Start hereSagan wrote that newborns' face recognition is "hardwired in our brains". A 1991 study did find newborns follow a face-like pattern further than a scrambled one. Yet a 2015 review concludes that experience shapes face processing too. So which of the book's claims hold up?
The Demon-Haunted World is Carl Sagan’s case for scientific thinking as a habit anyone can learn. It is also a warning about what happens to a society that loses it.
About the book
Random House published the book in 1995, a hardcover of 457 pages; the Ballantine paperback came out on 25 February 1997. One review describes it as twenty-five essays, several written with Sagan’s wife, Ann Druyan, and the publisher lists Druyan with him.
Sagan was a professor of astronomy and space sciences at Cornell University. He played a leading role in the Mariner, Viking, Voyager and Galileo spacecraft expeditions, according to his publisher. He died on 20 December 1996.
It is popular science for general readers. Its targets range from UFOs, crystals and faith healing to the face on Mars. It argues that scientific thinking matters to the health of democracy itself.
Science is a way of thinking
The book opens with an airport driver who wanted to talk about Atlantis, channelling and astrology. Sagan kept telling him the evidence was crummy and there was a much simpler explanation. He came away feeling that the driver’s natural curiosity had been let down: “all the science had gotten filtered out before it reached him.”
Try it: when a claim excites you, treat the curiosity as the start, not the end. Ask what a simpler explanation would look like.
A candle in the dark
Sagan worried about a public losing the habit of weighing evidence. In chapter 2 he writes, “I have a foreboding of an America in my children’s or grandchildren’s time”. He pictures a time when people are unable to tell what feels good from what is true.
He blames in part “the 30-second sound bites (now down to 10 seconds or less)” of television. He puts the fear in one image: “The candle flame gutters.”
Does it hold up? The sound bites did shrink, and the knowledge gaps he described have largely lasted, with some movement.
- Sound bites. A 2016 study in the Journal of Communication summed up the record, citing a 1992 study by Hallin. The average politician sound bite in US election news shrank from 43 seconds in 1968 to about 9 seconds in 1992. It has stayed near that level since. Neil Postman made the same complaint a decade earlier; see our summary of Amusing Ourselves to Death.
- The polls. In chapter 1 Sagan cites polls on what American adults did not know. Sixty-three percent were unaware that the last dinosaur died before the first human arose. Seventy-five percent did not know that antibiotics kill bacteria but not viruses, and 57 percent did not know that electrons are smaller than atoms.
- A generation on. The National Science Board tracks the same kind of questions. In 2018, Americans answered an average of 62% of its nine basic questions correctly. That is statistically similar to the historical average since 1992.
- Some movement. On antibiotics, the share holding the wrong belief has dropped over time, yet half of Americans still held it in 2018.
The Board’s 2020 report adds a caution. General measures of science knowledge typically have only small, though meaningful, links with how people make decisions. A quiz score is not the habit of mind his book is about.
Try it: when a short clip seems to settle a question, look for the longer version before deciding.
The baloney detection kit
Chapter 12, “The Fine Art of Baloney Detection”, gives the tools. In our words, with one line kept in his:
- Get independent confirmation of the “facts” wherever possible.
- Hear the experts argue. He encourages substantive debate on the evidence by knowledgeable proponents of all points of view.
- Weigh authority lightly. “Perhaps a better way to say it is that in science there are no authorities; at most, there are experts.”
- Spin more than one hypothesis, and try not to get overly attached to one just because it is yours.
- Quantify. If what you are explaining has some measure or numerical quantity, use it.
- Check every link. In a chain of argument, every link must work, including the premise, not just most of them.
- Use Occam’s Razor, a rule of thumb: when two hypotheses explain the data equally well, choose the simpler.
- Ask whether the claim could be shown false, at least in principle.
Does it hold up? The first tool matches how skilled fact-checkers work. In a 2019 study, fact-checkers judged an unfamiliar website by leaving it to see what other sources said about it. Whether such habits can be taught is a more modest story, set out in the table below.
Try it: before trusting a source, see what others say about it. Our SIFT guide turns this into four quick moves.
The fallacies to watch for
The same chapter lists the most common and perilous fallacies of logic and rhetoric. Two familiar ones are confusing correlation with causation, and suppressed evidence (half-truths).
One of his examples sticks. Observational selection is “counting the hits and forgetting the misses”, as he says the philosopher Francis Bacon described it. A state, for instance, boasts of the presidents it has produced “but is silent on its serial killers”.
