Certain legal principles have raging within them a kind of civil war. Important but abstract policy concerns pull the doctrine in one direction; common sense and practical experience pull it in another. Casualties ensue.
Perhaps no body of legal doctrine better exemplifies this sort of internal conflict than the general ban on character evidence, as embodied in Federal Rule of Evidence 404(a)(1). That rule states: “Evidence of a person’s character or character trait is not admissible to prove that on a particular occasion the person acted in accordance with that character or trait.” The rule has deep roots in the common law.
Rule 404(a)(1) follows a structure that is common among the exclusionary rules found in section 4 of the Federal Rules of Evidence. It identifies a category of evidence—here, evidence of someone’s character. And it describes a specific forbidden use of that evidence—here, inviting the fact finder to conclude the individual has a propensity to act consistent with that character.
The ban obviously applies to testimony that expressly speaks to someone’s character. So, for example, a prosecutor cannot, in the prosecution’s case-in-chief, call a witness to testify: “I’ve known Willie for many years and he’s criminally inclined.” That testimony comes up to you, grabs you by the lapels, shakes you, and screams: “I’m character evidence!”
But character evidence can also come in less obvious forms. The most common of these is “other acts” evidence. If, for example, a prosecutor offers proof of distinct but similar acts by the defendant in an effort to show the defendant likely also committed the crime charged, then that’s character evidence, too.
“Other acts” evidence does not necessarily come to us announcing that it is character evidence. Nevertheless, if the prosecutor offers such evidence to show that the defendant has a propensity to behave in a certain way, it meets the definition. A prosecutor therefore can’t try to prove that Willie robbed this bank by showing that he robbed 20 others and so has a predisposition to commit such crimes. That’s the reasoning the rule forbids.
Strong policy arguments support our law’s long-standing hostility toward character evidence. Such evidence brings with it the risk that the jury may convict the defendant because it views the defendant as a bad person, rather than because the prosecutor has proved beyond a reasonable doubt that the defendant committed the crime charged. And the jury may not worry about making a mistake, reasoning that there is little or no harm in taking a scoundrel off the street. Of course, such an error puts the wrong person in prison and leaves the perpetrator free to commit more crimes, but those issues may not deter the jury from thinking this way.
Proving character through other acts evidence raises additional concerns. It requires a series of mini-trials regarding each incident. Such a collection of detours may confuse the jury and distract it from its central task. And the exercise will unavoidably eat up trial time, delay the proceedings, and test the typically short attention spans of jurors.
In response to these policy concerns, our law adopts a fiction. It pretends that someone’s character and the acts that reflect it are irrelevant. It presents the defendant to the jury as if he had a clean slate. It asks the jurors to judge him as if he had just hatched from a very large egg.
The notion that the defendant comes to the jury pure and unsullied is what philosophers call a “useful fiction.” It prevents the concerns I noted above and serves the ends of justice. But it’s also a big fiction. It asks us to ignore what our common sense and practical experience tell us. Indeed, pretending character is irrelevant makes for a heavy lift.
Think about it: Every day, people make countless important decisions based on the idea that human beings have propensities to act in certain ways. I ask my Evidence students: Do you think you could argue to the hiring partner at a law firm that they have no right to review your grades and background credentials because that would amount to impermissible propensity reasoning? Or do you think that, on a first date, you could refuse to say anything about your past on the basis that it’s wholly irrelevant? These seem like reliable prescriptions for unemployment and dining alone.
If you want to watch a layperson’s jaw drop in disbelief, try explaining that a jury can’t consider Willie’s 20 other bank robberies as evidence that he had a propensity to rob this one. Tell the layperson that’s the law, and that person will probably say, with Mr. Bumble in Bleak House, that the law’s “an ass.” Little wonder: We’ve thought differently about things for a few thousand years. “Can the leopard change its spots?” the prophet asks in the Book of Jeremiah. “Neither can you do good who are accustomed to doing evil.”
