The Silence Penalty

I.     Introduction

For much of American history, criminal defendants could not testify.1 In fact, it was only a quarter century ago that the Supreme Court swept away the last vestiges of the testimonial prohibition, belatedly recognizing a criminal defendant’s constitutional “right to take the witness stand.”2 To justify its atextual ruling, the Court channeled “the considered consensus of the English-speaking world” that there could be “no rational justification for prohibiting the sworn testimony of the accused.”3 Legal commentators applauded. Despite regular appeals to historical intent and textual fidelity in other contexts, judges and academics across the ideological spectrum embrace the upstart constitutional right as an enlightened evolution, akin to the elimination of trial by ordeal.4

It was not always so. As reformers first ushered in an age of defendant testimony through statutes over a century ago, critics predicted dire consequences for the purported beneficiaries of the new right. Commenting on his state’s newly enacted statute in 1867, Massachusetts’ Supreme Court Justice Seth Ames argued that allowing defendant testimony would “destroy[] the presumption of innocence.”5 In light of jurors’ inevitably negative reaction to defendants who chose silence, Ames predicted defendants would have “practically no option at all”; the new right will “compel the defendant to testify” and “all will use it.”6

Judge Ames was prescient in some respects and spectacularly wrong in others. In particular, his prediction that “all” defendants would testify did not come to pass. In modern times, only about half of criminal defendants take the witness stand.7 Notably, refusing to testify is not limited to guilty defendants. Around 40% of defendants later exonerated by DNA evidence declined to testify at their initial trials.8 As this figure indicates, defendants with important stories to tell frequently sit silently while their attorneys plead their case.9

The remarkable prevalence of defendant trial silence can only be understood by reference to the consequences for those who do take the witness stand. While each case presents a variety of tactical considerations, the most concrete deterrent to testifying is a product of the evidence rules concerning prior crimes.10 In the modern era, most defendants who stand trial have a criminal record that predates the charged crime.11 This nation’s ongoing struggle with mass-incarceration suggests that the striking prevalence of trial defendants with prior convictions will only increase in the coming years.12

American evidence rules generally exclude evidence of prior crimes.13 But when defendants testify, their criminal record becomes eligible for admission as “impeachment.”14 Once a jury learns of a defendant’s record, it is more likely to convicta phenomenon labeled here the “prior offender penalty.”15  For example, famed DNA exoneree Ronald Cotton testified at trial to his innocence of a violent break-in and rape.16 He was then impeached with his past crimes, including “a prior conviction of assault on a female with intent to commit rape and a prior conviction of breaking and entering.”17 After hearing of Cotton’s earlier convictions, the jury disbelieved his (truthful) claim of innocence and convicted.18 

Since the choice to testify belongs to the defendant alone, a common reaction to the looming admissibility of prior crimes is to decline to take the witness stand. The customary defense tactic of remaining silent to avoid impeachment (or other harms) creates a new risk, however.19 When defendants do not testify, they suffer a different penaltylabeled here the “silence penalty.” It is the interplay between the “silence penalty” and the “prior offender penalty” that typically determines whether defendants testify, and how jurors react to that choice.20

The shaky historical pedigree and critical tactical importance of defendant testimony should make the topic a subject of spirited academic inquiry. Yet the literature is surprisingly thin. The legal and academic understanding of defendant testimony and trial silence has remained unchanged for decades. To the extent it is addressed at all, the modern discourse centers on a nebula of poorly-understood evidence doctrines and vacuous tactical folklore. As another commentator aptly summarized: “What has been largely missing from the debate are facts.”21

This Article seeks to reanimate the academic discourse on defendant testimony and highlight its importance to current criminal justice debates. The analysis focuses on empirical evidence, a key element of any serious discussion in the increasingly polarized criminal justice space. The data come from three sources. First, the Article (Part II) synthesizes the pertinent social science literature which contains a wealth of informative, experimental mock juror studies often overlooked by legal commentators. Next, the Article (Part III) presents the results of a new, 400-person mock juror simulation designed to fill a significant gap in the existing body of experimental data. Third, the Article (Part IV) explores observational data from real trials, drawing primarily from the well-known, but perplexing22 multi-jurisdictional study of felony trials conducted by the National Center for State Courts (NCSC). By tying together findings from the social science literature, the new mock jury simulation, and data from real trials, the Article is able to present (and support) a new, data-driven account of how jurors react to defendant testimony and its absence.

Analysis of the data summarized above unlocks a variety of insights. Perhaps the most intriguing of these is the underappreciated power of trial silence. As explained below, the evidence suggests that, broadly speaking, the “silence penalty” harms defendants nearly as much as the more universally-dreaded “prior offender penalty.” That is, a defendant who remains silent at trial suffers about the same damage to his acquittal prospects as a defendant who testifies and is “impeached” with a prior conviction.

The implications of this “parallel penalty” thesis (explored in Part V) are powerful for defendants, their attorneys and anyone concerned with the fairness and accuracy of our criminal justice system.23 For defendants and their attorneys, the tactical implications are clear. The lesson of the data reported in this Article is that defense attorneys should not counsel trial silence lightly. Instead, they should seek to have their clients testify absent a strong case- or defendant-specific factor that dictates silence.24 A criminal record may present such a factor, but in many circumstances, even the disclosure of a criminal record will be less damaging than the underappreciated silence penalty.

In a world where guilty pleas make up over 90% of convictions, trial tactics are only a part of the story.25 The real action in the criminal justice system happens pretrial and the parallel penalty dynamic operates there as well.26 The parallel penalties awaiting defendants with a criminal record at trial increase the pressure to forego trial and plead guilty. Defendants have only two trial options: testify or remain silent. If both options generate powerful penalties, guilty and innocent defendants will rationally bargain away an (illusory) presumption of innocence for a modicum of mercy.

Finally, the empirical evidence presented in this Article vindicates the handful of 19th century critics who objected to allowing defendant testimony in the first place. The critics’ reasoning does not map perfectly onto modern times, but the spirit of their critique rings true. For most defendants, including those with prior convictions and those who decline to testify, the right to testify probably does more harm than good.27 Distressingly, the data further suggest that this harm stems from juror assumptions about both defendant silence and prior convictions that, as far as the law is concerned, have no place in a criminal trial.28

II.     Deciding Whether to Testify or Remain Silent

Judge Ames’ forecast that “all” criminal defendants would testify falls flat in the modern era. But Ames was right that for many defendants the privilege to testify is more curse than blessing. As we will see, for most defendants the choice between testifying or remaining silent is an exercise in damage control.29 This Part isolates and analyzes the two options, with a particular focus on experimental evidence from the social science literature.

A.     The Prior Offender Penalty

Justice Benjamin Cardozo famously stated that “character is never an issue in a criminal prosecution unless the defendant chooses to make it one”; as far as the law is concerned, “a defendant starts his life afresh when he stands before a jury, a prisoner at the bar.”30 Unfortunately, Cardozo’s flowery sentiment only applies to silent defendants. When a defendant testifies, he becomes subject to cross-examination just like any other witness. The defendant-witness is “duty-bound to speak truthfully, entitled to the same privileges, and exposed to the same perils of impeachment, stress, embarrassment, and so on.”31

One longstanding method of impugning a witness’s sincerity is cross-examination regarding past crimes.32 Oliver Wendell Holmes explained the operative logic in 1884: A prior violation of the law indicates a “general readiness to do evil” and “from that general disposition” the jury may “infer a readiness to lie in the particular case.”33 As the evidence is admissible only with respect to the “witness’s character for truthfulness,”34 judges must instruct juries that the testifying defendant’s “prior convictions should only be used to judge [the defendant’s] credibility rather than his propensity to commit crimes.”35 When combined with the defendant’s right to testify, this evidentiary framework generates a peculiarity of the American trial system: The admissibility of a defendant’s criminal record typically depends on the defendant’s decision whether or not to take the witness stand. If the defendant testifies, his record will generally be admitted.36 If the defendant does not testify, it generally will be excluded.37

Social scientists have produced a valuable body of experimental research regarding the impact of prior convictions. This research reveals that “[j]urors are more likely to convict an accused if they receive information about previous convictions than if they do not.”38 That is no surprise. Admissible evidence generally damages one side or the other. Credibility impeachment with prior convictions should be no exception. Yet the research suggests that prior conviction impeachment does not operate as the law intends. Rather than relying on prior convictions as evidence of the defendant-as-witness’s character for truthfulness, jurors appear to rely on convictions as forbidden criminal propensity evidence“generalizing a defendant’s earlier bad act into bad character and taking that as raising the odds that he did the later bad act now charged.”39

If jurors used prior convictions as the law intends, past crimes that undermined the defendant’s truthful character, such as perjury, would be the most damaging to defendants’ chances of acquittal. Yet empirical research has shown that even when properly instructed, mock jurors convict most readily when presented with prior crimes that are similar to the charged crime, not, as the operating legal theory would predict, when presented with crimes related to dishonesty.40 The stronger salience of similar crimes as opposed to dishonesty crimes in generating guilty verdicts suggests that jurors use prior convictions to engage in legally forbidden criminal propensity reasoning.

A widely-cited study by Roselle Wissler and Michael Saks presented mock jurors with scenarios that included prior convictions for the same crime and prior convictions for perjury.41 The mock jurors were instructed on the proper use of the prior convictions through a variant of the standard instruction used in Massachusetts trials.42 Jurors in a mock murder case convicted 70% of the time when the defendant had a prior murder conviction, but only 50% of the time when the defendant had a prior perjury conviction.43 Other jurors given a mock auto theft case convicted 80% of the time when the defendant had a prior auto theft conviction, but only 70% of the time when the defendant had a perjury conviction.44 These results suggest that the mock jurors relied on the prior convictions as evidence of the defendant’s criminal propensities, not solely as impeachment.

The Wissler and Saks study, while extensively cited, suffers from some weaknesses. It relies on a relatively small sample size and its most powerful finding arises from an unusual scenario of a defendant with a prior murder conviction.45 Nevertheless, other researchers have reported the same result in similarly constructed mock juror studies.46 One portion of the new mock jury simulation presented in Part III tested for the same phenomenon, using a larger sample size and more typical crimes. Shoring up the findings referenced above, the mock juror simulation obtains similar results. The jurors in the simulation convicted more often when the defendant was impeached with a prior similar crime (82%) than when he was impeached with a dissimilar dishonesty crime (73%).47

Another social science study reviewed comments made by mock jurors during deliberations. It found that juror statements about prior convictions “tended not to center around the credibility issue” but instead “juries were more likely to discuss a prior conviction (particularly one for a similar crime) as a basis for inferring that the defendant is the type of person who is capable of committing the current offense.”48 Data from real trials points to the same conclusion. Analyzing data gathered from post-trial surveys with actual jurors, Valerie Hans and Theodore Eisenberg found that jurors’ self-reported assessments of the defendant’s credibility did not correlate with the admission of prior convictions.49 This suggests that real jurors are unimpressed with the legally sanctioned purpose for the admission of prior convictions.

Courts themselves acknowledge the danger that jurors will be unable to separate the proper from improper use of prior crimes evidence. The standard framework for assessing the admissibility of prior convictions as impeachment requires judges to consider, inter alia, “[t]he similarity between the past crime and the charged crime.”50 The introduction of similar crimes, the courts recognize, generates “inevitable pressure on lay jurors to believe that ‘if he did it before he probably did so this time.’”51 Although judges routinely admit prior convictions offered as impeachment, even for similar crimes,52 their recognition of this consideration stands as official acknowledgement of the difficulty jurors face in resisting the prohibited “criminal character” inference.

The academic literature includes one contrary claim that must be addressed. Two of the most prominent scholars to address prior convictions in recent years, Larry Laudan and Ronald Allen, argue that the empirical evidence regarding “how mock jurors handle prior crimes information” is “all over the map” and “contradictory.”53 They assert that a judge who believed that “prior crimes evidence has only a very limited effect on jury decisions” would find “plenty of studies to back up that prior disposition.”54 Laudan and Allen recognize that their claim runs counter to the (“clumsily cobbled together”) “conventional wisdom,” the views of academics, judges and “American evidence law.”55 Nevertheless, the prominence of these authors and the forcefulness of their assertion makes it necessary to examine their contention in some detail.

Laudan and Allen identify five studies that support the consensus described above—that when jurors learn of the defendant’s prior convictions, acquittal prospects drop.56 To make out the claim that the empirical evidence is nonetheless “all over the map,” Laudan and Allen cite four other studies that they believe point in the other direction.57 The four studies do not support Laudan and Allen’s contention. Three of the studies are easily dispensed with on the ground that they do not consider prior conviction evidence at all. Instead, in all three studies mock jurors learned that the defendant had made incriminating (or exculpatory) statements.58 The jurors were then instructed to disregard the statements because they were obtained through an illegal wiretap.59 These studies may be pertinent to the broader issue of jury instructions, but they concern a different species of evidence and altogether different considerations (e.g., police misconduct).60

Only one of the four studies Laudan and Allen cite for the proposition that the evidence on “how mock jurors handle prior crimes information” is “all over the map” concerns prior crimes. But the portion of that 40-year-old London61 study that supports Lauden and Allen’s contention is easily distinguished. The study presented a simulated rape trial of two co-defendants, and noted that mock jurors convicted at similar rates whether or not they heard about one of the co-defendant’s prior rape conviction. Importantly, the study’s authors introduced the prior conviction in an unusual manner. In real trials and in most simulations, prior crimes are introduced through unassailable sources such as an official record or the defendant’s own admission. Judges then instruct jurors to use the conviction only as impeachment, validating its accuracy. In the London study, however, the conviction was “let slip” in the other co-defendant’s testimonyand then the judge instructed the jurors to disregard it completely.62 This confounds the analysis. Jurors in the London study might have resisted the normal impact of prior convictions because they disbelieved the co-defendant’s offhand assertion that his accomplice had previously committed a rapeparticularly as each defendant stood to benefit by casting blame on each other.63

In sum, the prior conviction finding cited by Laudan and Allen that counters the broad consensus on prior crimes evidence can be explained as stemming from an atypical study design. The empirical evidence from mock juror experiments is one-sided and clear. The studies suggest that the introduction of prior conviction evidence substantially damages defendants’ chances for acquittal, primarily through a legally prohibited “criminal propensity” inference.64

B.     The Silence Penalty

The preceding section analyzed experimental evidence that illustrates the harm done to a defendant’s acquittal prospects when the jury learns of a prior conviction. To avoid this “prior offender penalty,” defendants can, and often do, choose to remain silent at trial.65 As one defense attorney manual explains, when it comes to defendant testimony, “[o]ftentimes . . . silence is golden. The client may be better off saying nothing and merely giving the appearance of a choir boy.”66 Yet, as this section reveals, the empirical evidence also demonstrates that remaining silent comes at a price.

