* Criminal Defense Lawyer, Cicchini Law Office, LLC.
** Professor and Chair of Psychology and Director of the Law and Justice Program, Beloit College.
Introduction
The Constitution protects us from criminal conviction unless the state can prove guilt beyond a reasonable doubt. 1 In re Winship, 397 U.S. 358, 364 (1970) (holding “the Due Process Clause protects the accused against conviction except upon proof beyond a reasonable doubt”). However, after defining reasonable doubt, many trial courts will then instruct jurors “to search for the truth” of what they think really happened. 2 See infra note 12 and accompanying text (giving examples of jury instructions from different states). Defendants have argued that such truth-related language reduces the state’s burden of proof to a mere preponderance of the evidence. That is, if the jury were to find the state’s case only slightly more convincing than the defendant’s, it would follow that, in a search for the truth, the jury would be obligated to convict.
Appellate courts, however, consistently reject this argument. 3 See Michael D. Cicchini & Lawrence T. White, Truth or Doubt? An Empirical Test of Criminal Jury Instructions, 50 U. Rich. L. Rev. 1139, 1158–59 & nn.60–63 (2016). Most appellate courts acknowledge that such truth-related language is inaccurate, highly disfavored, and could, in theory, lower the state’s burden of proof. 4 Id. at 1158 n.60. However, these courts then go on to conclude, without any empirical support, that such language probably does not cause any actual harm.
In our previous study and article, we put this judicial reasoning to the test. 5 Id. at 1150. In a hypothetical criminal case, we found that mock jurors who were properly instructed on reasonable doubt convicted the defendant at a rate of 16%. 6 Id. at 1155. However, mock jurors who received the identical case information and instruction and were also told “not to search for doubt” but instead “to search for the truth” convicted at a much higher rate of 29%. 7 Id.
In this Piece, we discuss the results of our new study wherein we first attempted a conceptual replication of our previous work and then attempted to identify a cognitive explanation for why truth-related language produces a higher conviction rate. Just as in our previous study, we found that mock jurors who were instructed “not to search for doubt” but instead “to search for the truth” convicted at a significantly higher rate than mock jurors who were properly instructed on reasonable doubt. Unlike our previous study, however, our new study also asked jurors a postverdict question about their subjective understanding of the burden of proof. Through this, we found that jurors who were first instructed on reasonable doubt and then told “not to search for doubt” but instead “to search for the truth” were nearly twice as likely to believe they could convict the defendant even if they had a reasonable doubt about his guilt. 8 See infra section II.C (describing the effect of different instructions on mock jurors’ responses to the postverdict question). Even more significant, jurors who held this mistaken belief (regardless of the group to which they were randomly assigned) actually convicted at a rate 2.5 times that of jurors who correctly understood the burden of proof. 9 See infra section II.C (describing the relationship between mock jurors’ responses to the postverdict question and their conviction rates).
Part I of this Piece details the burden of proof in criminal cases and examines the truth-related language that trial courts commonly tack on to the end of their reasonable doubt jury instructions. It then explains our previous study, including our study design and statistical findings. Part II, the heart of this Piece, examines our new study—a conceptual replication and extension of our previous work. In this Part, we outline our study objectives, formally state our hypotheses, discuss our study design, and explain our statistical findings.
Part III then explains the significance of our findings for trial judges, jury-instruction committees, and appellate courts. We also discuss the cognitive link between jury instructions and conviction rates—that is, truth-related language causes jurors to misunderstand the state’s burden of proof, which in turn causes jurors to convict even when they have a reasonable doubt about guilt. Based on our successful replication and new findings, we reiterate our argument from our previous article: In order to protect due process rights, courts should terminate their use of truth-based jury instructions. This Part also discusses the study limitations we corrected by virtue of this conceptual replication, as well as the study limitations that still exist but could be addressed by researchers in future studies.