Does it hold up? In studies, learning to name tricks helps people recognise them when tested. Whether that carries over to real news is still argued; the table below gives both sides.
Try it: when a run of hits looks impressive, ask how many misses went unreported. Our guide to reading a study has more questions in the same spirit.
The dragon in the garage
Chapter 10 opens with a thought experiment: “A fire-breathing dragon lives in my garage.” The story builds to its point: “Your inability to invalidate my hypothesis is not at all the same thing as proving it true.”
That is the kit’s last tool in story form.
Try it: ask what would change your mind. If nothing could, the claim is a dragon in the garage. Our logical fallacies cheat sheet covers the common traps, and when they are not traps.
Why we see faces and aliens
Sagan explains a lot of strange experience through ordinary brains. “As soon as the infant can see, it recognizes faces, and we now know that this skill is hardwired in our brains,” he writes. That, he argues, is why we see a man in the Moon and a face on Mars.
For alien abductions, he describes an experiment in which imagined abductions came out much like “the accounts that self-described abductees present”. He also notes that the researcher had cued his subjects.
Does it hold up? On abductions, largely; the face claim is weighed under “What holds up” below. The experiment Sagan describes was published by Alvin Lawson in 1977. Subjects were hypnotised and asked to imagine an abduction. Their stories “showed no substantive differences” from reported cases.
In a revised paper in 1978 Lawson stood by that, but argued there were also “crucial differences”, such as alleged physical effects and multiple witnesses.
Later studies point the same way as Sagan. People reporting recovered abduction memories were more prone than a control group to false recall and recognition. In one study of ten abductees, their claims were linked to apparent episodes of sleep paralysis. These are small samples, so they show how such memories can form, not how every report did.
Try it: for a strange experience, list the ordinary explanation next to the exciting one before deciding which the evidence favours.
Compassion, not contempt
Sagan is firm about evidence but gentle about people. “The chief deficiency I see in the skeptical movement is in its polarization: Us vs. Them,” he writes. He argues for an approach that starts by understanding why people believe. He thought an approach that starts from the human roots of pseudoscience “might be much more widely accepted”.
Try it: in a disagreement, ask what first drew the other person to the idea before you offer your view.
What holds up
What the sources show for each idea
Most of what later research can test holds up. The one place the book says more than the evidence allows is the newborn brain.
On faces, the early preference is real. A 1991 study replicated the finding that newborns follow a face-like pattern further than a scrambled one. Both the arrangement of the features and some aspects of the features themselves mattered for tracking in the first hour of life.
What it is made of is still argued. [A 2015 review sets two views side by side. One says newborns have a dedicated face detector. The other explains the preference by general attention biases that are not specific to faces.
The review concludes that both inborn predispositions and exposure to certain experiences shape how faces are processed.](https://doi.org/10.3389/fpsyg.2015.00969) So “hardwired” is stronger than the evidence allows.
Who it’s for, and who can skip it
It suits readers who want a warm, readable defence of scientific thinking and a practical checklist to go with it. Some examples have dated, and parts read as a tour of the pseudoscience of its day.
Readers who want only the tools can read chapter 12 and the table above.
Quick answers
What is The Demon-Haunted World about?
Carl Sagan's 1995 book argues that science is more than a body of knowledge: it is a way of thinking. It warns that a public without that habit is easy prey for pseudoscience, and it ties scientific thinking to the health of democracy.
What is Carl Sagan's baloney detection kit?
It is the set of tools in chapter 12, "The Fine Art of Baloney Detection". They include getting independent confirmation of the facts, spinning more than one hypothesis, quantifying, checking every link in an argument, and asking whether a claim could be shown false.
What is the dragon in the garage?
A thought experiment that opens chapter 10: "A fire-breathing dragon lives in my garage." The story builds to its point: "Your inability to invalidate my hypothesis is not at all the same thing as proving it true."
Is the face on Mars real?
It is a natural landform. Viking 1 photographed it on 25 July 1976; higher-resolution images from Mars Global Surveyor in April 1998, and more in 2001, confirmed it as an eroded hill whose face-like look came from the lighting.
Use this
Run one claim a week through Sagan's kit.
What would change my mind, and have I looked for it?
Quick quiz
Nothing is scored or saved. Recalling it once now is what makes it last.
