  
  Without JavaScript: knowing the story is false is actual malice; an honest error is not; deliberately avoiding the one check that would settle it can be evidence of reckless disregard; a poor general reputation for checking is not the test.
  The rule is not a licence to be wrong. In April 2023 Fox News agreed to pay $787.5 million to Dominion Voting Systems, which had sued over claims that its machines rigged the 2020 election. The judge had ruled in March that none of the statements about Dominion was true, and the remaining question for trial was actual malice. The case settled before a jury answered it.

  England, Europe and the long reach of a claim
  England and Wales take a different road. The statement is treated as false unless the publisher proves it true, and the claimant's status does not change the burden. The Defamation Act 2013 (assent 25 April 2013, mostly in force 1 January 2014) added several changes. A statement is not defamatory unless it has caused or is likely to cause serious harm to reputation (section 1), and there is a defence for publication in the public interest if the publisher reasonably believed it to be so (section 4), replacing the older Reynolds defence. Courts also lost most power to hear claims against defendants outside the UK unless England and Wales is clearly the most appropriate place (section 9).
  That last change answered "libel tourism": claimants with little link to England suing there because the law favoured them. The US Congress had passed the SPEECH Act in 2010 so that American courts would not enforce foreign libel judgments that fall short of US standards.
  Across the continent, the European Convention on Human Rights protects expression in Article 10, and the European Court of Human Rights gives politicians less shelter than private people, since they accept wider criticism (Lingens v. Austria, 1986). Many European systems treat reputation as part of personal dignity, weighed beside speech rather than tolerated narrowly.
  The sharpest modern problem is not losing but being sued. A SLAPP, a strategic lawsuit against public participation, is a claim brought mainly to drain a critic's money and nerve. The EU's anti-SLAPP directive (Directive (EU) 2024/1069, in force 6 May 2024) covers cross-border civil and commercial cases, and member states had to pass national laws by 7 May 2026. On 15 July 2026 the Commission opened infringement proceedings against 14 states for not having reported complete transposition. A non-binding Council of Europe recommendation (CM/Rec(2024)2) also covers criminal and administrative cases. In the United States, as of July 2026 forty states, the District of Columbia and Guam have anti-SLAPP statutes.
  
    Interactive · Why a weak claim still works
    Cost of defending the case
    Claimant's real chance of winning
    Damages if the claimant wins
    Share of costs recovered if you win
    Cost of settling and apologising
    
    
    Invented numbers in thousands of any currency. Expected loss from fighting = defence cost × (1 − chance of winning × share recovered) + chance of losing × damages. A model of incentives, not of any country's costs rules.
  
  Without JavaScript: with a 10% claim, defence costs of 150, damages of 50 and half of costs recovered, fighting is expected to cost about 87.5 against 20 to settle. If costs are fully recovered it drops to about 20.
  That is why reforms target costs and early dismissal, not only the rule of liability.

  Latin America, Africa and Asia: when speech is a crime
  In many countries defamation is also a crime, not only a civil wrong, and some keep older laws of desacato (contempt) that punish offending officials. The Inter-American Commission on Human Rights set out a regional benchmark in its 2000 Declaration of Principles. Principle 11 says that officials are "subject to greater scrutiny by society" and that desacato laws "restrict freedom of expression". Principle 10 says reputation should be protected through civil sanctions only, where the person is an official, a public person or someone who chose to enter a public debate, and the claimant should show intent to harm, knowledge of falsity or gross negligence.
  
    PlaceWhat happened
    ArgentinaThe Inter-American Court of Human Rights ruled for journalist Eduardo Kimel on 2 May 2008: his criminal conviction for criticising a judge's handling of a massacre inquiry violated his rights. Argentina's Law 26.551 (November 2009) removed prison for libel and slander on matters of public interest.
    Other Latin AmericaCosta Rica removed desacato in 2002 and Peru in 2003. Brazil's Supreme Federal Tribunal scrapped the 1967 Press Law in May 2009, though the penal code still has criminal defamation. Bolivia's constitutional court struck down a defamation-of-officials article in 2012. Chile and Colombia still had criminal defamation in CPJ's 2016 survey.
    AfricaIn Konaté v. Burkina Faso (5 December 2014) the African Court on Human and Peoples' Rights found that jailing a journalist for defamation violated freedom of expression. It divided 6 to 4 on the rest of the country's criminal defamation provisions: the ruling is against prison, not all criminal defamation.
    IndiaThe Supreme Court struck down section 66A of the Information Technology Act in Shreya Singhal (24 March 2015) as vague and over-broad, but upheld criminal defamation in Subramanian Swamy (13 May 2016). The offence now sits in section 356 of the Bharatiya Nyaya Sanhita, which replaced the old penal code.
    PhilippinesJournalist Maria Ressa was convicted of cyber libel on 15 June 2020 under the Cybercrime Prevention Act of 2012. She shared the 2021 Nobel Peace Prize with Dmitry Muratov.
  
  The list does not show a free West and an unfree rest. The United States ranks 64th on RSF's 2026 index, and a law on the books may be used often or hardly at all.

  Privacy: the second claim
  A story can be true and still be challenged. A privacy claim is about disclosing private facts, not falsity. The European Court of Human Rights treats private life as a right balanced against speech case by case. In the EU, the General Data Protection Regulation (adopted 2016, applicable from 25 May 2018) lets member states give journalism exemptions from many data-protection rules (Article 85), and the EU's "right to be forgotten" was first spelled out by the Court of Justice in the Google Spain case in 2014.
  The European Court of Human Rights has listed the factors it weighs, in Axel Springer AG v. Germany (2012): whether the story contributes to a debate of general interest, how well known the person is, their prior conduct, how the information was obtained, and the content and consequences of publication. Try them below. This is not a scoring system; a court weighs the factors together. The tally only shows which way each pulls.
  
    Interactive · Which way do the factors pull?
    
    
    
    Three invented stories. The five factors follow the Court's list; the tally is a teaching device, not a legal test.
  
  Without JavaScript: a story about a councillor hiding contracts with her own company hits all five factors for publication; one using a neighbour's hospital record hits none.

  Sources and prior restraint: before and after
  Two more questions run through media law. The first is sources: without a promise of confidentiality, insiders will not talk. In Goodwin v. United Kingdom (27 March 1996) the European Court of Human Rights held that ordering a magazine journalist to name his source under the Contempt of Court Act 1981 breached Article 10, because protecting sources is "one of the basic conditions for press freedom" and an order needs an overriding public interest. The Inter-American declaration (Principle 8) gives every communicator that right too. The EU's European Media Freedom Act (in force 7 May 2024, most provisions applying from 8 August 2025) aims to protect sources, including against spyware.
  The United States has no such constitutional rule. In Branzburg v. Hayes (1972) the Supreme Court held 5 to 4 that reporters have no First Amendment right to refuse to testify about confidential sources. Protection comes instead from statutes: forty states and the District of Columbia have shield laws, and 49 states and D.C. offer some form of protection. A federal PRESS Act passed the House unanimously in January 2024 but, to my knowledge, has not become law.
  The second is prior restraint: a court order stopping a story before it runs. The US treats it as the worst kind of restriction. In 1971 the government tried to stop The New York Times and The Washington Post from publishing the Pentagon Papers.
  
    Interactive · Predict, then compare remedies
    The Times began publishing on 13 June 1971. The government sought an injunction. What did the Supreme Court do?
    Allowed publicationStopped it for national securitySplit it: stopped part
    Pick one, then read the result.
    Three ways to remedy a harmful story
    Injunction firstDamages afterCriminal charge after
    
  
  Without JavaScript: New York Times Co. v. United States (30 June 1971) went 6 to 3 for the newspapers, 17 days after the first article, saying any prior restraint bears "a heavy presumption against its constitutional validity".

  Press freedom, measured
  Reporters Without Borders (RSF) publishes a World Press Freedom Index each spring, scoring 180 countries from 0 to 100 on political, economic, legal, social and safety indicators. The 2026 edition (spring 2026) reported the lowest global average in 25 years, with more than half of countries in the "difficult" or "very serious" bands for the first time, and the legal indicator falling in over 60% of states. The Committee to Protect Journalists (CPJ) counts people: 129 journalists and media workers killed in 2025, the most it has recorded, with Israel responsible for about two thirds of them, and 330 jailed on 1 December 2025, down from a record 384 a year earlier, with China (50), Myanmar (30) and Israel (29) holding the most.
  
    Interactive · Two ways to count
    RSF index score, 2026CPJ counts, 2025
    
    
    RSF scores for ten countries as reported for the 2026 edition; category bands are RSF's (good above 85, satisfactory 70 to 85, problematic 55 to 70, difficult 40 to 55, very serious below 40). CPJ counts are as published by CPJ in February 2026 (killed) and January 2026 (jailed). Indices disagree at the margins and do not measure the same thing.
  
  Without JavaScript: RSF 2026 scores run from Norway 92.72 (1st) to Eritrea 10.24 (180th), with the United States at 62.61 (64th) and the United Kingdom at 79.45 (18th).
  Read these with care. An index squeezes many judgements into one number. A count of killings records the worst events and says little about quieter pressure such as lawsuits and audits, or about self-censorship.

  What a graduate just sees
  
    Defamation law is a set of choices about who proves what, not one rule: check burden, status, state of mind and remedy.
    "Actual malice" is about what the publisher knew, not about spite.
    Where you can be sued matters as much as what you wrote.
    A weak claim can still win if defending it costs more than settling.
    Criminal defamation and desacato laws are a separate question from civil damages, and several regional bodies now call them out of line with free-expression standards.
    A true story can still raise privacy questions; courts weigh public interest against private life.
    Source protection comes from constitutions, courts, statutes or none, depending on the country.
  

  Where it shows up next
  
    Investigative Journalism (35/98): legal read-throughs, the records you can request and how you document a claim so it survives a lawyer's letter.
    Verification & Fact-Checking (35/94) and Journalism Ethics (35/95): the "I checked, and here is how" file is both an ethical practice and a legal defence.
    The Business & Future of News (35/99): SLAPPs fall hardest on small outlets.
    In the world: newsroom lawyers, anti-SLAPP clinics, media-freedom groups, platform moderation policies and every editor who has asked "can we prove that sentence?".
  

  If you want to go deeper
  
    📘 Textbook. Clay Calvert, Dan Kozlowski and Derigan Silver, Mass Media Law (McGraw-Hill, formerly by Don Pember and Clay Calvert; the 20th edition, 2018, is the latest I could confirm, so check for a newer one).
    📘 Gentler. Anthony Lewis, Make No Law: The Sullivan Case and the First Amendment (1991), the story of the case told as narrative.
    🌐 Free reference. Columbia University's Global Freedom of Expression database summarises cases from courts around the world, including Kimel, Konaté and Goodwin.
    🌐 Free material. Media Defence publishes legal training material for journalists and lawyers worldwide (check the site for current modules).
  

  
    The takeaway
    Media law answers who proves what when a story hurts someone, and the answers differ: the US asks officials to prove the publisher knew, England puts the burden on the publisher, much of the world still uses criminal law, and the SLAPP shows that costs can matter as much as rules. Sources and prior restraint get different protections too, and press-freedom data shows how far reality is from the text of a law.
    Try this: take any news story you read this week and ask four questions: who could sue, where, what would they have to prove, and what has the writer done to be able to prove it?
  

