This section discusses seven reasons why our counts of error are likely to be under—inclusive because they probably understate the number of death verdicts reversed due to serious error, and because they define as error—free some verdicts that were approved for execution by all three levels of review but later were shown to have condemned innocent people:

• Identifying all decisions finally reviewing capital verdicts during the 23—year study period is difficult. We undoubtedly missed some. Because decisions reversing death verdicts are harder to find than ones approving them, the decisions we missed likely include more reversals than affirmances. As a result the reversal rates we report are likely lower than the actual rates.

• Our conservative method of calculating error rates at the second, state post—conviction review stage assures that the reversal rates we report for that stage are lower than the actual rates.

• We do not count scores of reversed capital verdicts that occurred when the Supreme Court struck down entire capital—sentencing statutes. Instead, we count only reversals under statutes that generally satisfy the Court's "guided discretion" requirements.

• We count only reversals occurring on the federal courts' review of a first habeas petition, omitting a number of reversals in second or successive federal habeas proceedings.

• We use a restrained method of calculating overall reversal rates at the three review stages.

• By not counting death verdicts imposed during the study period that were still under review when the period ended, we understate error found at the federal habeas stage.

• We count only errors the courts actually discovered and relied on as a reason to reverse death verdicts. As a result, we omit many errors the courts found but chose to overlook, and some serious errors the courts failed to spot, including ones that put innocent people on death row.

1. Decisions reversing death verdicts are harder to find than ones approving them.

Information about decisions reviewing capital verdicts is unevenly available and harder to find when decisions reverse than when they affirm capital verdicts. If we missed cases, therefore—and we almost certainly missed some—they are more likely to be reversals than affirmances.

There is no official, publicly available list of the decisions in, or outcomes of, capital appeals for any state in the county. The only systematic way to find those decisions, therefore, is by using the names of death row inmates as search criteria in data bases collecting judicial decisions. Regrettably, neither the federal government nor any capital state has a list of all people sentenced to die between 1973 and 1995, or even "snapshot" lists of persons on death row at given moments during the period. The unofficial list of people on death row across the country kept by the NAACP Legal Defense Fund ("LDF") does provide snapshots of the people on death row, but only at four points each year. People on death row for only brief periods in between LDF's reporting dates are missed. The one complete list of a subset of death row inmates is LDF's roster of persons who were executed after their death verdicts were approved by all courts to which they appealed.

Under these circumstances, the best information for identifying court decisions reviewing capital verdicts is biased against finding decisions that overturn those verdicts and in favor of finding decisions that approve them. As noted, death row inmates whose verdicts were approved at all stages of review and were executed—including ones executed as late as the middle of 2000, but whose court cases were decided during our 23—year study period—are all counted in our study. Because each court decision approving a death verdict makes it more likely that the prisoner will be executed, and because a person's execution assures that we found all decisions approving his verdict, each decision reviewing a particular verdict makes it more likely that we discovered that and all other judicial affirmances of that verdict. Judicial decisions reversing death verdicts have the opposite effect: They make it certain that the inmate will not be executed under the verdict the court disapproved, and that the decision overturning it is not accessible via the single complete list of (executed) death row inmates.

Also, the longer someone is on death row before being removed by execution, reversal of the verdict or death from another cause, the more likely it is that he appears in one of LDF's periodic death row censuses and that we found the court decisions in his case. Because decisions approving verdicts either trigger the inmate's execution (which assures that all the affirmances in his case were found and counted) or extend the inmate's time on death row while he pursues the next level of court review, each court affirmance makes it more likely that we found the inmate and the decisions in his case. Decisions reversing death verdicts again have the opposite effect: They remove the inmate from death row, assuring he will not appear on later LDF censuses based on the faulty verdict.

Third, publicly available on—line legal data bases omit some unpublished decisions, but include all published ones, so that unpublished decisions are harder to find than published ones. Because at least some courts are more likely to use unpublished opinions to overturn than to approve death verdicts (perhaps because reversals are more controversial than affirmances, creating an incentive to make the former less visible to the public than the latter),¹⁰⁰ decisions available to researchers may be skewed in favor of those that affirm, and against those that reverse, death verdicts.

We worked hard to find every decision finally reviewing a capital verdict between 1973 and 1995, including by supplementing LDF's informal snapshot lists of death row inmates with lists and decisions collected intermittently by death penalty resource centers, lawyers and others in particular states, and by the Death Penalty Information Center in Washington, D.C.; keyword searches of legal and newspaper databases; and word of mouth. Using these methods, we believe we found most final decisions—certainly 90% or more. But we undoubtedly missed some, and those we missed are more likely to have reversed death verdicts than to have approved them.

