Unpacking the Doubt in “Beyond a Reasonable Doubt”: Plausible Alternative Stories Increase Not Guilty Verdicts

Abstract

Does introducing alternative suspects diminish belief in a defendant's guilt? Participants read a fictional murder trial transcript. In some conditions, the defense attorney described how one or more other people could have committed the crime. Accusing one alternative suspect dramatically reduced guilty verdicts. However, accusing two or three was not much better than accusing one. Theoretically, the story model and support theory can be interpreted as accounting for these results. Practically, a defense that argues not only that the defendant is innocent but also that some other individual(s) could have committed the crime is more likely to attain an acquittal.

ACKNOWLEDGEMENTS

We thank Natasha Olinger and Jessica Schneider for help conducting research, Anna Poppe for writing the trial transcript, and Gerald Clore and Gregory Mitchell for thoughtful comments. Portions of this research were presented at the 2005 American Psychology-Law Society Conference in La Jolla, California. HMDC's senior honors thesis under the supervision of BAS formed the pilot study for this research.

Notes

¹Other definitions are more lenient, such as “a doubt based on reason” (Tanford, 1990, p. 78).

²We borrow the phrase “Plan B” from the television show, “The Practice,” in which the defense lawyers turned against innocent witnesses when they became desperate because chances of an acquittal seemed remote. The (fictional) Plan B strategy is like the TODDI (“This Other Dude Did It”) defense known among real trial lawyers. A similar defense is the SODDI “Some Other Dude Did It” defense. With TODDI, lawyers point to a specific other possible perpetrator; with SODDI the perpetrator remains unidentified.

³Fox and Birke (2002) show that lawyers will show the unpacking effect when asked to predict trial outcomes from a list of possible outcomes. Van Boven and Epley (2003) demonstrate that mock-jurors will show the unpacking effect when specific types of health detriments allegedly caused by an oil spill have been listed.

⁴There is an iterative logic to why we believed that the 3-Plan B condition should show more subadditivity than the 1-Plan B conditions. With no Plan B suspects, “Not Guilty” means (loosely) someone other than Wilson did it. With one Plan B suspect, “someone other than Wilson” gets unpacked into “Garrett” and “someone other than Wilson or Garrett.” With two Plan B suspects, that latter phrase gets unpacked into “Meyer” and “someone other than Wilson or Garrett or Meyer.” Each step of unpacking should increase subadditivity.

Note. Conf = Confidence in verdict (1–7); Likely = Likelihood that defendant committed the crime (1–7); Prob = Probability that defendant committed the crime (0–100). Items in columns with different superscripts are significantly different.

Reading down the columns, numbers with different superscripts are significantly different from each other. Superscripts are not shown for the last two dependant variables because we analyzed them on the overall (i.e., all participants together regardless of verdict choice). See Figures 1 and 2.

Note. P = testified on behalf of the prosecution; D = testified on behalf of the defense. When there were multiple Plan B suspects, their testimony and the questions about them in the questionnaire appeared in the order listed above.

Reading down the 1-Plan B column, numbers with different superscripts are significantly different from each other. When there were two Plan B suspects, the mean probabilities (and standard deviations) were: For AG and JM: AG = 22.0 (11.0) and JM = 3.2 (12.6). For AG and JT: AG = 24.9 (14.0) and JT = 16.0 (9.9). For JM and JT: JM = 23.5 (14.3) and JT = 20.0 (12.4).

In the 3-Plan B condition, a within-subjects ANOVA showed F(2, 84) = 14.79, p < .0001; that is, suspects rated earlier received bigger ratings.

⁵When we designed this study, we thought that the Plan B strategy might “backfire” if participants thought the defense was grasping at straws. Had that obviously occurred, we would have compared these participants, who read the information in a trial format, to participants who learned the same information from a purportedly objective source (e.g., a newspaper).