David Crump


The Federal Rules of Evidence, and rules in the States, allow for impeachment of the testimony of a witness by proof of the witness's criminal convictions. If the witness is the criminal defendant, however, there are restrictions on this kind of impeachment. The theory is that the jury is supposed to use the evidence solely for impeachment and not to support an inference that the defendant has a propensity toward committing crimes. But intuition tells us that the jury is likely to be influenced toward the prohibited inference of guilt of the crime charged rather than devaluation of credibility alone.

The restrictive provisions are a compromise, and like many compromises, they include a component of illogic. The derogation of credibility that is intended to result is based on a set of inferences to the effect that people with criminal convictions are bad persons and therefore likely to do bad things, including giving false testimony. This reasoning is difficult to separate from a propensity toward crime. To reduce this effect, the judges usually give juries instructions that tell them not to use the conviction to infer guilt but only to diminish the defendant's credibility.

But what do people really think, when they are provided with the defendant's prior convictions? Researchers have conducted experiments to find out, and their results support the common intuition that people infer guilt rather than lesser credibility. The experiments are unavoidably imperfect, because they do not resemble the events in a courtroom with a defendant (and a victim) whose interests are at issue. Furthermore, the experiments have tended not to reproduce instructions from a judge. Therefore, this article reports on a new experiment that more closely represents the conditions in a trial.

The experiment, here, included evidence that was relatively strong in favor of inferring guilt, as one might expect in many cases today that are the result of prosecutorial screening. In such a case, the defendant's testimony is likely to reflect reasons for lesser credibility even without convictions. These conditions do not characterize every criminal trial, but they are present in many. Experimental subjects were given two slightly different scenarios. One included a prior conviction together with an instruction to disregard it except in considering credibility, and the other did not include the prior conviction.

The results were surprising. The group that was told of the prior conviction (and given the cautionary instruction) was Iess likely, not more likely, to find the defendant guilty. The counter-intuitive outcome, in other words, was that the experimental subjects who were told about the defendant's prior conviction tended to find him not guilty. As is often the case with psychological experiments, the results did not explain why this result occurred. The most salient reasoning expressed by the experimental subjects was that the judge's instruction made them less sure of what they were doing. This extra caution may have resulted from the intended message of the instruction, and it may also have been caused by the confusion inherent in the instruction, because it is impossible to both disbelieve the defendant and simultaneously avoid using this disbelief to infer his likely guilt. The experiment had limitations, of course, that resemble those of similar experiments, and the article suggests further research. It may be, for example, that an experimental design with less forceful evidence of guilt would reach a different result.

Included in

Evidence Commons