2. We understate error rates at the state post—conviction stage.

To calculate error rates, we worked to identify the number of death verdicts overturned as a result of reversible error at each relevant stage of review and then to divide that number by the number of verdicts reviewed at that stage:

Number Reversed/Number Reviewed = Reversal Rate

As we discuss in A Broken System, this approach proved impossible at the state post—conviction stage because of the high volume of unpublished opinions at that stage.¹⁰¹ To figure reversal rates at that stage, therefore, we tried to find all the death verdicts reversed at that stage (the numerator), while using a substitute figure derived from our direct appeal study to estimate the number of verdicts reviewed at that stage (the denominator). The substitute figure was the total number of death verdicts that cleared the first (direct appeal) stage of review and thus were available for state post—conviction review. Because we know for certain that, at any given time, a substantial number of death verdicts available for state post—conviction review (because they cleared direct appeal) are stuck in the delay—ridden review system and thus have not been finally reviewed by state post—conviction courts, we also know for certain that substituting the number of verdicts available for review as a proxy for the number actually reviewed inflates the latter. By systematically inflating the denominator, we systematically deflated the error rate at that review stage. (If the actual rate is, e.g., 3/5 (60%), inflating the denominator to, say, 3/6 (50%) or 3/7 (43%) produces a deflated estimate.)

The extent of our under—estimate of state post—conviction reversals is indicated by Nashville Tennessean reporter John Shiffman's study of Tennessee reversals of capital verdicts between 1980 and 2001.¹⁰² Shiffman identified all state post—conviction affirmances as well as reversals, enabling him to determine Tennessee's actual state post—conviction reversal rate during his 22—year study period—which overlaps but is not the same as our 23—years study period. The actual rate is 51% (35/68)—more than triple our conservatively estimated rate for Tennessee of 16%. If this difference held nationally, the 68% overall national error rate we calculated was instead 75%.¹⁰³

3. We exclude dozens of reversals caused by Supreme Court rulings overturning entire death statutes.

Our study period begins in 1973, the year after the Supreme Court in Furman v. Georgia¹⁰⁴ overturned all prior death statutes and verdicts in the nation, and ordered states to adopt new statutes satisfying the Court's revised standards. As a result, A Broken System counts none of the hundreds of death verdicts Furman reversed. We use 1973 as a starting point because it is then that states began using modern capital statutes designed to avoid the defects found by Furman.

Even so, we do not count all court reversals of death verdicts imposed under statutes passed after and designed to correct the problems found in Furman. In decisions in the mid— and late 1970s, the Supreme Court established rules under which it and other courts overturned entire capital statutes in Louisiana,¹⁰⁵ North Carolina,¹⁰⁶ Ohio¹⁰⁷ and at least 12 other states.¹⁰⁸ In doing so, the courts reversed many 1973—and—after death verdicts. We do not, however, count those reversals, though they occurred during the study period. We instead count only reversals occurring (1) during the study period but (2) after the relevant state adopted a death—sentencing scheme that passed Supreme Court muster. We omit these reversals to assure that our measure of serious error is constrained and reflects only the performance of statutes like those currently in effect across the nation.

We do not know how many 1976 reversals there were nationally. We know there were 20 in Louisiana.¹⁰⁹ Had we counted these reversals, Louisiana's direct appeal reversal rate would have risen from 47% to 55%, and its overall reversal rate would have gone from 63% to 69%.

4. We exclude reversals in second or successive federal habeas cases.

We count only federal habeas reversals occurring in an initial round of federal review. During the study period, federal courts reversed at least seven capital verdicts in second or successive habeas proceedings that were previously upheld in initial habeas cases. We thus count five death verdicts as having been affirmed at all three levels of review, even though they in fact were reversed during the study period in successive federal habeas proceedings.¹¹⁰ This again was an effort to be cautious: Because successive habeas reversals are considered controversial by some, we omit them.

In sum, our best estimate is that if we had not used the above four conservative ways of collecting and counting reversals, the overall national reversal rate would be between 75% and 80%.