  
    Sources & Further Reading
    Sources. The cases, statutes, directives and datasets named above: Sullivan (1964), Curtis, Gertz, Branzburg, New York Times v. United States (1971); the Defamation Act 2013; Goodwin, Lingens, Axel Springer (ECtHR); GDPR Article 85; Directive (EU) 2024/1069; Regulation (EU) 2024/1083; the IACHR Declaration of Principles (2000); Kimel; Konaté; Shreya Singhal; Subramanian Swamy; CPJ's 2016 survey of South American defamation laws; RSF's 2026 index; CPJ's 2025 killed and jailed counts. Details and links are in the fact-check log.
    
      📘 Lewis, Make No Law
      🌐 RSF World Press Freedom Index
      🌐 CPJ data


Step 8. Data Journalism: Finding the Story the Numbers Hold
-----------------------------------------------------------
Why: Finding stories in data, and charts that don't lie.
A spreadsheet never says "this is news". Data journalists learn to ask the question first, divide by the right denominator, distrust the extremes, and draw the chart a sceptical reader would trust.

A headline reads: "Dale has the most break-ins of any town in the region." It is true. Dale has 3,600 reported break-ins a year. It also has 3 million people, and the town at the top of any sensible safety ranking might be a village of 15,000 where 60 break-ins are a worrying rate. The headline was accurate and misleading, and the difference is a single division.
  Data journalism is the craft of finding and telling stories with numbers, and of not being fooled by them. It sits between reporting and statistics: you need a reporter's question, an analyst's care with the data, and a designer's honesty. Most of the work is not clever. It is checking definitions, dividing, comparing like with like, and showing what you did.
  All towns, series and tables in this piece are invented and labelled as such. Real cases are named only where the record is clear.

  
    The course in one box
    
      Who takes itJournalism students, usually in year 3, in courses called data journalism, computer-assisted reporting or computational journalism; newsrooms in many countries (the Guardian, ProPublica, La Nación, Caixin, Africa Check and others) run data teams.
      What it is forTurning datasets, records and statistics into accurate stories, and telling when a number or chart is not what it seems.
      Comes afterVerification & Fact-Checking (35/94) and Statistics (34/05).
      Leads toInvestigative Journalism (35/98), where data often finds the pattern, and The Business & Future of News (35/99).
      What this piece gives youYou will not be able to scrape a website or fit a model. You will see five habits that prevent most errors: the denominator, the small-numbers trap, the honest axis, the data audit and the split by group. Easier first reads: How to Spot Bullshit (17/03) and the statistics section of Statistics (34/05). This goes further in how a newsroom uses them.
    
  

  Start with a question the data can answer
  Data stories start with a question, not a file. In July 1967 Detroit had one of the worst riots in US history; the Detroit Free Press asked its young reporter Philip Meyer to test the popular theory that rioters were uneducated newcomers from the South. His survey of residents found that the theory did not fit the evidence. A paper that had a theory now had a number to hold it against. Meyer went on to write Precision Journalism (1973), the founding text of the field, arguing that reporters should use the methods of social science.
  Data work can also hold algorithms to account. In 2016 ProPublica reporters obtained risk scores used in courts in Broward County, Florida, matched them to later arrests, and reported that the tool wrongly labelled Black defendants who did not reoffend as high risk about twice as often as white ones. The software's maker disputed the conclusion, arguing that scores of a given level meant the same risk of reoffending for each group. Later academic work showed why both sides could be right: when groups have different underlying rates, no score can satisfy every reasonable definition of fairness at once. That is a data story at its best: a finding, a rebuttal and a clearer public question, not a verdict.
  The usual stages are the same everywhere: question → find the data (government records, freedom-of-information requests, scraping what is public, leaks) → clean it → analyse → check with reporting (call the people behind the dots) → show it. The last check matters most: a pattern in a spreadsheet is a lead, and the story comes from reporting on why.

  A number needs a denominator
  Raw counts mostly measure size. The biggest places have the most crimes, cases, deaths and complaints because they have the most people, so a ranking by count is often a ranking by population. The fix is a rate: events divided by the people (or households, or kilometres, or hours) exposed to them, then scaled to a readable unit such as per 100,000.
  rate = events ÷ population × 100,000Dale: 3,600 ÷ 3,000,000 × 100,000 = 120 per 100,000. Elm: 60 ÷ 15,000 × 100,000 = 400.
  Per-capita is not magic. It assumes everyone in the denominator is equally at risk (a town full of tourists or commuters is not), and a good story asks who is really in the base. But it is the first fix. Two related traps: a "50% increase" on a tiny base can be three cases, and a rise of 2 percentage points (from 10% to 12%) is a 20% relative rise. Say which one you mean.
  
    Interactive · Raw count or per-capita?
    Raw countPer 100,000 people
    
    
    Invented towns. The same six places rank in nearly the opposite order.
  
  Without JavaScript: by raw count Dale (3,600) is first and Elm (60) last; per 100,000 Elm (400) is first and Dale (120) last.

  Small places, wild rates
  Rates fix one problem and create another. A town of 1,000 people with one extra event has a rate that jumps by 100 per 100,000. The statistician Andrew Gelman has used a famous example: in US maps of kidney-cancer death rates by county, the lowest-rate counties and the highest-rate counties were both mostly small, rural ones. The reason is not something in the water or the air. Small samples are noisy, and noise produces extremes in both directions.
  The press has to be careful here. "Village has the nation's highest cancer rate" is exactly the story a rate-ranked list will keep producing, and the right response is to ask how many events it rests on. Good data teams use a minimum count, pool several years, or draw uncertainty bands. The simulation below gives every town the same true rate. Run it a few times.
  
    Interactive · Sixty towns, one true rate
    
    Run another year
    
    Simulated with a seeded random-number generator; events follow a Poisson distribution. The dashed line is the true rate. Axis is capped at 400.
  
  Without JavaScript: when every town has the same true rate, the highest and lowest observed rates both belong to towns far smaller than the typical one, because small counts swing.

  Charts that lie, and their honest versions
  Darrell Huff's How to Lie with Statistics (1954) and Edward Tufte's The Visual Display of Quantitative Information (1983) both ask whether a graphic's size on the page matches the size of the change in the numbers. Alberto Cairo's How Charts Lie (2019) is the modern guide. Four of the commonest tricks:
  
    The truncated axis. Start a bar chart at 410 instead of 0 and a 5% change looks like a ten-fold one. For bars, the length is the message, so bars should start at zero. For lines, a non-zero start can be fair if it is labelled; the test is whether the reader can tell.
    The cherry-picked window. Pick the three months that rise and call it a trend.
    Dual axes. Two lines on different scales will cross wherever you like.
    Hidden choices. Cumulative totals that can only rise, areas where only length matters, and maps that colour land, not people.
  
  
    Interactive · One year of invented complaints
    Axis from 0Axis from 410
    Window starts
    Window ends
    
    
    Invented monthly complaint counts. Try Jan–Mar with the axis from 410, then Mar–Oct, then the full year with the axis from 0.
  
  Without JavaScript: across the whole year complaints fell about 0.9% (430 to 426). January to March they rose 5.1%; March to October they fell 7.5%. A truncated axis makes every one of those look dramatic.
  The honest version shows the full period, starts the axis where the reader expects, says what the unit is, and puts the surprising number in the headline only when the chart shows it fairly. Newsroom style guides increasingly add a rule: if a chart was cut to fit, say where and why.

  Is the data what you think it is?
  Before analysing, interview the dataset. Who collected it, to count what, with what definition? Did the definition change? Many real "trends" are changes in counting: a new reporting system, a change in what counts as a case, a shift from paper to digital. Case counts of any disease depend on how many people are tested; crime figures depend on whether victims report; a country's statistics office may be under pressure, so cross-check with an independent source (the World Bank, UN agencies or a university) and say where they differ. Africa Check's work on official crime and inequality statistics is built on this habit.
  Then look at the raw rows: values of 999 or -1 (missing, coded as a number), a column mixing units, duplicates, and typing slips. These are the numbers that do the damage, because one stray 999 can drag an average far from the truth.
  
    Interactive · Audit the column
    Invented emergency response times in minutes. Press each fix and watch the average.
    
    #3: 999 means "missing"#6: 0.2 is in hours#8: drop the duplicate
    
  
  Without JavaScript: the raw average is 133.0 minutes; after treating 999 as missing, converting 0.2 hours to 12 minutes and dropping the duplicate, it is 11.3 minutes.

  Two groups, opposite stories
  The most famous warning is Berkeley. In the autumn of 1973 about 44% of men and 35% of women who applied to the graduate programmes were admitted, which looked like clear bias. Peter Bickel, Eugene Hammel and J. W. O'Connell (Science, 1975) split the numbers by department and found that, in most departments, women's admission rates were as high or higher. Women had applied more to departments that rejected more people. The combined figure was real, and what it meant needed the split.
  This is Simpson's paradox: a trend in the whole can reverse inside every group when the groups are different sizes. It is not a trick of arithmetic; it is a warning that a cause lives in the structure behind the totals. The choice of what to split by cannot be settled by the data alone. A reporter decides it by knowing how the process works, and says so. The same care applies to causation: two lines that move together (ice cream and drownings) are usually both driven by something else, here summer. See Statistics (34/05) for the general case.
  
    Interactive · Invented admissions
    CombinedDepartment ADepartment B
    
    
    Invented numbers in the shape of the Berkeley case. Dept A admits most applicants; Dept B admits few, and most women applied there.
  
  Without JavaScript: in Dept A men are admitted at 80% and women at 90%; in Dept B men at 20% and women at 30%. Combined, men are admitted at 70% and women at 40%.

  Showing your work
  Data journalism earns trust by being checkable. Good practice now includes a methodology note (where the data came from, how it was cleaned, what was excluded), the code or a spreadsheet where possible, and an honest statement of uncertainty. For a poll that means the margin of error; for a model it means a range, not a point. The best newsrooms also publish corrections to data stories as readily as to words. And they ask the human question the numbers cannot answer: if a chart shows that one neighbourhood has the highest rate of something, what is it like to live there, and does the person who lives there recognise it?

  What a graduate just sees
  
    "The most" usually means "the biggest": divide by the people exposed.
    The highest and lowest rates tend to belong to the smallest places; ask how many events sit behind each.
    A bar chart should start at zero; any other choice should be visible and justified.
    A window of a few months can show any trend you like; show the whole period.
    A jump in a series may be a change in counting, not in the world.
    A total can tell the opposite story to every one of its parts.
    A pattern is a lead. The story is the reporting that explains it.
  

  Where it shows up next
  
    Investigative Journalism (35/98): spotting the anomaly in a contract list or an ownership register before a reporter picks up the phone.
    Elections, Parties & Public Opinion (35/15): margins of error and how a poll is turned into a headline.
    The Business & Future of News (35/99): analytics, audience data and what a newsroom measures about itself.
    In the world: public-health dashboards, budget trackers, climate and air-quality maps, sports analytics, and every chart you scroll past.
  