I. Proof, Truth, and Doubt
A. The Burden of Proof in Criminal Cases
In 1970, the Supreme Court of the United States explicitly held that “the Due Process Clause protects the accused against conviction except upon proof beyond a reasonable doubt.” 10 In re Winship, 397 U.S. 358, 364 (1970). However, trial courts are given tremendous leeway in how they instruct jurors on this burden of proof. 11 See Victor v. Nebraska, 511 U.S. 1, 5 (1994) (“[T]he Constitution does not require that any particular form of words be used in advising the jury of the government’s burden of proof.”). And in so doing, many trial courts will conclude their instructtions not by telling jurors to examine the state’s evidence for reasonable doubt but instead by telling them to decide the truth of what they think really happened. In our previous article, we provided examples from 13 jurisdictions, including the following:
After giving an otherwise legally proper instruction on proof beyond a reasonable doubt, many courts will then instruct jurors that, when reaching their verdict, they should “[d]etermine what [they] think the truth of the matter is and act accordingly.” Similarly, other courts instruct jurors that, when reaching their verdict, they should “evolve the truth,” “seek the truth,” “search for the truth,” or “find the truth.” Some courts—again, after properly instructing jurors on the concept of reasonable doubt—will explicitly contradict themselves by further instructing jurors that “you should not search for doubt. You should search for the truth.” 12 Cicchini & White, supra note 3, at 1143 (footnotes omitted) (quoting United States v. Gonzalez-Balderas, 11 F.3d 1218, 1223 (5th Cir. 1994); United States v. Gray, 958 F.2d 9, 13 (1st Cir. 1992); United States v. Pine, 609 F.2d 106, 108 (3d Cir. 1979); Commonwealth v. Allard, 711 N.E.2d 156, 159 (Mass. 1999); State v. Dunkel, 466 N.W.2d 425, 430 (Minn. Ct. App. 1991) (emphasis omitted); State v. Avila, 532 N.W.2d 423, 429 (Wis. 1995) (emphasis added), overruled in part on other grounds by State v. Gordon, 663 N.W.2d 765 (Wis. 2003)).
Defendants have frequently challenged such truth-related language on appeal. One defense argument is that instructing the jury to determine, evolve, seek, find, or search for the truth of what they think happened diminishes the state’s burden of proof. That is, “‘seeking the truth’ suggests determining whose version of events is more likely true, the government’s or the defendant’s, and thereby intimates a preponderance of evidence standard.” 13 Gonzalez-Balderas, 11 F.3d at 1223 (emphasis added). More to the point, “truth is not the jury’s job.” 14 State v. Berube, 286 P.3d 402, 411 (Wash. Ct. App. 2012). Rather, “[t]he question for any jury is whether the burden of proof has been carried by the party who bears it. In a criminal case . . . [t]he jury cannot discern whether that has occurred without examining the evidence for reasonable doubt.” 15 Id.
Our nation’s appellate courts, however, have consistently rejected this argument. With only slight variations in their reasoning, appellate courts conclude that while such truth-related language is disfavored—and could, in theory, diminish the state’s burden of proof—it probably does no actual harm. 16 Cicchini & White, supra note 3, at 1158–59 & nn.60–63. One court, however, stated that such truth-related language “would be error if used in the explanation of the concept of proof beyond a reasonable doubt.” Gonzalez-Balderas, 11 F.3d at 1223. This, in turn, allows appellate courts to affirm defendants’ convictions and allows trial courts to continue to instruct juries to search for the truth of what they think really happened, rather than to examine the state’s evidence for reasonable doubt.
Based on the plain language of these truth-related jury instructions, we believed the courts’ thinking was quite obviously flawed; therefore, we decided to put their judicial reasoning to an empirical test.