Contemporary behavioral theory suggests that the character evidence rules may, in fact, be onto something. As philosopher Kwame Anthony Appiah discusses in his 2010 book Experiments in Ethics, character turns out to be a less stable and more elusive concept than we tend to think. This, too, aligns with our experience: We’ve all had occasions where somebody unfairly generalized about our character based on an isolated and unrepresentative incident.
Nevertheless, the temptation to consider character is strong, particularly when we remember that we process all evidence through the lens of relevance and that the standard for relevance is low and forgiving. As a general proposition, we default to the idea that the jury should have more information rather than less. And it feels hard to argue that character never sheds any useful light on whether someone did something or that prior acts tell us nothing whatsoever about present ones. Life as we find it in the real world teaches us otherwise.
In certain situations, we’ve concluded that the fiction that drives the character evidence exclusion becomes intolerable. So Federal Rules of Evidence 413, 414, and 415 (and their state counterparts) allow for the admission of propensity evidence against defendants in cases of sexual assault and child molestation. Supporters of these exceptions have argued that they are justified by higher recidivism rates among the perpetrators of these sorts of crimes, but experts disagree over whether the data support that conclusion. My own view is that we have these exceptions because, when it comes to such heinous crimes, we simply don’t have the intestinal fortitude to indulge the fiction on which the character evidence ban rests.
There are other exceptions as well, mostly driven by our understanding of the unique threats to life and liberty posed to defendants in criminal cases. So, under Federal Rule of Evidence 404(a)(2), we allow criminal defendants to put at issue their own character or the character of the victim—for example, to show that they are nonviolent by nature or that the victim was the first aggressor. Out of a sense of fairness, the rules further provide that—if a criminal defendant does so—the prosecutor may respond in kind.
This doctrinal maze includes an additional twist. As noted above, the character evidence rules preclude only the use of other acts for the forbidden purpose; that is, to show someone has a propensity to behave in a certain way. If other acts evidence is relevant to show something else, then the ban does not apply. In short: Evidence not used to show propensity is not character evidence.
Rule 404(b)(2) gives us an inexhaustive list of possible non-propensity uses for “other acts” evidence. It notes that such acts can be used to show things like motive and intent. A prosecutor who wants to offer evidence of prior altercations between the defendant and the victim, for example, isn’t using it to prove that the accused had a propensity toward violence; the prosecutor is using it to show that the defendant had a motive and the intent to punch the victim’s lights out.
Although such uses of “other acts” evidence technically avoid the forbidden propensity reasoning, they raise at least three concerns.
One concern is that the jury will resort to propensity reasoning anyway, will use the evidence to conclude that the defendant is a bad person, and will convict him regardless of his guilt for the crime charged. Of course, a judge may issue a limiting instruction under Rule 105 to try to avoid this problem, but skeptics will question whether such directives do any good. That concern seems especially keen under these circumstances, where propensity reasoning so closely aligns with our common sense and everyday experience.
Another concern is that a prosecutor will claim that she is offering the evidence for a 404(b)(2) purpose when she really isn’t. In State v. Kirsch, 662 A.2d 937 (N.H. 1995), for example, the prosecutor justified evidence of the defendant’s “other acts” on the basis that it went to the “motive” and “intent” of the accused. After careful review, the New Hampshire court concluded that the prosecutor’s argument came down to an allegation that the defendant was the kind of person who had the motive and intent to commit the crime. That’s propensity evidence, no matter how a prosecutor labels it.
The final concern is that some supposedly non-propensity uses of “other acts” evidence have propensity reasoning embedded within them, but it’s not obvious and it’s hard to detect. In this situation, the prosecutor has a good-faith argument that the evidence is relevant to an issue other than character. But a hard look at the rationale reveals that it serves as a kind of Trojan horse, smuggling character evidence into the case without appearing to do so. The allegedly non-propensity use of “other acts” evidence that has most often been accused of this sleight-of-hand is the so-called “doctrine of chances.”
Law students often learn about the doctrine of chances by reading only one case, but it’s a famous and fascinating one: the English court’s decision in Rex v. Smith, 11 Cr. App. R. 229 (1915). The case has attracted a great deal of attention over the past century, and a detailed history of the background facts can be found in The Brides in the Bath, Henry Brooke: Musings, Memories, and Miscellanea (July 28, 2016), and in the sources cited there. For our purposes, a truncated version will do.