There is little doubt that jurors view defendants’ silence with a cynical eye. A juror interview for the “Serial” podcast, a popular investigative series about a controversial murder conviction, illustrates the problem defendants face:

Reporter: “Did it bother you guys as a jury that [the defendant] Adnan [Syed] himself didn’t testify, didn’t take the stand?”

Juror: “[Y]eah, that was huge. We all kinda like gasped . . . we were all just blown away by that. You know, why not, if you’re a defendant, why would you not get up there and defend yourself, and try to prove that the State is wrong, that you weren’t there, that you’re not guilty? We were trying to be so open minded, it was just like, get up there and say something, try to persuade, even though it’s not your job to persuade us . . . .”67

The sentiment is not unusual. In public opinion surveys, about half of respondents say that a defendant who does not testify “is probably guilty” or “has something to hide.”68


Table 1: 2002 Public Opinion Poll69

Screen Shot 2018 02 05 at 4.41.43 PM


The apparently widespread belief that an innocent defendant would testify looms ominously over jury deliberations involving silent defendants.

The social science literature, and particularly research by psychology professor David R. Shaffer, supports the conclusion that jurors punish defendants for refusing to testify. Consistent with that of other researchers, Shaffer’s research reveals that mock jurors convict more readily when defendants appear to be withholding information.70 In one study, a group of mock jurors were told that a defendant, after testifying on direct, invoked his Fifth Amendment privilege not to incriminate himself in response to a question on cross-examination.71 These jurors convicted significantly more frequently.72 The study concludes that, “the defendant who gives the appearance of withholding crime-relevant information is likely to be viewed guilty and deserving of conviction.”73

Shaffer suspected that jurors would be less punitive when the defendant’s refusal to answer questions took the form of the invocation of the venerated Fifth Amendment right not to take the witness stand.74 With a co-author, he staged a sophisticated trial simulation where jurors were presented with either: (1) a defendant who did not testify; (2) a defendant who testified, but refused to answer a potentially incriminating question during cross-examination; or (3) a defendant who testified normally without refusing to answer any questions.75 The jurors were instructed, as applicable, that they could not hold the defendant’s invocation of the Fifth Amendment privilege against him.76 This time, Shaffer uncovered even more compelling evidence of a silence penalty. Two thirds of the juries that rendered verdicts in the scenarios where the defendant withheld information (either by declining to testify or refusing to answer a specific question) rendered a guilty verdict.77 By contrast, there were no guilty verdicts in the scenarios where the defendant did not invoke his Fifth Amendment privilege.78 Individual juror votes among the juries that could not reach a verdict followed the same pattern.79 Importantly, Shaffer found his surmise that jurors might respect the defendant’s choice not to testify at all (as opposed to testifying selectively) disproven. “[A]nalyses of the group verdicts, guilt ratings, and the verdicts of individual jurors revealed that defendants who declined to take the stand were judged just as harshly as their counterparts who refused to answer specific interrogation.”80

Another window into the impact of defendant silence comes from post-conviction case law. The law is stacked against challenges to juror verdicts as evidenced by legal doctrine that verdicts cannot be impeached by juror testimony regarding the content of their deliberations.81 This bar generally extends to “proof that one or more jurors held it against the accused that he failed to take the stand.”82 Still, challenges abound.83 A recent California case illustrates the genre. In Strand v. McDonald, a juror testified that the “defendant’s failure to testify was mentioned between six and twelve times during jury deliberations, by as many as half of the jurors. About one-quarter to one-third of the remarks included the view that ‘if [defendant] was innocent he would have testified.’”84 (Both the federal and state courts upheld the verdict.)85 Given the futility of these claims and the likely reluctance of jurors to broadcast their own malfeasance, the number of cases that raise the issue86 hints at an underlying reservoir of cases where jurors considered the defendant’s failure to testify as evidence of guilt.

The anecdotal and empirical evidence summarized above suggests that just as there is a significant “prior offender penalty,” there is also a powerful “silence penalty.” Although, as Shaffer notes, “how heavily” the silence penalty “weighs” remains an open question.87 The next Parts attempt to answer that question by analyzing the results of a new mock juror simulation and data from real trials.

III.     Comparing the Parallel Penalties: A Mock Juror Experiment

The mock juror studies summarized in Part II suggest that (1) jurors will convict more readily when they learn that a defendant has a prior criminal record; and (2) jurors will penalize defendants who do not testify. Yet the literature has a blind spot. Despite a bounty of mock juror studies, no previous study compares the effect of the introduction of a prior conviction against the effect of a failure to testifythe dilemma typically faced by criminal defendants. This Part presents the results of a trial simulation involving 400 mock jurors designed to do just that. As explained below, the simulation detected both a “silence penalty” and a “prior offender penalty.” Most interestingly, the simulation found the penalties to be roughly equal.

A.     Experimental Design

The mock juror experiment consisted of a simulated trial of a single defendant for breaking into a store and stealing jewelry. The simulation was designedand pilot testedto suggest guilt, but not conclusively. The goal was to construct a straightforward, realistic case that was close enough to engender disagreement, but still representative of a typical American criminal trial.88 (For example, in California the rate of conviction at trial in felony cases is reportedly higher than 80%; in Florida it is around 73%.)89

I recruited mock jurors using the Mechanical Turk employment marketplace. Researchers increasingly rely on Mechanical Turk for academic studies since it facilitates access to a broad range of willing research participants in a cost-and-time effective manner.90 As with jury service, Mechanical Turk users must be at least 18 years of age;91 eligibility was further limited to people located in the United States. A number of studies indicate that Mechanical Turk respondents are preferable in terms of representativeness and diligence to typical academic survey subjects.92 Mechanical Turk respondents, however, can skew younger, more female and more educated than the population at large. This would be problematic in a public opinion survey, but here we are looking at reactions to subtly-altered fact patterns distributed randomly to subsets of the survey group. It seems unlikely in this context that any potential differences between a Mechanical Turk sample and a typical jury pool would warp the results. In short, Mechanical Turk, like other survey tools, is not perfect, but in experiments such as this, can be “a reliable source of experimental data in judgment and decision-making.”93

Four hundred participants agreed to take the survey in exchange for a nominal fee approximating the (pro-rated) federal minimum wage.94 Respondents were routed to an automated, online survey interface that randomly presented one of four case scenarios.95 The scenarios were identical except with respect to whether the defendant testified and, if so, the type of impeachment presented. Respondents who encountered impeachment also received a jury instruction that the prior crime could be considered “only with respect to [the defendant’s] credibility.”96  In the scenarios where the defendant testified, his testimony added no new information. The defendant’s testimony was summarily described as being “consistent with that of” a defense alibi witness whose testimony (that he and the defendant were watching a baseball game at the time of the crime) appeared in all four scenarios.

The four scenarios presented to respondents were as follows:


Table 2

Screen Shot 2018 02 05 at 4.44.06 PM


Respondents answered a series of “reading check” questions to ensure that they were, in fact, reading and understanding their respective scenarios.97 At the conclusion of the presentation, the survey instructed respondents that “you should only find the defendant guilty if you believe the evidence establishes his guilt beyond a reasonable doubt.” Participants were then asked how they would vote “as a juror in a criminal case: not guilty or guilty.”

B.     Results

Jurors voted to convict in 73% of the cases. The following table reveals the breakdown of guilty votes by scenario, ordered by descending conviction percentage.


Table 3

Screen Shot 2018 02 05 at 4.45.07 PM


Approximately 100 distinct mock jurors voted in each scenario. As the above table shows, the conviction rate was highest for Scenario 4 where the defendant testified and was impeached with a prior robbery. The lowest conviction rate occurred in Scenario 2 where the defendant testified and was not impeached with any prior crimes. The other two scenarioswhere the defendant did not testify, or testified and was impeached with a criminal fraud convictionreturned similar conviction rates.

The results are consistent with the social science literature presented in Part II. The jury’s learning of prior convictions negatively impacted the defendant’s chances for acquittal. The jurors convicted most often (82%) when they learned that the defendant had a prior robbery conviction. The conviction rate was also elevated (73%) over the no record condition (62%) when the defendant was impeached with a “criminal fraud” conviction. Overall, jurors voted to convict 78% of the time in the two prior conviction conditions, but only 62% of the time when the same testifying defendant was not impeached with any prior crimes. This variance achieves statistical significance.98

These findings also support the critique developed in Part II that prior conviction impeachment does not operate in the manner that the law contemplates. If prior conviction impeachment speaks only to the defendant’s character for truthfulness, crimes of dishonesty would be most damaging. Here, the fraud conviction should have been most damaging since it is a crime that, unlike robbery, speaks directly to truthful character. “Criminal fraud” is one of a handful of offenses specifically referenced in the legislative history to Federal Rule of Evidence 609 and its advisory committee notes as directly “bearing on the accused’s propensity to testify truthfully.”99 Instead, in this experiment, the robbery convictionan offense that was likely seen by lay participants as similar to the charged jewelry store burglaryhad a larger negative impact. This suggests (consistent with the prior research discussed in Part II.A) that jurors indulged a forbidden, criminal propensity inference.

Finally, and most interestingly, the results support the “parallel penalty” hypothesis developed in the preceding sections. Consistent with Shaffer’s research described in Part II.B, the results reveal a clear “silence penalty.” Respondents convicted 76% of the defendants who remained silent, but only 62% of equally situated defendants who testified (but added no facts).100 Again the difference is statistically significant.101 Further, the weight of the silence penalty appears to be roughly equivalent to the “prior offender” penalty. Combining the two prior conviction conditions (Scenarios 3 and 4) results in a 78% conviction percentage. This is almost identical to the conviction percentage for Scenario 2 where the defendant did not testify and was not impeached with any past crime76%. These findings support a hypothesis that both remaining silent at trial and the admission of a defendant’s prior convictions substantially decrease the prospects for acquittal. And the operative “silence” and “prior offender” penalties appear to harm defendants to a similar degree.

IV.     Data from Real Trials

Mock juror studies only tell us so much. No matter how clear the patterns that emerge in juror simulations, doubts will persist as to whether these simulations accurately reflect juror decision making in actual trials. A clear picture of how real jurors react to defendant testimony requires data from real trials.

The dataset with the highest potential to unlock the mysteries of defendant testimony comes from the National Center for State Courts (“NCSC”).102 In 2000 and 2001, the NCSC surveyed attorneys, judges and jurors participating in felony cases at four sites: Los Angeles, Phoenix, the District of Columbia, and the Bronx.103 The surveys solicited a broad range of case-specific information as part of an inquiry into the frequency and causes of hung juries. The information included whether the defendant testified, admission of prior convictions, and case outcomes.104 The surveys also asked jurors and judges to assess the strength of the prosecution’s evidence.105

Analyzing data from real cases raises a number of challenges. The challenges range from specific quibbles with the NCSC data set to big picture questions about the daunting variability of criminal trials. With respect to specific quibbles, the NCSC data on “criminal record” is not ideal. It captures “criminal record” broadly but imprecisely, lumping together all convictions and arrests. The survey asked: “During the trial, did the jury become aware of the defendant’s criminal history (if any)?”106 Respondents answered: “Yes,” “No” and “Not applicable (no known arrests/convictions).”107 Important information that is not captured by this question is whether an arrest led to conviction and whether the conviction was for a petty offense (e.g., misdemeanor) that typically cannot be used as impeachment.108 More subtle ambiguity comes from data coding choices. If a jury acquitted a defendant of murder, but convicted of the lesser (but perhaps uncontested) offense of unlicensed gun possession, is that a conviction or an acquittal?109 What about hung juries? To minimize empirical objections and unintended variation, and facilitate comparison to other articles that analyze the NCSC data, this Article draws primarily on the published analysis of the NCSC data by Eisenberg and Hans. These prominent legal empiricists have already extracted the most pertinent data regarding defendant testimony from the NCSC data set for their article on the impact of prior convictions.110

Finally, no study of real cases can capture all the variables that influenced a jury verdict. Instead, the hope is that randomly pooled variation in a sufficient number of cases will permit the detection of broad patterns that reflect underlying realities. Definitive conclusions from the NCSC data alone will be impossible. Like the perceptions of judges and attorneys who work in the criminal justice system, and the mock juror simulations discussed above, the NCSC data simply offer another place to look in an effort to understand how jurors react to defendant testimony.