5. We use a cautious method of calculating overall error rates.

To arrive at an overall reversal rate of 68% and success rate of 32%, we combined the reversal and success rates at each of the three review stages, as follows:

.41 (reversed on direct review)

+ .10 (reversal rate on state post—conviction) x .59 (surviving direct appeal) =.06 of original group reversed on state post—conviction

.06
______
.47 (reversed cumulatively on state direct appeal and state post—conviction)

+ .40 (reversal rate on federal habeas corpus) x .53 (portion of original group surviving state review) = .21 of original group reversed on habeas corpus

.21
______
.68 (overall reversal rate)

Overall success rate = 1.00 — .68 = .32.¹¹¹

Using this method of calculating overall success and error rates treats many death verdicts as having been approved at earlier stages, though they eventually were reversed at a later stage. This has led to criticism that we greatly overstate the capital success rate, which instead should be calculated as the number of death verdicts that survived review at all stages (358 verdicts) taken as a proportion of death verdicts that were in the review process (4546 verdicts), or 8%. According to this criticism, the success rate we report (32%) is 400% greater than the actual success rate, and the failure rate we calculate (68%) is about 40% lower than the actual failure rate. This method of calculating overall success rates may have some appeal, but we reject it in order to be cautious in our judgments. While accurately indicating the exceedingly small proportion of death verdicts imposed during the study period that survived all court review in the period, it ignores the fact that some verdicts did not make it through the review process, not because they eventually were reversed, but because after being approved at an earlier stage, they got stuck in the system while awaiting review at a later stage. To take account of that fact, we analyze the combination of all verdicts' success and error rates at every stage at which a verdict was reviewed. This provides a more complete picture of how verdicts fare on review than looking only at the bottom line.¹¹² The effect of this cautious judgment, however, is a more conservative error rate than a different approach would calculate.

6. We understate the amount of error by not counting outcomes of verdicts imposed in the study period but still under review when the period ended.

To confine our reversal data to known outcomes and avoid extrapolation, we calculate reversal rates by counting only the results of death verdicts actually and finally reviewed at one or more review stages during the fixed study period ending in 1995. This prevents us from counting reversals, or estimating reversal rates, for the many death verdicts imposed during the study period that were not finally reviewed until after that period ended. This limitation probably leads us to understate the federal habeas reversal rate among verdicts imposed in the study period. During the study period, it took a year or two longer on average for flawed than for unflawed death verdicts to be finally reviewed at the federal habeas stage.¹¹³ As a result, a disproportionately high number of the delayed outcomes we did not count were reversals, while a disparately high number of the more timely outcomes we did count were affirmances. This almost certainly led us to understate the actual reversal rate among death verdicts imposed in the study period.¹¹⁴ (There is no analogous disparity in the time it takes for death verdicts to be approved or reversed at the first, direct appeal stage.¹¹⁵)

This point shows why it would be "absurd" (to borrow Professor Joseph Hoffmann's conclusion¹¹⁶) to assume that the outcomes our study did not count because they were delayed beyond the study period were all later approved by the courts. Given that 68% or more of the death verdicts imposed during the 1973—1995 period that were fully reviewed during the period were reversed, and given the above reason why 1973—1995 verdicts that were finally reviewed after 1995 are even more likely to have been reversed, it would be irresponsible to base any policy judgment on an assumption that verdicts imposed during the 1973—1995 period but reviewed after 1995 were all affirmed. We point this out because that unreasonable assumption is the premise of a method of calculating reversal rates that has been proposed.¹¹⁷ That method would divide the number of verdicts reversed during the study period by the number of verdicts imposed in the period, even if the verdicts were never reviewed during the study period. Doing so assumes that the only 1973—1995 verdicts that were ever reversed were those reversed in 1995 or before—i.e., absurdly, that the thousands of 1973—1995 verdicts that were finally reviewed after 1995 were all approved. Even when deflated in this irresponsible way, the overall reversal rate is still a "'depressing'" 40%.¹¹⁸

7. Our measure of serious error is conservative.

We count as errors only those defects that lead courts to conclude that capital verdicts are too flawed to be carried out and have to be retried. Such errors impose a heavy burden on the system. They squander resources spent on failed trials and lengthy appeals, waste still more resources on retrials, frustrate the expectations of victims, and undermine the system's deterrent and retributive goals as well as the integrity of the courts.¹¹⁹ All such error is serious, most of all, because it compromises the reliability of the verdict that the defendant is guilty and deserves to die.

For these reasons, the errors we count include only serious errors. But they do not include all serious errors—or even the most serious ones, because we count only errors courts spotted and chose to cure. We limit ourselves to these errors because they are serious enough to cause courts to reverse and order costly retrials, and because they are objectively identified by court decisions without requiring controversial judgment calls by us about what to count. This cautious approach assures, however, that we fail to count many disturbing errors, including the most disturbing ones: mistakes that put innocent people on death row but were missed by all phases of court review.

