Reading charts and their claims

Evidence file 01 / Clear presentation

03 / Clear presentationContext has been authorized for release.

Our first exhibit is not a trap. NASA's global temperature chart lets you inspect the annual observations, the longer-term pattern, and the uncertainty. A suspicious amount of useful information has been left in view.

Read the reference point, not just the direction

The vertical axis shows a temperature anomaly: a difference from the 1951–1980 average. Zero means that reference average—not freezing, and not “no temperature.” A value of +1°C means one degree warmer than that baseline. It does not mean the global temperature is one degree.

The black annual points preserve year-to-year variation. The red LOWESS line summarizes the longer-term pattern instead of replacing the observations. The gray band shows the estimated 95% uncertainty interval for annual values; it is not a forecast range or the variation in local weather. NASA's source page explains the smoothing and uncertainty calculations and supplies the data.

Why we approve this particular paperwork

The long time span makes it harder to mistake a short dip for the entire story. Units, baseline, observations, trend, and uncertainty are all inspectable. Even so, the image needs its accompanying explanation: “LOWESS” and “LSAT+SST” are not household vocabulary. Good is not the same as beyond improvement.

Authorized conclusion: This dataset shows long-term warming with annual variation and measurement uncertainty. The chart alone does not establish the causes, forecast next year's temperature, or tell you what happened in every town. Those require additional evidence. Our approval applies to the presentation—not a lifetime exemption from scrutiny for its publisher.

Data reference: GISTEMP Team (2026), GISS Surface Temperature Analysis, version 4, accessed September 13, 2026; Lenssen et al. (2024), A GISTEMPv4 observational uncertainty ensemble. NASA does not endorse Suscorp.

Evidence file 02 / Acknowledged design error

04 / Acknowledged design errorThe dogs did not change. The axes did.

The Economist's “Fit as a butcher's dog” chart compared average adult weight and neck size for dogs registered with the UK's Kennel Club. In the original, the two lines practically travel as a corporate delegation. In the redesign, the same data tell a less synchronized visual story.

Two rulers. One suspiciously tidy picture.

The original gives 18–21 kilograms and 42–45 centimetres the same vertical space. Both spans cover three units, but a kilogram is not a centimetre, and those changes are not proportionally equivalent. Adjusting two independent scales can bring lines together or pull them apart without altering a single observation.

Both measures do fall. What the near-overlap invites you to overread is that they change almost identically. Changing an axis does not change a calculated statistical correlation; it changes the visual impression. Nor does a falling average prove individual dogs got smaller. A changing mix of registered breeds could affect the average.

Why call this an error, rather than a plot?

Because we have more than the image. In Mistakes, we've drawn a few, visual journalist Sarah Leo describes the publication's misleading charts as unintentional, identifies this scale problem, and offers a correction. That is the publisher's account, not a mind-reading certificate. “Acknowledged design mistake” is better supported than either “secret deception” or “proven pure intentions.”

Authorized conclusion: Both averages declined in the displayed period. Their original visual near-match deserves less confidence than it first appears to demand. The redesigned chart is better for this comparison—not a perfect chart or proof of causation.

Sources: Sarah Leo, The Economist, March 27, 2019; Robert Kosara, Critiquing and redesigning, April 18, 2019. The example concerns historical registered-dog averages, not a current assessment of dog health.

Evidence file 03 / Documented misleading persuasion

05 / Documented misleading persuasionThe conclusion arrived before the evidence.

A bad chart does not automatically prove deception. This case has evidence beyond the chart: an agreed statement of facts in a criminal proceeding concerning OxyContin marketing. The distinction matters more than our corporate commentary.

Historical misleading claim—not medical guidance: The training diagrams below were used to promote a false impression that OxyContin had less abuse potential than short-acting opioids. They are reproduced to examine that claim, not endorse it.

A smooth line is not a safety study

The top diagram shows repeated rises into a zone labeled “Sedation, euphoria” and falls where pain returns. The bottom stays mostly within a “Pain relief” band. Before you inspect anything else, the titles have already called one use “inappropriate” and the other “appropriate.” The vertical axes provide no numerical blood concentrations.

A schematic can legitimately explain an idea. It cannot stand in for a measured comparison and then establish lower abuse potential. Here there are two separate leaps: this drawing represents how the drugs compare, and this shape establishes a safety advantage. Both need evidence. A pleasant curve is not a substitute.

What the record actually establishes

Paragraph 25 says that, during a December 1998 sales-manager training session sponsored by Purdue, a retained pharmacist used these diagrams instead of the actual clinical-data comparison and falsely stated that OxyContin's fewer peak-and-trough effects meant less euphoria and less abuse potential.

Paragraph 26 describes some new sales representatives being permitted in training to draw their own blood-level graphs to falsely represent those claims. Paragraph 27 records certain representatives using similar graphics and false claims with health-care providers between 1999 and June 30, 2001.

That documented use supports calling this misleading persuasion, rather than guessing motives from the line's appearance. It does not establish what every employee knew or an individual designer's private intentions. We do not need to enlarge the finding to make the lesson serious.

Authorized conclusion: These diagrams were used to sell a false impression of lower abuse potential, as documented in the company's agreed statement of facts. That conclusion rests on the record, not just on our dislike of the design. This historical analysis is not advice about an individual's treatment.

Primary source: United States v. The Purdue Frederick Co., Inc., No. 1:07-cr-00029-JPJ, Document 5-2, filed May 10, 2007, Attachment B, paragraphs 21–27. The exhibit is reproduced only for analysis and criticism; its promotional claims are not adopted by Suscorp.

08 / Supporting sourcesKeep the evidence attached.

• NASA GISS: graph, data, and uncertainty description — the temperature example. Dataset and methodological references appear alongside the exhibit.
• Sarah Leo, The Economist: Mistakes, we've drawn a few — the publisher's account of its design errors and revisions. Robert Kosara's contemporary discussion also reproduces the comparison.
• Purdue agreed statement of facts, paragraphs 21–27 (PDF) — the historical diagrams and documented misleading use.

Full-size images, attribution, context, and specific references remain beside each exhibit. The transit figures are invented; none of these sources is evidence of a real transit campaign.