Hindsight bias is a widespread and often unavoidable human fallacy. Relative to foresight observers, those with knowledge of the outcome of an action (e.g. in hindsight) believe that the given outcome has a much higher likelihood of happening. Hindsight bias plays a crucial role in civil trials, in which the defendant is often disadvantaged because jurors know the negative outcome of the defendant's behavior and are thus more likely to think he should have known about the risks associated with his behavior. Bifurcation (separate liability and damages phases) is proposed as a way to reduce or eliminate hindsight bias in the courtroom. The question though, is does bifurcation eliminate or reduce hindsight bias? This article presents a quick overview on hindsight bias and some detail on how hindsight bias applies to jurors in the courtroom. Finally, I discuss two empirical studies that tested the effectiveness of bifurcation as a tool in reducing hindsight bias in jurors.

Hindsight Bias
Fischoff (1975) identified a phenomenon in which individuals who were asked to make a judgment in hindsight differed markedly in their judgments than other individuals asked to make the same judgment in foresight. Specifically, he observed that individuals with outcome knowledge tended to exaggerate their ability to predict the inevitability of an event's outcome, such that they overestimated both the likelihood of the known outcome occurring and the ability to foresee the outcome (Stallard & Worthington, 1998). Fischoff coined the term "hindsight bias" to explain this human fallacy. He concluded that despite continually falling into the trap of hindsight bias, individuals are largely unaware of the effect that the outcome knowledge has on their perceptions. And even when they are told that they have this outcome knowledge and that they should try to ignore it, individuals are unable to avoid hindsight bias. Perhaps even more disheartening, both novices and experts alike (e.g. judges in court cases, surgeons in medical malpractice hearings) fall prey to hindsight bias (Harley, 2007). In addition, Fischoff (1975) found that we are more likely to remember event information that is consistent with the outcome and minimize, distort, or even forget information that is inconsistent with the outcome. Therefore, people pay much more attention to information that fits with the reported outcome, and marginalize information that is inconsistent with the reported outcome.
Hindsight Bias in the Courtroom
Jurors typically face the classic hindsight task. Jurors are put in a situation in which they know the outcome of some event, but are asked to judge whether the defendant's behavior at the time leading up to the accident was negligent. Jurors' duty is to assess whether the damage or injury caused by the defendant was foreseeable, and whether the defendant was aware or should have been aware of the risk (Harley, 2007). Put another way, when judging negligence the law asks jurors to attempt to recover a naive, ex ante view of once future events and to judge the defendant's conduct in light of what he should have known in the past (Hastie, Schkade, & Payne, 1999). Given what has been documented about hindsight bias, this should prove to be a much more difficult task than intended. Jurors are aware of the outcome when they make their determinations of negligence for the defendant. Jurors know how bad the injuries to the plaintiff are. Even in a bifurcated trial the jurors know that some sort of negative outcome occurred or there would be no trial. So the fact that jurors know the plaintiff has suffered an injury and that the plaintiff believes the defendant caused that injury makes it difficult for jurors to set aside that damaging information when judging whether the defendant's actions were negligent.
Hastie and colleagues (1999) examined the effect of hindsight bias in the courtroom. They put some participants in a foresight condition in which a railroad company was given a ban and told to stop operations, and it was up to the participants to decide whether the company should proceed anyway despite the ban. The rest of the participants were put in a hindsight condition in which the railroad company decided to continue operations despite the ban. A train crashed and spilled toxins into the surrounding ecosystem, causing extensive damage. Sixty-seven percent of hindsight participants said that the railroad company's actions were negligent, whereas only 33% of foresight participants said the railroad should not be allowed to operate. Thus, most foresight participants thought the company should continue operations, even though most hindsight participants thought that those actions were negligent. Hindsight participants were exposed to a negative outcome which made the defendant's behavior seem more negligent than it appears in foresight. In addition, foresight participants on average predicted a .34 probability of an accident, whereas hindsight participants on average predicted a .59 probability of an accident. Thus, the negative outcome information given to hindsight participants made it seem likely that an accident would happen, whereas those foresight participants who had not read about an accident were not prone to think that an accident was imminent. In this study, the authors found hindsight effects, even though hindsight participants were not exposed to any new ex post information regarding the defendant's negligence that foresight participants were not given.