B. The Original Study: Truth or Doubt?
In our previous study and article, we compared the conviction rates in a hypothetical criminal case. 17 Cicchini & White, supra note 3, at 1154–56. We recruited 200 study participants 18 Our previous study actually consisted of 300 participants; this number was reduced to 298 after excluding two participants who were not U.S. citizens. Id. at 1150–51. However, 98 of the mock jurors were randomly assigned to a separate group that received no reasonable doubt instruction of any kind. Id. at 1154. We discovered that mock jurors who were instructed on reasonable doubt and then told “not to search for doubt” but instead “to search for the truth” convicted at the identical rate as jurors who received no reasonable doubt instruction whatsoever. Id. at 1154–55. through Amazon’s Mechanical Turk 19 Amazon’s Mechanical Turk is an online platform for conducting social science research. See Amazon Mechanical Turk, http://www.mturk.com/mturk/welcome [http://perma.cc/8T3A-DJC5] (last visited Oct. 17, 2016). Several studies have found a high degree of similarity between the judgments and behaviors of Mechanical Turk “workers” and of participants recruited in more conventional ways, such as through university subject pools. See Winter Mason & Siddharth Suri, Conducting Behavioral Research on Amazon’s Mechanical Turk, 44 Behav. Res. Methods 1, 3–4 (2012), http://sidsuri.com/Publications_filfi/mturkmethods-print.pdf [http://perma.cc/69VN-THJ4]. for the purpose of testing the following hypothesis: “[W]hen truth-related language is added to an otherwise proper beyond a reasonable doubt instruction, the truth-related language not only contradicts but also diminishes the government’s burden of proof.” 20 Cicchini & White, supra note 3, at 1150.
To test this hypothesis, each study participant served as a mock juror and received the same case summary materials. More specifically:
Every mock juror read the same fact pattern in a hypothetical case of sexual assault of a child. The defendant in the case was alleged to have touched a fifteen-year-old child’s buttocks, over the clothing, for purposes of sexual arousal or gratification. The case summary began with an instruction on the charged crime, including its elements, followed by a 625-word synopsis of court testimony from three individuals: the alleged child victim, the child’s mother, and the defendant. The child’s accusation was not corroborated by an eyewitness or physical evidence. In essence, the case consisted, as most real-life sexual touching cases do, of an allegation and a denial. The case summary concluded with an 850-word transcript of the prosecutor’s and defense lawyer’s closing arguments, each arguing the points most favorable to their case. 21 Id. at 1151.
Before being asked to render a verdict, thes e 200 mock jurors were randomly assigned to one of two groups, each of which received a different instruction on the state’s burden of proof. 22 Id. at 1152. Jurors in the doubt-only group (N = 100) received a legally proper, 269-word burden-of-proof instruction that concluded as follows: “It is your duty to give the defendant the benefit of every reasonable doubt.” 23 Id. at 1152–53.
Jurors in the doubt-and-truth group (N = 100) received the same instruction except that the conclusion was changed to read as follows: “While it is your duty to give the defendant the benefit of every reasonable doubt, you are not to search for doubt. You are to search for the truth.” 24 Id. at 1153–54.
The doubt-only group (that received the legally proper instruction) convicted at a rate of only 16%. 25 Id. at 1155. However, the doubt-and-truth group (that was told “not to search for doubt” but instead “to search for the truth”) convicted at the much higher rate of 29%. 26 Id. More specifically:
This result is significant at the p < .05 level, with an exact p-value of 0.028 . . . . [T]he p-value measures the probability of a Type I error, i.e., obtaining a false positive. Therefore, we are more than 97% certain (1–p) that the difference in conviction rates between [the groups] is a real difference and did not occur by chance.
This finding provides strong empirical support for our . . . hypothesis that the truth-related language at the end of an otherwise proper reasonable-doubt instruction actually diminishes the government’s burden of proof. 27 Id.
We concluded that “[b]ecause ‘the Due Process Clause protects the accused against conviction except upon proof beyond a reasonable doubt,’ our findings provide strong evidence of a serious constitutional problem.” 28 Id. at 1157 (quoting In re Winship, 397 U.S. 358, 364 (1970)).
