Prejudice

Prejudice

The second prong of Strickland v. Washington asks whether counsel's deficient performance prejudiced the defense so as to deprive the defendant of a fair trial.¹ An unreasonable error does not support relief by itself. It must have produced an actual impairment of the defendant's chances of a favorable outcome.² The two prongs are therefore independent hurdles, and a court can find that counsel erred and still deny the claim. That is what happened in Strickland itself, where the Court held that counsel's conduct at and before the sentencing proceeding could not be found unreasonable and added that, even if it could, the defendant had suffered too little prejudice to set aside the death sentence.³ Kimmelman v. Morrison shows the same separation from the other direction. Counsel there missed the deadline for a suppression motion because of an ignorance of the law that no strategy could explain, so performance was deficient, yet the Court remanded for a decision on whether excluding the evidence would have affected the result.⁴ On an exam the trap is stopping once counsel's mistake is established, because deficiency alone generally earns no relief. Many otherwise strong claims end at this prong, since an error can be real while the result stays untouched by it.

The requirement has reasons of its own, and they begin with what the right to counsel protects. A lawyer's role is to supply the assistance needed to justify relying on the outcome of the proceeding, so a deficiency matters constitutionally only when it weakens that reliance.⁵ The nature of lawyers' mistakes points the same way. Attorney errors come in infinite variety, are as likely to be harmless in one case as harmful in another, and cannot be sorted in advance by their likelihood of causing damage. The government is not responsible for them and cannot prevent them, and because representation is an art, a step that is unprofessional in one case may be sound in another.⁶ No list of forbidden conduct could capture the errors that matter, so the workable question looks to the effect on the particular case.

Two further points explain where the standard sits. Finality supports requiring an effect, because a rule that unwound convictions over errors that changed nothing would erode the weight of judgments. The Court nonetheless treats that concern as somewhat weaker here than in a newly-discovered-evidence setting, since an ineffective-assistance claim asserts the absence of one of the crucial assurances that a result is reliable.⁷ History supplies the other point. Before Strickland, lower courts ranged from presuming prejudice upon a finding of incompetence to placing a heavy burden on the defendant, and Strickland answered with a middle course that demands a real effect without demanding proof that the outcome more likely than not would have changed.⁸

The allocation of the burden also separates this prong from harmless-error review. In Chapman v. California, the prosecutor commented on the defendants' failure to testify and the trial judge told the jury it could draw adverse inferences, conduct the state supreme court acknowledged violated the Constitution, but the state court affirmed on the theory that the error was harmless.⁹ The Supreme Court held that a federal constitutional error can be treated as harmless only if the court can declare a belief that it was harmless beyond a reasonable doubt, and it explained that the burden falls on the beneficiary of the error.¹⁰ Ineffective assistance reverses the burden, because the defendant must show a reasonable probability that the result would have been different.¹¹ The difference tracks who is responsible for the mistake. In Chapman state actors committed the error, so the side that benefited from it had to show it did not matter, whereas counsel's lapses are errors the government neither commits nor can prevent, so the defendant who invokes them must show their effect.¹²

The lawyer's conduct was the same in both trials, so what moved the answer was how much of the rest of the record supported the verdict, a point developed below under the totality of the evidence. That point is one of four separable ideas that make up this prong. The first is the quantum of proof, meaning how probable a different result must be. The assumption that the decisionmaker acted lawfully and impartially comes next, because it removes the lenient judge or the lawless jury from the calculation. After it comes the need to weigh the whole record rather than the error in isolation, and last is the permission to decide a claim on prejudice without first resolving deficiency. The place to begin is the quantum of proof, because every other idea in this prong is measured against it.

Reasonable Probability

The first of the four ideas in this prong is the quantum of proof, meaning how likely a different result must have been. Under Strickland v. Washington, the defendant must show a reasonable probability that, but for counsel's unprofessional errors, the result of the proceeding would have been different, and a reasonable probability is a probability sufficient to undermine confidence in the outcome.¹⁴ The formula reads like a compromise because it is one. The Court reached it by turning down a standard that was too easy to meet and another that was too hard, and each rejection carries a reason that shows how the test will behave on new facts.