With the above caveats, the NCSC data support many of the propositions already discussed. Only about half of defendants testified.111 Defendants with prior records were less likely to testify.112 And those who did testify were more likely to have their criminal records presented to the jury.113 Defendants who were members of racial minority groups testified less frequently, but this finding was only “marginally statistically significant” and can be attributed to “different rates of prior criminal records”; the few female defendants testified more frequently than male defendants.114 Interestingly, the reported evidentiary strength of the cases did not differ significantly between those cases where the defendants did and did not testify.115

A.     The NCSC “Paradox”

The NCSC data present what other commentators have described as a “seeming paradox” and “puzzling” result.116 The paradox begins with the not-so-surprising finding that juries convicted at a much higher rate if the defendant had a criminal record (76%) than if the defendant had no record (56%).117 The perplexing aspect of the data is that juries convicted defendants with prior crimes at approximately the same rate whether their convictions were revealed to the jury or not.118 Thus, the NCSC data’s surprise for researchers was: “It is whether or not the defendant has a criminal recordnot whether the jury learns about itthat has the greatest influence on the acquittal/conviction decision.”119

While others have flagged this curious finding,120 no one has produced a compelling explanation for it. Larry Laudan and Ronald Allen, who make this finding the centerpiece of a provocative article, grapple with the question most directly.121 They conclude that the NCSC data show that the purportedly devastating impact of the admission of prior crimes evidence is a myth that has bedazzled attorneys, policymakers, and scholars for decades.122 “Admitting evidence of prior crimes,” they argue, “apparently leads to few additional convictions.”123 They surmise that jurors “are generally able to infer who has priors” regardless of whether the prior crimes are made known to them in the evidence.124

For ease of reference, I will call Laudan and Allen’s hypothesis the “jury sophistication” hypothesis. In light of their hypothesis, Laudan and Allen disparage efforts through evidence rules and litigation strategy to keep prior convictions out of evidence as “self-defeating,” and label academic criticism of the admission of prior crimes “unnecessary hyperbole.”125 If Laudan and Allen’s interpretation is correct, volumes of scholarly articles and judicial opinions, along with a number of evidence ruleswhat Laudan and Allen label “the criminal justice system’s fetish about excluding prior crimes”126must be rewritten; and attorneys have for decades been pointlessly fighting over, and tailoring trial strategies to, the admission of prior convictions.127

As explained below, Laudan and Allen’s ground shaking juror-sophistication hypothesis is not the best explanation for the NCSC data. Their theory is initially undermined by the experimental data analyzed in Part II.128 This experimental evidence reveals that when mock jurors are told that the defendant has a prior record, they convict more readily. This finding is difficult to reconcile with Laudan and Allen’s claim that real jurors are indifferent to being told about prior crimes evidence. As discussed earlier, Laudan and Allen’s response to this challengethat the mock juror evidence is “all over the map”is incorrect.129 Further, Laudan and Allen’s hypothesis contradicts the near-universal views of practitioners, judges, academics and policymakers that informing jurors of prior convictions powerfully impacts defendants’ prospects.130 Everyone else may be wrong, of course, but the existence of such a robust countervailing consensus raises a red flag.

There are a number of other possible explanations for why defendants with a criminal record fare worse at trial regardless of juror awareness of the prior record. This Article posits the “parallel penalty” hypothesis as the best explanation,131 but others come to mind as well. For example, a criminal record correlates with lower income.132 Lower income means defendants will be unlikely to afford a (perhaps superior) private attorney. But as paid attorneys appear in less than 18% of criminal cases,133 this “free-attorney hypothesis” can only be a partial explanation. Further, elite public defender offices in large cities like those studied in the NCSC survey (e.g., the District of Columbia’s Public Defender Service or the Bronx Defenders) may actually outperform retained counsel.134

Another explanation could be the relationship between criminal history and an increased likelihood of pretrial detention.135 Defendants held pending trial generally fare worse than those released.136 But again, this “pretrial detention hypothesis” seems at best a partial explanation. As explained below, neither these explanations nor the one posited by Lauden and Allen convincingly account for the NCSC data.

The “parallel penalty” hypothesis presented in this Article represents the best explanation for the “puzzling” NCSC data. Under this hypothesis, the equivalent conviction rates for defendants with criminal records do not result from jurors detecting hidden convictions, or variations in lawyer quality or pretrial release. Rather, unrevealed convictions inflict harm indirectly by causing defendants to remain silent at trial, leading to a “silence penalty.” If this alternative manifestation of a criminal record harms defendants roughly as much as the introduction of prior convictions, the NCSC data fall neatly into place.

The “parallel penalty” hypothesis is not as jarring to the conventional wisdom as Laudan and Allen’s juror sophistication hypothesis, but it nonetheless demands a deeper appreciation of the importance of trial silence to American jurors. The balance of this Part analyzes specific slices of the NCSC data to determine whether the “parallel penalty” hypothesis or the alternatives better explain the data. (Each subsection below includes an introductory table to highlight the NCSC data slice to be discussed.)

B.     Defendants Without Priors: Testifying v. Non-Testifying


Table 4

Screen Shot 2018 02 05 at 10.56.49 PM 


We begin with the data for defendants without prior convictions. This comparison provides the most straightforward evidence that a silence penalty and not jury sophistication about hidden criminal records, or other factors that correlate with criminal records, explains the “seeming paradox.” Table 4 shows that when defendants without prior convictions testified, they were convicted 41% of the timea startlingly low percentage considering the high overall conviction rates.138 When defendants without prior convictions did not testify, their conviction percentage skyrocketed to 70%.139

Thus, for defendants without prior convictions, testifying coincided with an almost doubling of the chances of acquittal. As this variation emerges among defendants without a criminal record, the “pretrial detention” and “free-attorney” hypotheses cannot explain it. Of course, case-specific facts undoubtedly play a key role and many variables are not captured in the NCSC study. Still, the data presented in Table 4 constitute circumstantial evidence that a “silence penalty” exists in the American trial system. Like the mock jurors discussed in Part III, jurors in real cases appear to be more willing to convict defendants who remain silent at trial.

C.     Defendants with Priors: Testifying v. Non-Testifying


Table 5

 Screen Shot 2018 02 05 at 11.00.15 PM


The analysis for defendants with prior convictions is more complex but again supports the parallel penalty hypothesis. Just as the silence penalty hurts defendants without prior records who remain silent, it should harm defendants with prior convictions who decline to testify. But the silence penalty will not be as readily identified for these defendants. Testifying strongly correlates with the introduction of prior convictions.141 This means defendants with prior convictions can typically only avoid the silence penalty by (testifying and) suffering a prior offender penalty instead. Thus, the powerful benefit from testifying that appears in Table 4 should not reappear in Table 5. Any benefit prior offender defendants gain by avoiding the “silence penalty” will be offset by a “prior offender penalty.”

Rather than reflecting the benefits of testifying, data broadly comparing prior offenders who testify to those who do not should reflect the relative powers of the silence and prior offender penalties. (Recall that evidentiary strength did not differ significantly between cases with testifying and non-testifying defendants.)142 A possibility suggested by the juror simulation discussed in Part III is that the two penalties are roughly equivalent.143 If so, we would expect that the conviction percentages for defendants with priors will remain about the same whether they testify or remain silent. This fits the NCSC data. Table 4 reflects that defendants with priors who testified were convicted about 77% of the time.144 Defendants with priors who declined to testify were convicted about 72% of the time.145

Concededly, the rough equivalence of the conviction rates between prior offenders who testify and those who remain silent also supports Laudan and Allen’s juror sophistication hypotheses. It could be, as they argue, that jurors are somehow correctly divining that these silent defendants have prior convictions and penalizing accordingly. Similarly, the results could also support the free-attorney and pretrial detention hypotheses. Under these hypotheses, no matter what tactic this subset of defendants choose, their disproportionate inability to hire private counsel and secure pretrial release dominates acquittal prospects. Note, however, that contrary to the empirical evidence presented so far, all of these competing hypothesis only fit the NCSC data if there is no silence penalty. Under these alternative theories, silence has no negative effect. Table 5 reflects that defendants with prior crimes who testified were convicted at least as often as those who remained silent. Thus, while each of the posited theories can claim some support from this slice of the NCSC data (Table 5), the parallel penalty hypothesis (i.e., a rough equivalence of the silence and prior offender penalties) provides a stronger explanation for the equivalent conviction rates. Juries are quick to convict defendants with prior records if they do not testify or if they testify and are impeached with their prior crimes, and the respective penalties in either scenario are roughly the same.

The results hold if we look at a purer (if smaller) comparison of the prior offender and silence penalties by comparing defendants with a prior record who do not testify—and whose convictions are not disclosed—with defendants who do testify and whose convictions are disclosed. Juries convicted defendants who successfully hid their prior record by remaining silent in 71.4% of cases.146 Juries convicted defendants who testified and had their record disclosed in 77.8% of cases.147 Interestingly, the NCSC data here closely tracks the trial simulation described in Part III, where mock jurors convicted non-testifying defendants 76% of the time and convicted impeached defendants 78% of the time.148

D.     Non-Testifying Defendants: Priors v. No Priors


Table 6

Screen Shot 2018 02 05 at 11.01.37 PM 


A silence penalty should harm defendants equally regardless of whether they have a criminal record. If, however, jurors are adept at detecting undisclosed criminal records, as Laudan and Allen propose, silent defendants with criminal records will fare worse than those without. Similarly, if income effects of a prior record or likelihood of pretrial release drive the data, we should see a broad criminal-record-related variance between silent defendants. Instead, the NCSC data summarized in Table 6 strongly supports the parallel-penalty hypothesis.

Table 6 shows that the conviction rate for defendants with a prior record who declined to testify (72%)150 is almost identical to the conviction rate for defendants without a prior record who declined to testify (70%).151 This is what we would expect if a silence penalty influences the outcome. As the jury generally remains ignorant of any prior convictions when the defendant declines to testify, both sets of non-testifying defendants should suffer the same single penalty (the silence penalty). The jury never hears of the non-testifying prior offenders’ criminal records and so the conviction rates of the two groups of non-testifying defendants should be roughly equal. The NCSC data (Table 6) fits, revealing an almost identical likelihood of conviction. To jurors, these non-testifying defendants are broadly indistinguishable, and they suffer similar outcomes. If jurors were somehow detecting prior convictions, as Laudan and Allen claim, we would expect a different result. Similarly, if defendants with prior records were suffering worse outcomes across the board due to subpar attorneys or pretrial detention, the data should look different.

E.     A Touch of Discordant Data

The parallel penalty hypothesis appears to fit the NCSC data better than the alternatives, but it is not a perfect fit. Assuming the silence penalty is held constant, the parallel penalty theory predicts that the “prior offender penalty” would only appear when the jury is aware of the defendant’s prior record. Yet in one slice of NCSC data, jury awareness of a prior record does not have the anticipated effect on trial outcomes. Among testifying defendants with prior convictions, the conviction rate was only slightly lower (76.6% vs. 77.8%) when their convictions remained unknown to the jury.152 All things being equal, the parallel penalty hypothesis predicts that the conviction rate for prior-offender defendants who testify and are not impeached would be relatively lower since those defendants should not suffer the “prior offender penalty.” This is the one cell of the NCSC data that the parallel penalty hypothesis struggles to explain.153 As we are dealing with real cases, the most likely explanation is that case-specific factors in this relatively modest sample of unusual cases (where defendants testify despite a criminal record, but are nevertheless not impeached with that record in cross-examination) skew the expected percentages.

Indeed, this curious finding drew the attention of Eisenberg and Hans, who scoured the NCSC data for evidence of the impact of the disclosure of prior convictions.154 They determined that the relatively similar conviction rates in this data slice obscured important variance in evidence strength. With evidentiary strength factored in, the prior offender penalty resurfaces as predicted: “[J]ury knowledge of prior criminal history is significantly associated with conviction in weak [prosecution] cases and not significantly associated with conviction in strong cases.”155

Eisenberg and Hans’s further exploration of this data helps the case for the parallel penalty theory and also highlights an important point. The silence and prior offender penalties push juries toward convictions, but they are by no means dispositive. Facts matter most.156 The smaller the number of cases, the higher the risk that unmeasured distinctions across case categories will overwhelm any prior offender or silence penalty.

In sum, while one narrow slice of the NCSC data does not support the parallel penalty hypothesis, the variance can be explained. Deeper statistical analysis of this slice by Eisenberg and Hans hints at variation in case characteristics that skew the observed percentages. Consequently, the parallel penalty hypothesis remains an attractive explanation of the NCSC data overall.

The parallel penalty theory fits the NCSC data fairly wellno small feat given the daunting variability of criminal trials. The parallel penalty hypothesis is also consistent with the extant experimental evidence summarized in Parts II and III, and refines (rather than defies) the decades-old collective wisdom of judges, practitioners, academics and American evidence rules.157 Although there is undoubtedly more to this complex story, the parallel penalty hypothesis fits the empirical data better than the alternatives and provides a promising theoretical framework for how American jurors react to defendant testimony and its absence.

V.     Implications

The previous Parts set out the theoretical and empirical case for the parallel penalty hypothesis. This Part discusses the implications of these findings. It begins with tactical implications primarily of interest to practitioners and judges, and then moves to broader implications of chief concern to policymakers and legal scholars.

A.     Defendants Should Testify More Often

The empirical data presented in Parts II, III, and IV support two aspects of the conventional wisdom regarding trial tactics. The data reinforce the widely-held view that juries rely on impermissible propensity reasoning to convict when informed that the defendant has previously broken the law (the “prior offender penalty”). The data also suggest that juries punish defendants for remaining silent at trial with a “silence penalty.” Finally, and most strikingly, the data suggest something not captured in the conventional wisdomthat the silence penalty is roughly as damaging as the prior offender penalty.