In that case, one George Joseph Smith was charged with murdering his fifth wife, Bessie Mundy, by drowning her in a bathtub. Even taken in splendid isolation, the circumstances surrounding her death seemed suspicious: Bessie had inherited a significant amount of money; Smith pressed her to make a will from which he would benefit; he then reported to her doctor that his wife was having “fits”; and, shortly thereafter, he claimed to have found her drowned in her bathtub.
Subsequent events raised additional questions. After Bessie passed, Smith married his sixth wife, Alice Burnham. She similarly made a will in his favor, and he took out an insurance policy on her life. When Smith’s downstairs landlady saw water trickling through her ceiling, she rushed upstairs to find him cradling his wife in the tub, another victim of a tragic “fit.”
About a year later, under the assumed name of John Lloyd, Smith married Margaret Lofty. The morning after the wedding, she went to a solicitor and executed a will leaving everything to her husband. After returning home and writing a letter to her family extolling her groom’s virtues, she made the unfortunate decision to take a bath. In short order, the landlady heard Smith call for help and ran upstairs to find him lifting his drowned wife from the tub. The marriage had lasted roughly 24 hours.
Smith was arrested and put on trial for the murder of Bessie Mundy, the first of his drowned wives. For understandable reasons, the prosecutor sought leave to offer evidence regarding the deaths of Alice and Margaret. For equally understandable reasons, the defense objected.
The trial judge acknowledged that the prosecutor couldn’t offer the evidence in order to show that Smith had a criminal character; that is, that he had a propensity to kill his wives in bathtubs to collect under their wills and insurance policies. But the judge ruled that this prohibition did not prevent the prosecutor from offering the evidence for some other relevant purpose, such as to rebut the defense argument that Bessie’s death was an accident. The judge accordingly allowed the evidence, and Smith was convicted and sentenced to death.
For years, scholars have debated whether such evidence involves the sort of propensity reasoning the rules forbid, even if it doesn’t appear to do so at first glance. Among others, evidence scholar Edward J. Imwinkelried argues that it does not. Professor Imwinkelried notes that the theory of admissibility here, the doctrine of chances, rests on “a solid non-character theory of logical relevance” based on “probability reasoning.” See Edward J. Imwinkelried, A Brief Essay Defending the Doctrine of Objective Chances as a Valid Theory for Introducing Evidence of an Accused’s Uncharged Misconduct, 50 N.M. L. Rev., no. 1, 1, 15 (Winter 2020). In his view, the relevance of the evidence depends not on any assumptions about the defendant’s tendency to behave in a certain way but on the sheer statistical improbability of these three events occurring accidentally. Courts have widely and approvingly cited Imwinkelried’s work on this issue.
A competing view argues to the contrary. Evidence scholar Paul Rothstein is a leading voice among those who contend that the doctrine of chances necessarily entails propensity reasoning, even if not obviously. See, e.g., Paul F. Rothstein, Comment: The doctrine of chances, brides of the bath and a reply to Sean Sullivan, 14 Law, Probability & Risk, no. 1, 51–66 (Mar. 2015). Professor Rothstein’s argument is complex and nuanced and resists easy summary, but a thought experiment he proposes helps explain why he thinks this is so.
Imagine two hypothetical groups of husbands, each with 100 members. Group A comprises some number of husbands (out of 100) who had a rich wife accidentally drown in the tub and who also had other wives drown under similar circumstances. Group B comprises some number of husbands (out of 100) who intentionally drowned their rich wives in the tub and who also had other wives drown under similar circumstances.
For the evidence of other acts to be relevant in this case, Professor Rothstein argues, we must have some reason to think that the number of husbands in group B will be greater than that in group A—but it’s not at all clear that we do. To the contrary, all things being equal, it seems likely that the numbers would be the same or close. After all, why wouldn’t they? To find the number in group B will be meaningfully larger than the number in group A, the jury must import an additional concept: The husbands in group B have a propensity to kill their wives.