1. About how long was the average politician sound bite on US election news in 1992?
Answer: About 9 seconds.Read it again ↑
2. What is the point of Sagan's dragon in the garage?
Answer: Failing to disprove a claim is not the same as proving it true.Read it again ↑
3. In the 2019 study, how did professional fact-checkers judge an unfamiliar website?
Answer: By leaving it to see what other sources said about it.Read it again ↑
How we checked this · Last checked October 2026
- Quotations from the book come from the publisher's page, with its free excerpt of chapter 1, and from short passages of the 1995 Random House first edition found through Open Library's search-inside, a line or two at a time.
- Publication details come from the Open Library record of the 1995 edition and the publisher's page; the description of the book as twenty-five essays comes from one review.
- Each study and report on the page was read at least in its abstract.
- Lawson's 1977 and 1978 papers are known to us through Philip Klass's 1981 account in The Skeptical Inquirer.
Sources
- C. Sagan, The Demon-Haunted World: Science as a Candle in the Dark (Random House, 1995): chapter 1 as published on the publisher’s page; other short passages confirmed through Open Library search-inside of the 1995 Random House first edition; Open Library record of the 1995 edition.Primary
- R. T. Carroll, review of the book, The Skeptic’s Dictionary.Popular
- S. Wineburg and S. McGrew, “Lateral Reading and the Nature of Expertise”, Teachers College Record (2019).Scholarly
- P. Abrami and colleagues, “Strategies for Teaching Students to Think Critically”, Review of Educational Research (2015); L. Oswald and colleagues, Psychological Science (2026).Scholarly
- Y. Wang and colleagues, PNAS Nexus (2025); A. Modirrousta-Galian and P. Higham, Journal of Experimental Psychology: General (2023); A. Simchon and colleagues, “A Signal Detection Theory Meta-Analysis of Psychological Inoculation Against Misinformation”, Current Opinion in Psychology (2026).Scholarly
- National Science Board, Science and Engineering Indicators: “Public Familiarity with S&T Facts” (2020) and “Public Attitudes” (2022).Scholarly
- E. M. Rinke, “The Impact of Sound-Bite Journalism on Public Argument”, Journal of Communication (2016), author’s preprint.Scholarly
- M. Johnson and colleagues, “Newborns’ preferential tracking of face-like stimuli and its subsequent decline”, Cognition (1991); F. Simion and E. Di Giorgio, “Face perception and processing in early infancy”, Frontiers in Psychology (2015).Scholarly
- European Space Agency, “Cydonia – the face on Mars”.Primary
- P. Klass, “Hypnosis and UFO Abductions”, The Skeptical Inquirer (1981), for Lawson’s 1977 and 1978 papers.Popular
- S. Clancy and colleagues, “Memory distortion in people reporting abduction by aliens”, Journal of Abnormal Psychology (2002); R. McNally and S. Clancy, “Sleep paralysis, sexual abuse, and space alien abduction”, Transcultural Psychiatry (2005); R. McNally and colleagues, Psychological Science (2004).Scholarly
Keep the questions, date the examples
Sagan's kit is a set of tools, not a list of answers. Our view: the tools have aged better than some of his examples. Practise one until you use it without thinking.
Work one claim through the kit
- Write the claim down. One sentence, in plain words. A vague claim is hard to confirm and hard to rule out.
- Put a number on it. Sagan's advice: if it has a measure, use it. 'Many people' and 'most people' are different claims.
- Find the weakest link. List each step from evidence to conclusion, premise included. Check the one you're least sure of first.
- Find who disagrees. Look for knowledgeable people who argue other points of view, and read their best case.
- Ask for the evidence. When a claim leans on a famous name, ask what that person's evidence is.
This week: test a claim you'd like to be true
Sagan warns against getting attached to a hypothesis just because it is yours. Pick one claim you'd like to be true. Look for independent confirmation, and ask what result would show it false. Write one line on what you found.
Talk with a believer without contempt
- Start with why. Sagan argued for an approach that starts by understanding why people believe. Ask what first drew them to the idea.
- Respect the curiosity. Sagan felt the airport driver's natural curiosity had been let down. Treat interest in big questions as common ground.
- Agree on evidence first. Before you argue, ask what would count as evidence for each of you.
- Offer, don't rule. Put the simpler explanation as a question: 'Could it be this?' Our view: a verdict tends to invite a defence.
- Share your method. Show how you check a claim, not only what you concluded. They can use the method on the next claim.
Turn careful thinking into work
- Check the reasoning in reports or funding bids before they go out: sources, numbers and every link.
- Write plain explainers of new science stories, saying what was measured and how sure the researchers are.
Ideas to explore, not financial or business advice. No result is promised.