The lenient end of the range is a showing that the errors had some conceivable effect on the outcome. The Court refused that standard because virtually every act or omission of counsel would satisfy it, and not every error that conceivably could have influenced the outcome undermines the reliability of the result.¹⁵ A prejudice requirement that every error passes sorts nothing, so the second prong would disappear and every proven deficiency would end in relief. The Court gave the same answer to a proposal that the defendant need only show the errors impaired the presentation of the defense. Any error, if it is an error, impairs the presentation of the defense, so the proposal offered no principle for deciding which impairments are serious enough to set aside a result.¹⁶

The demanding end of the range is a requirement that the defendant show the errors more likely than not changed the outcome. The Court granted that this outcome-determinative standard has real strengths: courts know how to apply it, it reflects the importance of finality in criminal cases, and it matches the familiar test for new-trial motions based on newly discovered evidence.¹⁷ It still declined to adopt the standard, and the reason lies in what the borrowed test assumes. The strict newly-discovered-evidence standard presupposes that the original proceeding already had every essential element of a presumptively accurate and fair trial. An ineffective-assistance claim alleges the opposite, that one of the crucial assurances of reliability was missing. In that setting finality concerns are somewhat weaker, so the appropriate standard is somewhat lower, and a result can be unreliable even though the errors cannot be shown by a preponderance to have determined it.¹⁸

Between those poles sits the reasonable-probability test, which the Court took from two existing materiality standards: the one governing exculpatory information the prosecution fails to disclose, from United States v. Agurs, and the one governing testimony made unavailable when the government deports a witness, from United States v. Valenzuela-Bernal.¹⁹ Borrowing gave courts a standard they already knew how to administer. Kyles v. Whitley later applied the same formulation to suppressed evidence and confirmed its meaning: the defendant need not show by a preponderance that the outcome would have been an acquittal, and the word reasonable matters, because the question is whether the verdict remains worthy of confidence.²⁰

Because the test asks about confidence, no fixed percentage applies. A reasonable probability does not require a 51 percent chance that the outcome would have changed, or any other particular likelihood, and it is enough that the mistake undermined confidence in the outcome.²¹ The standard sits below a preponderance, so a defendant who cannot show that a different result was more likely than not can still prevail.²²

The first defendant can satisfy the threshold even though acquittal was less than even odds, because a real chance of acquittal on a thin identification case is enough to shake confidence in the verdict. The one-in-three figure only pictures a chance that is less likely than not, and no court computes a number. The second defendant has shown only a conceivable effect, which is the showing the Court rejected because virtually every error can meet it. The lawyers' mistakes were comparable in kind, so the fact that moved the answer was whether the error connected to a genuine prospect of a different result.

Students lose points on this prong in two opposite ways. One error is demanding proof that the verdict more likely than not would have changed, which imports a burden the Court expressly refused to impose.²³ The other is accepting any plausible-sounding effect as enough, which collapses the prong into a requirement that every error satisfies. The analysis can be run in four steps that avoid both mistakes.

The fourth step does the real work, because it replaces a numerical question that courts do not ask with the confidence question they do ask. How the strength of the rest of the record bears on that question is developed below under the totality of the evidence.

Objective Decisionmaker Assumption

The reasonable-probability test asks whether the result would probably have been different, and that question is incomplete until it says different in the eyes of whom. A defendant could point to the actual judge or jury and argue that a lenient, sympathetic, or inattentive decisionmaker would have acted otherwise if counsel had done better. Strickland forecloses that argument by instructing the reviewing court to presume that the judge or jury acted according to law, and to assess the likelihood of a better result on the assumption that the decisionmaker was reasonably, conscientiously, and impartially applying the standards that govern the decision.²⁵ This is the objective decisionmaker assumption, the second of the four ideas in the prejudice prong. It excludes arbitrariness, whimsy, caprice, and nullification (a jury's refusal to apply the law to the facts it has found), so a defendant has "no entitlement to the luck of a lawless decisionmaker."²⁶ The presumption is stated for claims that do not themselves attack the sufficiency of the evidence, which is a separate kind of challenge.

Each part of the assumption has its own reason. Excluding the lawless decisionmaker follows from what the prejudice requirement protects, namely reliance on the outcome of the proceeding.²⁷ A result that could have changed only through caprice is not made unreliable by counsel's lapse, and counting that possibility would turn the prejudice showing into a reward for the chance that a decisionmaker might ignore the law. Excluding the particular decisionmaker's idiosyncrasies, such as an unusual propensity toward harshness or leniency, solves a problem of proof. What one judge or juror was inclined to do is rarely in the record, and establishing it would require evidence about the actual process of decision or about a judge's sentencing habits, which Strickland says should not be considered.²⁸ Allowing that evidence would turn every claim into an inquiry into the mind of a decisionmaker who cannot meaningfully be examined, and identical lawyer errors would earn relief in one courtroom and be denied in another for reasons that have nothing to do with the lawyer. A single hypothetical standard of decisionmaker keeps attention on the reliability of the proceeding.