  If you want to go deeper
  
    📘 Textbook. Philip Meyer, Precision Journalism (Rowman & Littlefield, fourth edition 2002), and Alberto Cairo, The Truthful Art (New Riders, 2016), for analysis and chart design.
    📘 Gentler. Alberto Cairo, How Charts Lie (Norton, 2019), and Darrell Huff, How to Lie with Statistics (1954).
    🌐 Free text. The Data Journalism Handbook (free online, edited by Jonathan Gray, Liliana Bounegru and Lucy Chambers, 2012), with a newer open-access volume, The Data Journalism Handbook: Towards a Critical Data Practice (Amsterdam University Press, 2021).
    🌐 Free course. Khan Academy's Statistics and Probability (free); the Knight Center for Journalism in the Americas also runs free online courses in data journalism, in several languages.
  

  
    The takeaway
    A number is not a story until it is compared with something. Divide by the right base, distrust the extremes from small places, draw the chart a sceptic would accept, audit the data before using it, and look inside the total. Then call the people the numbers describe.
    Try this: find a "top ten" list in the news and look up the population of each place; does the order survive dividing?
  

  
    Sources & Further Reading
    Sources. Meyer, Precision Journalism and the Detroit Free Press 1967 riot survey (the paper won a Pulitzer Prize in 1968 for its riot coverage); ProPublica, "Machine Bias" (23 May 2016) and the maker's published response; Kleinberg, Mullainathan and Raghavan (2016) on incompatible fairness criteria; Gelman and Nolan on the kidney-cancer map and Wainer, "The Most Dangerous Equation" (2007); Bickel, Hammel and O'Connell, Science 187 (1975); Huff, Tufte, Cairo. All towns, series and tables here are invented; every number was computed.
    
      📘 Cairo, How Charts Lie
      📘 Meyer, Precision Journalism
      🌐 DataJournalism.com


Step 9. Elections, Parties & Public Opinion: Rules That Turn Votes into Power
-----------------------------------------------------------------------------
Why: Polls, elections and public opinion.
The same ballots can give one party a landslide or three parties a coalition. Which one happens depends on arithmetic you can learn in an afternoon, and on what polls can and cannot see.

At the United Kingdom's general election on 4 July 2024, the Labour Party won 33.7% of the votes and 411 of the 650 seats, about 63% of the House of Commons. Reform UK won 14.3% of the votes and 5 seats, under 1%. Turnout was 59.7%. Nobody cheated and nobody miscounted. That is simply what the rule, one winner per constituency, does with votes.
  In a country that uses proportional representation, the same votes would have produced a very different parliament. And in both cases, in the months before the vote, pollsters had tried to tell the country what would happen using a few thousand phone calls or online panels. This piece is about those two machines: the one that turns votes into seats, and the one that tries to measure votes before they are cast. Plus the oldest question of all: why people vote the way they do, or at all.
  This describes how elections and polling work. It does not say which party or system is better; each electoral system trades something off, and every example is given to show a mechanism.

  
    The course in one box
    
      Who takes itEvery political science major, usually in year 2, and students of sociology, economics, law and journalism. Programmes such as MIT's Course 17, Oxford's Politics in PPE, Sciences Po and Harvard Government teach electoral systems, parties and public opinion in a core course, often with a methods course beside it.
      What it is forUnderstanding how rules, parties and attitudes combine to decide who governs, and how to read a poll or a result without being fooled.
      Comes afterComparative Politics (35/13) and Probability & Statistics (34/05).
      Leads toPublic Policy (35/16) and Research Methods in Political Science (35/18).
      What this piece gives youYou will not be able to run a campaign or design a survey. You will be able to predict what an electoral rule does to the number of parties, say why a poll can be off, and understand why a two-point lead is not a certainty. Easier first reads: How Voting Systems Shape Outcomes (18/03), Democracy: The Many Flavors (18/02) and The Political Spectrum, Fairly Presented (18/05).
    
  

  Seats are made by rules: plurality versus proportional
  Every electoral system answers one question: how do votes become seats? There are three big families. In plurality systems (the United Kingdom, Canada, India, the United States for most offices), each district elects one person, and whoever gets the most votes wins, even with less than half. In proportional representation (PR) systems (the Netherlands, Spain, Brazil's lower house, South Africa, Israel and most of continental Europe), a party's share of seats follows its share of votes, using a formula. Mixed systems combine the two (Germany, New Zealand, Mexico).
  Two PR formulas are common. The D'Hondt method (Victor d'Hondt, 1878; used for example in Spain and Portugal) gives seats one at a time to the party with the highest quotient votes ÷ (seats won + 1). Sainte-Laguë (André Sainte-Laguë, 1910; used for the German Bundestag and in New Zealand) divides by 1, 3, 5 and so on, which is a little kinder to small parties. A threshold (5% in Germany) excludes parties that fall below it.
  Political scientists measure the damage a system does to proportionality with the Gallagher index (Michael Gallagher, 1991): the square root of half the sum of squared gaps between each party's vote share and seat share. Zero is a perfect match. They also count how many parties matter with the effective number of parties (Laakso and Taagepera, 1979): one divided by the sum of squared shares. Four equal parties give 4; one party with 90% gives about 1.2. Below, ten districts, four parties, and an invented set of votes show all three systems side by side.
  
    Interactive · One set of votes, three rules
    Shift voters between A and B in every district (points)
      
    PR threshold (national vote %)
      
    
    
    Invented data for ten single-member districts and four parties; a PR system with 10 seats nationally is compared with 10 single-member districts. Shifting moves votes from A to B (negative: B to A) in each district, never below zero. Party D is strong in just one district.
  
  At the default setting, party A has 33.5% of the votes and wins 7 of 10 seats under plurality, while party C, with 24.2% spread evenly, wins none. The PR systems give A 4 seats (D'Hondt) or 3 (Sainte-Laguë) and C 2 or 3. Slide the shift and watch a small change in votes swing several seats. That sensitivity is the mechanism behind the "landslide on a minority of votes" in the UK example. Raise the threshold to 15% and see D drop out. Real systems add many details (district size, open or closed lists, bonus seats), and each has its defenders: plurality gives clear single-party governments and a local member; PR gives fairer representation and more coalitions.

  Duverger's law: why plurality squeezes parties
  The French jurist and sociologist Maurice Duverger noticed, in Les partis politiques (1951), that countries with plurality elections in single-member districts tend to have two main parties, while PR tends toward many. He gave two reasons. A mechanical effect: a party that wins 15% of the votes everywhere wins no seats, as party C just did. And a psychological effect, which is the stronger one in modern work: voters see this and avoid "wasting" a vote on a party that cannot win, so they back their favourite of the two who can. The tendency is not a law; India and Canada have regional parties that win locally, and Duverger's logic works district by district rather than nationally.
  You can see the psychological effect in a small model. Voters sit on a left-to-right line (more are near the middle). Candidate A stands at −1.0 and candidate B at 0.3. Candidate C's position is yours to set. Voters sincerely back the closest candidate. Then the voters of whichever candidate is last in the polls start to doubt: a share of them switch to their preferred of the other two. Try a plurality vote, and then a two-round runoff, in which the top two meet again in a second round.
  
    Interactive · Strategic voting and the spoiler
    
      One round (plurality)
      Two-round runoff
    
    Candidate C's position (A is at −1.0, B at 0.3)
      
    Share of the last-place candidate's voters who switch
      
    
    
    Voters are normally distributed around 0 (the median voter). A at −1.0, B at 0.3. Plurality: most votes wins. Runoff: the top two meet again and voters back the closer one. "Closest to the median" is the candidate who would beat each rival head to head (the Condorcet winner).
  
  With C at 0.8 and nobody switching, A wins with 36% under plurality, even though B, the candidate closest to the middle, would beat either rival in a head-to-head vote. Let half of C's voters switch, and B wins. A runoff produces the same outcome without anyone needing to guess: C's voters back B in round two. Move C to 0.5, next to B, and it is B that gets squeezed out. This is why parties that share a space worry about "splitting the vote", why many countries pair presidential elections with runoffs (France, Brazil, Argentina, Chile, Colombia), and why electoral rules change how parties form.

  Why people vote as they do
  Anthony Downs's An Economic Theory of Democracy (1957) gave the field its simplest model. If voters sit on a line and each backs the closer candidate, two vote-seeking candidates are pulled toward the position of the median voter, the voter with half the electorate on either side. Duncan Black proved the underlying result in 1948. Try it, and then add the realistic twist: a voter who feels too far from both candidates stays home.
  
    Interactive · The median voter, and what pushes candidates apart
    Candidate 1 position
      
    Candidate 2 position
      
    Voters' average position
      
    Voter stays home if candidates are farther than
      
    
    Voters follow a bell curve (standard deviation 1) around the chosen average. A setting of 6 means nobody abstains.
  
  Drag candidate 1 toward the middle: they gain, and the other candidate must follow. If both end at the same spot, the vote splits evenly. Then lower the "stays home" slider: the candidates at the extremes lose voters who prefer neither, which is why turnout and the pull toward the centre trade off. Real elections do not collapse to the middle, for many reasons: parties have activists and loyal supporters, voters care about more than one issue, and in multiparty systems each party aims at its own group.
  Political scientists also ask what drives the vote in practice. The Michigan school's The American Voter (1960, Campbell, Converse, Miller and Stokes) found that party identification, a long-term emotional attachment often learned in the family, predicts the vote better than issue positions. Philip Converse (1964) found many voters held opinions that were not tightly connected across issues. Morris Fiorina (1981) emphasised retrospective voting: reward or punish the government for how things went. Christopher Achen and Larry Bartels (Democracy for Realists, 2016) argued that voters respond to events they should logically ignore, and cited a study suggesting that shark attacks on the New Jersey shore in 1916 cost Woodrow Wilson votes. That claim did not hold up: Anthony Fowler and Andrew Hall (2018) re-examined every fatal shark attack in the US and presidential returns from 1872 to 2012 and found little compelling evidence of an effect. Achen and Bartels have replied that their broader argument stands. It is a good example of why the field now checks striking findings against more data.

  Turnout: why anyone votes at all
  Downs raised an awkward problem. Voting costs time, and the chance that one vote decides an election is tiny. William Riker and Peter Ordeshook (1968) wrote the voter's calculation as R = pB − C + D: the chance of being decisive times the benefit, minus the cost, plus a sense of duty or expression (D). Without D, almost nobody would vote. A rough size: even in the most favourable case, an election dead-level on average, the chance that one vote is the tie is about √(2 ÷ (π N)) for N voters. For a million voters that is about 0.08%, or 1 in 1,250, and real elections are rarely so close, so the true number is far smaller.
  Yet hundreds of millions vote. Turnout differs sharply between countries, and rules explain part of that. In Australia, voting at federal elections has been compulsory since 1924, and turnout has rarely fallen below 90%. Belgium, the oldest compulsory system (1893), has about 90%. Argentina adopted compulsory voting in 1912, Brazil in 1932, and Peru, Bolivia and Uruguay also enforce it. Fines, or small bureaucratic penalties such as the difficulty of getting a passport, are the enforcement. In the UK in 2024, with voluntary voting, turnout was 59.7%. Scholars debate whether compulsory voting changes results or only adds reluctant voters, and whether people grow into the habit.