I see things differently. From my perspective, the critical point is that the prosecutor’s primary use of “other acts” evidence in these cases is not to prove the charge but to rebut a defense, most commonly a claim of accident. The evidence therefore puts to the jury a relatively simple question: Is the defendant’s argument consistent with your experience of how the world works? Prosecutors raise that sort of question all the time, including with respect to evidence that has nothing to do with the doctrine of chances.
To answer “no” here, the jury doesn’t need to engage in propensity reasoning, explicit or implicit. Nor, for that matter, does it need to conduct a formal statistical analysis. Rather, the jurors just need to conclude that “something’s up”—that, based on their lived experience, “accident” seems like an improbable explanation for the events described. The something’s-up standard meets the relevance test because, again, that test requires so little; the evidence is relevant if it has any tendency to make any fact of consequence to the determination of the case more or less probable. Here, the evidence does so; it raises an eyebrow regarding the defendant’s explanation of what happened.
Of course, “something’s up” doesn’t tell us “what’s up.” The fact that the jury concludes that something’s up in no way relieves the prosecutor of the burden of proving the defendant had the motive, means, opportunity, and intent to commit the crime. And a not insignificant risk exists that when the jury seeks to answer the “What’s up?” question, it will engage in propensity reasoning, just as Professor Rothschild predicts. But, as discussed above, that risk exists even when everyone would agree that the use of the evidence in question does not involve propensity reasoning and is permissible under Rule 404(b)(2).
This debate may seem eye-rollingly academic. It may feel as if scholars have spilled a disproportionate amount of ink over a doctrine that is most closely associated with a 100-year-old case from another country that involves wildly idiosyncratic facts. Every time I teach the doctrine of chances, some tortured student raises a hand and asks if it ever actually comes up in practice.
The doctrine of chances may not be everyday fare for most litigators, but it matters; in fact, it matters a lot. Professor Imwinkelried notes, “In federal practice, Rule 404(b) generates more published opinions than any other provision in the federal rules.” Insofar as the doctrine of chances helps us understand Rule 404(b) more broadly, it makes a critical contribution.
Furthermore, the doctrine of chances gets more airtime in trials today than one might think. Imwinkelried contends that “in the past few decades, the doctrine of chances has become one of the most common weapons for prosecutors seeking to introduce evidence of an accused’s uncharged conduct.” Indeed, the doctrine of chances is alive and well and appearing at a federal or state courthouse near you.
In what kinds of cases is the doctrine used? Well, in the occasional murder case, of course. But Imwinkelried underscores the more common example of child abuse prosecutions. “Numerous courts,” he observes, “have allowed the government to introduce evidence of other injuries the child suffered while in the accused’s custody to negate the accident claim. The courts reason that the larger the number of injuries suffered by the child, the less probable it is that all the injuries are accidental.” In the natural order of things, kids get hurt; but when the number of injuries exceeds our ordinary experience, there are reasons to think that something’s up.
Prosecutors have also frequently invoked the doctrine in arson cases. Consider, for example, People v. Mardlin, 487 Mich. 609, 790 N.W.2d 607 (Mich. 2010). In that case, the defendant in an arson prosecution claimed that he didn’t intentionally set the fire for which he was charged. The prosecution offered evidence showing that, in the 12 years preceding that incident, the defendant had been associated with four other fires—each of which also inured to his benefit.
The Michigan Court of Appeals held that the trial court erred in admitting the evidence, but the Michigan Supreme Court reversed, noting that the court of appeals went astray in concluding that a “high degree of similarity” between past acts and the crime charged is necessary for evidence of those acts to be admissible. “To the contrary,” the Supreme Court declared, “unusually frequent events—and particularly purported accidents—associated with the defendant and falling into the same general category of incidents are admissible to disprove lack of accident or innocent intent with regard to the charged event.”
The Michigan Supreme Court’s opinion contains a detailed analysis of the doctrine of chances and makes for fascinating reading. The court stressed the commonsense basis of the doctrine, among other things echoing this memorable quotation from a Seventh Circuit decision: “The man who wins the lottery once is envied; the one who wins it twice is investigated.” And the court emphasized the special utility of the doctrine in arson cases, “where unusually frequent individual fires, which could appear to be accidents when viewed in isolation, may constitute the most probative objective evidence that the defendant intentionally set the fire underlying the arson charge.”