II. The New Study: A Conceptual Replication
A. Objectives and Hypotheses
First, our main objective is to test the reliability of our previous finding by replicating the study. In order to do this, we designed and conducted a conceptual replication rather than a direct replication. A conceptual replication retests the original hypothesis but intentionally varies specific features of the original methodology. 29 See, e.g., Stefan Schmidt, Shall We Really Do It Again? The Powerful Concept of Replication Is Neglected in the Social Sciences, 13 Rev. Gen. Psychol. 90, 91 (2009). A benefit of conceptual replication is that it addresses one of the common weaknesses of psychological research: limited generalizability. 30 Generalizability refers to the extent to which the results of a study can be applied from the study sample to a larger population or from the specific circumstances of the study to other circumstances. For a discussion of external validity (i.e., generalizability) and threats to external validity, see Thomas D. Cook & Donald T. Campbell, Quasi-Experimentation: Design & Analysis Issues for Field Settings 70–80 (1979).
More specifically, in our original study as discussed in section I.B, each of the 200 mock jurors read the same case summary about a child sexual-assault allegation. All of the material was held constant between the two test conditions. This consistency allowed us to isolate the effect of the variable being tested: the closing mandate “not to search for doubt” but instead “to search for the truth.” However, with such standardization comes limited generalizability. As we cautioned, “we cannot say that the impact of the doubt-and-truth instruction would be identical when applied to different cases.” 31 Cicchini & White, supra note 3, at 1162.
A conceptual replication allows us to address this limitation by testing our hypothesis under a different set of circumstances. As we discuss below in section II.B, our new study has a larger sample size, a different fact pattern, and includes stronger evidence of the defendant’s guilt. We also provided mock jurors with a shorter underlying instruction on reasonable doubt. However, the variable being tested—the mandate “not to search for doubt” but instead “to search for the truth”—is the identical language that we tested in our previous study.
Second, in addition to replicating our study, we also extended our study so as to identify a cognitive link between the change in the test conditions (i.e., adding truth-related language to one group’s burden-of-proof instruction) and the change in juror behavior (i.e., a higher conviction rate). In order to accomplish this secondary objective, we added an additional, postverdict question to our test materials. We discuss this and our overall study design below. 32 Infra section II.B. First, we will formally state our hypotheses.
Our first hypothesis is that when truth-related language is added to an otherwise proper reasonable-doubt instruction, the truth-related lanCguage will diminish the state’s burden of proof—i.e., mock jurors will convict at a higher rate. Our second hypothesis is that mock jurors who receive the truth-related language at the end of their reasonable-doubt instruction will subjectively interpret their instruction to permit conviction even if they have a reasonable doubt about the defendant’s guilt.
B. Study Design
To test these hypotheses, we recruited 250 study participants—a 25% increase in the sample size of our original study—through Amazon’s Mechanical Turk. 33 See supra note 19 and accompanying text (discussing Amazon’s Mechanical Turk). These 250 participants served as mock jurors and rendered a verdict in a hypothetical criminal case. To ensure data quality, we monitored the participants and immediately rejected those who completed the task in fewer than three minutes; we replaced them with new participants in order to maintain our desired sample size. Each participant was required to be an adult and a U.S. citizen. After data collection was completed, we discovered that one participant was not a U.S. citizen and one failed to render a verdict; their data were discarded, leaving us with a sample of 248 mock jurors.
Our sample was large and diverse. Participants hailed from 42 different states. Fifty-two percent of participants were female. Participants’ ages ranged from 19 years to 73 years; the mean (average) age was 35.8 years, and the median age (50th percentile) was 32 years. The ethnic composition of the sample was also diverse: 74% non-Hispanic whites, 10% African Americans, 5% Hispanics, 5% Asian Americans, 5% mixed race, and 1% other. Fifty-six percent of the participants reported at least a four-year college degree, while an additional 35% have completed some college. Thirteen percent reported having prior jury experience.
Every mock juror read the same fact pattern, which involved two adults interacting at a party and concluded with an accusation of a misdemeanor fourth-degree sexual assault—i.e., the defendant’s sexual touching of the alleged victim without her consent. The case summary began with an instruction on the charged crime, including its elements, followed by an 887-word summary of the trial evidence. The evidence consisted of testimony from two witnesses—the accuser and the defendant—and a factual stipulation entered into between the prosecutor and defense lawyer. 34 We included a factual stipulation for two reasons. First, it allowed us to shorten the summaries of the witnesses’ testimonies by removing “identity” as an issue in the case. Second, the data used in this study were obtained as part of a larger data-collection effort that included a third group. The inclusion of the factual stipulation allowed us to test an additional hypothesis that is not related to this study but may form the basis for a future article.