  Polling and its errors
  A poll asks a few hundred or thousand people and infers the opinions of millions. The inference works because of a result you met in Statistics: if you draw a truly random sample of size n from a population in which a share p holds an opinion, your estimate will land within about 1.96 × √(p(1 − p) ÷ n) of the truth 95% of the time. For p = 50% and n = 1,000, that is ±3.1 points; for n = 4,000 it is ±1.55. Halving the margin costs four times the sample. And the margin on the lead between two candidates is roughly double.
  The catch is the word "truly random". The classic failure is the Literary Digest poll of 1936, which mailed ballots to about 10 million people drawn from its own subscribers, telephone directories and car registrations, got 2.38 million back, and predicted a landslide win for Alf Landon. Franklin Roosevelt won with 60.8% of the vote. A huge sample did not help, because it was a biased one, and the people who strongly disliked Roosevelt were more likely to answer. A modern version is non-response: people who do not pick up the phone differ from those who do. The American Association for Public Opinion Research (AAPOR) task force on the 2020 US election found that national polls overstated Joe Biden's lead by about 3.9 points and state polls by 4.3, the largest national polling error in about 40 years, and could not pin the cause on a single factor (non-response was the leading suspect). The UK's 2015 polling inquiry, led by Patrick Sturgis, blamed unrepresentative samples too.
  Simulate it. Each click draws 1,000 polls from a population where one candidate really has the share you choose. Then add a bias, the tilt that non-response introduces.
  
    Interactive · A thousand polls
    Sample size per poll
      
    True share for candidate X (%)
      
    Bias from non-response (points, against X when negative)
      
    Draw 1,000 new pollsHuge but biased
    
    
    Each poll samples n people at random from a population in which X's true share, plus the bias, is the chance of a "yes". Margin of error is 1.96 √(p(1−p)/n) around the true share.
  
  Leave the bias at zero: about 95% of polls land within the margin of error, and with a 52% truth and n = 1,000 around one poll in ten calls the wrong winner. Now press "Huge but biased": five thousand people per poll, bias of 4 points. The margin of error shrinks to under 1.4 points, yet almost no poll lands within it and nearly all call the wrong winner. Sampling error shrinks as samples grow; bias does not. Good pollsters weight their samples to match known population figures, and aggregators average many polls, but a shared bias passes through both.

  Polarisation, measured
  "Polarisation" is used for several different things. Ideological polarisation is how far apart people's positions on issues are. Affective polarisation is how much people dislike and distrust the other side. They need not move together. The most-cited cross-country study, by Levi Boxell, Matthew Gentzkow and Jesse Shapiro, measured affective polarisation (negative feeling toward opposing parties) in twelve OECD countries over four decades. The United States showed the largest rise; five other countries rose by less, and six fell. In other words, there is no worldwide trend that everyone shares, and the American pattern is the exception more than the rule, though the causes are debated.
  Other measures exist: the V-Dem project, for example, publishes an expert-coded indicator of political polarisation for each country-year, and survey projects ask people how warmly they feel toward the parties. Different indices count different things, so check what an index measures before comparing countries. Researchers also disagree about causes: some point to the sorting of parties by identity, some to media, some to electoral rules (plurality tends to create two big opposing camps; PR, more cross-cutting coalitions), and none of these explains everything.

  What a graduate just sees
  
    The electoral rule is part of the result. Ask how votes become seats before arguing about who "won".
    Plurality tends toward two parties (Duverger), and small shifts in votes can swing many seats.
    Proportional systems track votes closely, and the formula (D'Hondt, Sainte-Laguë) and threshold decide who is just in or out.
    A party that splits a camp can hand the seat to the other side; runoffs and ranked ballots exist to fix that.
    Candidates are pulled toward the median voter, and held away from it by activists and abstention.
    For one voter the chance of deciding is tiny, which is why turnout is partly duty and partly rules.
    A poll's margin of error covers only sampling luck. Bias, from non-response or an unrepresentative sample, is a separate and often bigger risk.
    "Polarisation" can mean disagreement or dislike, and measures differ across countries.
  

  Where it shows up next
  
    Public Policy (35/16): coalitions, parties and veto players shape what gets proposed and passed.
    Research Methods in Political Science (35/18): sampling, bias and causal claims, the tools behind every poll and every "voters do X" finding.
    Probability & Statistics (34/05): margin of error, sample size and bias are the same ideas you meet there.
    In the world: campaign and polling firms, election administration and observation (the OSCE, the Carter Center), journalists who report polls, civil servants who must govern with a coalition, and anyone who has to read a result and a forecast sensibly.
  

  If you want to go deeper
  
    📘 Textbook. Michael Gallagher and Paul Mitchell (eds.), The Politics of Electoral Systems (Oxford University Press, 2005). Country-by-country coverage of how each system works.
    📘 Textbook. Arend Lijphart, Patterns of Democracy (Yale University Press, 1999; 2nd edition 2012). Compares 36 democracies on electoral systems, party systems and executives.
    📘 Gentler. Nate Silver, The Signal and the Noise (Penguin, 2012), a readable account of forecasting, including election forecasts and why they miss; and Achen and Bartels, Democracy for Realists (Princeton, 2016), a harder read but written for general audiences, for the case that voters are less issue-driven than we assume.
    🌐 Free course. MIT OpenCourseWare 17.50 Introduction to Comparative Politics covers electoral systems, party systems and voting alongside regimes. For electoral systems in particular, the ACE Electoral Knowledge Network (run by International IDEA and other partners) has a free encyclopaedia with an Electoral Systems section.
  

  
    The takeaway
    Votes become power through a rule, and the rule shapes how many parties exist, who splits whose vote and how far a small swing travels. Voters are pulled by party identity, by how things went, and by where the candidates stand, and polls measure them well only when the sample is random: size shrinks luck, not bias.
    Try this: next election you hear about, find the rule first (one winner per district or proportional?) and the margin of error on any poll. Then ask how far the "lead" is from the margin, and what share of votes the winner needed to win.
  

  
    Sources & Further Reading
    Sources. UK 4 July 2024 general election: Labour 33.7% of votes and 411 of 650 seats; Reform UK 14.3% and 5 seats; turnout 59.7% (as compiled on Wikipedia from the House of Commons Library and Electoral Commission; checked against the 650-seat total). Duverger, Les partis politiques (1951); Laakso and Taagepera, "Effective Number of Parties", Comparative Political Studies 12(1), 1979; Gallagher, "Proportionality, Disproportionality and Electoral Systems", Electoral Studies 10(1), 1991; Downs, An Economic Theory of Democracy (1957); Black (1948); Campbell et al., The American Voter (1960); Converse (1964); Fiorina, Retrospective Voting in American National Elections (1981); Riker and Ordeshook, "A Theory of the Calculus of Voting", APSR 62(1), 1968; Achen and Bartels, Democracy for Realists (2016); Fowler and Hall, "Do Shark Attacks Influence Presidential Elections?", Journal of Politics 80(4), 2018. Compulsory voting: Australia since 1924, Belgium since 1893, Argentina 1912, Brazil 1932 (as summarised on Wikipedia's compulsory-voting page). Literary Digest 1936: about 10 million mailed, 2.38 million returned; Landon predicted at 57.08%, Roosevelt won 60.8% (Wikipedia, with non-response as the main cause). AAPOR Task Force on 2020 Pre-Election Polling (2021): national polls overstated Biden's lead by 3.9 points, state polls by 4.3, the largest national error in about 40 years. UK polling inquiry: Sturgis et al., report for the British Polling Council and Market Research Society (2016). Boxell, Gentzkow and Shapiro, "Cross-Country Trends in Affective Polarization" (NBER Working Paper 26669). V-Dem Democracy Report 2026 as reported. Every number an interactive shows was computed in node.
    
      🌐 ACE Electoral Knowledge Network
      🌐 MIT OCW 17.50 Introduction to Comparative Politics
      📘 Gallagher & Mitchell, The Politics of Electoral Systems
      📘 Lijphart, Patterns of Democracy


Step 10. Investigative Journalism: How a Tip Becomes a Story That Holds
-----------------------------------------------------------------------
Why: How investigations are built, from a tip to publication.
Investigations are slow, documented and legally checked. The famous ones show the method: build every claim from evidence, give people a chance to answer, protect the people who helped, and accept that consequences come on their own timetable.

.opt{display:flex;gap:10px;align-items:flex-start;padding:8px 6px;border-bottom:1px solid var(--line);font-family:-apple-system,sans-serif;font-size:.92rem}
  .opt input{margin-top:.25em;min-width:22px;min-height:22px}
  .opt label{flex:1;min-width:0}

  
  
  
  

  In January 2002 the Boston Globe published a story about a single priest, John Geoghan, who had been moved between parishes despite complaints about him. The paper had spent months in court getting sealed documents opened, and months more checking them. Within a year the Archbishop of Boston had resigned and the story had become a worldwide reckoning with how institutions protect themselves.
  That arc, a tip, a long grind, a story that holds up under attack, is what separates investigative journalism from a scoop. A scoop is first. An investigation is proved: someone is accused of something serious, and the evidence is laid out so that a hostile reader, a lawyer and a court can all examine it. This piece takes real, well-documented cases and shows the method. It is descriptive: how newsrooms work, not legal advice and not a guide to anything risky.

  
    The course in one box
    
      Who takes itJournalism students, usually in year 4, often as a capstone or a workshop with a working newsroom. Postgraduate programmes and the Global Investigative Journalism Network's trainings teach it in dozens of countries.
      What it is forPlanning, proving and publishing accountability stories about people with power, safely and fairly.
      Comes afterVerification & Fact-Checking (35/94) and Media Law & Press Freedom (35/96).
      Leads toThe Business & Future of News (35/99), since investigations are expensive and the economics decide who can afford them.
      What this piece gives youYou will not be able to run an investigation, which takes mentors and years. You will see the eight stages, why evidence is built the way it is, how cross-border teams work, and what the law and personal risk look like. Easier first read: Propaganda, Misinformation and How to Resist It (18/09) and Free Speech: the Arguments, the History and the Lines (18/13).
    
  

  The shape of an investigation
  Every serious investigation goes through roughly the same stages, whatever the country or the subject. The names differ between newsrooms, but the logic is constant: you begin with a sentence you can disprove, collect more evidence than you need, and stop only when each claim can survive a cross-examination.
  Hunter's Story-Based Inquiry, the manual UNESCO published in 2009 for reporters in many countries, makes the first move explicit: start with a hypothesis, a plain statement of what you think happened, then plan the work as a search for evidence that would confirm or kill it. A reporter who only collects what supports the idea has not investigated.
  
    Interactive · The eight stages
    
    
    Next stage
    The "in practice" lines are drawn from the cases told below.
  
  Without JavaScript: tip, hypothesis, gather, corroborate, reply, legal review, publish, follow up.