The surprising power of the silence penalty should give pause to the many defendants without a prior record who demand a trial but then decline to take the witness stand (nearly 40% of trial defendants in the NCSC data).158 By testifying, these defendants could avoid both the silence penalty and the prior offender penalty. Declining to testify, by contrast, puts them in the same position as a defendant with prior convictions. This is a major blow to acquittal prospects and one that (tactically speaking) should be avoided if at all possible.

Of course, case and defendant-specific factors can overwhelm general trends. Defense counsel may believe that juries will react so negatively to their client’s appearance on the witness stand that any silence penalty pales in comparison. Noted defense advocate and law professor Barbara Babcock recognizes all the familiar reasons a defendant without prior convictions might not take the stand, including “that he has no defense . . . . or maybe he is unattractive, even scary, or slow and obtuse so that he could hurt, rather than help himself as a witness.”159 But drawing on her experience as a public defender, Babcock emphasizes that “few defendants who fail to testify win their cases” and notes the critical role for defense lawyers in this equation.160 “[Defendants] who are well-defended rehearse their testimony; the better defended they are the more they rehearse.”161 Babcock further urges counsel to place their client’s testimony in context when necessary. For example, counsel can argue to the jury that the defendant never “deviate[d] from her basic testimony in this case: she is not guilty” despite the “unequal contest” between “my client, with her sixth grade education, who has never before spoken to an audience in public” and the “government prosecutor with her decades of learning, and years of experience.”162 “[T]his argument alone, can make it worth the defendant’s taking the stand, even if his testimony is weak in substance and halting in style.”163 The empirical data marshalled above support Babcock’s contention. Absent “strong affirmative” case or defendant-specific reasons to avoid the witness stand, defendants without prior convictions should testify to avoid a powerful silence penalty.164

Defendants with prior convictions should similarly consider the power of the silence penalty before reflexively embracing the widely-accepted tactic of refusing to testify to hide one’s criminal record. Failing to testify may reduce the impact of a prior conviction, but it only does so by exposing the defendant to a damaging silence penalty. In cases where the defendant’s testimony would add salient facts to the jury’s deliberations or where the prior conviction is not likely to generate negative propensity reasoning (e.g., a different type of crime than the charged crime), testifying may well be the best tactic. Even when impeachment results, the data suggest that, at worst, defendants end up in about the same position they would have occupied if they declined to testify.165 Rather than reflexively silencing defendants, counsel should begin with a presumption that their clients will benefit if they take the witness stand.

Ethical constraints also operate. A lawyer cannot sponsor witness testimony that she knows to be untrue.166 That said, “[t]he conventional defense view also holds that a lawyer ‘knows’ only if the client has told him so categorically.”167

Again we must pause to consider a discordant sentiment by an expert commentator. As already noted, experiments by one of the leading researchers on defendant testimony, psychologist David Shaffer, highlight the damage done to defendants when they do not testify. Shaffer is, nevertheless, “hesitant to advise against” remaining silent at trial because of the results of a study by sociologist Martha Myers.168 Shaffer’s reluctance is worth exploring because, as discussed below, it appears to be based on a misreading of Myers’s data.

Shaffer claims that Myers’s research on Indianapolis trials in the mid-1970s determined that “82 percent of the defendants in her sample did not testify,” and that she “found that defendants who testified were more likely to be convicted than those who did not.”169 As this stunning 82% number suggests, Shaffer misreads Myers’s study. (Recall the comparable figure for non-testifying defendants from the NCSC study was 50%; and the percentage from the famous but dated 1950s Kalven and Zeisel study is 18%).170 The variable in Myers’s study that Shaffer is interpreting as defendant testimony is labeled, “X3 Testimony of Defendant and/or Accomplices.”171 Right away a problem is apparent: A variable intended to measure whether the defendant presented exculpatory testimony should not also include testimony from “[a]ccomplices.” Myers did find that the X3 variable and two other measures correlate strongly with convictions: “[J]uries were more likely to convict if: the defendant or accomplice made a statement about his involvement in the crime or lack thereof (X3); a weapon was recovered (X6); and a large number of witnesses was specified in the indictment or information (X7).”172 But as this excerpt indicates, the X3 variable primarily captures not defendant testimony, but prosecution evidence, such as admissible pretrial statements of the defendant and accomplices to police.173 Myers describes the variable as a defendant (or accomplice) making a “statement,” codes it as “[n]one” or “[o]ne or more statements,” and includes it among other pro-prosecution evidentiary variables such as recovery of a weapon.174 Further, Myers fails to editorialize on the finding at any point.175 One would expect Myers to highlight and discuss an earth-shattering finding that defendant testimony typically backfires, strongly increasing the likelihood of conviction. It is not surprising, however, that Myers did not discuss the finding if she understood it to be mundanethat pretrial statements to police “about [the defendant’s] involvement in the crime” increase the likelihood of conviction.176 In sum, the finding Shaffer interprets as showing that defendants who testify are convicted more often says something very different. Shaffer and others who shy away from recommending that defendants testify based on Myers’s results are misreading her study.

B.     The Ineffectiveness of Legal Doctrine Governing Defendant Testimony

The impacts of the silence and prior offender penalties described in the preceding sections reach far beyond trial tactics. As discussed in the next subparts, the empirical data summarized in Parts II through IV constitute a powerful indictment of the current legal framework.

One powerful implication of the existence of substantial silence and prior offender penalties is that legal doctrine designed to eliminate these penalties is ineffective. Two critical legal rules are in play: (1) jurors may not draw an adverse inference from the defendant’s refusal to testify;177 (2) when defendants do testify and are impeached with prior convictions, juries cannot consider the evidence as showing a criminal propensity.178 The empirical data analyzed above indicate that the legal system fails to effectively enforce both rules. Jurors are drawing adverse inferences from the defendant’s failure to testifythe “silence penalty.” And jurors improperly consider prior convictions offered solely as credibility impeachment as evidence of criminal propensitiesthe “prior offender penalty.” The penalties turn up consistently in experimental studies with mock jurors (see Parts II and III). They also appear to be powerful enough that they can be observed across a broad run of cases in data from real trials (see Part IV). To the extent the criminal justice system cares about enforcing its own rules, the data presented in this Article generate cause for concern. In this critical context, it appears that Justice Robert Jackson’s cynical critique rings true: “The naive assumption that prejudicial effects can be overcome by instructions to the jury all practicing lawyers know to be unmitigated fiction.”179

C.     Distortions of Jury Factfinding

The data marshalled above also suggest that jurors indulge rough and often incorrect proxies for guilt that may interfere with their search for truth. The damage begins with the startling number of defendants who demand trial but decline to testify, depriving factfinders of a witness “who above all others may be in a position to meet the prosecution’s case.”180 When defendants do not testify, juries not only lose a testimonial resource, they are also tempted to draw an inference of guilt from that silence. Although prohibited by law, the inference may be warranted in certain circumstancesand thus not detrimental to the search for truth.181 But in many cases defendants decline to testify to avoid prior conviction impeachment. In that frequent scenario, jurors misread the silence signal; the defendant’s silence indeed suggests guilt, but it is guilt of a prior offense, not the offense for which he is on trial. Finally, when defendants do testify and are impeached with prior convictions, juries appear to slip into a form of propensity reasoning that causes them to overlook evidentiary weaknesses in the prosecution’s case.182

The dangers described above come into sharp relief in studies of convicted defendants who were later exonerated by DNA evidence. A comprehensive survey by John Blume found that 39% of later-exonerated defendants did not testify in their own trials; another study puts the number at 47%.183 These numbers are moderately lower than the generic 50% non-testifying rate in the NCSC study,184 suggesting that factual guilt is a factor, but not a powerful determinant of a defendant’s decision to testify. The factor that does seem determinative is a prior record. Blume reports that 91% of the later-exonerated defendants who did not testify “had prior convictions that potentially could have been used for impeachment purposes.”185 The fear of impeachment is well founded. Blume reports that “[i]n every single case in which a defendant with a prior record testified, the trial court permitted the prosecution to impeach the defendant with his or her prior convictions.”186

The stories of the exonerated defendants in Blume’s study parallel the themes presented in this Article. Legal rules inflate the number of silent defendants, and then distort jury factfinding by ensuring that no matter what course impeachable defendants take (testify or remain silent), juries receive a strong, if legally proscribed, pro-conviction signal. It is no surprise then that Blume reaches a conclusion that sadly resonates with the findings described above: “[T]he current rules of evidence contribute to wrongful convictions.”187

D.     Incentivizing Guilty Pleas and Exacerbating Discriminatory Impacts

Two of the most troubling aspects of the modern criminal justice system are the sky-high prevalence of guilty pleas188 and the disproportionate impact of criminal punishment upon racial minorities.189 The historical data is unfortunately too sporadic to support definitive conclusions, but it hints that the parallel penalty dynamic plays a provocative and largely hidden role in each of these phenomena.

The historical link between the parallel penalties and guilty pleas is most provocative. The parallel penalties highlighted in this Article first became possible in the late 1800s as states enacted laws permitting defendant testimony. Legal historian George Fisher posits an early connection between defendant testimony rights and guilty pleas, noting that the “dramatic conversion to a plea bargaining regime” in Middlesex County, Massachusetts “started about a decade after defendants first began to take the witness stand.”190 Data from other state jurisdictions similarly support a trend of increased plea bargaining starting “in the latter part of the nineteenth century”the same time defendant testimony laws emerged.191

But the parallel penalty dilemma may not have surfaced acutely with the onset of defendant testimony laws if the defendants standing trial were less likely to have a criminal record. In a time when jury trials more commonly concerned offenders with no official record, it would make senseas Judge Ames predictedfor defendants to routinely take the witness stand. A defendant with a clean record can deftly sidestep the parallel penalty dilemma by testifying.

As the population of potential offenders with criminal records increases, the dynamic changes. When more and more defendants have a criminal record, fewer can avoid the parallel penalty dilemma by testifying. Foregoing trial altogether through a guilty plea becomes the most rational tactic.

Again, the historical evidence is by no means conclusive, but the apparently increasing prevalence of defendants with prior records, and decreasing prevalence of defendant testimony suggests the promise of future inquiries. The analysis begins with a comparison of the NCSC data to the other broad survey of American jury decision making, Kalven and Zeisel’s fabled empirical examination of American trials. In 1954–1955, Kalven and Zeisel surveyed criminal trials in a variety of jurisdictions and found that 47% of trial defendants had a criminal record and 82% testified.192 By 2000–2001, the NCSC study reports that 76% of defendants had a criminal record and only 50% of defendants testified.193 The samples do not match up precisely, of course, but the data hint that the pool of defendants who can (and do) avoid the parallel penalty dynamic by testifying is shrinking. The Bureau of Justice Statistics similarly reports that the percentage of arrestees with a criminal record is increasing over time.194 This makes intuitive sense in light of the ballooning population of former felons. The pool of potential defendants with a criminal record is deeper than it has been at any other point in history. In 1974, for example, 1.8 million American adults had previously served time in state or federal prison.195 By 2001, that number was 5.6 million.196

The parallel penalty-based pressure to plead guilty becomes particularly acute as more and more defendants face trial burdened with a prior record. These defendants can no longer cleanly avoid the silence penalty by testifying. And indeed guilty plea rates appear to have shot up once more in recent decades alongside the growing population of prior offenders. As a general matter, the proportion of guilty pleas increased in the past 35 yearsthe same time frame as the mass incarceration explosion.197 Causation is likely impossible to show as the variables are overlapping and interrelated. But the historical data we have fit an unsettling narrative. The parallel penalty dilemma makes trial unattractive for defendants with a criminal record, pushing these defendants to forego trial and plead guilty. And as defendants with criminal records increasingly become the norm rather than the exception, the parallel penalty dilemma inevitably contributes to a steady increase in guilty pleas and a corresponding decrease in trials.

There is another important piece to this story. The increasing prevalence of criminal records is not equally distributed among defendants.198 In the NCSC study, 71% of minority defendants had criminal records compared with 55% of white defendants.199 This disparity is consistent with general population trends. While in 2010 the percentage of the non-African American adult population with a prior felony conviction reached a high of 6%, that number was 25% for adult African Americans.200 This means that the pernicious effects of the parallel penalty dilemma disproportionately impact African-American defendants. The cycle is self-perpetuating. Every new conviction leads to a decreased likelihood of success in a subsequent trial and a stronger incentive to plead guilty.

VI.     Conclusion

The Supreme Court and commentators’ celebration of the defendant’s right to testify is premature. Empirical evidence on the state of defendant testimony paints a picture more deserving of lamentation than glee. The landscape is so gloomy that it supports the 19th-century critics who counterintuitively proclaimed that defendants are better off when prohibited from testifying at trial.201 The right to testify may, in fact, play a supporting role in some of the most troubling aspects of modern American criminal justice, including the proliferation of guilty pleas and the disproportionate imprisonment of racial minorities.

The root of the problem is that the American criminal justice system has strayed from the presumption of innocence and the fabled insistence that “a defendant starts his life afresh when he stands before a jury, a prisoner at the bar.”202 The data summarized above suggest that juries consider trial silence to be incriminating, and draw legally improper criminal character inferences from prior convictions admitted as “impeachment.” As a result, the only defendants who truly enjoy a presumption of innocence at trial are the relatively few defendants without admissible prior crimes who elect to testify. By contrast, the bulk of criminal defendants, particularly those with a prior record, face a choice between two damaging options, a choice they can only avoid by forgoing trial and pleading guilty. In essence, the venerated right to testify may primarily operate to push prior offenders quickly through the justice system by limiting the attractiveness of trial and thereby incentivizing guilty pleas.