There were no eyewitnesses to the alleged sexual assault. The accuser immediately reported the incident to law enforcement. The defendant denied the allegation. Both the accuser and the defendant testified and admitted to consuming alcohol during the party at which the sexual assault allegedly occurred. The defendant, however, also admitted to consuming other drugs earlier in the day and admitted to a prior, unrelated instance of untruthful conduct. In order to shorten the overall length of the case summary materials, we did not include closing arguments from the lawyers. We did, however, instruct the jury that the definition of “evidence” includes the testimony of witnesses as well as the factual stipulation.
Before being asked to render a verdict of guilty or not guilty, the 248 mock jurors were randomly assigned to one of two test conditions, each of which received a different jury instruction on the state’s burden of proof. Group 1 (N = 124) received a legally proper, 94-word jury instruction that explained the presumption of innocence, placed the burden of proof on the state, and identified the burden of proof as beyond a reasonable doubt. This doubt-only instruction, in its entirety, reads as follows:
The defendant is presumed to be innocent of the charge. This presumption continues during every stage of the trial and your deliberations on the verdict. It is not overcome unless from all the evidence in the case you are convinced beyond a reasonable doubt that the defendant is guilty as charged. The government has the burden of proving the guilt of the defendant beyond a reasonable doubt.
This burden of proof stays with the government throughout the case. The defendant is never required to prove his innocence or to produce any evidence at all. 35 Comm. on Fed. Criminal Jury Instructions for the Seventh Circuit, Pattern Criminal Jury Instructions for the Seventh Circuit No. 2.03 (1998).
Group 2 (N = 124) received an identical jury instruction, with one exception. The instruction given to Group 2 concluded with this additional mandate: “While it is your duty to give the defendant the benefit of every reasonable doubt, you are not to search for doubt. You are to search for the truth.” This doubt-and-truth instruction, in its entirety, reads as follows:
The defendant is presumed to be innocent of the charge. This presumption continues during every stage of the trial and your deliberations on the verdict. It is not overcome unless from all the evidence in the case you are convinced beyond a reasonable doubt that the defendant is guilty as charged. The government has the burden of proving the guilt of the defendant beyond a reasonable doubt.
This burden of proof stays with the government throughout the case. The defendant is never required to prove his innocence or to produce any evidence at all.
While it is your duty to give the defendant the benefit of every reasonable doubt, you are not to search for doubt. You are to search for the truth. 36 Id. (emphasis added). This jury instruction includes the Seventh Circuit instruction in its entirety. Id. The additional, truth-related language added to the end of this instruction is the identical language tested in our original study and is taken from Wisconsin’s pattern jury instruction on the burden of proof. Wis. Criminal Jury Instructions Comm., Wisconsin Jury Instructions—Criminal No. 140 (2016). This language is similar, and often identical, to the truth-related language used in the 13 different jurisdictions we identified in our original study and article. See Cicchini & White, supra note 3, at 1143 nn.13–18.
After rendering their verdicts, mock jurors were asked to answer a question about how they interpreted their burden-of-proof instruction. The question posed to all jurors was as follows: “You received an instruction from the judge explaining the prosecutor’s burden of proof. Which of the following do you believe is the most accurate summary of the judge’s instruction?” Jurors were instructed to “choose only one answer.” Their options were as follows:
A. If I have a reasonable doubt about the defendant’s guilt, I must not convict the defendant.
B. Even if I have a reasonable doubt about the defendant’s guilt, I may still convict the defendant if, in my search for the truth, the evidence shows the defendant is guilty.