  Evidence: two sources, or one document you can check
  The Watergate reporting is the best-known American case. On 17 June 1972 five men were arrested inside the Democratic National Committee's offices in Washington. Two young Washington Post reporters, Bob Woodward and Carl Bernstein, with editor Ben Bradlee, followed the money from the burglars to President Nixon's re-election campaign. A government source they called "Deep Throat" was identified in 2005 as Mark Felt, then the FBI's deputy director.
  The honest version has several parts. The Post's work kept the story alive, won the 1973 Pulitzer Prize for Public Service, and was not infallible: an October 1972 story about a grand jury and a White House aide overstated what a source had said, and the administration used the error to attack the paper. The consequences came from elsewhere: a judge, John Sirica, who pressed the burglars; a Senate committee whose 1973 hearings revealed the existence of White House tape recordings; and the Supreme Court's 8–0 order of 24 July 1974 that Nixon hand them over. He resigned on 9 August 1974. Journalism supplied evidence and attention. Institutions supplied the penalty.
  In 1971 the New York Times and then the Washington Post published the Pentagon Papers, a secret 47-volume history of the Vietnam War of about 7,000 pages that Defense Secretary Robert McNamara had commissioned in 1967. Daniel Ellsberg, who had worked on it, copied it and gave it to reporters; the Times began publishing on 13 June 1971. The government won a temporary injunction, a rare prior restraint on a newspaper, and the Supreme Court lifted it on 30 June in New York Times Co. v. United States (6–3). Ellsberg and his colleague Anthony Russo were prosecuted, and the case was dismissed in 1973 because of government misconduct. The lesson for method: the document was real and the paper checked it, yet the real fight was legal.
  That is why the bar is high. A common newsroom rule is that a serious claim needs either a document you can authenticate plus a person who can explain it, or two independent human sources. A single anonymous source is a lead. The board below uses that rule on an invented story.
  
    Interactive · Can we publish it?
    Invented: a town contract. Tick what you have found; see which claims are ready.
    
    Reply: Councillor Reyes and Tidewell were sent the findings and given time to respond
    
    A teaching rule (document + person, or two documents; an anonymous tip never alone; named people asked to reply), not a legal standard. All names are invented.
  
  Without JavaScript: a claim needs a checked document plus a person who can explain it, or two documents. A single anonymous email is a lead. Claims about named people also need their reply.

  Follow the money: ownership and the Panama Papers
  A great many investigations come down to one question: who really owns this? Companies can own companies, and a chain of "holding companies" in different countries can hide the person at the end. The effective share is the product of the shares in each link, and control is a different thing.
  your share = 70% × 70% × 70% × 70% = 24.0%Four links at 70%: you control every company in the chain, yet fall below a 25% disclosure line
  
    Interactive · Who owns Tidewell?
    You own this share of Holdco A
    A owns this share of B
    B owns this share of C
    C owns this share of Tidewell
    
    Invented companies. Many beneficial-ownership rules, such as the EU's anti-money-laundering directives, use a 25% line; countries differ and some use lower ones.
  
  Without JavaScript: with four links of 70%, your effective share is about 24%, below a 25% disclosure line, but you control every company in the chain because 70% is more than half at each step.
  The Panama Papers showed how this works at scale. In 2015 an anonymous source, calling themself "John Doe", gave the German newspaper Süddeutsche Zeitung about 11.5 million documents (2.6 terabytes) from Mossack Fonseca, a Panama law firm that created offshore companies. The paper brought in the International Consortium of Investigative Journalists (ICIJ), founded in 1997, and journalists in about 80 countries worked on the files for a year. They began publishing on 3 April 2016. Iceland's prime minister, Sigmundur Davíð Gunnlaugsson, resigned on 5 April after the files linked him to an offshore company; the firm closed in 2018; and the project shared the 2017 Pulitzer Prize for Explanatory Reporting. Holding an offshore company is often legal. The story was about secrecy, tax avoidance and who benefits, and where the use was improper, the evidence had to say so.

  Many newsrooms, one story: leaks, networks and Lava Jato
  A data leak of millions of files cannot be read by one newsroom, and the people with the most to lose can sue or threaten in the place the paper is based. So investigative journalists now share. ICIJ's networks led to the Paradise Papers (2017) and the Pandora Papers (2021). Africa has produced its own: the Luanda Leaks (January 2020, about 715,000 files on the finances of Angola's Isabel dos Santos, with the Platform to Protect Whistleblowers in Africa) and the South African GuptaLeaks (2017, a large cache of emails published by amaBhungane and Daily Maverick), which fed the public inquiry into "state capture". In Ghana, Anas Aremeyaw Anas's undercover team filmed football officials taking cash in "Number 12" (2018); his methods are debated by press-ethics scholars, and a colleague, Ahmed Hussein-Suale, was shot dead in January 2019.
  In Brazil, Operation Car Wash (Lava Jato) began in March 2014 as a police investigation into money laundering and grew into one of the largest corruption cases in history, centred on the oil company Petrobras and contractors including Odebrecht, which admitted bribing officials in twelve countries across the Americas and Africa. Here the press was often a recipient: reporters published leaked plea statements and police files. In 2019 The Intercept Brasil published leaked chat messages that, in its reading, showed the judge, Sergio Moro, advising prosecutors; Moro and the prosecutors disputed what they showed. In 2021 Brazil's Supreme Court annulled Lula's convictions on jurisdiction grounds and a panel later found Moro had been partial. That chain shows both halves of the craft: journalists exposing corruption, then journalists examining the investigators who exposed it.
  In Asia, Reuters reporters Wa Lone and Kyaw Soe Oo were arrested in Myanmar in December 2017 while investigating the killing of ten Rohingya men and boys at Inn Din village. They were sentenced to seven years under the Official Secrets Act, released in May 2019 after a presidential amnesty, and shared the 2019 Pulitzer Prize for International Reporting. In the Philippines, Maria Ressa of Rappler shared the 2021 Nobel Peace Prize with Russia's Dmitry Muratov for work on freedom of expression; she was convicted of cyber libel in 2020, a verdict she appealed.

  Right of reply and legal review
  Two steps happen late and decide whether the story survives. The right of reply: before you publish a serious claim, you tell the person or company what you will say and give them reasonable time to respond. It is fair, it sometimes turns up the missing fact, and in several legal systems it matters for the defence. Then legal review: a lawyer reads every sentence and asks whether it can be proved, whether it is fact or opinion, and whether it serves the public interest.
  The rules differ by country. This table is an overview of how some systems work, not legal advice.
  
    PlaceThe shape of the rule
    United StatesTruth is a defence. For public officials the plaintiff must also show "actual malice" (knowing falsehood or reckless disregard), from New York Times v. Sullivan (1964).
    England and WalesDefamation Act 2013: a claimant must show serious harm; defences include truth, honest opinion and publication in the public interest, which protects reasonable belief on a matter of public interest.
    EuropeArticle 10 of the European Convention on Human Rights protects the press; the European Court of Human Rights treats public-interest reporting and the watchdog role as weighty. The EU's 2024 anti-SLAPP directive targets abusive lawsuits.
    Latin AmericaThe Inter-American system opposes desacato (insulting officials) laws and criminal defamation for public-interest speech; practice varies by country, and several states have repealed such laws.
  
  
    Interactive · What would the desk do with each sentence?
    Invented story, using the evidence above.
    
    Choose for each sentence.
    A drill in newsroom habits, not a legal test: the rules differ by place and the facts of each case matter.
  
  Without JavaScript: do not state a crime you cannot prove; publish what public records show as written; attribute allegations to a named source and include the denial; and keep conclusions out of news copy.

  Protecting sources, and the people who do this work
  Without sources there are no investigations, and the promise to keep a name secret is the most serious one a journalist makes. Protection comes from several places and none is complete.
  
    Law. In the United States there is no general federal reporter's privilege: Branzburg v. Hayes (1972) held that reporters must answer grand-jury questions, though most states have shield laws. In 2005 the New York Times reporter Judith Miller spent 85 days in jail rather than name a source. In Europe, Goodwin v. United Kingdom (1996) held that forcing a journalist to reveal a source violates Article 10 unless an overriding public-interest need exists, and the EU's European Media Freedom Act (2024) includes source protection. Brazil's Constitution safeguards the secrecy of the source, and the Inter-American Commission's 2000 Declaration of Principles says every journalist has a right to keep sources confidential.
    Practice. Don't promise what you can't deliver; agree terms (on the record, background, off the record) before the information is given; keep as few records as possible; and many newsrooms run secure drop-boxes, such as SecureDrop, with their own instructions for sources.
    Risk to reporters. Daphne Caruana Galizia, a Maltese journalist who reported on Panama Papers links to Maltese politicians, was killed by a car bomb on 16 October 2017. The Slovak reporter Ján Kuciak and his fiancée Martina Kušnírová were shot dead in February 2018 while he was investigating tax fraud. Networks such as Forbidden Stories now continue murdered or jailed journalists' work so killing the reporter does not kill the story. The Committee to Protect Journalists and Reporters Without Borders keep dated counts; see Media Law & Press Freedom (35/96).
  

  What happens after publication
  People expect an exposé to be followed by justice. The record is more varied. Some stories end within days; others take years and are won by other institutions; and some cost the reporters more than the targets. The chart counts days from one event to another in the real cases above, not measures of success.
  
    Interactive · Days from story to consequence
    
    
    
    Dates from public records; the endpoints differ in kind (a court ruling, a resignation, a release), so compare the pattern and not the numbers.
  
  Without JavaScript: Panama Papers 2 days (publication to the Icelandic prime minister's resignation), Pentagon Papers 17 days (to the Supreme Court ruling), Spotlight 341 days, Reuters in Myanmar 511 days (arrest to release), Watergate 783 days.
  The causal link is also debated. The press rarely acts alone: the Panama Papers resignation followed street protests in Reykjavik, Watergate needed the courts and Congress, and the Boston story drew on lawsuits. The honest claim is that investigations supply evidence that other institutions can act on, when those institutions are willing.

  What a graduate just sees
  
    An investigation starts as a sentence that could be wrong, and goes looking for the evidence that would show it.
    One anonymous source is a lead. A checked document plus a person is a claim.
    Ownership multiplies down the chain, and control and ownership are different.
    Big leaks are shared work: no single newsroom reads, checks and survives them alone.
    The target gets to answer before publication, and a lawyer asks "can we prove it?"
    Source protection varies by country and can fail, so promise carefully.
    Credit for consequences is shared with courts, legislators, whistleblowers and the public.
  

  Where it shows up next
  
    The Business & Future of News (35/99): investigations cost months of pay, so non-profits, foundations and subscriptions decide how many are done.
    Media Law & Press Freedom (35/96): the defamation, privacy and source-protection law behind every step.
    Public Policy (35/16) and Criminal Law & Procedure (35/07): what happens when a story reaches a regulator or a court.
    In the world: auditors, prosecutors, anti-corruption agencies, whistleblower lawyers and any citizen reading a "Papers" story.
  

