The Challenge of Measuring Botched Executions
We don't know how big the problem is—and that's part of the problem.
From its first use onward, lethal injection has presented troubling signs that it wasn’t meeting its promise of what Justice Antonin Scalia called a “quiet“ and even “enviable“ death. Exactly how prevalent the problem is, however, has been a subject of serious debate. From advocates’ insistence that only 1% of lethal injection executions are botched, to experts’ calculations as high as 70%, it’s difficult to pin down exact numbers. There have been three major attempts to quantify the harm inflicted by lethal injection, each with its own limitations.
What Counts As a “Botched“ Execution?
While the Roberts Court has routinely deluded itself about the suffering involved in lethal injection, they’re correct that it may be impossible for execution methods to completely avoid pain. Lethal injection necessarily involves some pain; the “liiiiiiittle poke“ people like me tell you about in the back of an ambulance feels a lot bigger when you’re on the other end. Yet no one would argue this makes lethal injection an especially inhumane method of execution, and it’s likely less so than the (equally necessary) tightness of a coarse rope around one’s neck.
There’s also the inherent problem of measuring how much pain actually happens; as anesthesiologist Joel Zivot (whose work in the field we’ll talk about later) told me, “The easiest pain to bear is someone else's. Pain is subjective and can't be measured objectively.” The fact that we’re not even measuring that subjective pain (the people who could tell us about undergoing the procedure are dead at the end of it) but rather looking for objective proxies for it compounds the issue. What constitutes a “botched“ execution, therefore, will always be subject to debate.
Studies from Thiopental Were Alarming but Flawed
Early attempts to quantify suffering in lethal injection produced jarring results. A research letter by Korianis et al, printed in Lancet in 2005, found that 43% of autopsied lethal injection subjects had insufficient serum levels of thiopental to guarantee unconsciousness; case law at the time required that lethal injection protocols essentially guarantee unconsciousness outside an “isolated mishap.”1
The following issue featured a letter to the editor2 by Mark Heath et al, an anesthesiologist whose work on the definition of consciousness was cited by the original letter. Heath noted that the prior study failed to account for how postmortem redistribution of thiopental (its absorption by other tissues out of the bloodstream after death) would happen in the often lengthy times between a prisoner’s execution and autopsy; thiopental is highly “lipophilic“ and readily dissolves in fat. Though Heath praised the letter’s highlighting other issues with lethal injection—and later testified he had never seen a lethal injection protocol he could certify was humane—he concluded that the earlier letter’s data couldn’t support the authors’ conclusions.3
At any rate, these conclusions are no longer relevant to contemporary executions, which do not rely on thiopental—though its supposed reliability was favorably cited in the controlling case law.
The Sarat Study Presents Valuable but Vague Data
Amherst4 law professor Austin Sarat, quite possibly the godfather of botched execution studies, reviewed publicly available data, witness statements, and media reports surrounding lethal injection executions between 1980 and 2010, concluding that more than 7% were botched. Sarat’s findings suggest that lethal injection is far and away the most botched execution method (the historical average is a little over 3%) even before his caveat:
“identifying a botched lethal injection is somewhat problematic because the medicalization of the process and the three-drug protocol5, which until recently has been the standard, work to prevent the body from registering signs of suffering.”
Sarat’s methodology, however, casts a fairly wide net: he considers any presence of “unanticipated problems or delays that caused, at least arguably, unnecessary agony for the prisoner or that reflect gross incompetence of the executioner” to be a “botched” execution. This vague criteria has led to criticism about some of his case studies, some fairer than others. No one is speaking of a death chamber curtain opened too early when they refer to botched executions; likewise, merely comparing the timelines of physician-assisted suicide and lethal injection paints a very narrow picture of what’s going on in the latter.