The study was approved by Beloit College’s Institutional Review Board. 37 An institutional review board (IRB) is a committee that reviews and approves research that involves human participants. See Hazel Glenn Beh, The Role of Institutional Review Boards in Protecting Human Subjects: Are We Really Ready to Fix a Broken System?, 26 Law & Psychol. Rev. 1, 25–26 (2002). IRBs ensure that researchers protect the rights and welfare of human subjects. See id.
C. Findings
To test our first hypothesis—that adding truth-related language to the end of an otherwise proper reasonable-doubt instruction will diminish the government’s burden of proof—we must compare the conviction rates of Groups 1 and 2.
In Group 1, which received the doubt-only instruction, only 28 of 124 mock jurors returned verdicts of guilt for a group conviction rate of 22.6%. In Group 2, which received the doubt-and-truth instruction, 41 of 124 mock jurors returned verdicts of guilt for a group conviction rate of 33.1%. That is, the conviction rate among jurors who were told “not to search for doubt” but instead “to search for the truth” was almost 50% higher than the conviction rate for jurors who were simply instructed to evaluate the state’s case for reasonable doubt.
This result is significant at the p < .05 level, with an exact p-value of 0.033. The p-value measures the probability of a Type I error—i.e., the risk of obtaining a false positive when testing a hypothesis, given the two sample sizes and the difference in conviction rates between the two groups. In plain language, we are more than 96% certain (1–p) that the observed difference in conviction rates between Groups 1 and 2 is a real difference and did not occur by chance. 38 A statistical test for the difference between two proportions produced a Z-score of -1.84. In a one-tailed test, the p-value is 0.033. Researchers use a two-tailed test (also called a two-sided test) when they cannot predict if a test variable will increase or decrease scores. We used a one-tailed test because we had empirical evidence (from our first study) that truth-instructed jurors would convict at a higher rate, not a lower rate. For a full discussion of when to use one-tailed and two-tailed tests, see Arthur Aron & Elaine N. Aron, Statistics for Psychology 199–202 (3d ed. 2003).
After mock jurors rendered their verdict, they were asked to report how certain they were (on a 10-point scale) that they had made a correct decision. There were no statistically significant differences in levels of certainty between the doubt-only group and the doubt-and-truth group. In fact, both group means were essentially 6.6 (fairly certain) on the 10-point scale. 39 We also uncovered several subsidiary findings not directly related to the main purpose of our study: (a) women (34%) were more likely than men (22%) to vote guilty (p < .04); (b) there were no statistically significant relationships between a participant’s verdict and his or her age, education, ethnicity, or prior jury experience; and (c) mock jurors who voted guilty were significantly more certain than other jurors were that they had made the correct decision (a mean score of 7.5 versus a mean score of 6.3 on a 10-point scale, p < .001).
Participants also answered an attention-check question that tested their recollection of the elements of the charged crime. The question included 5 potential elements, only 3 of which were correct. The attention-check results were encouraging. Nearly 92% of participants correctly identified the elements of the charged crime. 40 Our standard for a correct answer was high; the answer of a mock juror who identified the correct elements of the charged crime, but also an incorrect element, was classified as “incorrect.” Those mock jurors who voted not guilty were correct 93% of the time, while those who voted guilty were correct 87% of the time. This difference is not large enough to be statistically significant, but it suggests that those mock jurors who paid closer attention to the legal elements of the charge (fourth-degree sexual assault) were less likely to convict.
To test our second hypothesis—that mock jurors receiving the doubt-and-truth instruction would subjectively interpret it to permit conviction even if they had a reasonable doubt about the defendant’s guilt—participants were reminded that they had received an instruction from the judge about the state’s burden of proof. Participants were then asked to indicate the most accurate summary of the judge’s instruction by choosing either answer A or answer B. As indicated above, A is the correct interpretation of the constitutionally mandated burden of proof and B is the incorrect interpretation, as it permits conviction even when there is a reasonable doubt about the defendant’s guilt.