There are no easy solutions. A criminal justice system that relies on lay jurors will inevitably impose a “silence penalty” on defendants who refuse to testify.203 So long as American evidentiary rules simultaneously impose a “prior offender” penalty on defendants who do take the witness stand, these parallel penalties will predictably tilt the criminal justice scales toward conviction.

The critical point is that, for all its accolades, the modern right to testify is not just a lofty principle. It is part of a complex web of legal rules that, while vulnerable to criticism, have proven stubbornly resistant to change.204 Under the sway of those rules, the right to testify appears to be broadly harming criminal defendants and undermining the criminal justice system itself. In that light, the Supreme Court’s claim that there could be “no rational justification for prohibiting the sworn testimony of the accused”205 seems profoundly out of touch. There is certainly a rational basis for prohibiting sworn defendant testimony. It can be found in the data that reflect the stark realities of our flawed criminal justice system.

  1. [1]. See McGautha v. California, 402 U.S. 183, 214 (1971) (“[A]t the time of framing of the Fifth Amendment and for many years thereafter the accused in criminal cases was not allowed to testify in his own behalf ... .”).

  2. [2]. Rock v. Arkansas, 483 U.S. 44, 49 (1987); cf. State v. McKenzie, 303 A.2d 406, 413 (Md. Ct. Spec. App. 1973) (“The right to testify is not constitutional, but statutory.”).

  3. [3]. Rock, 483 U.S. at 50 (quoting Ferguson v. Georgia, 365 U.S. 570, 582 (1961)).

  4. [4]. See United States v. Dunnigan, 507 U.S. 87, 96 (1993) (explaining in a 9-0 decision that “[t]he right to testify on one’s own behalf in a criminal proceeding is ... a right implicit in the Constitution”); Riggins v. Nevada, 504 U.S. 127, 144 (1992) (Kennedy, J., concurring) (“It is well established that the defendant has the right to testify on his own behalf, a right we have found essential to our adversary system.”); Ferguson, 365 U.S. at 581–83 (collecting commentary from judges and academics); Daniel J. Capra & Joseph Tartakovsky, Why Strickland is the Wrong Test for Violations of the Right to Testify, 70 Wash. & Lee L. Rev. 95, 147 (2013) (arguing for more robust protections of the right to testify, saying that “[w]e should simply speak of an independent ‘right to testify,’ an undisputed guarantee ‘implicit’ in the Due Process, Self-Incrimination, and Compulsory Process Clauses”); Richard Friedman, Character Impeachment Evidence: Psycho-Bayesian (!?) Analysis and A Proposed Overhaul, 38 UCLA L. Rev. 637, 666 (1991) (“[T]he right to testify in one’s own defense, although of far more recent vintage than some other rights of a criminal defendant, must now be considered as one of the most fundamental in our jurisprudence.” (footnote omitted)); Raymond J. McKoski, Prospective Perjury by a Criminal Defendant: It’s All About the Lawyer, 44 Ariz. St. L.J. 1575, 1642–43 (2012) (describing the right to testify as a “cherished constitutional right[]” that is “engrained in the fabric of the legal system”).

  5. [5]. Testimony of Persons Accused of Crime, 1 Am. L. Rev. 443, 444 (1867). For a comprehensive description of Judge Ames’s (and others’) objections and a discussion of the attribution of the quoted document to Ames, see George Fisher, Plea Bargaining’s Triumph 104 & n.31 (2003).

  6. [6]. Testimony of Persons Accused of Crime, supra note 5, at 444, 446.

  7. [7]. See Theodore Eisenberg & Valerie P. Hans, Taking a Stand on Taking the Stand: The Effect of a Prior Criminal Record on the Decision to Testify and on Trial Outcomes, 94 Cornell L. Rev. 1353, 1373 tbl.2 (2009) (summarizing findings from the broad study of felony trials); Gordon Van Kessel, Adversary Excesses in the American Criminal Trial, 67 Notre Dame L. Rev. 403, 483 (1992) (“[I]nquiries with trial lawyers and judges lead me to believe that the extent of defendant refusals to testify is considerable—from one-third to well over one-half in some jurisdictions.”).

  8. [8]. See infra note 183 and accompanying text (discussing DNA-exoneration studies that reveal a high percentage of innocent defendants who declined to testify).

  9. [9]. See John H. Langbein, The Criminal Trial Before the Lawyers, 45 U. Chi. L. Rev. 263, 283–84 (1978) (“It is one of the great peculiarities of modern Anglo-American procedure, on which Continental observers often remark, that we have so largely eliminated the accused as a testimonial resource.”); Anna Roberts, Impeachment by Unreliable Conviction, 55 B.C. L. Rev. 563, 575 (2014) (summarizing the negative consequences of silent defendants); Christopher Slobogin, Lessons from Inquisitorialism, 87 S. Cal. L. Rev. 699, 707 (2014) (arguing that the American adversary system can “contribute to inaccuracy during trial” by “prevent[ing] the factfinder from hearing from the defendant, despite the fact that the defendant is probably the single most important source of information about events relating to the offense”); Jay Sterling Silver, Truth, Justice, and the American Way: The Case Against the Client Perjury Rules, 47 Vand. L. Rev. 339, 357 (1994) (arguing that even defendant testimony that includes lies provides useful information to juries); Van Kessel, supra note 7, at 482 (criticizing legal rules that encourage defendants to remain silent, resulting in the “loss of evidence from the most important witness in the case”).

  10. [10]. See infra Part II.A.

  11. [11]. See Brian A. Reaves, Bureau of Justice Statistics, Felony Defendants in Large Urban Counties, 2009 - Statistical Tables 8 (reporting that 75% of suspects charged with a felony had a prior arrest, 60% had a prior felony arrest, 60% “had at least one prior conviction,” and 43% “had at least one prior felony conviction”). In the NCSC study of felony trials in four jurisdictions, 76% of the defendants standing trial had some kind of criminal record. Eisenberg & Hans, supra note 7, at 1371 tbl.1. In another study of 201 Indianapolis jury trials between 1974 and 1976, that author reported that “most defendants had prior convictions.” Martha A. Myers, Rule Departures and Making Law: Juries and Their Verdicts, 13 L. & Soc’y Rev. 781, 790 (1979). Myers does not report the percentage, but instead provides an average of 2.7 prior convictions per defendant. Id. at 786 tbl. 1. Kalven and Zeisel reported that in 47% of the trials in their sample from 1954–1955, the defendant had a prior record. Harry Kalven, Jr. & Hans Zeisel, The American Jury 33 n.1, 145 (2d ed. 1971) (providing the dates of the survey and the percentage of defendants with a record).

  12. [12]. See Steven Raphael & Michael A. Stoll, Why Are So Many Americans in Prison?
    3–8, 51 tbl. 2.4 (2013) (presenting statistical analysis of the expanding incarcerated population over time and average time served).

  13. [13]. See Fed. R. Evid. 404; infra Part II.A.

  14. [14]. See Fed. R. Evid. 609; infra Part II.A.

  15. [15]. See infra Part II.A.

  16. [16]. See Cynthia E. Jones, “I AM Ronald Cotton”: Teaching Wrongful Convictions in a Criminal Law Class, 10 Ohio St. J. Crim. L. 609, 609 n.1 (2013) (“The case of Ronald Cotton is probably the most famous DNA exoneration.”).

  17. [17]. State v. Cotton, 394 S.E.2d 456, 458 (N.C. Ct. App. 1990).

  18. [18]. See id.

  19. [19]. See infra Parts II.B, III.B and IV.B–D; cf. Stephen J. Schulhofer, Some Kind Words for the Privilege Against Self-Incrimination, 26 Val. U. L. Rev. 311, 331 (1991) (describing defendants’ refusal to take the witness stand as “an everyday staple of trial practice”).

  20. [20]. See infra notes 86–87 and accompanying text (summarizing the findings of Part II).

  21. [21]. John H. Blume, The Dilemma of the Criminal Defendant with a Prior Record—Lessons from the Wrongfully Convicted, 5 J. Empirical Legal Stud. 477, 478 (2008); Eisenberg & Hans, supra note 7, at 1356 (“Limited empirical analysis exists of defendants’ decisions to testify or of the effect of prior criminal records on trial outcomes in real jury trials.”); Larry Laudan & Ronald J. Allen, The Devastating Impact of Prior Crimes Evidence and Other Myths of the Criminal Justice Process, 101 J. Crim. L. & Criminology 493, 496 (2011) (arguing that “many judges and legal scholars have been largely indifferent to, or unaware of, the empirical evidence”). The three articles cited in this footnote are exceptions, analyzing empirical data on the impact of prior conviction impeachment.

  22. [22]. See Laudan & Allen, supra note 21, at 493, 497 (noting the “seeming paradox” of the NCSC’s findings in this context).

  23. [23]. See infra Part V.A.

  24. [24]. See infra notes 160–64 and accompanying text.

  25. [25]. See Missouri v. Frye, 566 U.S. 134, 143 (2012) (“Ninety-seven percent of federal convictions and ninety-four percent of state convictions are the result of guilty pleas.”).

  26. [26]. See infra Part V.D.

  27. [27]. See infra Part V.D.

  28. [28]. See infra Part V.B.

  29. [29]. Cf. F. Lee Bailey & Kenneth J. Fishman, 2 Criminal Trial Techniques § 44:1 (2d ed. 2017) (discussing the decision to testify and negative jury reactions from a defendant’s choice not to testify).

  30. [30]. People v. Zackowitz, 172 N.E. 466, 468 (N.Y. 1930); cf. Michelson v. United States, 335 U.S. 469, 475–76 (1948) (discussing how the law generally ignores the defendant’s character).

  31. [31]. See Capra & Tartakovsky, supra note 4, at 155.

  32. [32]. See Fed. R. Evid. 404(a)(3), 609. Before laws permitting defendant testimony, defendants “rarely had to fear that their past convictions could become evidence against them.” Fisher, supra note 5, at 105.

  33. [33]. Gertz v. Fitchburg R.R. Co., 137 Mass. 77, 78 (1884) (“The evidence has no tendency to prove that [the witness] was mistaken, but only that he has perjured himself, and it reaches that conclusion solely through the general proposition that he is of bad character and unworthy of credit.”).

  34. [34]. Fed. R. Evid. 609.

  35. [35]. United States v. Richardson, 515 F.3d 74, 79 (1st Cir. 2008) (highlighting trial court’s instruction to “only consider these prior convictions for purposes of assessing [the defendant’s] credibility”); United States v. Stanley, 94 F. App’x 984, 986 (4th Cir. 2004).

  36. [36]. See Kalven, Jr. & Zeisel, supra note 11, at 147 (reporting that the jury hears about the defendant’s record in 72% of the cases when defendant takes the stand and in 13% of the cases when the defendant does not); Jeffrey Bellin, Circumventing Congress: How the Federal Courts Opened the Door to Impeaching Criminal Defendants with Prior Convictions, 42 U.C. Davis L. Rev. 289, 330 (2008) (tracing doctrinal developments that led state and federal courts to “routinely permit prior conviction impeachment of criminal defendants”); Blume, supra note 21, at 490 (reporting that in survey of exonerated defendants, “[i]n every single case in which a defendant with a prior record testified, the trial court permitted the prosecution to impeach the defendant with his or her prior convictions”). Eisenberg and Hans report that the NCSC data show only about half of the defendants with prior records who testified were impeached with their prior records and conclude that “the results suggest that judges, as evidentiary rules require, balance the relevance of a prior criminal record with the possible prejudice to the defendant.” Eisenberg & Hans, supra note 7, at 1373. A more likely explanation is that some of what constituted a “prior record” for the NCSC data did not qualify as impeachment. See infra Part III.A (describing limitations in NCSC prior record data).

  37. [37]. See Fed. R. Evid. 404(b); Michelson v. United States, 335 U.S. 469, 475 (1948); Eisenberg & Hans, supra note 7, at 1375 tbl.4 (reporting that in only 12 out of 137 cases in which defendants with prior convictions did not testify were the convictions were revealed).

  38. [38]. David R. Shaffer, The Defendant’s Testimony, in The Psychology of Evidence and Trial Procedure 124, 131 (Saul M. Kassim & Lawrence S. Wrightsman eds., 1985); see also A. N. Doob & H. M. Kirshenbaum, Some Empirical Evidence on the Effect of s. 12 of the Canada Evidence Act Upon an Accused, 15 Crim. L.Q. 88, 95 (1972) (concluding form mock juror study that jurors learning of prior conviction increased likelihood of conviction and that judicial “instructions to disregard the evidence will not counteract the damaging ‘halo’ effect of the previous convictions”); Eisenberg & Hans, supra note 7, at 1358 (“Most of the experimental studies show that knowledge of a defendant’s criminal record has statistically significant biasing effects on jurors’ guilt perceptions and verdicts.”); Edith Greene & Mary Dodge, The Influence of Prior Record Evidence on Juror Decision Making, 19 L. & Hum. Behav. 67, 71–73 (1995) (summarizing mock juror study that revealed that likelihood of conviction increased from 17% to 40% when jurors told about prior burglary conviction in bank robbery trial); Valerie P. Hans & Anthony N. Doob, Section 12 of the Canada Evidence Act and the Deliberations of Simulated Juries, 18 Crim. L.Q. 235, 243 (1976) (summarizing mock juror studies that found that jurors were more likely to convict in burglary case when told of prior burglary conviction and concluding that “[p]resence of record, then, appears to reliably increase the probability that a defendant will be found guilty by a jury, regardless of the evidence”); Kathryn Stanchi & Deirdre Bowen, This Is Your Sword: How Damaging Are Prior Convictions to Plaintiffs in Civil Trials?, 89 Wash. L. Rev. 901, 911 (2014) (“Most studies show that admission of a defendant’s prior conviction leads to more guilty verdicts in criminal trials, regardless of whether the jurors receive a limiting instruction.”).