Although courts have successfully and sensibly applied the doctrine of chances in many contemporary cases, some decisions send warning signals. In a few cases, the problem wasn’t that the court applied the doctrine of chances—the problem was that it didn’t. Courts have sometimes invoked other non-propensity rationales where the doctrine of chances seems like a significantly better fit. People v. Catlin, 26 Cal. 4th 81, 26 P.3d 357 (Cal. 2001), as modified (Sept. 26, 2001), may present an example of such a case. In Catlin, the defendant was charged with the murder of his fourth wife and his mother, both of whom died of poisoning by paraquat, a highly toxic chemical used as an herbicide. The prosecution offered evidence indicating that the defendant’s fifth wife had died of the same cause. The prosecutor also showed that paraquat poisoning is rare, that the defendant had access to this poison through his work, and that the defendant benefited financially from all of these deaths.
This seems like a good case for applying the doctrine of chances: To the extent the defendant claimed the poisonings were accidental and coincidental, our common sense and experience argue otherwise. In admitting the evidence, however, the trial court focused primarily on arguments that the evidence showed a “common scheme or plan” and “identity.” Both of these are non-propensity rationales specifically identified in Rule 404(b)(2),but neither aligns well with these facts.
The notion that this evidence reflects a common scheme or plan on the part of the defendant seems like a stretch. The defendant’s conduct in Catlin looks less like some grand, unified, carefully plotted enterprise and more like a series of similar acts done for similar reasons. If that’s enough to steer around the character evidence ban, then it would seem that a wily prosecutor could do so in any case in which the other acts bear any similarity to the charged one. Affording prosecutors that much latitude could effectively gut the rule.
It seems even harder to understand how the evidence shows “identity.” Perhaps the argument is that the defendant must be the one who poisoned the victim because the odds of accident are remote. But that sounds like a doctrine of chances argument gone wrong, one that asks the jury to take the improbability of accident as affirmative proof of the defendant’s guilt. It confuses “something’s up” reasoning with “what’s up” reasoning.
In other contemporary cases involving the doctrine of chances, the warning sign comes from a prosecutor’s effort to invoke the doctrine even though the evidence was pretty obviously being used to show the defendant’s propensity to act in a certain way. As discussed above, this problem can arise with respect to any supposedly non-propensity use of “other acts” evidence; there’s nothing special about the doctrine of chances in this regard. As we might expect, some courts have proven very good at spotting the forbidden propensity reasoning through the camouflage; others, not so much.
Consider, for example, State v. Skillicorn, 367 Or. 464, 479 P.3d 254 (Or. 2021). In that case, the defendant was charged with first-degree “criminal mischief” and other crimes based on an incident of reckless driving. The prosecutor’s theory of the case was that the defendant intentionally drove his truck into his girlfriend’s car after a heated argument. The defendant admitted that he hit the car but claimed he did so because he lost control of his truck after it malfunctioned.
To rebut this claim, the prosecutor offered evidence that the defendant had driven recklessly after a prior argument with his girlfriend. The prosecutor contended that the evidence showed that when the defendant got angry, he “acted out,” belying his claim that this collision was accidental. Over the defendant’s objection, the trial court admitted the evidence; the jury returned a guilty verdict.
The appellate court reversed. Noting that “[i]t is often difficult to apply the doctrine of chances correctly,” the court observed that “just because a factfinder is using a series of events to assess the odds that the charged event was an accident does not mean that the factfinder is employing doctrine-of-chances reasoning. The factfinder could be employing propensity reasoning.” The court concluded that this is exactly what happened here: Under the prosecutor’s theory, the relevance of this “other acts” evidence depended “solely on a propensity inference—that if the defendant drove dangerously once, he had a propensity to drive dangerously again.”
