On January 21st of 1959, a freighter travels down the darkened waters of the Buffalo River, headed with the river’s westerly flow.  It’s 10:45 p.m.  She moves effortlessly with the current, her boilers completely cool.  She moves without propulsion, without a captain, without direction.  She is the MacGilvray Shiras, more than 400 feet in length and filled with wintering wheat.  On this Wednesday night, she’s been set adrift by a broken ice jam, let loose upriver by a rainy thaw. The Shiras is captained only by the current and towing the fleet of thick and jagged ice that had just pulled her from her moorings.  In a few minutes from now, her stern will grab the bow of the Michael K. Tewksbury, pulling it from the dock like a sword from its scabbard.  By the will of the Buffalo River, they are both heading for the open waters of Lake Erie.  It’s nearly 11:17 p.m.  Less than 100 yards downriver, late-arriving bridge tenders scramble upon the Michigan Avenue Bridge like ants.  Their eyes darting upriver, down to a set of controls that will raise the bridge, and back upriver again.  Two massive ships are emerging from the darkness ahead of them, a dusky white blizzard of ice on the water behind them.  The bridge is down.  It won’t be raised in time.[1]

Oh, how I love teaching this case in my Torts class!  It’s a jaw-dropping cascade of errors, from the negligent mooring of the Shiras, to the failed attempts to drop an anchor onto the riverbed, to the bridge tenders who weren’t tending the bridge.  These two drifting ships smash the unraised bridge and fray its beams into ribbons.  They wedge into a perfect V—from shore, to bridge, to shore—that collects all the flowing ice, piling it from the depths up into the peaks of a jagged dam.  The Buffalo River floods for miles, willing itself over the land to continue its trek to Erie.

Everyone sues everyone.

Given the monthly bent of this blog, you might imagine how excited I was when I learned that an entire study borrowed from this set of gloriously horrendous facts in order to test one of our brain’s most notorious biases.[2]

Let’s change the facts just a bit and think about the case being tried to a jury.  Let’s say that only the city of Buffalo had been sued because, months before the winter, the city decided it didn’t need a bridge tender at all.  The city looked at the likelihood of rising water producing dams at the bridge and decided that it didn’t need to spend an estimated $100,000 to hire bridge tenders for the winter months.  When the jury gets the case, they see very clearly the consequence of failing to hire the bridge tenders.  They see all the tattered scraps of bridge submerged in the Buffalo River and the snarl of river traffic that results.  The jury is then asked to decide if the failure to hire tenders for the winter was reasonable care given the chances of ice jams.  To the jury judging with the bias of hindsight, however, the “chance” of an ice jam at the bridge seems to be 100%.  It has already happened.  But what if the actual chance of a catastrophic, weather-fueled flood by ice jam was only, say, 4%?  The only thing known to the city when it decides to pass on bridge tenders is how unlikely a floating battering-ram of river boats and ice seems to be.

And so we arrive at the study.  In this carefully crafted experiment, three groups of participants were shown three very similar slide-show videos.  The first group saw a slide-show video depicting in-person, pro-and-con arguments made to the city’s urban planning committee on whether to hire bridge tenders.  In this presentation—as well as all the others—experts testified for each side on the incidences of high water on the river. These experts didn’t give percentages of the likelihood of dangerously high water on the river.  Instead, they relayed the number of times the river had risen to dangerous levels during a span of decades. The testimony of the experts is anecdotal and not drastically different from one another. The second group saw the same photos of the same people in the first narrated slide show, but the proceeding is now presented as a trial. This time, the city commissioner from the committee meeting was dressed as a judge.  The narrations of the people arguing pro-and-con in the first slide show video were tweaked slightly to become closing arguments in a trial. This time, the characters in the slide show argued as lawyers for and against tort liability for an ice flow that had already destroyed the bridge.  The experts gave the same testimony as before. The third group saw this same courtroom-trial version of the slide show, but in this set of slides, the judge appeared at the end and cautioned the jury (the test subjects of the study) on the problems of predicting the likelihood of an event once you know the event has already happened.[3]  These instructions are called “debiasing” instructions, meant to counteract the effect of the so-called “hindsight bias,” described in the footnote below[4]