  If you want to go deeper
  
    📘 Textbook. Brant Houston and the IRE, The Investigative Reporter's Handbook (Bedford/St. Martin's, several editions), a standard US text from Investigative Reporters and Editors; check for the latest edition.
    📘 Gentler. Carl Bernstein and Bob Woodward, All the President's Men (1974), the Watergate story told as a detective story; and Daniel Ellsberg, Secrets: A Memoir of Vietnam and the Pentagon Papers (2002).
    🌐 Free text. Mark Lee Hunter and colleagues, Story-Based Inquiry: A Manual for Investigative Journalists (UNESCO, 2009), free to download in several languages.
    🌐 Free course. The Global Investigative Journalism Network (gijn.org) has free guides and resource centres in many languages, and the Knight Center for Journalism in the Americas runs free online courses on investigative reporting.
  

  
    The takeaway
    An investigation is a claim built so that a hostile reader cannot knock it down: a hypothesis, evidence from documents and people, a fair chance to reply, a lawyer's check, and protection for those who helped. The landmark cases show the method, and also that consequences depend on courts, legislators and public pressure as much as on the story.
    Try this: read the methods note of any recent "Papers" or investigation story and list which of the eight stages you can see in it.
  

  
    Sources & Further Reading
    Sources. Watergate dates (17 June 1972; Nixon's resignation 9 August 1974; United States v. Nixon, 24 July 1974; Felt revealed 2005) and the Pentagon Papers dates, volumes and the Supreme Court ruling of 30 June 1971 (6–3); the Panama Papers' size, publication date, Iceland resignation, closure of Mossack Fonseca and 2017 Pulitzer; the Boston Globe's 2003 Pulitzer for Public Service and Cardinal Law's resignation (13 December 2002); the Operation Car Wash record, Odebrecht's admissions in twelve countries and the 2019 Intercept Brasil publications; Reuters' account of Wa Lone and Kyaw Soe Oo; ICIJ's accounts of the Luanda Leaks; Branzburg v. Hayes (1972), Goodwin v. UK (1996), New York Times v. Sullivan (1964), UK Defamation Act 2013; the Inter-American Declaration of Principles (2000); Brazil's Constitution art. 5; CPJ and RSF. The companies, councillor and story in the interactives are invented.
    
      📘 Woodward and Bernstein, All the President's Men
      🌐 Global Investigative Journalism Network
      🌐 ICIJ


Step 11. The Business & Future of News: Who Pays for Finding Out
----------------------------------------------------------------
Why: Who pays for news, platforms, algorithms and AI.
Reporting costs money, and for a century someone else paid: advertisers, mostly. That arrangement broke. What replaces it decides which stories exist.

Imagine a town with one newsroom of eight reporters. Each year the readers pay a small amount for the paper, and local shops and car dealers pay far more to put adverts beside the stories. Readers got the news cheaply because someone else was buying their attention. Then the shops moved to search and social media, the classified ads moved to free websites, and the car dealers followed. The readers were still there. The money was not.
  That is the central fact of the business of news. A story costs the same to report whether ten people or ten million read it, and for most of the twentieth century the price of finding out was hidden inside the price of advertising. This piece is about what has taken its place: subscriptions, platforms, public money and, lately, software that writes.
  All figures are dated, from the sources named, and change often. The models are invented illustrations, not forecasts, and this is not investment advice.

  
    The course in one box
    
      Who takes itJournalism majors, usually in year 4, and media-management and business students. Schools teach it as a capstone on "media economics" or "the future of news"; it is where the craft meets the budget.
      What it is forUnderstanding who pays for reporting, why that is fragile, and what the options are, so a journalist can reason about their own employer.
      Comes afterJournalism Ethics (35/95), because every revenue model creates a conflict of interest, and Digital Marketing & Advertising (34/62), which explains how ads are bought and why last-click attribution misleads.
      Leads toIt closes the Journalism major, and pairs with Media Law & Press Freedom (35/96) and Social Psychology (35/25).
      What this piece gives youYou will not be able to run a media company. You will see the arithmetic of ads and subscriptions, what the 2026 data says, and what is still open. Easier first reads: How to Read the News Without Losing Your Mind (06/04), How Large Language Models Actually Work (29/02) and The Internet (29/04).
    
  

  Who pays: four possible payers
  News has four possible payers, and most outlets mix them: audiences (subscriptions, single copies, memberships), advertisers, the public (licence fees, taxes, grants) and patrons (foundations and donors). Each buys something different. Audiences buy the product. Advertisers buy the audience. The state, when it pays well, buys a public good and, when it pays badly, buys influence. Donors buy a cause.
  In the United States, newspaper advertising was worth about $49.4 billion in 2005, according to the Pew Research Center's newspaper fact sheet, and about $9.8 billion in 2022. Over the same years circulation revenue barely moved, from about $10.7 billion to $11.6 billion. Advertising had been four fifths of the money. It is now less than half.
  
    Interactive · Where US newspaper revenue came from
    20052022
    
    
    US newspapers only, from the Pew Research Center fact sheet (advertising peaked in 2005; 2022 was the latest year listed when checked). Other countries' histories differ.
  
  Without JavaScript: advertising fell from about 82% of newspaper revenue in 2005 to about 46% in 2022, while circulation revenue rose about 8%.

  Advertising: why one page view is worth so little
  A print advert sat beside a story that a local reader would read; the paper had a monopoly on that reader. On the web the same reader can be reached on a thousand sites, and an advert is sold through an automatic auction (see Digital Marketing & Advertising, 34/62). The price is quoted per thousand views, the CPM, and it is low. A site that earns a few dollars per thousand page views needs enormous traffic to pay one reporter.
  revenue = page views × ads per page × share sold × (CPM ÷ 1,000)the whole of display-ad economics in one line
  
    Interactive · How many reporters does an ad-funded site pay for?
    Page views per month
    Ads per page
    Price per thousand views (CPM)
    Share of ad slots actually sold
    
    Invented numbers in dollars. One reporter is assumed to cost $6,000 a month in all (salary and overheads), a round figure for the exercise.
  
  Without JavaScript: with 2 million page views, 4 ads a page, 60% sold and a $3 CPM, a site earns about $14,400 a month, enough for roughly 2.4 reporters.

  Subscriptions: the leaky bucket
  The alternative is to ask readers directly. A subscription is steadier than advertising, but it behaves like a bucket with a hole in it. New subscribers pour in; some leave each month (churn); the number you have settles where gains match losses. If you add A subscribers a month and lose a fraction c of them, the bucket tends to A ÷ c, and the average subscriber stays about 1 ÷ c months.
  
    Interactive · Fill the bucket for three years
    New subscribers per month
    Monthly churn
    Price per month
    
    
    Invented numbers; assumes constant sign-ups and churn, which real audiences do not have. Reporter cost as above, $6,000 a month.
  
  Without JavaScript: 1,000 new sign-ups a month with 4% monthly churn tends to 25,000 subscribers; after 36 months there are about 19,250, and at $8 a month that is about $154,000 a month.
  Cutting churn is worth as much as winning readers. The largest example is The New York Times, which reported about 12.80 million digital-only subscribers (about 13.35 million counting print) at the end of the second quarter of 2026, with digital-only average revenue per subscriber of $9.94. Such scale is rare. The Reuters Institute's Digital News Report 2026 (published 16 June 2026; 48 markets, 97,520 respondents) found that across 20 countries where paywalls are common only 17% pay for online news, roughly unchanged from 18% in 2025. The spread is wide: Norway 40%, Sweden 32%, the United States 16%, the United Kingdom 10%.
  Subscriptions do not fix everything. They pay best for national and specialist titles, and they ask readers to pay for news they could get free elsewhere. A paywall also shuts out people who cannot pay, which is a public-interest cost that publishers weigh differently.

  Platforms and algorithms: renting your audience
  For years the front door of the news was a platform. The 2026 report's headline finding: for the first time, social and video networks (54%) are the most widely used route to news, ahead of TV (52%) and news websites and apps (51%). Some 27% of respondents got news from news-focused creators or influencers, and trust in news overall fell to 37%, the lowest since the survey began in 2015. Some 42% said they sometimes or often avoid the news.
  A newsroom that depends on a platform is renting. The platform sets the algorithm, and can change it. Governments have tried to make platforms pay. Australia's News Media Bargaining Code (2021) was followed by Canada's Online News Act (2023), after which Meta began blocking news for Canadian users in August 2023, and Google reportedly agreed to a payment of about C$100 million a year. Whether such measures reduce dependence is debated.
  
    Interactive · What if a platform turns the tap down?
    Share of visits that come from platforms
    Fall in platform referrals
    Ad revenue per 1,000 visits
    Visitors who become subscribers
    
    A site with 1,000,000 visits a month. Invented numbers, and a simplification: it assumes lost visits are like all others, which they often are not.
  
  Without JavaScript: with 40% of 1,000,000 monthly visits from platforms and a 40% fall in referrals, a site loses 160,000 visits, about $3,200 a month in ads at $20 per thousand and about 80 new subscribers a month at 0.05%.

  News deserts: when no one is left to count
  When the money leaves, local news goes first. Northwestern University's Medill School tracks it in the United States. Its 2025 report (released October 2025) counted 213 counties as news deserts (no local news source), up from 206, and 136 newspaper closures in the past year, more than two a week. Newspaper jobs fell 7% in a year and more than three quarters since 2005, and about 50 million Americans live where there is one local source or none. Medill's earlier 2024 report said the country had lost more than a third of its newspapers since 2005. Counts differ between sources and are estimates.
  Nothing equivalent exists everywhere: the data are strongest for the US and parts of Europe, which is itself a finding. What is lost is not only information. The people who attend council meetings, read budgets and ask hospitals questions are mostly local reporters, and few others are paid to do it.

  Public-service broadcasters: another way to pay
  Many countries decided long ago that news is a public good, and fund it publicly. The BBC is funded mainly by the UK television licence, £180 a year from 1 April 2026 (£174.50 before), about 49p a day. Its charter expires at the end of 2027, and the government opened a review in December 2025 that includes how the BBC is funded. NHK in Japan is financed by receiving fees from households with television sets. The ABC in Australia is funded mainly by the federal government in multi-year funding rounds. In the United States, Congress rescinded about $1.1 billion for the Corporation for Public Broadcasting in July 2025, and the corporation began winding down, with staff remaining into January 2026.
  Scholars and politicians disagree about the best model. Defenders argue that public funding insulates news from the market and reaches people advertisers ignore. Critics argue that anything the state funds can be held hostage to the state, and that it crowds out private outlets. Each model has a different weakness: a fee set by government invites pressure at renewal; a grant invites budget cuts; membership alone can starve a station.
  
    Interactive · Four ways to fund a broadcaster
    
    
    Descriptive summaries; each real broadcaster mixes sources and rules change. Per-day figure: £180 ÷ 365.
  
  Without JavaScript: the BBC is licence-fee funded (£180 a year from April 2026); NHK is funded by receiving fees; the ABC by multi-year government funding; US public media by membership, underwriting and now far less federal money.

