One Case, Two Files: Jurors Who Read the Record Voted Differently

Seventeen public jurors decided Liebeck v. McDonald's, Retried. They did not read the same case. All four jurors who found McDonald's liable had read the trial record.

Each juror was randomly assigned one of two files and could not see the other, or know that another existed. Nine received the story as the public tells it, roughly one hundred and twenty words: a woman bought coffee, spilled it, and won nearly $2.9 million. Eight received the trial record: coffee held at 180 to 190 degrees by written policy, more than 700 prior reports of burns between 1982 and 1992, an $800 response to a request for about $20,000 covering medical expenses and related losses, and third-degree burns requiring skin grafts.

The result, in counts because the panels are small: jurors given the popular story found the company liable 0 of 9. Jurors given the trial record split 4 to 4. Across both files, 4 of 17 jurors found liability, which is the 24 percent shown on the case page. That figure is an average across two different files and should not be read as this court's public verdict on the case; the two counts above are the finding.

One correction, published here rather than behind a link. After voting closed, the court found an error in its own popular-story file: it described Ms. Liebeck as riding in the car when the record shows the car was parked. The error makes her look more careless, which pushes toward no liability, and that is the direction the popular-story panel broke. Part of the 0 of 9 may therefore be the error rather than the file's brevity. The file was not quietly edited, because correcting a research file after votes are cast turns an experiment into an advertisement. The full accounting, including the notice to the affected jurors and the change to the court's verification process, is in the record addendum published with the case.

With that flaw on the table: under random assignment, a split this lopsided would arise by chance roughly three times in a hundred. Seventeen jurors is a small panel, nine and eight are smaller still, and one arm carries the error above. The finding is worth reporting. It is not established.

The 1994 Albuquerque jury, which heard the full record, found McDonald's liable, assigned Ms. Liebeck 20 percent of the fault, and awarded $200,000 in compensatory damages, reduced to $160,000 by her share, alongside $2.7 million in punitive damages that the trial judge cut to $480,000. The parties later settled on confidential terms.

The court draws the one conclusion the record supports: change what the jury reads and you may change the case.

The court's majority opinion and the dissent publish today. The opinions: truejury.com/verdict/liebeck-v-mcdonalds/. The experiment record and addendum: truejury.com/experiment/liebeck-v-mcdonalds/. Methodology: truejury.com/methodology/.

The TrueJury Project retries contested cases before public jurors and publishes the results as written opinions with the underlying vote counts, misses and flaws included.