The Zivot Autopsy Study Presents the Most Thorough—And Most Alarming—Results
Dr. Zivot later compiled his own study, combining Sarat’s thoroughness with the attempted medical rigor of the earlier thiopental study. The resulting 2022 preprint examines 43 autopsies of lethal injection subjects from eight states—a slightly smaller sample than the 49 used in the thiopental study, but one that included a much wider variety in execution cocktails. Moreover, Zivot turned from signs of anesthetic in the blood to physiological signs consistent with suffering—in this case, pulmonary edema, the buildup of fluid in the lungs that makes breathing difficult if not impossible.6
The results were alarming: 82% of executions involving midazolam (a benzodiazepene that’s still the primary sedative in six states despite known deficiencies) and 66% of those involving pentobarbital (a barbiturate similar to thiopental) resulted in the development of pulmonary edema, with the lungs reaching a mean weight of nearly 150% the expected value. Moreover, “froth“ was discovered in the airways of several subjects, a finding only possible if the subjects were struggling to breathe.
The numbers are staggering, but don’t necessarily lead to a conclusion that all positive cases were botched executions—a conclusion which, for his part, Zivot scrupulously avoids. The preprint does not note the number of cases of frothy airways, the best sign that subjects had been struggling to breathe, and does note that these are based on autopsy reports rather than firsthand experience (though the authors suggest, and I’m inclined to agree, that this would have led to false negatives if anything). Zivot also didn’t have access to the patients’ medical histories, calling the suggestion that the sample was predisposed to pulmonary edema “extraordinary;“ I’m again inclined to agree, but a population subjected to years of poor diets and sedentary living might be predisposed enough to make the conclusions hard to extrapolate from. Finally, the preprint stage is not subjected to peer review, reducing at least the visibility of responses if not the number.
Dr. Zivot’s work is the best in its class, but death penalty abolitionists should be as careful as he is in deciding what it means.
The States Don’t Care to Know If Executions Are Botched
I’d be remiss not to point out that part of the reason the question is unknowable is that states don’t want you to know. Death penalty states have engaged in ever-tightening secrecy about their drug sources, providers, and visibility to witnesses since activist and foreign governmental pressure led to the end of thiopental in executions. The largest executioner in raw terms, Texas, doesn’t even perform autopsies, stating that “we know how they died.“ Alabama takes this see-no-evil response to baffling heights, saying that witnessed punctures to an inmate’s bladder don’t constitute a “problem“ and denying access to execution autopsies for its under the pretense of continued investigation even after they’d already called it a “textbook” execution.
With a higher approval rating than the death penalty itself, the myth of “humane” lethal injection may be the only thing keeping the death penalty above water. Whether it’s 7% or 82%, we know it’s much higher than prison officials want you to believe.
While one may consider this a given, 2019’s Bucklew v. Precythe only requires that methods not “intensif[y] the sentence of death” with a “superaddition of terror, pain, or disgrace.”
For clarity: a research letter is like a quick-and-dirty study, which isn’t peer-reviewed or considered particularly reliable, while a letter to the editor is what it sounds like.
Korianis responded that the time between execution and autopsy, unknown at the time he wrote the original letter, had since been published, and did not justify a retraction. This final letter includes a thought I hadn’t seen stated so clearly before:
It is perilous for Heath and colleagues and Groner to extrapolate from clinical experience to lethal injection. No anaesthesiologist practices bolus dosing of thiopental in potentially sedative-hypnotic, resistant, otherwise healthy, unpremedicated, profoundly hyperadrenergic individuals anticipating a peculiarly painful and public death. Modelling thiopental efficacy and pharmacodynamics in executions is further compounded by the profound physiological derangements associated with pancuronium and potassium administration.
A previous version of this post misidentified Sarat’s institution. I regret the error and thank Dudley Sharp for the correction.
As we’ve discussed before, the “three-drug protocol” includes a paralytic, which paralyzes the subject’s ability to control movement or breathing. At the time, all states used the same three-drug protocol, though a significant minority of states now employ protocols which don’t use a paralytic.
A less rigorous review by NPR found similar results.
Thank you for writing on this important issue. The death penalty is a bizarre tool used in few nations. Historically it was far more popular, but many countries have abandoned the practice. Without questioning its efficacy one can still think about the executions themselves. I fear that for many painful executions may be seen as a feature and not a bug despite what appear to be clear violations of constitutional rights. There are also all kinds of underlying assumptions behind each camp here.