  AI in newsrooms: faster, and wrong in new ways
  Newsrooms use software for transcripts, translation, tagging and drafting headlines. The risk is in the other direction too: readers now ask chatbots for the news. The Reuters Institute found that 10% of people used an AI chatbot for news each week in 2026, up from 7%, with 16% among under-35s and just 4% in the UK. A study by the European Broadcasting Union and the BBC, published on 22 October 2025, had journalists at 22 public-service media organisations in 18 countries check over 3,000 answers from ChatGPT, Copilot, Gemini and Perplexity about the news. About 45% had at least one significant issue; 31% had serious sourcing problems and 20% major accuracy problems. The share ranged from 30% for the best assistant to 76% for the worst.
  Those assistants will change, and the figures are from one study at one date. But the lesson is stable: a small error rate compounds across many answers, and an answer that reads confidently is not checked.
  
    Interactive · How errors add up
    Answers with a significant issue
    Answers you rely on
    Share of issues a human check catches
    
    
    Assumes each answer fails independently, which real errors do not; the 45% default is the EBU/BBC study's overall figure.
  
  Without JavaScript: at a 45% error rate, relying on five answers gives about a 95% chance that at least one has a significant issue; if a human check catches 80% of issues, that falls to about 38%.
  The business questions are as open. Publishers split between licensing and suing: The New York Times sued OpenAI and Microsoft in December 2023, while other publishers have signed licensing deals (check the current state of both). If chatbots answer questions with a summary of the news, fewer readers reach the sources that paid for it, which could undermine the supply they depend on. Nobody yet knows.

  What a graduate just sees
  
    A story costs the same to report however many people read it, so reach matters less than who pays.
    Every revenue source buys something: audiences buy the product, advertisers the audience, the state a public good or influence.
    A subscriber base is a bucket: sign-ups divided by churn tells you where it settles.
    Page views earn pennies, so ad-funded news needs huge scale.
    Platform traffic is rented; a change in an algorithm is a change in budget.
    When local news closes, the stories that vanish are the ones nobody knew to ask about.
    No funding model is neutral: each has its own way of being captured or starved.
    A small error rate compounds, so the cost of AI is the checking.
  

  Where it shows up next
  
    Media Law & Press Freedom (35/96): the lawsuits and SLAPPs that small outlets cannot afford.
    Journalism Ethics (35/95): sponsored content, donor influence and the wall between sales and newsroom.
    Digital Marketing & Advertising (34/62): how the money on the other side of the screen works.
    In the world: newsroom product teams, membership drives, non-profit newsrooms, public-media boards and every freelancer pricing a story.
  

  If you want to go deeper
  
    📘 Textbook. Robert G. Picard, The Economics and Financing of Media Companies (Fordham University Press, 2nd edition, 2011), a standard media-economics text; the data in it is dated, the frameworks are not.
    📘 Gentler. Margaret Sullivan, Ghosting the News (Columbia Global Reports, 2020), a short account of local news loss, and Julia Cagé, Saving the Media (Harvard University Press, 2016) on ownership models.
    🌐 Free data. The Reuters Institute's Digital News Report 2026, free each June, with country pages for every market.
    🌐 Free course. The Knight Center for Journalism in the Americas offers free online courses for journalists, in several languages (check the current list).
  

  
    The takeaway
    News costs the same to make however many read it, and advertising used to pay. It no longer does, so outlets rely on subscriptions that only a few readers pay for, on platforms they do not control, on public money that comes with politics, and on software that brings its own errors. Whoever pays shapes what exists.
    Try this: pick one outlet you read and find out how it is paid for (paywall, ads, donations, public money). Then ask what story it would find hardest to run.
  

  
    Sources & Further Reading
    Sources. Reuters Institute, Digital News Report 2026 overview (16 June 2026); Pew Research Center newspaper fact sheet; Medill State of Local News 2025 (October 2025) and 2024; The New York Times Company, second-quarter 2026 results; UK government and press reports of the 2026 licence fee; EBU/BBC News Integrity in AI Assistants (22 October 2025); reports on the Corporation for Public Broadcasting's wind-down; reports on Canada's Online News Act. Links and checks are in the fact-check log.
    
      📘 Picard, The Economics and Financing of Media Companies
      🌐 Reuters Institute Digital News Report 2026
      🌐 Pew Research Center, Newspapers fact sheet


Step 12. Social Psychology: How Other People Change What We Do
--------------------------------------------------------------
Why: Why people believe what they believe.
The classic experiments say that ordinary people go along, obey and look away. The re-checked versions say something more careful, and more interesting: the situation matters a great deal, and so does what people think the situation is.

In 1964 a young woman was murdered in Queens, New York, and a newspaper told the country that thirty-seven neighbours watched and did nothing. The story became the opening page of nearly every social psychology textbook. It has also, in the decades since, turned out to be wrong in most of the details that made it shocking.
  That arc, a vivid story, a famous experiment, a tidy lesson, then a slow correction, is a fair summary of the whole subject. Social psychology asks how the presence of other people changes what one person thinks, says and does. Its best ideas survive. Many of its best-known demonstrations were less clean than the textbooks said. A graduate of this course keeps both: the ideas, and the habit of asking how a classic study was actually done.

  
    The course in one box
    
      Who takes itPsychology majors, usually in year 2; it is one of the core areas in the British Psychological Society's accreditation list and in the American Psychological Association's guidelines for the undergraduate major. Sociology, business and law students often take it as an elective.
      What it is forExplaining how groups, authority, first impressions and beliefs about other people shape behaviour, and how to tell a solid finding from a famous one.
      Comes afterResearch Methods in Psychology (35/19), which is where you learn why a single study proves little.
      Leads toOrganizational Behavior (34/53) and Consumer Behavior (34/58), and it sits beside Behavioral Economics (34/46) and Developmental Psychology (35/24).
      What this piece gives youYou will not be able to run an experiment. You will know six classic findings, what each actually showed, what its critics say, and how well it has held up. Easier first reads: The Biases That Fool Everyone (17/02) and Why You Understand Less Than You Think (17/10).
    
  

  Conformity: Asch and the line that was obviously wrong
  Solomon Asch's question was simple: will people state something they can plainly see is false, just because a group does? In his experiments (reported from 1951), a volunteer sat with several others, all but one of them secretly working with the experimenter, and judged which of three lines matched a standard line. The answer was obvious. On 12 of the 18 trials the confederates, one after another, gave the same wrong answer, and the real participant answered near the end.
  Alone, people made fewer than 1% errors. In the group, about 36% of answers on the critical trials went along with the wrong majority. About three quarters of participants conformed at least once, and about a quarter never did. That last number is the one textbooks tend to leave out: most answers were still correct, and a survey of textbooks in 2015 found that only one of twenty major texts reported that most responses defied the majority.
  The size of the group matters up to a point. A single dissenting-looking companion barely moves people; two or three make a real difference; beyond about three or four the effect levels off. The tool below draws that curve and lets you run a toy group of fifty simulated volunteers.
  
    Interactive · How many confederates does it take?
    People in the room giving the wrong answer
      
    
    Run 50 simulated volunteers
    
    Curve is approximate and rounded, following the commonly reported shape of Asch's group-size results; secondary sources differ by a point or two. The 50 volunteers are a toy model with a made-up spread of suggestibility (some people never conform), not data.
  
  Without JavaScript: with three or more people giving the wrong answer, about a third of answers conformed in Asch's experiments; with one, only a few per cent.
  Why did they conform? Asch interviewed them. Some doubted their own eyes, some knew the group was wrong but did not want to stand out, and a few came to see the lines the group's way. Psychologists later split this into informational influence (the group probably knows something) and normative influence (disagreeing has a social cost). John Turner and colleagues argued that the data fit a third idea better: we follow those we see as "people like us". Conformity also varies by place and time. A 1996 meta-analysis by Bond and Smith, covering 133 studies in 17 countries, found more conformity in more collectivist societies and a decline in the United States over the decades after the 1950s.

  Obedience: Milgram and the shock machine
  In 1961–62 at Yale, Stanley Milgram told volunteers they were helping a study of learning. A "teacher" read word pairs to a "learner" in the next room and gave an electric shock for each mistake, rising in 15-volt steps to 450 V. The learner, an actor, protested and eventually fell silent. If the teacher hesitated, the experimenter in a grey coat gave a fixed series of prods, the first being "Please continue." In the best-known condition, 26 of 40 men (65%) went all the way to 450 V.
  The number was shocking, and so was Milgram's reading of it: ordinary people will harm a stranger when an authority tells them to. But Milgram ran many variations, and they are the real finding. Obedience was not a trait of the people; it moved with the situation, from nearly everyone to hardly anyone.
  
    Interactive · One experiment, many results
    
    
    
    Percentages who went to the maximum 450 V, as reported for Milgram's variations (about 40 participants each). Telephone instructions are often given as 20.5% or 21%.
  
  Without JavaScript: baseline 65%; learner in the same room 40%; learner's hand forced onto the plate 30%; instructions by telephone about 21%; run-down office 47.5%; two fellow "teachers" refuse 10%; a fellow "teacher" who delivers all the shocks 92.5%.
  The critics are serious. Gina Perry, who went through Milgram's archive for her 2012 book Behind the Shock Machine, argued that the published account did not match what she found, including that only about half of participants fully believed the shocks were real, and that two thirds of those who did believe disobeyed. Alex Haslam and Stephen Reicher argue that people did not simply obey: they went on when the prod appealed to science ("The experiment requires you to continue") and stopped when it sounded like a bare order. Their reading is "engaged followership": participants identified with the study's goal, not with the learner. A 2009 partial replication by Jerry Burger, which stopped at 150 V for ethical reasons, found rates close to Milgram's at that point. The shock is real; the "people are robots" reading is not.

  Attribution: why we blame the person and excuse ourselves
  You see a driver cut across three lanes. You think: what a jerk. When you do it, the exit was suddenly there and the mirror was blocked. Attribution theory, started by Fritz Heider in the 1950s, studies how we explain behaviour: by something in the person (a disposition) or something in the situation. The consistent human habit is to over-read the person. Lee Ross named it the fundamental attribution error in 1977.
  The classic demonstrations: in 1967 Edward Jones and Victor Harris had people read an essay praising or attacking Fidel Castro, and told them the writer had been assigned the side. Readers still guessed the writer leaned that way. In Ross, Amabile and Steinmetz's 1977 quiz game, students were assigned by coin toss to write hard questions or to answer them. The questioners looked cleverer to everyone, including the answerers, though the role had made it so.
  Harold Kelley's 1967 covariation model gives a tidy rule for reasoning well. Ask three questions about the behaviour: does everyone do it (consensus), does this person do it to everything (distinctiveness), does this person do it every time (consistency)?
  
    Interactive · Sam laughs at a comedian. Why?
    Consensus: do other people laugh at this comedian?
      Yes, most doNo, hardly anyone
    Distinctiveness: does Sam laugh at other comedians too?
      No, only this oneYes, at nearly all
    Consistency: does Sam laugh at this comedian every time?
      Yes, every timeNo, only tonight
    
    Kelley's three textbook patterns are shown; the other five combinations are ambiguous and the model says so. Later studies (McArthur, 1972) found that people tend to under-use the consensus information.
  
