- Narration: Bryan DePuy reads Darrell Huff's mid-century prose with clean clarity, matching the book's brisk, no-nonsense tone without over-performing.
- Themes: Statistical manipulation, media literacy, numerical deception
- Mood: Wry, educational, and surprisingly brisk
- Verdict: A seventy-year-old primer on statistical manipulation that has lost none of its diagnostic power and gains something in audio form thanks to a narrator who keeps the pace moving.
Darrell Huff's How to Lie with Statistics, narrated by Bryan DePuy in this audiobook edition, is one of those rare texts where the age of the material is almost irrelevant. First published in 1954, it reads with the confidence of a book that knew it had found a permanent subject. Every graph that misleads, every poll that buries its sample size in fine print, every headline that rounds a percentage to the nearest convenient integer, Huff catalogued the techniques before most of us had the vocabulary to name them, and DePuy's narration brings the sardonic wit of the original text to life with measured ease.
The book's central argument is not that statistics lie but that people use them to lie, and that the lie succeeds because most readers lack the tools to identify what has been done to the numbers. Huff's project is to hand you those tools in plain language, one trick at a time. The tone is never contemptuous toward those being misled, Huff seems genuinely more interested in the mechanics of deception than in assigning blame, and that neutrality makes the lessons more transferable than a more polemical approach would.
Huff's method is the worked example. He doesn't define "the semi-attached figure" in the abstract and move on, he shows you an ad claiming that a particular cigarette has "less nicotine" without specifying less than what, then walks you through exactly why that construction is designed to mislead while remaining technically defensible. He does the same with truncated bar graphs, with the difference between mean and median income, with correlation presented as causation. By the time you reach the end of this three-hour recording, you have seen the full toolkit laid out and tested.
What's remarkable is how durable the specific examples remain. The techniques Huff describes, the carefully chosen sample, the loaded question in a survey, the misleading visual scale, appear in virtually every political news cycle today, often in more sophisticated form. One reviewer noted the book "has aged like fine whisky," and that holds: the statistical fallacies Huff dissects have grown more elaborate, not less common, in the seventy years since publication. The chapter on the "well-chosen average", where Huff shows how the mean, median, and mode of the same dataset can produce three different "true" numbers depending on which one the presenter finds most convenient, remains one of the clearest explanations of that deception available anywhere in popular literature.
Why Listen to How to Lie with Statistics
Bryan DePuy's narration is efficient and well-matched to the material. Huff's prose has a mid-century journalistic snap, short sentences, dry humor, the occasional wink, and DePuy doesn't bury it under interpretive flourish. He reads it straight, which is the right call for a text that does its own work. At three hours, this is a particularly good fit for audio; the chapters are short enough that you retain the structure of each argument as a discrete lesson, and the listening never feels like a commitment that demands full concentration over many sittings.
The audio format also surfaces something about the book's original intent. Huff was not writing for statisticians; he was writing for newspaper readers, which is to say for anyone who encounters numbers in daily life. Hearing DePuy deliver the analysis conversationally reinforces that this is not a technical text. No mathematical background is expected or needed. What is required is the willingness to slow down when numbers appear and ask the questions Huff teaches you to ask: Who collected this data? How large was the sample? What is the number being compared to? What is being left out? The final chapter, Huff's own guide to "How to Talk Back to a Statistic", pulls all of those questions into a usable checklist that is the most immediately actionable section in the recording.
What to Watch For in How to Lie with Statistics
Listeners expecting a modern data science primer will find the technology dated. Huff's examples draw from print newspapers, advertisements, and early-postwar surveys. There are no examples involving social media metrics, algorithmic recommendation data, or A/B testing, none of which existed in 1954. The conceptual framework, however, transfers directly. The principle that a graph with a truncated y-axis magnifies apparent differences applies whether the underlying data comes from a 1952 Gallup poll or a 2025 platform analytics dashboard.
Some reviewers have noted that the examples, while effective, are occasionally repetitive, Huff makes some of his points more than once using slightly different examples. In audio form this is less disruptive than it might be in print, since the rhythm of listening naturally smooths over what would feel like backtracking on the page. The book also concludes with a chapter on how to read statistics defensively, which is the most immediately practical section and earns its place as a summary of everything that preceded it.
Who Should Listen to How to Lie with Statistics
This audiobook is well suited to anyone who reads news regularly and wants a reliable framework for interrogating numerical claims, not just students, but working adults in any field where data gets cited to make decisions. It works as background for journalism students, as a refresher for people who studied research methods years ago and have since forgotten the vocabulary, and as a companion for anyone reading heavier books on political polling or media epistemics. Bill Gates has recommended it as ideal listening, which signals the audience correctly: this is for generalists who want a short, durable filter for the statistics they encounter daily. Those looking for current examples or computational depth should supplement with more recent material; those who need the foundational vocabulary will find it delivered here with unusual clarity and speed.
Is this audiobook based on the full text of the original 1954 book?
Yes, this audio edition covers the complete text of Huff's original book, including all of his chapters on sampling, graphical deception, correlation versus causation, and how to defend yourself against statistical manipulation.
Do you need a math background to follow this audiobook?
No mathematical background is necessary. Huff wrote explicitly for general readers, and DePuy's narration maintains that accessible register. The concepts are illustrated through examples rather than formulas.