The same tendencies can surface at two points in a claim, and the doctrine treats them differently. Strickland recognized that a judge's propensity toward harshness or leniency may actually have entered into counsel's choice of strategy, and to that limited extent it may bear on the performance inquiry, yet the same factors are irrelevant to prejudice.²⁹ The division follows from what each prong asks. Performance asks what a reasonable lawyer would have done with what the lawyer knew at the time, and a reasonable lawyer takes account of who will decide. Prejudice asks whether the proceeding can still be trusted, and that cannot turn on who happened to be deciding. The rule also cuts both ways. A defendant who argues that the sentencer was known to be unusually harsh, and that a gentler one would have decided differently, establishes nothing, because prejudice is assessed as though the sentencer were impartial and law-abiding.³⁰

The assumption is not peculiar to ineffective-assistance claims. United States v. Bagley carried the same reasonable-probability formulation into the setting of evidence the prosecution suppressed,³¹ and a dissent in Kyles v. Whitley described the corresponding premise there as an assumption of jury conduct that is rational and obedient to the law.³² The idea reaches its sharpest edge, though, when the question is what the law entitles the defendant to receive.

In Lockhart v. Fretwell, counsel failed to make an objection that, under the case law prevailing at the time, would have prevented a death sentence, and that case law was later overruled.³³ The defendant would have avoided death had counsel objected, yet the Court held he was not unconstitutionally prejudiced.³⁴ Prejudice there depended not only on how counsel's deficiency affected the sentence but also on whether the sentence was fundamentally unfair or unreliable, and the defendant had lost only the opportunity to have the sentencing court make a mistake.³⁵ The link to the decisionmaker assumption is direct. Strickland denies a defendant the luck of a decisionmaker who ignores the law, and Fretwell denies the benefit of a decisionmaker who would have applied a rule the law no longer recognizes. The Nutshell account puts the principle generally: prejudice cannot rest on the loss of a right the law does not give, which is also why a claim that counsel blocked perjured testimony failed as a matter of law in Nix v. Whiteside.³⁶ Fretwell is an unusual limit and not a second general test, since Strickland's two parts govern virtually all claims and the exception applies only where treating the difference in outcome as prejudice would be unjust.³⁷

Students go wrong most often by treating the particular decisionmaker's reputation as proof that prejudice is present or absent, when the inquiry asks only what a reasonable, law-following decisionmaker would have done. The second slip is reading Fretwell as an escape hatch for any outcome a court thinks was fair, which it is not, since the lost outcome must have depended on legal error. Both mistakes are avoided by a fixed order of questions.

The third step seldom changes the answer, because most claims rest on rights the law plainly gives. Its job is to catch the rare claim in which the only thing lost was a windfall.

Totality of the Evidence

The reasonable-probability threshold needs a body of facts to be applied to, and that body is the whole record. A court hearing an ineffectiveness claim must consider "the totality of the evidence before the judge or jury," so the question is never how damaging the lawyer's lapse looks standing alone.³⁹ The court asks what the decisionmaker actually had before it, then asks how the proceeding would have looked with the error corrected against that background. The earlier discussion settled how much doubt a defendant must raise, and this inquiry supplies the material that doubt is measured against. The strength of the state's case therefore drives the outcome, and an identical error can earn relief on one record and none on another.

The reason follows from what the prejudice prong protects. Prejudice asks whether confidence in the result is undermined, and confidence in a verdict depends on how much the verdict rested on the evidence the error touched. A conviction supported by several independent strands of proof is hardly shaken when one strand is weakened, because the remaining strands still carry it. A conviction that hung on a single contested witness can be shaken by one omission, because nothing else holds it up. Measuring against the whole record also keeps a court from judging each error separately, where a dramatic-looking lapse might seem harmful even though the case it touched would have ended the same way.