This finding by Hastie and his colleagues is not an uncommon finding. The authors found that the hindsight participants, who are akin to jurors, were not good at setting aside their outcome information when asked to make negligence decisions and estimates of probabilities, even when they were instructed to do so. This is a prime example of a phenomenon known as "fusion", which is when jurors use legally inappropriate information (e.g. outcome information such as the extent of the damage) when making an unrelated legal decision (e.g. negligence) (Ellis, 2002). Thus the courtroom is perhaps a flawed venue for making determinations of negligence because jurors are put in the privileged spot of knowing outcome information that the defendant could not possibly take into account at the time of the accident. But there is no feasible or practical way to eliminate this problem. Bifurcation has been one of the primary methods proposed to attempt to lessen, if not entirely eliminate, hindsight bias in the courtroom.
The typical civil trial is presented to the jury in a unitary format, in which the jury hears liability-related evidence (e.g. the defendant's conduct) and damages-related evidence (e.g. the outcome of the accident including the plaintiff's injuries) all at once. The jury then makes its decision regarding liability. If the jury finds the defendant liable, jury members then discuss appropriate damages amounts. The criticism often raised with unitary trials is that they bias the liability verdict towards the plaintiff and away from the defendant. This is because the damages-related evidence in a unitary trial is likely to be strongly pro-plaintiff. It is the presence of this evidence that often helps the plaintiff win on liability (Horowitz & Bordens, 1990).
In contrast to the unitary format, in a bifurcated format the jury first hears only liability-related evidence (e.g. the defendant's conduct) and then makes its decision on liability. Only if they find the defendant liable do they then hear damages-related evidence (e.g. the extent of the negative outcome). Advocates of bifurcation claim that it reduces the probability that jurors will utilize damages-related evidence when deciding liability (Horowitz & Bordens, 1990). In addition, bifurcation helps by separating the two components of the case to facilitate independent liability and damages decisions by reducing confusion about which evidence is to be used for which decision and by minimizing the biasing effect of evidence regarding one component on the decision regarding the other (Wissler, Rector, & Saks, 2001). Hence, it reduces fusion, which may be due to hindsight bias. So the empirical question is then, do bifurcated trials reduce hindsight bias and help defendants prevail on liability?
Both Horowitz and Bordens (1990) as well as Smith and Greene (2005) found that defendants were more likely to prevail in bifurcated than in unitary trials. So it certainly seems possible then that bifurcation is an effective tool to reduce hindsight bias. While it is a worthy goal to reduce hindsight bias, it would be a much better goal to eliminate it entirely. In explaining how bifurcation helps the defendant win on liability, Smith and Greene (2005) conclude that bifurcation eliminates the possibility of hindsight bias because it removes from jurors' consideration the very information that biases their decisions in the first place (e.g. outcome-related information).
But is the above assertion premature? In a bifurcated trial, jurors are aware that some negative outcome has occurred, even if they do not know the full extent of it. Why else would they be called to serve on a jury? They know that the defendant's actions allegedly led to some sort of negative outcome. Since they know that a negative outcome occurred, hindsight bias might not be eliminated but instead merely reduced in a bifurcated trial. In order to test whether bifurcation actually eliminates hindsight bias, a bifurcated trial condition must actually be compared to a foresight condition, which would be a true baseline condition because foresight individuals do not have any outcome knowledge and hence no propensity to be affected by hindsight bias.
Does Bifurcation Actually Work?