In Group 1, which received the doubt-only instruction, only 15% of participants selected answer B; that is, only 15% believed they could convict the defendant if they had a reasonable doubt about guilt. However, in Group 2, which received the doubt-and-truth instruction, 28% selected answer B; that is, 28% believed they could convict the defendant even if they had a reasonable doubt about guilt. This difference is highly significant (p = 0.01).
Perhaps more importantly, when analyzing the responses across both Groups 1 and 2, a juror’s understanding of the burden-of-proof instruction was an incredibly strong predictor of his or her verdict. Of those participants who selected the legally correct answer A—that they could not convict if they had a reasonable doubt about the defendant’s guilt—only 21% voted guilty. Of those who selected the legally incorrect answer B—that they could convict despite their reasonable doubt about the defendant’s guilt—54% voted guilty. This difference is highly significant (p < .001).
III. Implications and Limitations
A. Discussion: An Even Stronger Case Against Truth
Our first finding confirms our hypothesis that adding truth-related language to the end of an otherwise proper reasonable-doubt instruction diminishes the state’s burden of proof. That is, the jurors in Group 1, who were instructed simply to evaluate the state’s evidence for reasonable doubt, convicted at a rate of 22.6%. However, the jurors in Group 2, who were instructed “not to search for doubt” but instead “to search for the truth,” convicted at a rate of 33.1%—a conviction rate nearly 50% higher than Group 1’s rate. This replicates the finding in our original study, which also revealed a statistically significant gap in conviction rates when testing the identical hypothesis. 41 Cicchini & White, supra note 3, at 1155 (finding conviction rates of 16% in the doubt-only group and 29% in the doubt-and-truth group, with a p-value of 0.028).
Our second finding in this study is, in some ways, even more compelling. We hypothesized that jurors who received the doubt-and-truth instruction would be more likely to subjectively interpret the burden of proof to permit conviction even if they had a reasonable doubt about the defendant’s guilt. What we found was that in Group 1 (doubt only), only 15% of jurors believed they could convict the defendant if they had a reasonable doubt about guilt. However, in Group 2 (doubt-and-truth), 28%—nearly double—believed they could convict the defendant even if they had a reasonable doubt about guilt.
Even more striking, when analyzing the responses of all participants across groups, jurors who mistakenly believed they could convict, even when they had a reasonable doubt about guilt, found the defendant guilty 54% of the time. This conviction rate is more than 2.5 times the conviction rate (21%) of jurors who correctly understood the burden of proof—a highly significant difference.
These findings suggest that we have identified a cognitive mechanism that explains why the truth-related language produces a much higher conviction rate. Specifically, the truth instruction (TI) produces in jurors a mistaken belief (B) about the legally mandated burden of proof, and jurors base their verdicts (V) on that mistaken belief.
That is, in our original study we demonstrated the impact of the truth-related jury instruction on jurors’ conviction rates, but we did not attempt to explain why, in a cognitive sense, the truth-related language led so many jurors to find the defendant guilty. In this study, however, we have demonstrated empirically that TI → B and B → V. The mistaken belief B is the intermediate cognitive mechanism that explains the impact of TI on V. In plain language, telling jurors not to focus on doubt but instead “to search for the truth” leads them to form an incorrect understanding of the state’s burden of proof. This misunderstanding, in turn, leads many jurors to vote guilty, even when the state has not met its burden.
Because “the Due Process Clause protects the accused against conviction except upon proof beyond a reasonable doubt,” 42 In re Winship, 397 U.S. 358, 364 (1970). courts that tack truth-related language on to their burden-of-proof instructions are creating a serious constitutional problem. Therefore, jury-instruction committees and individual trial judges should eliminate truth-related language from those instructions. 43 Cicchini & White, supra note 3, at 1158. Further, such a simple change should not be controversial. No court claims that truth-related language is necessary or even valuable. Rather, such language is merely tolerated based on the assumption that it probably does no actual harm 44 Id. at 1158–59. —an assumption that we have debunked by demonstrating that truth-instructed jurors convict at a higher rate.