  Without JavaScript: low consensus, low distinctiveness and high consistency points to the person; all three high points to the comedian; low consensus, high distinctiveness and low consistency points to the circumstances of tonight.
  Culture matters here too. In a well-known study Michael Morris and Kaiping Peng (1994) found that American and Chinese participants explained the same animated fish scene differently, Americans leaning on the fish's inner traits and Chinese participants on the group around it. It fits the larger point that much of social psychology was built on university students in a few rich countries, which Joseph Henrich, Steven Heine and Ara Norenzayan called WEIRD samples (Western, Educated, Industrialised, Rich, Democratic) in 2010.

  Attitudes and persuasion: when behaviour changes belief
  Common sense says attitudes drive behaviour. A good deal of the time it runs the other way, and Leon Festinger's cognitive dissonance theory (1957) says why: holding two clashing thoughts is uncomfortable, and the cheapest fix is often to change the belief, not the act. The famous test (Festinger and Carlsmith, 1959) had 71 male Stanford students do dull peg-turning, then asked some to tell the next person it was fun, for either $1 or $20. Which group came to like the task more?
  
    Interactive · Predict, then reveal
    Afterwards, who said the boring task had really been enjoyable?
    
      The $1 groupThe $20 groupNeither: no difference
    Pick one, then read the result.
  
  Without JavaScript: the $1 group rated the task as more enjoyable than the $20 group, the opposite of what simple reward would predict.
  The $1 group did, and the $20 group only slightly more than people who were never asked to lie. Dissonance theory says $1 is too little to justify the lie, so the mind adjusts the belief; $20 is justification enough. The theory has had rivals (Daryl Bem's self-perception account) and refinements, and the "free-choice" variant, in which people supposedly like what they chose more, has been criticised for a statistical flaw. But the core pattern, that acts shape beliefs especially when we feel responsible for them, is one of the more durable ideas in the field.
  On persuasion itself, Richard Petty and John Cacioppo's elaboration likelihood model (1986) says there are two routes: a central one, where people weigh arguments when they are motivated and able, and a peripheral one, where cues such as a confident speaker or a famous face do the work. Strong arguments matter most when people are paying attention. When they are not, packaging does. And attitudes predict behaviour only loosely: Richard LaPiere's 1934 trip with a Chinese couple found almost every hotel and restaurant served them, while most of the same businesses later said by mail that they would not. What people say and what they do can differ.

  Groups and prejudice: how little it takes to make "us"
  In the early 1970s Henri Tajfel divided schoolboys in Bristol into two groups by a trivial rule (they were told it was a preference for paintings by Klee or by Kandinsky) and then asked each boy to hand out points to anonymous members of the two groups. The boys knew none of them and would gain nothing. Yet they favoured their own group, and often chose options that gave their group fewer points in total as long as the gap over the other group got bigger. This minimal group paradigm underlies social identity theory: part of who we are comes from the groups we belong to, and we want our group to look good against others.
  
    Interactive · Hand out the points
    You are in the "blue" group. Choose one column: points for an anonymous blue member (top) and an anonymous "green" member (bottom).
    
    Choose a pair.
    An invented matrix in the spirit of Tajfel's, not his actual numbers. Real participants were schoolboys, and what they picked on matrices designed to pull the strategies apart is what the study reported.
  
  Without JavaScript: in the matrix, giving 13 points to both groups gives each 13; giving 7 to your group and 1 to the other leaves a gap of 6 but costs your group 6 points.
  Muzafer Sherif's Robbers Cave study (Oklahoma, 1954) is the famous other half: twenty-two boys at a summer camp, split into two groups, competed for prizes and soon despised each other, then made peace when forced to cooperate on shared goals. The conclusion that competition breeds hostility and common goals reduce it is widely taught. The critique is that the story is tidier than the record: earlier versions of the study in 1949 and 1953 did not produce the same result (in one, the boys turned on the experimenters), and Michael Billig has argued that the staff were effectively a third group steering events. What has held up better is Gordon Allport's 1954 contact hypothesis: contact between groups reduces prejudice under conditions such as equal status and shared goals, supported by a large meta-analysis of hundreds of studies by Thomas Pettigrew and Linda Tropp.
  Some newer popular ideas deserve caution. The Implicit Association Test is widely used to measure unconscious bias, but its scores are only modestly stable from one session to the next and predict individual behaviour weakly, and changing a score has not reliably changed behaviour. It may tell you about group-level patterns better than it tells you about a person.

  Bystanders: the story and the evidence
  The Kitty Genovese story, as taught for decades, said thirty-eight people watched a murder from their windows and did nothing, out of apathy. In 2007 Rachael Manning, Mark Levine and Alan Collins went back to the record: far fewer people had a clear view, many who heard noise did not realise what was happening, and at least one person called the police. In 2016 the New York Times itself called its original reporting flawed and said it had grossly exaggerated the number of witnesses.
  The better evidence is laboratory. In 1968 John Darley and Bibb Latané had students talk by intercom, one at a time, with others in separate rooms; one of the "others" seemed to have a seizure. Students who believed they were the only listener reported it before the fit ended 85% of the time. With one other possible listener the figure was 62%, and with four others, 31%. Latané and Darley called the mechanism diffusion of responsibility: with others around, each person feels less personally responsible.
  
    Interactive · One person, or the whole room?
    
    
    
    The 85, 62 and 31 are from Darley and Latané (1968), the proportion reporting before the fit ended. The "someone helps" row is arithmetic that assumes each listener acts independently, which real groups, who also watch each other, will not.
  
  Without JavaScript: each person is less likely to act in a group (85%, 62%, 31%), but if five listeners each had a 31% chance, the chance that at least one acts is about 84%, close to the lone person's 85%.
  Look at what the arithmetic says. The effect is real for each individual, but the group as a whole may still produce help about as often as one person would. That is why later work added caveats. A 2011 meta-analysis by Peter Fischer and colleagues found the bystander effect strong in non-dangerous emergencies and weaker in dangerous ones, and weaker when bystanders are not strangers. A 2020 study by Richard Philpot and colleagues of more than 200 real CCTV recordings of street conflicts in the UK, the Netherlands and South Africa found that in over 90% someone intervened, and more bystanders meant more chance of help. The bystander effect is real, but people are not generally apathetic.

  What held up
  The Open Science Collaboration (2015) tried to repeat 100 psychology studies; about 36% of the repeats were statistically significant against 97% of the originals, and the repeats showed effects about half the size. Social psychology fared worse than cognitive psychology (about a quarter replicated against about a half). Treat the list below as a rough report card, not a verdict on any one study.
  Held up well:solid Asch-style conformity in its broad direction; anchoring-like context effects; the contact hypothesis; the bystander effect in non-dangerous settings.
  Held, but smaller or messier:shrank Milgram's obedience (the rate depends on setup, and the interpretation changed); cognitive dissonance in its simplest forms; the fundamental attribution error (strong in the lab, culture-bound in size).
  Failed or collapsed:failed Many social-priming effects, such as words about old age making people walk slower; power posing changing hormones; ego depletion. See Research Methods in Psychology (35/19).
  Disputed:disputed The Stanford Prison Experiment, which lasted six days in 1971, said guards turned cruel by role. Thibault Le Texier (2018) reported from the archive that the guards were coached on how to behave, so it is better read as a demonstration of what people do when told what is expected than as evidence of a natural tendency.

  What a graduate just sees
  
    Most people in an Asch room still gave the right answer; the effect is real and also smaller than the legend.
    Obedience swings from 10% to 90% depending on the setup, so ask what the situation is, not what kind of person someone is.
    When judging someone else, check the situation first; when judging yourself, you already do.
    Acts change beliefs: if you were paid little for something you did, you will find reasons it was worth it.
    A trivial label is enough to start "us and them"; shared goals are better at undoing it.
    In a crowd each person feels less responsible, but a named person asked directly usually helps.
    A famous study is a lead, not a fact: look for the replications, the sample, the critics.
  

  Where it shows up next
  
    Organizational Behavior (34/53): group decisions, conformity in meetings, authority and safety culture.
    Consumer Behavior (34/58) and Behavioral Economics (34/46): social proof, defaults, norms, and which of the "nudges" are weaker than advertised.
    Clinical Psychology (35/27): stigma, group therapy and attribution styles in depression.
    In the world: jury deliberation, online pile-ons, workplace safety, military ethics, and how emergency courses teach "you, in the red coat, call 999".
  

  If you want to go deeper
  
    📘 Textbook. Elliot Aronson, Timothy Wilson and Robin Akert, Social Psychology (Pearson, many editions), or David Myers, Social Psychology (McGraw-Hill). Both are standard undergraduate texts; check for a recent edition that covers replication.
    📘 Gentler. Gina Perry, Behind the Shock Machine (2012) on Milgram, and Gina Perry, The Lost Boys (2018) on Robbers Cave.
    🌐 Free text. The free Noba psychology textbooks have a module on Conformity and Obedience, written for introductory courses.
    🌐 Free course. MIT OpenCourseWare 9.00SC Introduction to Psychology (Fall 2011) is the free introductory course.
  

  
    The takeaway
    Other people shape what we say, do and believe, and the situation can matter more than character. But the best-known demonstrations are more qualified than their reputations: most people resisted in Asch's rooms, Milgram's results depended on the setup, the Genovese crowd was exaggerated, and several famous effects shrank on re-testing.
    Try this: the next time someone's behaviour annoys you, write down two situational explanations before you settle on a trait.
  

  
    Sources & Further Reading
    Sources. Asch, "Opinions and Social Pressure," Scientific American 1955, and the 1951 monograph (about 36% of critical responses conforming; about 74% at least once; control error under 1%); a 2015 survey of twenty textbooks on Asch (and an earlier one in 1990); Turner on referent informational influence; Bond and Smith, Psychological Bulletin 1996. Milgram, Journal of Abnormal and Social Psychology 1963 and Obedience to Authority (1974): variations as summarised in Wikipedia, "Milgram experiment"; Perry 2012; Haslam, Reicher and Birney on engaged followership; Burger, American Psychologist 2009. Jones and Harris 1967; Ross 1977; Ross, Amabile and Steinmetz 1977; Kelley 1967; McArthur 1972; Morris and Peng 1994; Henrich, Heine and Norenzayan, Behavioral and Brain Sciences 2010. Festinger and Carlsmith, 1959 (71 male Stanford students); Petty and Cacioppo 1986; LaPiere 1934. Tajfel, Billig, Bundy and Flament, "Social categorization and intergroup behaviour," European Journal of Social Psychology 1971; Sherif et al., Intergroup Conflict and Cooperation: The Robbers Cave Experiment (1961); Billig; Allport, The Nature of Prejudice (1954); Pettigrew and Tropp, 2006. Darley and Latané, Journal of Personality and Social Psychology 8 (1968), 377–383; Manning, Levine and Collins, American Psychologist 2007; the New York Times acknowledgement of 2016; Fischer et al., Psychological Bulletin 2011; Philpot et al., American Psychologist 2020. Open Science Collaboration, Science 2015; Le Texier, American Psychologist 2019. Calculations on this page (the 84% and 85% group figures, the toy simulation) were run in node. This piece describes research and is not a guide to any individual's behaviour.
    
      📘 Perry, Behind the Shock Machine
      📘 Aronson, Wilson and Akert, Social Psychology
      🌐 Noba Project