  39. [39]. Old Chief v. United States, 519 U.S. 172, 180 (1997).

  40. [40]. Shaffer, supra note 38 (“[S]everal investigators have found the evidence of a prior record influences jurors’ verdicts without affecting their assessments of the defendant’s credibility as a witness.”).

  41. [41]. Roselle L. Wissler & Michael J. Saks, On the Inefficacy of Limiting Instructions: When Jurors Use Prior Conviction Evidence to Decide on Guilt, 9 L. & Hum. Behav. 37, 39 (1985).

  42. [42]. Id. at 40 & n.5.

  43. [43]. Id. at 43 tbl.2.

  44. [44]. Id.

  45. [45]. The study reports findings from 20 subjects per condition. Id. at 39–40 (160 persons assigned to 8 distinct conditions).

  46. [46]. See E. Gil Clary & David R. Shaffer, Another Look at the Impact of Juror Sentiments Toward Defendants on Juridic Decisions, 125 J. Soc. Psychol. 637, 645 (1985) (finding that mock juries convicted markedly more often in an armed robbery case if the defendant had a prior robbery conviction than if he had a prior conviction for counterfeiting); Sally Lloyd-Bostock, The Effects on Juries of Hearing About the Defendant’s Previous Criminal Record: A Simulation Study, 2000 Crim. L.R., 734, 744 (summarizing results of mock juror study: “[a] recent similar conviction produces the most guilty verdicts”).

  47. [47]. See infra Part III.B.

  48. [48]. Clary and Shaffer, supra note 46; Shaffer, supra note 38.

  49. [49]. Eisenberg & Hans, supra note 7, at 1387 (relying on juror credibility ratings to conclude, “we do not find evidence that criminal records affect defendant credibility... . In cases in which defendants testified, criminal record was not significantly associated with the degree of believability.”).

  50. [50]. United States v. Mahone, 537 F.2d 922, 929 (7th Cir. 1976); see also Bellin, supra note 36, at 313 (describing the Mahone-based framework courts utilize for evaluating the admissibility of prior convictions).

  51. [51]. Gordon v. United States, 383 F.2d 936, 940 (D.C. Cir. 1967).

  52. [52]. See supra note 36 (identifying sources); see also Bellin, supra note 36, at 322–35 (citing examples and highlighting pro-admission bias of five-factor framework that governs admissibility of prior conviction impeachment).

  53. [53]. Laudan & Allen, supra note 21, at 500–01 (describing “congeries of conflicting results”).

  54. [54]. Id. at 501.

  55. [55]. Id. at 494–96 (arguing that the “conventional wisdom” along with “aspects of American evidence law—and considerable American evidence scholarship” is all erroneously based on a “widely-shared set of beliefs” including that “[t]elling jurors about the prior crimes of a defendant dramatically increases their disposition to convict”).

  56. [56]. Id. at 500–01 (identifying Doob & Kirshenbaum, supra note 38; Greene & Dodge, supra note 38; Lloyd-Bostock, supra note 46); Hans & Doob, supra note 38; Wissler & Saks, supra note 41.

  57. [57]. Id.

  58. [58]. Bruce Rind et al., Effect of Crime Seriousness on Simulated Jurors’ Use of Inadmissible Evidence, 135 J. Soc. Psychol. 417, 419 (1995) (testing jurors’ ability to ignore “incriminating wire tap evidence”); William C. Thompson et al., Inadmissible Evidence and Juror Verdicts, 40 J. Personality & Soc. Psychol. 453, 456 (1981) (testing jurors’ ability to disregard a “tape recording of a conversation between the defendant and [a] bookmaker”); see also Thomas R. Carretta & Richard L. Moreland, The Direct and Indirect Effects of Inadmissible Evidence, 13 J. Applied Soc. Psychol. 291, 295 (1983) (testing jurors’ ability to disregard “a taped telephone conversation” involving, again, a bookmaker—in this example, the wiretap was not illegal, but the evidence was deemed inadmissible because the defendant was not the target of the wiretap).

  59. [59]. See supra note 58 and accompanying text.

  60. [60]. Laudan & Allen, supra note 21, at 500–01 (incorrectly describing the Thompson et al. study as concerning “inadmissible priors evidence,” Rind et al. study as concerning “prior crimes information,” and Carretta and Moreland study as finding similar results in two groups, one told about prior crimes and the other not).

  61. [61]. L.S.E. Jury Project, Juries and the Rules of Evidence, 1973 Crim. L. Rev. 208, 210 (1973) (noting that solicitations to participate were mailed to “government and commercial offices in central London”).

  62. [62]. Id. at 212–13; cf. Rind et al., supra note 58, at 418 (cautioning about generalization of the Sealy and Cornish’s findings because “many factors in addition to seriousness differed between the theft and rape cases”).

  63. [63]. L.S.E. Jury Project, supra note 61, at 218.

  64. [64]. Michael J. Saks & Barbara A. Spellman, The Psychological Foundations of Evidence Law 168 (Linda J. Demaine ed., 2016) (“The available empirical evidence is unanimous in finding that, notwithstanding judicial instructions to the contrary, most people travel the forbidden path of using prior crimes evidence to make substantive inferences about the likelihood that the testifying defendant committed the current crime charged.”).

  65. [65]. See infra Part III.B.

  66. [66]. See 1A Criminal Defense Techniques § 24A.03 (Matthew Bender ed., 2017); cf. Bailey & Fishman, supra note 29 (noting that “classically it was thought that a defendant should avoid testifying unless absolutely necessary”).

  67. [67]. See Serial Podcast, Episode 8: The Deal with Jay, Genius, (last visited Oct. 15, 2017).

  68. [68]. Fox News/Opinions Dynamics Poll (Feb. 2002) (survey of 900 registered voters by telephone, asking, “[w]hen someone cites his or her Fifth Amendment right in refusing to testify, do you generally think the person is probably guilty [50%], or the person is simply exercising a right [36%]?”); Decision Quest, Fair Juror Survey (Sept. 1999) (survey of 1,000 people by telephone, asking, “[w]hen a defendant in a criminal case does not testify it means he or she probably has something to hide [50%] or it does not mean he or she has something to hide [38%]”); Gallup Poll (Apr. 1957) (personal interviews of 1,654 people, asking, “[w]hen you hear of a person using the fifth amendment, do you generally think he is guilty [48%], or not [16%]?”). The balance of each percentage break down consists of “I don’t know” or “No opinion.” All of the polls cited above were located using iPoll. Cornell Univ., Search U.S. Questions with iPoll, RoperCenter, (last visited Oct. 15, 2017) (finding Fox News poll using ID “USODFOX.021402.R11”; Decision Quest poll with ID “USBRUSKN.99JURY.R03A”; and Gallup Poll with ID “USGALLUP.57-581.Q035”). Thanks to John Blume for the inspiration to search iPoll and the citation to the Fox News poll. See Blume, supra note 21, at n.1.

  69. [69]. Fox News/Opinion Dynamics Poll, supra note 68.

  70. [70]. See generally David R. Shaffer & Thomas Case, On the Decision to Testify in One’s Own Behalf: Effects of Withheld Evidence, Defendant’s Sexual Preferences, and Juror Dogmatism on Juridic Decisions, 42 J. Personality & Soc. Psychol 335 (1982); Shaffer, supra note 38.

  71. [71]. E. Gil Clary & David R. Shaffer, Effects of Evidence Withholding and a Defendant’s Prior Record on Juridic Decisions, 112 J. Soc. Psychol. 237, 242 (1980).

  72. [72]. Id. at 243.

  73. [73]. Id. at 245; cf. Hans & Doob, supra note 38, at 245 (noting that, in another study, the “presence of record increased the salience of the negative evidence against the defendant”).

  74. [74]. Shaffer & Case, supra note 70, at 336.

  75. [75]. Id. at 339.

  76. [76]. Id.

  77. [77]. Id. at 341.

  78. [78]. Id.

  79. [79]. Id. at 342.

  80. [80]. Id. at 344.

  81. [81]. E.g., Fed. R. Evid. 606.

  82. [82]. 3 Christopher B. Mueller & Laird C. Kirkpatrick, Federal Evidence § 6:17, at 81 (3d ed. 2007) (“Despite the constitutional right not to testify ... the Rule bars proof that one or more jurors held it against the accused that he failed to take the stand.”).

  83. [83]. See, e.g., United States v. Torres-Chavez, 744 F.3d 988, 998 (7th Cir. 2014) (“join[ing] every other circuit court to consider the issue” in holding that juror statements about considering the defendant’s failure to testify cannot be used to impeach a verdict under Rule 606(b) and citing cases from the 3d, 5th, 6th, 8th, 9th and 10th Circuits); United States v. Kelley, 461 F.3d 817, 831 (6th Cir. 2006) (rejecting claim for new trial based on juror statement to press that, “[i]f [the defendants] were innocent, they would have testified”); United States v. Rutherford, 371 F.3d 634, 638–39 (9th Cir. 2004) (same); Pendergast v. Newland, 29 F. App’x 459, 463 (9th Cir. 2002) (same); United States v. Rodriquez, 116 F.3d 1225, 1226 (8th Cir. 1997) (same); United States v. Tran, 122 F.3d 670, 673 (8th Cir. 1997) (same); United States v. Martinez-Moncivais, 14 F.3d 1030, 1036 (5th Cir. 1994) (rejecting challenge based on fact that “one of the jurors stated in a post-trial affidavit that two other jurors suggested that if the defendant had been innocent he would have taken the stand in his own defense”); United States v. Friedland, 660 F.2d 919, 927–28 (3d Cir. 1981) (same); Coleman v. Sisto, No. 2:09–cv–0020, 2012 WL 6020095, at *8 (E.D. Cal. Dec. 3, 2012) (same); Rowland v. Chappell, 902 F. Supp. 2d 1296, 1335 (N.D. Cal. 2012) (same); United States v. Stewart, No. 4:07CR–10–M, 2008 WL 2038897, at *1 (W.D. Ky. May 12, 2008) (same); United States v. Hollingsworthmata, 72 M.J. 619, 622–23 (A. Ct. Crim. App. 2012) (discussing Military Rule of Evidence 606(b) and concluding the same); see also United States v. Stewart, No. 91–00070, 1994 WL 547811, at *4 (E.D. Pa. Oct. 3, 1994) (rejecting allegation of ineffective assistance of counsel based on counsel’s failure to seek new conviction based on juror’s statement that defendant’s failure to testify demonstrated his guilt); United States v. Edwards, 486 F. Supp. 673, 674 (S.D.N.Y. 1980) (rejecting claim for new trial based on anonymous phone call alleging “that the jury deliberations were in large measure focused on discussions concerning Mr. Edwards’ failure to testify”).

  84. [84]. Strand v. McDonald, No. 2:12–cv–1237, 2013 WL 5755059, at *4 (E.D. Cal. Oct. 23, 2013) (alteration in original).

  85. [85]. Id. at *1.

  86. [86]. See supra note 83 (collecting cases).

  87. [87]. Shaffer, supra note 38, at 145.

  88. [88]. See Eisenberg & Hans, supra note 7, at 1386 (theorizing that the effect of a criminal record “occurs primarily in cases in which the evidence is not overwhelming” and noting that this “resonates with Kalven and Zeisel’s classic finding that extralegal factors have the most impact primarily in close, as opposed to clear, cases”).

  89. [89]. Judicial Council of Cal., 2013 Court Statistics Report: Statewide Caseload Trends 2002–2003 Through 2011–2012, at 47 fig.35 (2013),; Trial Courts Statistics, Florida Courts,
    TrialCourtStats.aspx (last visited Oct. 15, 2017) (reporting state-wide 73% rate for convictions after jury trial for 2015, excluding cases resolved by plea); see also Barbara Allen Babcock, Introduction: Taking the Stand, 35 Wm. & Mary L. Rev. 1, 12 (1993) (“Felony conviction rates after trial are around eighty percent ... .”); Laudan & Allen, supra note 21, at 501–02 (criticizing mock juror studies for relying on cases that present “ambiguous and non-decisive” evidence and noting “that most real criminal cases are not borderline cases”); Neil Vidmar et al., Should We Rush to Reform the Criminal Jury, 80 Judicature 286,
    288–89 (1997) (surveying jury conviction rates).

  90. [90]. See Christopher T. Robertson et al., Perceptions of Efficacy, Morality, and Politics of Potential Cadaveric Organ-Transplantation Reforms, 77 L. & Contemp. Probs. 101, 108 n.30 (2014) (“Mturk, a human-subject population ... is increasingly utilized for social science research.”).

  91. [91]. See, e.g., 28 U.S.C. § 1865 (2012) (federal jury service qualifications).

  92. [92]. See Matthew R. Ginther et al., The Language of Mens Rea, 67 Vand. L. Rev. 1327, 1342 n.48 (2014) (explaining that “[m]ultiple studies have validated results using Amazon’s Mechanical Turk on a variety of assessments, especially when compared to samples of convenience” and citing numerous studies); Danielle N. Shapiro et al., Using Mechanical Turk to Study Clinical Populations, 1 Clinical Psychol. Sci. 213, 214 (2012) (“Cross-sample investigations comparing MTurk to other methods of data collection have demonstrated that data obtained from its workers are similar to data collected from more traditional subject pools (e.g., college undergraduates or community samples derived from college towns) in a variety of research domains, including ... basic biases in decision making.” (citation omitted)).