Uncounted error falls into two categories: errors that appellate courts find but choose to ignore (so called "harmless," "non—prejudicial" and "waived" error), and errors—even egregious ones—that appellate courts fail to discover.

a. Errors courts found that we did not count.

Start with the first category. Suppose a capital defendant's appointed lawyer is addicted to alcohol or drugs, which he ingests repeatedly during the trial, preventing him from providing a decent defense. Or suppose the lawyer fails before trial to interview any witnesses or otherwise to investigate the case. The Supreme Court has said that this behavior falls below the minimum the Constitution requires.¹²⁰ But there are many cases involving this or equally poor representation by capital defense attorneys that courts did not reverse, and that we did not count as serious error, because the defendant could not additionally prove, years after the fact—as the controlling law requires—that a better lawyer would probably have changed the outcome of the trial.¹²¹ We likewise did not count numerous death verdicts where police and prosecutors suppressed evidence suggesting that the defendant was innocent or did not qualify for the death penalty or presented evidence or made arguments the law forbids; nor did we count cases where trial judges barred defense evidence or cross—examination of accusing witnesses the law permits, or gave instructions misinforming jurors about the law governing their determination of guilt or innocence, and sentence.¹²² We declined to count these many errors because reviewing courts ruled them "not prejudicial" (meaning the defendant could not prove that, if the error had not been committed, the outcome of the case probably would have been different) or "harmless" (meaning the error had no or no "substantial" effect on the outcome).¹²³ In our federal habeas study, we tabulated the number of death verdicts that courts refused to reverse because any error was "harmless." (We did not, however, tabulate findings of "no prejudice."¹²⁴) Harmlessness findings were made in 18% of the cases in which death verdicts were approved on federal review.

Courts declined to reverse, and thus we declined to count, still other errors occurring at capital trials that evidently were prejudicial but were allowed to stand because the defense lawyer incompetently forgot to object. When lawyers negligently fail to object to errors, courts usually will not correct the error, even if it was serious and would have required reversal if the lawyer had done his job.¹²⁵ Such "waivers" were found in 32% of the habeas cases in our study in which relief was denied. Overall, federal courts chose to overlook error as harmless, waived or both in 44% of the habeas cases in which verdicts were upheld.¹²⁶ Including just a third of these uncounted errors identified at only one of three review stages would raise the overall national reversal rate from 68% to 73%.

b. Errors courts failed to find.

Between 1973 and the first week of January 2002, 99 men and women sentenced to die were later exonerated by official findings that they were not guilty of the capital offense.¹²⁷ In each case, the original trial verdict reached a conclusion that either was factually wrong because the defendant was innocent, or legally wrong because there was not enough evidence for a reasonable grand jury to indict or a trial jury to convict. In 63 of the 99 exonerations, the mistake was discovered only after the highest state court had upheld the capital verdict at least once. On 35 of those occasions, either a state, or a state and a federal, post—conviction court affirmed the verdict a second or third time before the mistake was discovered. And in some cases, all three levels of courts missed the error, clearing the way for execution, before someone else discovered the mistake. In most of the 99 known cases of factually or legally innocent people sentenced to die, therefore, our method of defining serious error led us to count at least one, and often two or three, judicial decisions as finding no error.¹²⁸ Our cautious and objective measure of error—mistakes appellate courts actually discovered and held serious enough to require reversal—assures that we undercount the actual amount of serious error.

c. Four illustrative cases in which stringent rules limiting reversals led courts to approve the capital verdicts of innocent men despite a full set of appeals.

How can innocent men and women be convicted of a capital crime and sentenced to die? And how can the mistakes escape detection by multiple courts that approved the prisoners' execution? Four typical cases provide an answer: The courts define error serious enough to require reversal so cautiously and under—inclusively that they often hold known errors—even ones that put innocent people on death row—to be harmless, not prejudicial or waived. Because we use the same judgments to define serious error, our counts of error are also cautious and under—inclusive.

i. Lloyd Schlup was convicted and sentenced to die by Missouri for killing another inmate in prison. After the Missouri Supreme Court on direct appeal, the trial court and Missouri Supreme Court a second time on state post—conviction, and a United States District Court and the United States Court of Appeals on habeas rejected his claims that errors in his case had led to his conviction for a crime another prisoner committed, thus clearing him to be executed, a prison videotape and a guard's testimony about the time of the events revealed by the tape confirmed, as Schlup had always said, that he was in another part of the prison when the killing occurred. Fourteen years after his arrest, Schlup agreed to a settlement of the case so his conviction of capital murder could be withdrawn.

