When is prejudice prejudicial? That is the question the Supreme Court justices heard on Wednesday in the case of Duane Buck. In 1996, Buck murdered his ex-girlfriend in front of her children along with a man he thought she was sleeping with. At his death penalty trial in Texas in 1997, the prosecutor asked a defense expert whether it was correct that “the race factor” that he was “black” increased Buck’s future dangerousness. The expert answered, “Yes.” Such determination of future dangerousness by a jury makes a person eligible for the death penalty in Texas and in other states. Buck was sentenced to die.

“What competent counsel would put that evidence before a jury,” Justice Ruth Bader Ginsburg asked at Wednesday’s oral arguments. Texas Solicitor General Scott Keller answered that the state was “not defending [the] defense counsel’s actions,” and he pointed out that the justices must find “prejudice” on the part of the jury, or that the evidence sufficiently affected the outcome reached by the jury in order to determine that Buck’s death sentence should be overturned. He argued there was no harm to the defense from psychologist Walter Quijano’s racist and inflammatory testimony—essentially that it’s likely the jury would have given Buck the death penalty even without that testimony.

It is a little rich that the justices must now consider such fine questions of impossible-to-determine probabilities, when it is the Supreme Court’s doing that this expert was testifying about made-up probabilities in the first place. Buck’s case—and specifically the model of putting people to death depending on vague considerations, such as whether or not a jury thinks a convict may be violent in the future—demonstrates that the entire project of modern death sentencing is prejudiced beyond repair.

Much of the arguments before the Supreme Court on Wednesday revolved around technicalities that will decide whether Buck lives or dies. Was there prejudice or a measurable effect on the outcome? Was there a substantial likelihood that jurors would have reached a different conclusion without the testimony? Was the issue properly preserved or important enough to permit an appeal? If the court ultimately does not grant Buck relief, it may be due to those intricate legal standards.

Sure, Keller said, if the prosecutor used that testimony to get the death penalty, that would be “highly prejudicial” and unconstitutional. In five other cases where the prosecution called that expert, who gave the same race-based testimony, the state agreed the sentences should be reversed. Here, though, it was the defense lawyer that called the expert. Yet as Justice Elena Kagan put it, that seems like it would have even more of an influence on the sentence. “When the defendant’s own lawyer introduces this, the jury is going to say, well, it must be true,” she noted.

So it seems pretty obvious that this prejudice was prejudicial. What the lawyers and the justices did not address, though, is that the entire future dangerousness exercise in death penalty cases is in and of itself one in prejudice. In 1983, the court ruled in Barefoot v. Estelle that experts can testify in death penalty cases about future dangerousness, despite the fact that studies had shown such predictions were so error-prone that they were wrong as much as two-thirds of the time. At the time, the justices ruled that having such an expert aid a jury in determining life or death was better than nothing. Such testimony was not “entirely unreliable.”

In reality, such testimony was wrong more often than not. A series of Texas experts, from the infamous James Grigson, nicknamed Dr. Death, to Quijano, testified time and time again as to long-term future dangerousness and sent hundreds of people to the death chamber. Yet leading organizations like the American Psychiatric Association have rejected the use of such testimony as scientifically unsound and far too error-prone.

Buck, in fact, has been a model prisoner these past 20 years, without so much as minor prison infractions. Such expert testimony is even more flawed today, when life without parole is an option in states like Texas. Judges ask jurors to consider the future “threat to society” of someone who will, if not sentenced to death, serve their entire life in the society of a maximum-security prison. Neither side should be able to put on junk evidence, particularly when life and death is at stake.

Jurors are asked to sentence someone to death based on an impossible task that the Supreme Court and the lawmakers in death penalty states have set out for them. Jurors are still forced to try to predict future violence without good information—something that expert psychologists cannot even be expected do. Jurors are asked to do the impossible under all death-sentencing schemes; other statutes pose similarly fraught questions, such as whether a murder was especially “heinous” or “atrocious.” Time and time again, we see that the death penalty is given not to the worst murderers, because those are hard to identify, but to the defendants with the most unfortunate social backgrounds and the worst lawyers.

Ultimately, the Buck case encapsulates the modern American death penalty: infected with racial prejudice, scientifically unsound, arbitrary, and tainted by incompetent defense lawyering. No wonder public support for the death penalty just reached its lowest levels in four decades. Indeed, things have even changed in Harris County, Texas, where Buck was convicted, which used to lead the country in death sentences. Last year there were no new death sentences there.

Even if the procedural rules in this case are onerous, finding a clear effect of such race bias is an obvious call. The larger question, though, will linger after Buck’s case is settled—how much longer can the justices allow prejudiced and flawed decision-making to support death sentences?

The testimony that led to Buck’s death sentence was racist and prejudicial, but the entire enterprise of the death penalty itself is unconstitutional.

This article originally appeared in Slate.