The appellate court’s decision is noteworthy for the depth of its discussion of the doctrine of chances. That doctrine, the court observed,
is a theory of relevance based on the objective improbability of the recurrence of uncommon events. The idea underlying the doctrine is that, at some point, it becomes unlikely that each event in a series of similar events can have the same uncommon cause; therefore, if the number of events in a series claimed to have the same uncommon cause exceeds the number that can reasonably be expected to have that cause, a factfinder can infer that not all of the events actually have that cause.
The prosecutor’s argument in this case didn’t fit that description; it was simply an argument that because the defendant had deliberately driven recklessly before, it was more likely that he deliberately drove recklessly on this occasion.
The final warning sign looks very much like the one that worries Professor Rothstein: In some cases, courts have failed to notice propensity reasoning that is embedded in a doctrine of chances argument but embedded in a subtle and non-obvious way. Propensity reasoning came flying in under the judicial radar, without so much as a blip on the 404(a)(1) screen.
The high-profile case of People v. Spector, 194 Cal. App. 4th 1335, 128 Cal. Rptr. 3d 31 (2011), as modified (May 5, 2011), as modified on denial of reh’g (May 26, 2011), may provide an example. In that case, famed record producer Phil Spector was charged with shooting and killing one Lana Clarkson at his California mansion. Spector claimed that Clarkson had grabbed a gun he was holding, put it in her mouth, and pulled the trigger—either accidentally killing herself or in order to commit suicide.
To rebut Spector’s claim, the prosecutor introduced evidence from five women concerning seven incidents in which Spector had assaulted them with a weapon. The prosecutor offered a variety of theories in favor of the admissibility of this evidence, including the doctrine of chances. The trial court allowed the evidence over Spector’s objection, and the appellate court affirmed. The courts’ endorsement of the doctrine of chances argument is perhaps understandable, given that the facts sound vaguely reminiscent of the serial crimes against women in Rex v. Smith.
A close look at the appellate court’s reasoning, however, raises questions about applying that doctrine to these facts. The court declared:
To the extent the circumstances of Clarkson’s death were similar to those prior armed assaults, the latter evidence, without any improper reliance on Spector’s bad character or propensities, tended to prove the objective improbability that Clarkson had either committed suicide or killed herself accidentally. This is because the evidence tended to show, by operation of the doctrine of chances, the unlikelihood that this time it was the woman, not Spector, who reached for a gun.
Reasonable minds can disagree, but this sounds disquietingly like propensity reasoning. Who reached for the gun? Spector did. Why do we think he’s the one who did it? Because that’s what he has a history of doing; because that’s who he is; because that’s his character.
If all of the other incidents offered into evidence involved shootings that Spector claimed were accidental, then this might be a good case in which to apply the doctrine of chances. Such a coincidence of extraordinary events would trigger the sense of implausibility on which the doctrine depends. Our common sense and experience would tell us that something’s up.
But those weren’t the facts. Each of the prior incidents differed from the others in meaningful ways, and Spector didn’t try to explain them all away by claiming they were accidental. The notion that those acts rendered it implausible that Clarkson shot herself only seems to make sense if we believe they show that Spector had a propensity to engage in such conduct. According to the character evidence ban, that’s exactly how we’re not supposed to think about them.
A different argument addressed by the court may have greater appeal. The court pointed out that Spector’s defense put the character of the victim at issue by suggesting that Clarkson was reckless or self-destructive. Under Rule 404(a)(2)(B), this defense arguably opened the door to evidence of Spector’s character, too. On this analysis, the evidence was not only admissible; it was admissible to show Spector’s propensity toward recklessness and violence.
I’ll end with a thought I share with my Evidence students when we wrap up our discussion of the doctrine of chances. I tell them that the doctrine of chances plays the odds. And I tell them that the overwhelming odds in cases like Rex v. Smith and People v. Spector are that the court is going to figure out some way to deem this kind of evidence admissible.
The court has to do so because in those cases common sense and experience win their war with theory and policy. It has to do so because in those cases the force of the pragmatic “something’s up” standard becomes morally and intellectually irresistible. It has to do so because in those cases the useful fiction of the character evidence ban darkens into an intolerable lie.
Their job, I tell them, is to figure out how the court is going to do it.