How did three levels of reviewing courts approve this miscarriage—leading A Broken System to count Schlup's verdict among the 32% in which no serious error occurred? The answer lies in the harmless error, no—prejudice and waiver rules noted above. On direct appeal, Schlup objected to the admission of photos supposedly showing that a guard who falsely identified Schlup as the assailant could see the site of the killing from his guard station. The photos had not been "authenticated" by anyone who could say they showed the view of the crime scene from the guard post, rather than from a different vantage point. Authentication is a legal requirement some call a technicality, and the Missouri Supreme Court treated it as such: "The fact that Maylee [the guard who said Schlup was the killer] . . . did not testify that the photos depicted his exact vantage point," the court said, although an error, was harmless, so that "[t]he trial court did not abuse its discretion in admitting the photographs." The Missouri Supreme Court then concluded—as the jurors also apparently did—that the photos strongly "corroborate[d] Maylee's testimony by demonstrating that he could have witnessed the murder from his station."¹²⁹ In fact, the photographs did not show what Maylee could see from his post; contrary to his trial testimony, the guard could not and did not see Schlup at the scene. But because of the Missouri high court's reluctance to reverse based on "technical" error, Maylee's flawed identification sent a man to death row for a crime someone else committed.

This same treatment of uncorrected (and so, by us, uncounted) error continued on state post—conviction review. There, the Missouri Supreme Court chose to ignore another, this time non—technical, error because it was not "prejudicial." The known error was the prosecutor's "fail[ure] to disclose exculpatory evidence" tending to show the defendant's innocence. State lawyers failed to reveal that the warden of the prison where the killing occurred "had evidence that another individual may have committed the murder, and the warden [told police] he did not believe appellant would intentionally hurt someone." Although prosecutors are required to disclose exculpatory evidence, their failure to do so is ignored if the defendant fails to show that the prosecutors' withholding of exculpatory evidence probably changed the trial outcome. Applying this exception, the Missouri high court chose to ignore the error, calling mere "rumor" the warden's belief about what occurred in his prison and the information making him think another man was the killer.¹³⁰ The warden was correct, of course. But the courts refused to cure the error (and we did not count it) for lack of "prejudice."¹³¹

The same thing happened on federal habeas review. There, Schlup showed that his trial lawyer incompetently failed to interview or call three known alibi witnesses. The court did not dispute that the lawyer failed to give Schlup decent legal help, but the court chose to ignore the error because it was not shown to be prejudicial. Accepting the lawyer's claim—though he never talked to the three witnesses—that their testimony that Schlup was not near the killing would be "repetitive or . . . damaging," the court ruled that Schlup had not shown that the denial of his right to counsel had probably led to the wrong outcome. As a result, the error went uncorrected by all three stages of court review (and uncounted by us), and the three alibi witnesses went unnoticed by the judicial system until the videotape and supporting testimony finally showed that Schlup was with those witnesses, away from the killing, when it occurred.

The procedure Schlup used to prove he was not guilty after all three regularly available review stages failed him no longer exists. Congress decided to abolish it in 1996.¹³²

ii. Earl Washington's death verdict is also counted by us as error—free because it was affirmed at all three stages of court review. A recent press account describes Washington's conviction and death sentence, despite his innocence:

"Did you stab a woman in Culpeper?" the state police detective asked. The illiterate farm worker nodded.

"Was this woman white or black?"

"Black."

A few questions later, Special Agent C. Reese Wilmore tried again. "Was she white or black?"

This time Earl Washington Jr. said, "White." That answer launched the biggest mistake ever made by Virginia's judicial system—and landed Washington on death row.

It wasn't until Oct. 2 [2000]—17 years after that police interview—that new DNA tests cleared Washington of the 1982 rape and slaying of Rebecca Lynn Williams. Recent interviews with Washington and Williams's widower as well as dozens of police officers, judges and lawyers involved in the case turned up warnings that went unheeded along the way:

• Police and prosecutors moved forward with a case based almost entirely on a statement full of inconsistencies from an easily persuaded, somewhat childlike special—education dropout. Washington told investigators he "stuck her . . . once or twice," but Williams bled to death from 38 stab wounds. He said she was alone. But there was a baby in a playpen and a toddler roaming the small apartment. The defense made no mention of most of these inconsistencies during the trial.

• A judge ruled that the statement was admissible after hearing from a state mental health expert that a man with an IQ of 69 was competent to waive his rights to a lawyer during initial questioning—even though Washington still doesn't know what the words "waive" and "provided" mean.