You should be able to see what the scientists were after here.  How many people would say the city should hire bridge tenders while guessing whether an accident might happen? How many would say the city should have hired tenders once they already knew an accident happened?  And, finally, whether a jury’s view of the city’s liability would change with some counseling on the problems of hindsight judgments?  Can you see also the stakes involved for plaintiffs’ and civil defense lawyers?  In theory, and if justice is blind, what is or is not ordinary care for bridge tending should be the same before and after a bridge is smashed to bits.  If the chances of a smashed bridge are determinably low, it’s either ordinary care to forego tending the bridge or it isn’t.  What actually happens later is irrelevant.

So, how did the experiment turn out?  Let’s ease our way back into the icy waters of the Buffalo River.  First, the two groups viewing the accident as jurors in hindsight gave significantly higher estimates of the chances of a high-water flood event than the foresight group.[5]  While just 24% of the test subjects without hindsight knowledge said the city should hire tenders for the upcoming winter, a whopping 56.9% of those in both post-accident hindsight groups said the city should have hired tenders.[6]  (And with those two results in view, all of our civil defense lawyers reading this are now scooping up their sagging jaws.  Perhaps they’re telling themselves they could have predicted that disparity.)  The real question, then, is how the third, “debiased” test group fared on the related question of liability.  Would the special jury instructions from the scientists purge the hindsight bias and normalize the liability judgments of the other hindsight jurors?  No.  The debiasing instruction had no significant effect.  57.7% of the un-debiased jury found the city liable.  56% of the debiased jurors cautioned about judging with hindsight also ruled against the city.[7]

The authors expected hindsight bias.  They predicted that the debiasing would work better than it did and they were wrong.  These things can happen, right?  However, nobody involved—not with three crystal balls, two decks of tarot cards, and a partridge in a pear tree—could predict this last finding of the study.  All the study participants were asked to rate the performance of the “con” arguers; that is, the city manager opposing the hiring of tenders and the city’s defense attorney arguing against liability.  I’ll let you read this portion of the study for yourself:  “…hindsight participants rated the opponent/defense attorney’s performance significantly higher than did the foresight participants…and [this result] is in the opposite direction from what the hindsight bias might predict.  Hindsight participants felt that the defense attorney did a better job, despite expressing more disagreement with [the defense attorney’s] position.”[8]  The authors were at a loss on how to explain this result.[9]

There is enough data on hindsight effects out there to sink both the Shiras and the Tewksbury to Erie’s depths.  As far as hindsight bias relating to trial outcomes, however, the authors cite one article by scholars who proposed bifurcating a trial in such a way that a jury would judge negligence without knowing about the actual harms that resulted.[10]  As trial lawyers—as plaintiffs’ lawyers at least—we shudder at the thought of telling the jury a story of the night of January 21st, 1959 in which we can hardly tell the jury about the night of January 21st, 1959.  No freighters.  No ice.  No harried race to raise an untended bridge.  This bifurcation idea might raise for you interesting questions of fundamental fairness which you can ponder by the fireside this December.  With that, I’ll bring this ship back to the dock.  From me and Jules to you and yours, we wish you a happy and healthy holiday season.


[1] See Petitions of the Kinsman Transit Company, 338 F.2d 708 (2nd Cir. 1964) for a fuller and less-stylized recitation of the facts.

[2] Kim. A. Kamin and Jeffrey J. Rachlinski, Ex Post ≠ Ex Ante: Determining Liability in Hindsight, 10 Law and Human Behavior, no. 1, 1995, at 89-104, https://psycnet.apa.org/doi/10.1007/BF01499075

[3] See Id. at 94-97 for a thorough account of the experimental methods employed.

[4] Id. at 92.  The hindsight bias is a term that labels the human tendency to see events as more predictable after the event has already happened.

[5] Id. at 98.

[6] Id.

[7] Id.

[8] Id. at 100.

[9] Id.

[10] Id. at 92, citing D.B. Wexler and R.F. Schopp, How and when to correct for juror hindsight bias in mental health malpractice litigation: Some preliminary observations, 7 Behavioral Sciences and the Law, 1989, at 485-504.

Leave a Comment