  93. [93]. Gabriele Paolacci et al., Running Experiments on Amazon Mechanical Turk, 5 Judgment
    & Decision Making 411, 416 (2010); see sources cited in supra notes 90–92.

  94. [94]. After pilot testing established a rough estimate of how long the survey took, I established a payment rate that translated to a little more than the federal and Virginia minimum wage. This rate is high by Mechanical Turk standards, and resulted in an almost immediate acceptance of the survey by the maximum number of respondentssomething that likely assisted with obtaining a random sample. Payment of minimum wage is not standard even for academic requesters. See Shapiro et al., supra note 92 (describing the $2.25 per hour pay provided as “above average for MTurk”); Jill D. Weinberg et al., Comparing Data Characteristics and Results of an Online Factorial Survey Between a Population-Based and a Crowdsource-Recruited Sample, 1 Soc. Sci. 292, 298 (2014) (describing $3 per hour pay provided as “relatively high by MT standards”). For quality control purposes, users could not have more than a 5% failure rate on previous Mechanical Turk tasks. Weinberg et al., supra (same qualification). Open-ended comments at the conclusion of the survey indicated that the respondents enjoyed the survey experience and took it seriously.

  95. [95]. This is a common process that delivers the data directly to the researcher, rather than leaving it on the Mechanical Turk survey platform. See Weinberg et al., supra note 94.

  96. [96]. Qualtrics Survey Software, Survey Preview, California v. Ronald Willis 5 (Aug. 25, 2015) (survey print preview on file with author). The instruction paralleled the standard jury instruction in this context. See United States v. Stanley, 94 F. App’x 984, 986 (4th Cir. 2004) (“[P]rior convictions should only be used to judge ... credibility rather than ... propensity to commit crimes.”).

  97. [97]. Again, this is common practice. See Weinberg et al., supra note 94, at 294 (“[R]esearchers routinely embed ‘comprehension checks’ in surveys ... .”). Seven of the 400 respondents (1.75%) were excluded from the final tally for failing to demonstrate minimal comprehension of the factual scenario; the disqualification threshold was established in advance.

  98. [98]. Significant at p <0.05 and p <0.01 with two-tailed hypothesis. To test for statistical significance, I used a Z-test for two population proportions, following Robert A. Hanneman et al., Basic Statistics for Social Research 294–96 (2013), and confirmed my calculations with an online statistical significance calculator available at the “Social Science Statistics” web site. Z Score Calculator for 2 Population Proportions, Social Science Statistics,
    Default.aspx (last visited Oct. 15, 2017). I report the statistical significance only for skeptics. The experiment is sufficiently intuitive that the reported percentages speak for themselves. Like most experimental results, however, they are powerful only in the context of the other supporting evidence reported throughout this paper. See D.H. Kaye, Is Proof of Statistical Significance Relevant?, 61 Wash. L. Rev. 1333, 1344–45 (1986) (“There is no strictly objective basis, in science or in anything else, for believing that a proposition is true simply because the evidence for it is ‘statistically significant’ at the .05 level.”).

  99. [99]. Fed. R. Evid. 609 advisory committee’s note to 1990 amendment (quoting H.R. Rep. No. 93-1597, at 9 (1974) (Conf. Rep.)) (providing a list of such crimes).

  100. [100]. The fact scenario informed the mock jurors that the defendant is not required to testify, but did not specifically instruct them to disregard his decision not to. See Carter v. Kentucky, 450 U.S. 288, 300 (1981) (requiring such an instruction “when requested by a defendant”). In theory, such an instruction could have mitigated the damage. The evidence is, however, to the contrary. See Blume, supra note 21, at 488 (“The jury is likely to disregard an instruction that this inference [of guilt from silence] is not permissible.”); Shaffer, supra note 38, at 147 n.1 (citing empirical study that “found the juries instructed not to draw adverse inferences from a defendant’s Fifth Amendment plea were no more likely to convict the accused than were juries receiving no judicial commentary on the meaning and use of the privilege”); J. Alexander Tanford, The Law and Psychology of Jury Instructions, 69 Neb. L. Rev. 71, 72 (1990) (“[R]esearch by psychologists demonstrates that instructing jurors to disregard the silence will not accomplish the task.”); id. at 89 (“An admonition will not reduce the likelihood that jurors will draw adverse inferences from the defendant’s silence, but will tend to aggravate its prejudicial impact.”). Some defense counsel actually view the instruction as harmful “because it calls attention to the defendant’s silence.” Coleman v. Brown, 802 F.2d 1227, 1235 (10th Cir. 1986) (noting that “counsel admitted that attorneys often decide not to request such an instruction because it calls attention to the defendant’s silence” in rejecting claim of ineffective assistance based on failure to request instruction); Michaelwicz v. State, 186 S.W.3d 601, 624 n.12 (Tex. App. 2006) (noting that “[d]efense counsel frequently complain when the no adverse inferences instruction is given in the jury charge” and citing cases). Given the evidence that the instruction is ineffective and the danger of providing it in this scenario (where it would reveal the purpose of the simulation to participants, who would wonder why the scenario noted the defendant’s failure to testify if they were not to consider it), I did not include one.

  101. [101]. Significant at p <0.05 with two-tailed hypothesis. See supra note 98 for methodology and a disclaimer regarding the import, in this context, of a finding of statistical significance.

  102. [102]. The other major study of real trials in this context is “The American Jury” study by Kalven and Zeisel. The study conflates defendant testimony and prior record into one variable making it unhelpful for current purposes. Kalven, Jr. & Zeisel, supra note 11, at 159 n.17, 179 tbl.56. Nevertheless, as this choice indicates, Kalven and Zeisel come very close to assuming the equivalence that this Article hypothesizes. Id. (explaining that they grouped those with a revealed record with those who declined to take the stand because the jury penalizes defendants for “either the record of which the jury learns or the suspicion of a record because of the refusal to take the stand”). They also report cases where judges specifically note the jury’s learning of the defendant’s record through the evidence as a reason for conviction, and the authors add that their study “lends support to the legal tradition which so closely guards the disclosure of a prior record in a criminal case.” Id. at 389–90.

  103. [103]. Paula L. Hannaford-Agor et al., Nat’l Ctr. for State Courts, Are Hung Juries a Problem? 29 (2002).

  104. [104]. Id. at 1–3 (describing purpose of study). To review the questionnaires, see generally Paula L. Hannaford-Agor et al., Nat’l Inst. Of Justice, Evaluation of Hung Juries in Bronx County, New York, Los Angeles County, California, Maricopa County, Arizona, and Washington, DC, 2000–2001: Codebook and Data Collection Instrument (2003).

  105. [105]. Hannaford-Agor et al., supra note 103, at 73–74, 85.

  106. [106]. Id. at 67.

  107. [107]. Id.; see also Eisenberg & Hans, supra note 7, at 1365 (acknowledging that “we lack information about the nature of defendants’ prior crimes”); id. at 1389 (“[W]e had information only about the presence of a defendant’s criminal record, not its type.”).

  108. [108]. See Fed. R. Evid. 609; cf. Blume, supra note 21, at 490 n.50 (critiquing the same finding on the ground that the data used in the Eisenberg and Hans study “did not permit a determination of whether the prior conviction could have been used for impeachment purposes”). In addition, the study’s authors reported that over 25% of the responses did not answer the question at all. Paula L. Hannaford-Agor et al., supra note 104.

  109. [109]. Givelber and Farrell count all convictions as wins for the prosecution. Daniel Givelber & Amy Farrell, Not Guilty: Are the Acquitted Innocent? 176 n.41 (2012). Neither Laudan and Allen, nor Eisenberg and Hans appear to state how they coded mixed results, suggesting that they similarly viewed a conviction for any offense as a “conviction” even if the ultimate result reflected some corresponding acquittals.

  110. [110]. See Eisenberg & Hans, supra note 7.

  111. [111]. Id.

  112. [112]. Id. at 1371 & tbl.1.

  113. [113]. Id. at 1378 (“A defendant’s testifying is the only factor substantially and significantly contributing to whether the jury learned of the defendant’s prior criminal record.”) The NCSC data does not suggest as strong a relation between defendant testimony and impeachment as pertinent legal doctrine might suggest, perhaps due to the catch-all nature of the criminal history question. See Paula L. Hannaford-Agor et al., supra note 104, at 7; Eisenberg & Hans, supra note 7, at 1365 (acknowledging that “we lack information about the nature of defendants’ prior crimes”); id. at 1389 (“[W]e had information only about the presence of a defendant’s criminal record, not its type.”); cf. Blume, supra note 21, at 490 n.50 (critiquing same finding on the ground that the data used in the Eisenberg and Hans study “did not permit a determination of whether the prior conviction could have been used for impeachment purposes”).

  114. [114]. Eisenberg & Hans, supra note 7, at 1372.

  115. [115]. Id. (“[T]here are no significant differences in the strength of evidence between cases of defendants who did or did not testify.”); id. at 1378 (reporting regression model results attempting to determine “whether a defendant testifying is associated with the strength of the evidence as reported by juries” and “whether the strength of the evidence, as reported by judges, is associated with whether a defendant testified” and finding “no significant association in either relation”).

  116. [116]. Laudan & Allen, supra note 21, at 493, 499.

  117. [117]. Id. at 504 tbl.2.

  118. [118]. Id. at 498.

  119. [119]. Daniel Givelber, Lost Innocence: Speculation and Data About the Acquitted, 42 Am. Crim. L. Rev. 1167, 1190 (2005); see also Eisenberg & Hans, supra note 7, at 1380 (noting that with respect to defendants with prior records, “conviction rates did not noticeably differ between defendants who testified and those who did not”).

  120. [120]. Eisenberg & Hans, supra note 7, at 1380; Givelber, supra note 119 at 1190.

  121. [121]. Laudan & Allen, supra note 21.

  122. [122]. See id. at 496 (“One way or another, all of these hypotheses undergirding the conventional wisdom about prior crimes evidence are empirically testable. More than that, they have already been tested and most stand refuted or, at least, rendered highly implausible.”).

  123. [123]. Id. at 498.

  124. [124]. Id. at 508–09, 519 (“[J]urors will generally know when the defendant has no prior record and can usually infer when he is a serial felon.”); id. at 521 (“[J]urors generally figure out which defendants have prior convictions ... .”); id. at 522 (“[J]urors can readily infer that a defendant is a serial felon even when no priors are admitted ... .”).

  125. [125]. Id. at 515 (“Refusing to admit prior crimes evidence for fear that jurors will over-interpret its significance or derive some propensity inferences from it is, in the current system, self-defeating.”); id. at 498 (“[A] jury’s learning of prior crimes directly through the evidence is not the inflammatory, unfairly prejudicial, conviction-ensuring information it is often depicted as being... . [R]ailing against the admissibility of prior crimes on the grounds that they unfairly disadvantage defendants with criminal records is unnecessary hyperbole.”); id. at 499 (“[N]ot much depends on the admissions of the priors.”).

  126. [126]. Id. at 507.

  127. [127]. Id. at 498 (“[T]he strenuous efforts of legal experts and defense attorneys to restrict the admissibility of prior crimes evidence seem misplaced.”).

  128. [128]. See supra notes 53–64 and accompanying text.

  129. [129]. See supra notes 57–60 and accompanying text.

  130. [130]. See Givelber, supra note 119, at 1189 (recognizing “the universal perception” that the jury’s hearing of a criminal record negatively impacts the defendant’s prospects); Laudan
    & Allen, supra note 21, at 494 (recognizing consensus).

  131. [131]. See infra Part IV.E.

  132. [132]. See Gary Fields & John R. Emshwiller, As Arrest Records Rise, Americans Find Consequences Can Last a Lifetime, Wall St. J. (Aug. 18, 2014, 10:30 PM), (reporting on studies that reflect correlation between arrests record and lower income).

  133. [133]. Caroline Wolf Harlow, Bureau of Justice Statistics, Defense Counsel in Criminal Cases 1 (2000) (reporting that, in 1996, 17.6% of defendants in state felony cases in the 75 largest counties retained a private attorney).

  134. [134]. See Jeffrey Bellin, Attorney Competence in an Age of Plea Bargaining and Econometrics, 12 Ohio St. J. Crim. L. 153, 153, 155 (2014) (discussing reputation of D.C. Public Defender Service and study that found Philadelphia public defenders outperformed court-appointed, private attorneys); Brooks Holland, Holistic Advocacy: An Important but Limited Institutional Role, 30 N.Y.U. Rev. L. & Soc. Change 637, 637 (2006) (describing “Bronx Defenders [as] a well respected public defender office in Bronx County”).

  135. [135]. Thomas H. Cohen & Brian A. Reaves, Bureau of Justice Statistics, State Court Processing Statistics, 1990–2004: Pretrial Release of Felony Defendants in State Courts 6 (2007) (reporting that defendants with a prior arrest or prior conviction “had a lower probability of [pretrial] release”).

  136. [136]. Id. at 7 tbl.5 (reporting that 69% of held defendants and 46% of released defendants were subsequently convicted of a felony offense). One explanation for this finding is that judges consider the strength of the evidence against a defendant in deciding whether to grant release. See, e.g., 18 U.S.C. § 3142(g)(2) (2012) (including “the weight of the evidence against the person” among factors the judge must consider in determining conditions of release).

  137. [137]. This table is derived from Laudan & Allen, supra note 21, at 516 tbl.6. Laudan and Allen do not provide the number of defendants from which they obtained these percentages. To obtain an approximation, I drew the number of defendants with these characteristics from Eisenberg & Hans, supra note 7, at 1371 tbl.1.

  138. [138]. See Laudan & Allen, supra note 22, at 516 tbl.6 (finding same).