• No eyewitness or physical evidence put Washington at the scene. His blood type did not match a semen stain, and police instructed the state lab not to test key hair evidence. A judge rejected defense efforts to test the hair, and the defense lawyers never told the jury about the mismatched blood types.

• Six courts rejected the inmate's claims of innocence, including a panel of federal judges who determined that Washington's trial attorney had failed to meet minimal standards but upheld the conviction anyway. Virginia's appeals judges . . . ruled that Washington's confession was properly admitted and the blood evidence was inconclusive.

* * * * *

In October, Gov. James S. Gilmore III (R) pardoned Washington after more sophisticated genetic testing found no trace of him at the scene.

* * * * *

Although state officials have reopened the investigation, Williams's widower, Clifford, feels betrayed by Culpeper authorities, who assured him that Washington was the right man and now won't talk to him, he says.

"What do they have to hide? Why won't they talk about it?" he asked in a recent interview. "I went for nearly 18 years believing Washington did it. Now I don't know what to think."¹³³

According to another news report:

Genetic material found on Williams's battered body did not match [Washington], her [the victim's] husband or any man in the state's DNA data bank of convicted felons. But lab tests done on a blue blanket at the crime scene found the DNA of a convicted rapist [who was never punished for the 18—year—old offense], Gov. Gilmore said in a statement.¹³⁴

These accounts again show that the court standards for judging serious error—the same ones we use here and in A Broken System—were too forgiving to spot the errors leading to Washington's false conviction. Answers to more specific questions about the case compel the same conclusion.

How could the courts have ruled that a retarded man, whose memory of the events clashed with the known facts on several crucial points, could understand his rights and validly confess? Here is what the Virginia Supreme Court said:

On appeal, the defendant argues . . . that he made no waiver of his right to counsel [when he made his alleged confession] on May 22, 1983, and that he was, in any event, incapable of making a voluntary and intelligent waiver of his constitutional rights. . . . These contentions lack merit. The record clearly shows that on at least three occasions . . . [Washington] gave his questioners clear indications that he understood and waived his rights, both orally and in writing.¹³⁵

Washington's inexperienced trial lawyer had a copy of a blood report showing that all the semen evidence at the crime scene had a blood type different from Washington's. He decided the report wasn't important and never told the jury about it.¹³⁶ At first, the U.S. Court of Appeals thought this might be incompetent representation, and ordered a hearing:

[Washington's] allegation [that his lawyer was incompetent] was supported by 2 affidavits. One, by an . . . expert in the field, opined that the laboratory reports of the blood type and PGM [enzyme] type of the semen stains, as compared to Washington's, excluded Washington as the depositor of the semen. The other, by his trial counsel . . . stated that counsel had received the laboratory reports but did not recognize their arguably exculpatory nature.

The district court rejected this claim of ineffective assistance without an evidentiary hearing on alternative grounds: that counsel's conduct, as alleged, did not fall outside the range of acceptable professional conduct, and that in any event there was no reasonable probability that the outcome of the proceeding would have been different but for the challenged conduct. . . .

If, as Washington alleged, his counsel failed to offer available evidence which in a significant way drew his factual guilt in issue, counsel's performance obviously fell below an objective standard of reasonable professional conduct, unless some cogent tactical or other consideration justified it. . . . The allegation that the laboratory reports indicated Washington's blood type as O with PGM type of 2—1 whereas four samples of the semen stains on the blanket from the crime scene showed blood type A with PGM type of 1, was undisputed. The allegation that this disparity of types indicated that Washington could not have been the depositor of the semen in the stains was supported by the . . . affidavit of a . . . qualified expert that was not disputed by opposing expert opinion or other evidence.

[As for the ruling that] there was no reasonable probability, given the evidence of Washington's guilt, that the result of the proceeding would have been different had the challenged conduct not occurred, . . . we believe the district court could not properly make that assessment without an evidentiary hearing . . . . [Unless shown otherwise at a hearing] . . ., the exculpatory quality of the forensic evidence . . . made it reasonably probable that had it been laid before the jury, it would at the least have created in that body a reasonable doubt as to guilt or resulted in the recommendation of a lesser sentence reflecting that doubt.