Further, our second study has not only replicated the result of our first study, but it has also identified a cognitive mechanism that serves as a bridge or link between the legally defective, truth-based instruction and the jurors’ higher conviction rate. This makes an even more compelling case for the removal of truth-related language from burden-of-proof jury instructions.
B. Study Limitations and Further Testing
In our original study and article, we identified five potential limitations that researchers may wish to address in future studies. 45 Id. at 1159–65. In this new study, we have addressed two of those five ourselves.
First, by conducting a conceptual replication, rather than a direct replication, we have expanded the generalizability of our findings. In our original article we cautioned that “we cannot know the extent to which this effect will also be observed in other cases with different fact patterns.” 46 Id. at 1161. Therefore, in this new study we changed the fact pattern. Instead of a delayed report by a child accuser, we used an immediate report by an adult accuser. We also incorporated more evidence of guilt than in our first study, including the defendant’s drug use on the day of the incident and the defendant’s prior, unrelated instance of untruthful conduct. Both of these pieces of evidence tend to diminish the credibility of the defendant’s testimony.
In addition to changing the fact pattern, we also changed other parts of the case summary to further expand the generalizability of our findings. We eliminated closing arguments of the lawyers on both sides. We added an instruction telling jurors that “evidence” includes the testimony of witnesses, which was designed to correct any misconception that physical evidence is required in order to convict. We also changed the underlying jury instruction on reasonable doubt. Instead of the lengthy, 269-word doubt-only instruction from our original study, 47 Id. at 1152–54. we used a much shorter, 94-word doubt-only instruction. 48 As explained earlier in this Piece, states are given tremendous leeway when instructing juries on reasonable doubt. And while the two doubt-only instructions used in our two studies are dramatically different in length and content, both are legally proper. What remained unchanged from our original study, however, was the closing mandate (for one of the two groups) that “[w]hile it is your duty to give the defendant the benefit of every reasonable doubt, you are not to search for doubt. You are to search for the truth.” 49 Cicchini & White, supra note 3, at 1153–54.
Second, we also corrected the problem of some participants’ inadequate attention level. In our previous study, we did not reject any study participants and included their data regardless of the amount of time they spent on the task. In this new study, however, we monitored the incoming data and rejected the work of study participants who spent fewer than 3 minutes on the task. These participants were replaced before we concluded the data-collection process. Our attempt to ensure quality responses was apparently successful, as reflected in the results of our attention-check question. In this new study, nearly 92% of mock jurors were able to correctly identify the legal elements of the charged crime. In our previous study, 84% of mock jurors correctly identified the legal elements. 50 Id. at 1156.
This leaves three potential study limitations for other researchers to address in the future: our use of the case summary method, the lack of juror deliberations, and participant bias. As we explained in our original study, however, the case summary method may actually be the best method for testing the impact of a jury instruction, as it eliminates extraneous variables, such as witnesses’ race and ethnicity, from the equation. 51 Id. at 1160–61. With regard to mock-juror deliberations, there is mixed evidence as to their value. 52 Id. at 1162–63. Finally, with regard to participant bias, this problem mirrors the problem with real-life juries and, for purposes of controlled studies like ours, is mitigated by the random assignment of participants to test conditions. 53 Id. at 1164–65.
Conclusion
In our previous study and article, we demonstrated that mock jurors who were first instructed on reasonable doubt and then told “not to search for doubt” but instead “to search for the truth” convicted at a much higher rate than mock jurors who received a legally proper reasonable-doubt instruction. In this new study—a conceptual replication and extension of our previous work—we replicated the results of our original study and identified a cognitive explanation for the difference in conviction rates: Mock jurors who were told “not to search for doubt” but instead “to search for the truth” were nearly twice as likely to mistakenly believe they could convict the defendant even if they had a reasonable doubt about guilt. Further, jurors who held this mistaken belief actually voted to convict the defendant at a rate that was 2.5 times that of jurors who properly understood the burden of proof.
Our original study, our successful replication of that study, and our new empirical findings regarding the cognitive explanation for juror behavior all combine to provide powerful evidence that truth-related language in jury instructions diminishes the constitutionally mandated burden of proof.