  139. [139]. See id. (finding same). This data comes from Laudan and Allen’s own analysis of the NCSC data sets. Id. at 516 n.71. Laudan and Allen do not provide information on the evidentiary strength of the cases within this subset, but we do know from Eisenberg and Hans’s analysis of the same data that there was no significant difference overall in the perceived evidentiary strength between cases in which the defendant testified or remained silent. Eisenberg & Hans, supra note 7, at 1371.

  140. [140]. These numbers come from Eisenberg & Hans, supra note 7, at 1381 tbl.8.

  141. [141]. See sources cited supra notes 36–37.

  142. [142]. Eisenberg & Hans, supra note 7, at 1371 tbl.1, 1381 tbl.8.

  143. [143]. See supra Part III.B.

  144. [144]. Eisenberg & Hans, supra note 7, at 1381 tbl.8 (76.6% and 77.8% conviction rates for two subsets of defendants with prior convictions who testified).

  145. [145]. Id. (reporting 71.4% and 72.7% conviction rates for two subsets of defendants with prior convictions who did not testify).

  146. [146]. Id.

  147. [147]. Id.

  148. [148]. Supra Part III.B.

  149. [149]. Eisenberg & Hans, supra note 7, at 1381 tbl.8; supra Table 4.

  150. [150]. Eisenberg & Hans, supra note 7, at 1381 tbl.8 (reporting 71.4% and 72.7% conviction rates for two subsets of defendants with prior convictions who did not testify).

  151. [151]. Laudan & Allen, supra note 21, at 516 tbl.6.

  152. [152]. Id. at 506 tbl.5; Eisenberg & Hans, supra note 7, at 1381 tbl.8.

  153. [153]. Laudan and Allen also claim support from the indistinguishable conviction rates between non-testifying defendants with prior convictions whose convictions were disclosed (72.7%) and those whose convictions were not disclosed (71.4%). Laudan & Allen, supra note 21, at 512. But this result should be dismissed as an artifact of a small sample size. Because “juries rarely learn of criminal records unless defendants testify,” the first percentage comes from an idiosyncratic subset of only 11 cases Eisenberg & Hans, supra note 7, at 1379, 1381 tbl.8. As Laudan and Allen acknowledge, “[a] strong word of caution is in order ... . [when] we focus on issues that segment th[e] sample down to smaller and smaller subsets.” Laudan & Allen, supra note 21, at 505 n.45.

  154. [154]. See Eisenberg & Hans, supra note 7, at 1381–83.

  155. [155]. Id. at 1383.

  156. [156]. Cf. id. at 1380 (“Studies of jury behavior indicate that the strength of the evidence dominates decision making.”).

  157. [157]. Laudan and Allen also cite a study of 201 Indianapolis jury trials to support their conclusions, but that study offers more support for the parallel penalty hypothesis. Laudan
    & Allen, supra note 21, at 507 (citing Myers, supra note 11, at 792). Consistent with the NCSC data, Myers found that a record of prior convictions increased the likelihood of conviction. Myers, supra note 11, at 793 tbl.2. Laudan and Allen emphasize that this supports their juror sophistication hypothesis because “in most cases jurors never learned about the priors directly through the evidence.” Laudan & Allen, supra note 21, at 507. Myers does not report that the prior convictions were rarely admitted, and Laudan and Allen don’t include a citation for their assertion; perhaps they, like Shaffer, are misreading Myers’s study to suggest that 82% of defendants did not testify (and then assuming that therefore “most” jurors never learned of the convictions). See infra notes 168–76 and accompanying text. Even if Laudan and Allen’s assumption is accurate, however, Myers’s findings support the parallel penalty hypothesis as well: Defendants with prior convictions are convicted more frequently because they are either impeached or do not testify. This appears to be how Myers interprets her own data, summarizing her findings as follows: “Jurors were more likely to convict if the defendant had numerous prior convictions ... , and thus may have been potentially discreditable as a witness.” Myers, supra note 11, at 792–93; see also id. at 795 (“[Jury] rulings tended to be adverse where the defendant was discredited or discreditable.”).

  158. [158]. Eisenberg & Hans, supra note 7, at 1371 tbl.1.

  159. [159]. Babcock, supra note 89, at 14­–15; see also Barbara Allen Babcock, The Duty to Defend, 114 Yale L.J. 1489, 1514 (2005); cf. Jeffrey Bellin, Reconceptualizing the Fifth Amendment Prohibition of Adverse Comment on Criminal Defendants’ Trial Silence, 71 Ohio St. L.J. 229, 269–80 (2010) (exploring reasons defendants may not testify).

  160. [160]. Babcock, supra note 89, at 12, 14–15; cf. 3 Anthony G. Amsterdam, Trial Manual for the Defense of Criminal Cases § 390 (3d ed. 1974) (“[T]he defendant probably ought to testify unless there are reasons of substance why he should not.”).

  161. [161]. Babcock, supra note 89, at 15.

  162. [162]. Id.

  163. [163]. Id.

  164. [164]. Cf. Amsterdam, supra note 160.

  165. [165]. There are other adverse consequences to testifying, such as sentencing enhancements in certain jurisdictions, if the judge finds that the defendant lied on the stand. See Jeffrey Bellin, Improving the Reliability of Criminal Trials Through Legal Rules that Encourage Defendants to Testify, 76 U. Cin. L. Rev. 851, 868–80 (2008) (exploring disadvantages of testifying).

  166. [166]. See Model Rules of Prof’l Conduct r. 3.3 & cmt. (Am. Bar Ass’n 2014).

  167. [167]. Babcock, supra note 159, at 1513; cf. Silver, supra note 9, at 423–24 (advocating elimination of ethical limits on presentation of client’s testimony).

  168. [168]. Shaffer, supra note 38, at 144 (relying on Myers’s study as basis for the hesitancy despite his own findings).

  169. [169]. Id. at 127, 144. Eisenberg and Hans also appear to read the study this way, explaining that Myers found that “in just 36 [of 201] jury trials, the defendant (or accomplices) provided statements or testimony about involvement or lack of involvement in the crime.” Eisenberg
    & Hans, supra note 7, at 1364.

  170. [170]. Kalven and Zeisel reported the opposite—that 82% of defendants testified—in their study of trials in 1954–1955. Kalven, Jr. & Zeisel, supra note 11, at 144. This dated study is actually closer to the date of Myers’s 1974­–1976 study than Myers is to the NCSC (2000–2001) study, making Shaffer’s interpretation of Myers finding even more unlikely.

  171. [171]. Myers, supra note 11, at 786 tbl.1; Shaffer, supra note 38, at 127.

  172. [172]. Myers, supra note 11, at 792.

  173. [173]. This explains why Myers grouped accomplice and defendant statements together. Pretrial statements by defendants and accomplices recorded by police will generally be incriminatory. (Myers gathered data from the prosecutor’s “file folder” and “police arrest records.”). Id. at 785. By contrast, grouping trial testimony of accomplices and defendants makes little sense, as trial testimony by defendants will always be exculpatory whereas testimony by accomplices is often incriminating.

  174. [174]. Id. at 786 tbl.1 (emphasis added).

  175. [175]. Id. at 787.

  176. [176]. I contacted Myers on this point and she agreed with my interpretation of her findings. E-mail from Martha Myers, Emeritus Professor, Univ. of Ga., to Jeffrey Bellin, Cabell Research Professor of Law, William & Mary Law Sch. (July 16, 2015, 5:50 PM) (on file with author). Professor Myers graciously allowed me to relate her agreement with my interpretation of her study. Id.

  177. [177]. Carter v. Kentucky, 450 U.S. 288, 300 (1981) (“[T]he Fifth Amendment requires that a criminal trial judge must give a ‘no-adverse-inference’ jury instruction when requested by a defendant to do so.”); Griffin v. California, 380 U.S. 609, 615 (1965) (stating that the Fifth Amendment precludes “either comment by the prosecution on the accused’s silence or instructions by the court that such silence is evidence of guilt”).

  178. [178]. Old Chief v. United States, 519 U.S. 172, 180–81 (1997).

  179. [179]. Krulewitch v. United States, 336 U.S. 440, 453 (1949) (Jackson, J., concurring) (citation omitted).

  180. [180]. Ferguson v. Georgia, 365 U.S. 570, 582 (1961); Silver, supra note 9, at 357 (arguing that when defendants are silenced, “the factfinder is deprived ... of the opportunity to learn directly about the defendant’s credibility and her version of the facts” and even in the case of lying defendants, is “deprived of the truthful information that typically accompanies falsehoods”).

  181. [181]. See sources cited in supra notes 9, 176.

  182. [182]. See Eisenberg & Hans, supra note 7, at 1387 (noting that evidentiary strength ratings for convictions were lower for cases in which the defendant’s record was revealed); supra Part IV.E.

  183. [183]. Brandon L. Garrett, Convicting the Innocent: Where Criminal Prosecutions Go Wrong 160 (2011) (reporting that 53% of “exonerees took the stand at trial to claim their innocence”); Blume, supra note 21, at 489.

  184. [184]. See supra note 111 and accompanying text.

  185. [185]. Blume, supra note 21, at 490; see also Garrett, supra note 183, at 162 (“Most of those who did not testify likely chose not to because they had prior convictions that could be used to discredit them.”).

  186. [186]. Blume, supra note 21, at 490.

  187. [187]. Id. at 479; see also Friedman, supra note 4 (reaching a similar conclusion).

  188. [188]. See Missouri v. Frye, 566 U.S. 134, 143 (2012) (“Ninety-seven percent of federal convictions and ninety-four percent of state convictions are the result of guilty pleas.”).

  189. [189]. See generally Michelle Alexander, The New Jim Crow: Mass Incarceration in the Age of Colorblindness (2012).

  190. [190]. See Fisher, supra note 5, at 109; see also Raymond Moley, The Vanishing Jury, 2 S. Cal L. Rev. 97, 107 (1928) (presenting historical data on plea bargaining).

  191. [191]. See Malcolm M. Feeley, Plea Bargaining and the Structure of the Criminal Process, 7 Just. Sys. J. 338, 350 (1982).

  192. [192]. Kalven, Jr. & Zeisel, supra note 11, at 144–45, 145 n.12.

  193. [193]. See Eisenberg & Hans, supra note 7, at 1371 tbl.1. For an exploration of the similarities and differences between the two studies, including the narrow urban focus of the NCSC study versus the broader scope of Kalven and Zeisel, see Theodore Eisenberg et al., Judge-Jury Agreement in Criminal Cases: A Partial Replication of Kalven and Zeisel’s The American Jury, 2 J. Empirical Legal Stud. 171, 178–79 (2005).

  194. [194]. The Bureau of Justice Statistics reports that the percentage of felony arrestees “with a felony conviction record increased from 36% in 1990 to 43% in 2009.” Reaves, supra note 11, at 9. One would expect that the presence of a criminal record impacts the prosecutor’s calculus as to whether to pursue a case to trial.

  195. [195]. Thomas P. Bonczar, Prevalence of Imprisonment in the U.S. Population, 1974–2001, at 1 (2003).

  196. [196]. Id.

  197. [197]. Raphael & Stoll, supra note 1212, at 3–8 (discussing the increasing incarceration rate); William J. Stuntz, The Pathological Politics of Criminal Law, 100 Mich. L. Rev. 505,
    566–67 nn.230–31 (2001) (surveying statistics suggesting similar trend in state courts); Ronald F. Wright, Trial Distortion and the End of Innocence in Federal Criminal Justice, 154 U. Pa. L. Rev. 79, 90 (2005).

  198. [198]. See Montré D. Carodine, “The Mis-Characterization of the Negro”: A Race Critique of the Prior Conviction Impeachment Rule, 84 Ind. L.J. 521, 551 (2009).

  199. [199]. Eisenberg & Hans, supra note 7, at 1371–72 (counting “blacks and Hispanics” as minority and excluding from both categories Asian defendants and those described as “other”).

  200. [200]. See Sarah Shannon et al., Growth in the U.S. Ex-Felon and Ex-Prisoner Population, 1948 to 2010, at 7 (unpublished manuscript),

  201. [201]. A minority of defendants, approximately 31.5% of defendants in the NCSC data, unequivocally benefit from the venerated right. See Eisenberg & Hans, supra note 7, at 1371 tbl.1, 1375 tbl.4 (the 31.5% percentage reported above consists of defendants with no record who testify (49) plus defendants with records who testify but the jury did not learn of their record (55) divided by total defendants (330)).

  202. [202]. People v. Zackowitz, 172 N.E. 466, 468 (N.Y. 1930).

  203. [203]. See supra Part II.B.

  204. [204]. A few small states dispense with the prior offender penalty. See Blume, supra note 21, at 483 n.20 (referencing Montana, Hawaii and West Virginia as the only States that prohibit or severely restrict prior conviction impeachment); see, e.g., Haw. R. Evid. 609(a) (“[T]he defendant shall not be questioned or evidence introduced as to whether the defendant has been convicted of a crime ... .”).

  205. [205]. Rock v. Arkansas, 483 U.S. 44, 50 (1987) (citing Ferguson v. Georgia, 365 U.S. 570, 582 (1961)).


Professor, William & Mary Law School.

I wish to thank Adam Gershowitz, Paula Hannaford-Agor, Chris Slobogin, and Neil Vidmar, as well as the participants in both the Vanderbilt Criminal Justice Roundtable and the Neighborhood Criminal Justice Roundtable at William & Mary, for helpful comments on earlier drafts of this paper. The Institute of Bill of Rights Law at William & Mary Law School funded the empirical study reported in this paper; the College of William & Mary Protection of Human Subjects Committee approved the study. Rachel Sollecito provided valuable research assistance.