[T]he evidence of guilt presented to the jury . . . was not without its difficulties . . . . The evidence consisted essentially of a confession obtained by interrogation almost a year after the crime, from a mildly retarded person upon whom suspicion had not earlier focused during the crime's investigation, and who was not indeed suspected when the critical interrogation which elicited his inculpatory statement was commenced, apparently blindly, while he was in custody in connection with an unrelated crime.¹³⁷

After holding a hearing, the lower court ruled that the lab report indeed showed the semen stains did not match Washington, but concluded that the lawyer's error in failing to tell the jury about the report should be ignored because it was not prejudicial. On appeal, the higher court agreed—revealing the strictness of the courts' (and our) definition of error serious enough to require reversal:

We cannot say the district court erred in concluding that petitioner was not prejudiced by [his lawyer's failure to introduce] the forensic evidence. ... Even assuming that petitioner had presented the stained blanket and his experts at trial, the prosecution still had a strong case against petitioner [based on "Washington's confession to the crime"]. . . . [G]iven the case's strength, we cannot say that inconclusive forensic evidence would have overcome it.¹³⁸

iii. Anthony Porter is another retarded victim of a flawed capital trial who spent 17 years on (Illinois's) death row for a crime another man committed. His death verdict also was upheld at all three stages of court review, and so is counted by us as error—free. But as the courts knew all along, Porter's trial was in fact marred by two major problems—a biased juror and an incompetent lawyer. The courts held the errors unimportant given the supposedly strong evidence of guilt.

On his first appeal, Porter pointed out that one of the jurors who voted to convict and condemn him had failed to tell the judge, when asked directly, that she knew the mother of one of the murder victims. Once on the jury, the woman urged the other 11 to "vote guilty right then . . . before any discussion was had on the evidence." The Illinois Supreme Court ruled there was no prejudice because Porter's trial lawyer showed only that the juror "knew the victim's mother as someone who attended the same church that she attended,"¹³⁹ but did not show that "the relationship between the juror and the victim's mother" was close. When Porter's new lawyer, on his second appeal, supplied the missing information—that the juror and the victim's mother were good friends—the court again chose to ignore the error, saying the error was waived by the first lawyer's incompetent failure to discover the information.¹⁴⁰

That incompetence went even further. Due to a dispute with Porter over his fee, the lawyer refused to interview or call five witnesses (including three close relatives of the victims) who said a man named Alstory Simon had killed the victims in a fight over drugs. Without disagreeing that the lawyer incompetently failed to investigate evidence identifying a different killer, the second reviewing court ruled the error non—prejudicial—again showing how narrow the courts' (and our) measure of serious error is:

Even assuming counsel performed incompetently in not generating the proposed testimony, sufficient prejudice did not result to support the claim. . . .

Prejudice is measured by looking at findings unaffected by error and accounting for the error's effect on remaining findings to answer whether the decision would "reasonably likely" have been different. The assessment "must exclude the possibility of arbitrariness, whimsy, caprice, 'nullification,' and the like." The showing of prejudice must be a strong one. [That standard was not met here, because t]he evidence against defendant [Porter] was considerable.¹⁴¹

Both errors again went unremedied at the federal habeas stage of review. As for the biased juror and several other errors, the federal district court wrote:

Porter['s lawyer] did not [properly] raise several of his asserted grounds for relief in the Illinois courts . . .; as such, those arguments are procedurally barred. "In all cases in which a state prisoner has defaulted his federal claims in state court . . ., federal habeas review of the claims is barred ." . . . Under these standards, the following claims now raised by Porter are procedurally barred: use of allegedly perjured testimony, use of constitutionally unfair procedures, and denial of an adequate hearing on the extent of juror bias . . . .¹⁴²

As for his trial lawyer's incompetent failure to interview five witnesses who identified Alstory Simon as the killer, the federal district court again illustrated how difficult it is to show that even clearly below—standard lawyering is prejudicial enough to be reversible error (and, thus, to be counted by us as serious error). To overturn a capital conviction, the court said, a

"defendant must show that there is a probability that, but for counsel's unprofessional errors, the result of the proceeding would have been different." . . . [D]eficient performance, by itself, "does not warrant setting aside the judgment of a criminal proceeding if the error had no effect on the judgment." Porter has not made the requisite showing . . . [because he] was convicted by a jury which heard considerable evidence that Porter committed the crimes.¹⁴³

A federal appellate court agreed, rejecting Porter's claim that he was prejudiced by his lawyer's admitted incompetence with a rhetorical question that speaks volumes about how hard it is to satisfy the courts' (and our) test for serious, reversible error:

Porter asserts that his counsel should have presented evidence that Alstory Simon and Inez Johnson were responsible for murdering Green and Hilliard. Porter has offered a number of affidavits and sworn statements by people in the neighborhood stating, among other things, that Simon and Johnson went to the park that night with Green and Hilliard, that Simon had just been released from the penitentiary and had a financial dispute with Hilliard regarding drug dealing, that Hilliard was seen arguing in the park that night with a man who was not Porter, . . . that Simon threatened someone who asked Johnson what had happened at the park[, and that Inez Johnson had been overheard admitting that she and Simon committed the killings]. None of this evidence was offered at trial, although the State concedes that Simon and Johnson were in the park with [the victims] at some point on the night of the murders.

