Jury instructions are not just formal language read at the end of trial. They tell jurors what legal standard governs the case. When the instructions change, the same facts may produce a different verdict.
The Commonwealth v. Berkowitz trial provides a useful example. The case involved a rape prosecution in Pennsylvania where the victim did not consent and repeatedly said “no,” but the defendant did not use force or threaten force. At the time, Pennsylvania followed a traditional definition of rape that required force or threat of force. Professor Dan Kahan used the case to study how different legal definitions of rape affect jurors’ judgments.
Kahan’s study was not originally designed as a trial-error study. It was part of a broader inquiry into how people understand consent, coercion, and criminal responsibility. But the data are valuable for trial-error analysis because they show how individual jurors respond when the legal rule changes.
My research in Measuring Fairness asks the next question: How much does that change in individual juror preferences affect the probability that a jury would convict?
From Individual Votes to Jury Verdicts
Kahan’s research found that different jury instructions produced different individual-level verdict preferences. In the Berkowitz study, 65% of respondents favored conviction under one instruction, compared with 53% under the other.
That 12-point difference is meaningful on its own. But courts do not ordinarily ask whether a jury instruction changes some individual jurors’ views. They ask whether the instruction likely affected the verdict.
That is where the method developed in Measuring Fairness adds value. By connecting jury-pool preferences to jury verdict probabilities, the analysis estimates the likely effect of different instructions on the final outcome. In Berkowitz, applying Pennsylvania jury rules, the analysis estimates that a “no means no” instruction would increase the probability of conviction by .257, or about 26 percentage points, compared with the common-law definition of rape.
That result shows why jury instructions can be outcome-determinative. The difference in individual-level preferences was 12 percentage points, but the difference in verdict probabilities was more than twice as large because the case fell in the range where many juries would be closely divided. In close cases, moving even one or two jurors can substantially alter the probability of conviction.
Why This Case Study Matters
The Berkowitz case study demonstrates the importance of measuring the effect of jury instructions. Legal definitions do not merely describe the law in the abstract. They shape how jurors classify conduct, evaluate consent, assess culpability, and decide guilt.
This case also shows the versatility of the broader method. I did not conduct Kahan’s original study. But Measuring Fairness makes it possible to use individual-level experimental results to estimate jury-level verdict probabilities. That translation is crucial because legal standards for harmless error, prejudice, and fair trial analysis usually turn on the probability of a different outcome.
The analysis does not imply that every change in jury instructions is a trial error. States may define crimes differently. But when a court gives the wrong instruction, or when litigants dispute whether an instructional error was harmless, empirical analysis can help estimate whether the mistake likely changed the verdict.
This is the kind of problem Fair Trial Analysis was built to address. The organization’s work on jury instructions, trial errors, and verdict probability analysis reflects a broader commitment to rigorous, transparent, and practically useful methods. The goal is not to favor one side in litigation. It is to help lawyers and courts determine whether a legal error likely affected the outcome.
The larger point is that jury instructions are part of the causal structure of a trial. When the legal rule changes, the verdict may change too. Fairness requires a method capable of measuring that effect.