Two studies were conducted to test whether bifurcation reduces or eliminates hindsight bias. The study materials were adapted from Kamin and Rachlinski (1995). In the first study, college students read one of three versions of a civil trial transcript. In the foresight condition, participants acted as a town council member and read about a town council administrative hearing in which the town was deciding whether or not to enact a law to require the employment of a bridge operator during the winter months. Specifically, the participants were asked about the probability of a preventable flood in any given year as well as whether they would vote to enact the law. In the hindsight bifurcated condition, participants acted as mock jurors in a civil trial and read a transcript in which the town was sued by a bakery owner for not hiring the bridge operator, and as a result the river overflowed and destroyed his bakery. Jurors in the bifurcated condition only read about liability-related information and were not exposed to any information about how severe the outcome of the flood was. In the hindsight unitary condition, participants again acted as mock jurors in the same civil trial. The only difference was that these jurors were exposed to outcome-related information, and thus learned about the full extent of the outcome. Jurors in the two hindsight conditions were asked about the probability of a preventable flood in any given year as well as whether they found the city liable.
Although there were no significant differences for the probability of a preventable flood between the three different conditions, there were interesting differences in the dichotomous liability question – the question that would determine liability in a civil trial. Mock jurors in the hindsight unitary condition almost uniformly (95%) found the city to be liable. Mock town council members in the foresight condition were much more lenient on the city. Only 41% thought that the city should enact the law to require the employment of a bridge operator. Thus, the majority of foresight participants did not agree to hire the bridge operator, something that most hindsight unitary participants found negligent. The critical question is, what did the hindsight bifurcated participants think?
While there was a significant difference between the hindsight unitary and hindsight bifurcated conditions, there was not a significant difference between the hindsight bifurcated and foresight conditions. Only 54% of hindsight bifurcated participants found the city to be liable for not hiring a bridge operator. Since there was a statistically significant difference between the hindsight unitary and the hindsight bifurcated conditions but not between the hindsight bifurcated and the foresight conditions, these results suggest that bifurcation actually eliminated hindsight bias. Mock jurors in the hindsight bifurcated condition were not as influenced by the outcome information (since they did not learn the full extent of it as hindsight unitary mock jurors did), and consequently they were less inclined to find the defendant liable.


  Foresight Condition Hindsight Bifurcated Condition Hindsight Unitary Condition
Decisions Finding the City Liable (%) 41%a 54%a 95%b


Table 1. Percentage of participants across conditions finding the defendant liable. Note that the difference between the first two conditions is not statistically significant, whereas the third condition is statistically significantly different from the first two conditions.
After finding that bifurcation effectively eliminated hindsight bias, the next question is what is the mechanism or process through which bifurcation reduces the bias? Several researchers have found that more severe injuries to the plaintiff (e.g. more severe outcomes) lead to a higher likelihood of liability verdicts against the defendant (Bornstein, 1998; Greene, Johns, & Bowman, 1999). In order for a case to actually go to trial, the injury to the plaintiff or the damage to a property is likely to be pretty severe. So in a unitary trial the outcome that jurors hear about has the potential to be bad. And since severe injuries exacerbate hindsight bias compared to milder injuries, a unitary trial is likely to lead to greater hindsight effects and hence a higher percentage of verdicts against the defendant. In a bifurcated trial, on the other hand, jurors know nothing about the severity of the outcome. If jurors are imagining a less severe outcome in a bifurcated trial, then it follows that outcome knowledge would have less of a detrimental effect on liability verdicts, leading to smaller hindsight effects and hence fewer liability verdicts against the defendant.
A second study was conducted to test this assumption. The goal was to manipulate the severity of a flood in the minds of the mock jurors. Mock jurors were placed in one of two conditions. In the "typical" condition, mock jurors first read a short cover story about a flood that resulted in minimal damage. They then read the bifurcated trial transcript from the first study. Before making their liability verdicts, they were asked to write about the "most likely" outcome that would have happened to the plaintiff's store as a result of the flood. In the "severe" condition, mock jurors first read a short cover story about the recent Nashville floods and the detrimental impact they have had on that city. They then read the bifurcated trial transcript from the first study. Before making their liability verdicts, they were asked to write about the "most severe" outcome that would have happened to the plaintiff's store as a result of the flood in the transcript.