. . . [But h]ow much credence can we reasonably give to third—hand information when it contradicts two eyewitnesses and a police officer who put Porter right at the scene of the crime?¹⁴⁴

These decisions cleared the way for Porter's execution, which was hours away when he received an emergency reprieve on the ground that he might be too retarded to understand why he was being executed. In the ensuing pause, some Northwestern undergraduate students, as a class project, tracked down Alstory Simon in Milwaukee where he had fled after Porter's arrest. The result—when someone finally followed—up on the leads Porter's lawyer had incompetently ignored—was Simon's taped confession to the killings. Porter was released. Simon pleaded guilty to killing the two people,¹⁴⁵ and is believed to have killed a third person after Porter's arrest.¹⁴⁶

The Chicago Tribune's report on the Porter case again illustrates how high the courts (and thus A Broken System) set the bar for establishing serious, reversible error in capital cases:

It took two days to put Anthony Porter behind bars and send him on his way to Illinois' Death Row. It took nearly 17 years to set him free.

Between those bookends of Porter's incarceration, the criminal justice system failed him at several critical turns, according to police and court records as well as interviews.

When initially investigating the crime, for instance, police never seriously considered other suspects, and they discounted Porter's alibi.

Witnesses who could have exonerated him lied, although some say they were coerced by police. And others who knew the real details of the crime kept silent, even when they knew an innocent man faced execution.

Although the justice system is supposed to ensure that everyone—even the destitute—is provided an attorney to defend himself, the reality is that Porter's lack of financial resources meant he received only the most basic defense, even though he was facing the most serious punishment.

By his trial attorney's own admission, efforts on Porter's behalf were spare. . . .

After Porter's conviction, judges in state and federal courts—including the U.S. Supreme Court—turned away more than a half—dozen of Porter's appeals and other filings, dismissing arguments raised on grounds ranging from ineffective counsel to claims of innocence.

[A]s Porter's case moved through the courts, Chicago police and the Cook County state's attorney's office saw their work validated. Questions of innocence were denied by higher courts, and the procedural appeals were turned away. An appeal that examined whether a juror was biased was unsuccessful.

"We had good claims," said Daniel Sanders, Porter's appellate attorney. "It's just because of the tough rules in the court that we kept losing."¹⁴⁷

iv. Frank Lee Smith's recent exoneration for a 1985 Florida rape—murder followed the same distressing pattern, but with a tragic twist:

DNA evidence has exonerated Death Row inmate Frank Lee Smith of the rape and murder of an 8—year—old Broward County girl.

But he died 11 months ago.

Another man, Eddie Lee Mosley, is now the main suspect in Shandra Whitehead's 1985 death, police and prosecutors said. DNA tests have also linked Mosley to the murder of another Fort Lauderdale child . . . police said Thursday. . . .

Smith died of cancer on Jan. 30 while his attorneys and family fought to prove his innocence. . . . The victory that came with this week's FBI release of the DNA test results was bittersweet, said . . . the Tallahassee attorney hired by Smith's family to try to clear his name.

"The state prosecutors had resisted testing while Frank Lee Smith was alive and pursuing his appeals," said the attorney . . . . "Once he was dead, they relented and became more cooperative about letting us get the tests done."

Smith's sister . . . and his aunt . . . broke down and cried earlier this week when they heard that DNA tests conducted by the FBI had exonerated him.

"They knew from the very beginning he was innocent," [the lawyer] said. The family believed in Smith's innocence, he said, because he was convicted on such scant evidence—the word of a witness who later recanted and said she was pressured by police . . . .¹⁴⁸

The Florida Supreme Court also recognized flaws in Smith's trial and the weakness of the evidence against him. But the court relied on the harmless error, no—prejudice and waiver rules, and the strict standards for court relief to affirm Smith's capital conviction and sentence—which in turn required us to count Smith's verdict as error free. In its opinion, the Florida high court wrote:

As these cases show, state and federal courts do not reverse death verdicts for weak or technical reasons. Instead, their decisions (and thus our test for serious error) run in the opposite, highly cautious direction: Absent clear proof of error with a proven effect on the verdict, even doubts about guilt do not lead courts to reverse capital verdicts.