Significant differences were found between the "typical" and the "severe" conditions. Whereas only 52% of mock jurors in the "typical" condition found the city liable, 76% of mock jurors in the "severe" condition found the city to be liable for failing to hire the bridge operator. Therefore, it appears that when jurors are imagining a milder outcome, they are less inclined to find for the plaintiff on liability than when they are imagining a more severe outcome. This implies that bifurcation may be effective in reducing hindsight bias and hence leveling the field between the plaintiff and the defendant because jurors may not be imagining as grave of an outcome as the one they may hear about in a unitary trial. But when jurors are induced to imagine a severe outcome in a bifurcated trial, verdict rates rise to more closely resemble those from a unitary trial.
Closing Remarks and Recommendations
Despite its potential for reducing the amount of liability verdicts against the defendant, not much empirical research has focused on the effectiveness of bifurcation at meeting its intended goal. The aim of bifurcation is to reduce hindsight bias, which is what often hurts defendants. The research reviewed above demonstrates that bifurcation essentially eliminates hindsight bias so that liability verdicts were not statistically significantly different from a foresight baseline. Also, the mechanism through which bifurcation works is potentially through the less severe outcomes that jurors are imagining in a bifurcated trial.
The advice for attorneys is quite simple. A civil defense attorney should advocate for a bifurcated trial, unless of course if liability evidence is very strong against the defendant in which case the defendant will likely lose on liability no matter which trial format is used. Bifurcation will reduce, or even eliminate, jurors' hindsight bias and thus lead to less liability verdicts. Furthermore, during the bifurcated trial the defense attorney should carefully attempt to create expectations of mild outcomes that could have happened as a result of the accident. On the flip side, the plaintiff's attorney should advocate for a unitary trial. It will exacerbate hindsight bias and thus lead to more liability verdicts against the defendant. During the bifurcated trial the plaintiff's attorney should carefully attempt to create expectations of severe or extreme outcomes that could have happened as a result of the accident.
Matt Groebe, M.A. is a graduate doctoral student in social psychology at Miami (OH) University in Oxford, Ohio. His research interests are in the areas of juror and jury decision-making, small groups research, and other legal applications of social sciences research. He hopes to enter the field of trial consulting after obtaining his doctorate degree. In his spare time, he likes working out, playing with his dog Wrigley, and watching his beloved Chicago sports teams.
Bornstein, B. (1998). From compassion to compensation: the effect of injury severity on mock jurors' liability judgments. Journal of Applied Social Psychology, 28(16), 1477-1502.
Ellis, L. (2002). Don't find my client liable, but if you do…: defense recommendations, liability verdicts, and general damage awards. Unpublished doctoral dissertation. University of Illinois at Chicago.
Fischhoff, B. (1975). Hindsight is not equal to foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance, 1(3), 288-299.
Greene, E., Johns, M. & Bowman, J. (1999). The effects of injury severity on jury negligence decisions. Law and Human Behavior, 23(6), 675-693.
Harley, E.M. (2007). Hindsight bias in legal decision making. Social Cognition, 25(1), 48-63.
Hastie, R., Schkade, D.A., & Payne, J.W. (1999). Juror judgments in civil cases: Hindsight effects on judgments of liability for punitive damages. Law and Human Behavior, 23(5), 597-614.
Horowitz, I.A. & Bordens, K.S. (1990). An experimental investigation of procedural issues in complex tort trials. Law and Human Behavior, 14(3), 269-285.
Kamin, K.A. & Rachlinski, J.J. (1995). Ex post does not equal ex ante: Determining liability in hindsight. Law and Human Behavior, 19(1), 89-104.
Smith, A.C. & Greene, E. (2005). Conduct and its consequences: Attempts at debiasing jury judgments. Law and Human Behavior, 29(5), 505-526.
Stallard, M.J. & Worthington, D.L. (1998). Reducing the hindsight bias utilizing attorney closing arguments. Law and Human Behavior, 22(6), 671-683.
Wissler, R.L., Rector, K.A., & Saks, M.J. (2001). The impact of jury instructions on the fusion of liability and compensatory damages. Law and Human Behavior, 25(2), 125-139.