What the court weighs depends on what the claim attacks. If the alleged deficiency goes to the finding of guilt, the defendant must show that competent counsel would have taken steps that might have led to an acquittal or to a conviction on a lesser charge, and the court examines the strength of the evidence of guilt that was introduced and whether a competent lawyer would likely have cast doubt on it.⁴⁰ If the deficiency goes only to the sentence, the same method runs on a different record, and the court asks whether, had counsel performed adequately, there is a reasonable probability that a lower sentence would have been imposed.⁴¹ Sentencing claims carry refinements of their own, which are taken up below under Beyond the Trial.

Two decisions show a strong record absorbing an error, and they differ in what was being weighed. In Strickland v. Washington the defendant pleaded guilty to three capital murders, and his lawyer presented little mitigating evidence at sentencing. The Court found counsel's choice a reasonable strategic one, and added that even if it were not, the aggravating circumstances were overwhelming and the omitted mitigating evidence would not have altered the outcome.⁴² That was a sentencing record, so the weighing set what the missing evidence would have added against the aggravation already before the sentencer. In Berghuis v. Thompkins, counsel failed to request an instruction limiting how the jury could use the outcome of an accomplice's separate trial. The Court assumed that failure was deficient and still found no prejudice, because the surviving victim identified Thompkins as the shooter and a surveillance photograph supported the identification, a friend testified that Thompkins confessed, and the details of that confession were corroborated by evidence that Thompkins stripped and abandoned the van after the shooting.⁴³ That was a guilt record, so the weighing set the missing instruction against the evidence of guilt, and the court of appeals had erred by neglecting that other evidence.⁴⁴

The pattern generalizes. Where the evidence of guilt is overwhelming and the defense lawyer made only a few relatively minor mistakes, the prejudice prong will be decisive against the defendant.⁴⁵ Both decisions also reached prejudice while assuming or leaving open the quality of the lawyering, which is why courts so often begin with prejudice, a practice the next subsection explains.

The opposite pole appears where the state's case is thin. Kyles v. Whitley arose from suppressed evidence rather than a lawyer's error, but it carried the same reasonable-probability formulation into that setting, and it shows how a court weighs a record that rests on a few witnesses.⁴⁶ The state's case there was essentially eyewitness testimony, and disclosure of the witnesses' earlier statements would have substantially reduced or destroyed the value of two of them, leaving a markedly weaker prosecution and a markedly stronger defense.⁴⁷ The Court also insisted on a cumulative evaluation of the evidence rather than a series of independent evaluations of individual items.⁴⁸ It treated the question as one of confidence in the verdict instead of sufficiency, so a defendant need not show that, after discounting the inculpatory evidence, too little would remain to convict.⁴⁹

The lawyer's error is the same in both trials, and what moved the answer is the record around it. In the first trial the omitted impeachment would have weakened one strand of proof among several, so a court weighing the whole record can say that confidence in the verdict survives. In the second the impeachment reaches the only evidence on identity, so the same omission goes to the foundation of the conviction and confidence may well fail.

Two mistakes recur on this point. The first is evaluating the error in isolation from the rest of the record, asking only whether the omitted impeachment, instruction, or mitigation was impressive in itself, when the cases above weigh it against everything else the decisionmaker had. The second runs the other way and concludes that prejudice is absent because enough evidence remained to convict. The test asks whether confidence in the verdict survives, and a defendant on a thin or contested record can meet it without showing that the remaining proof would be insufficient.⁵⁰

The second step is where claims are usually won or lost, since a defendant's best argument is often that the state's case was less corroborated than it looked. The fourth step ties the analysis back to the confidence standard from the reasonable-probability discussion, so no numerical threshold enters at this stage.

Order of Analysis

Both prongs of the test must be satisfied, but nothing in the test dictates which one a court takes up first. Strickland v. Washington said that a court need not decide whether counsel's performance was deficient before examining the prejudice the defendant suffered, and that if a claim is easier to dispose of for lack of sufficient prejudice, that course should be followed.⁵² The Court described its standards as guides rather than mechanical rules, because the ultimate focus is the fundamental fairness of the proceeding whose result is challenged.⁵³ The order in which the prongs are stated therefore says nothing about the order in which they must be decided. The prejudice prong developed in the preceding subsections can be the whole case.

The permission to skip deficiency has its own logic. Judging performance is demanding work: a court must reconstruct the circumstances of counsel's challenged conduct, evaluate it from counsel's perspective at the time, and indulge a strong presumption that the conduct was reasonable.⁵⁴ When the record shows that the claimed error could not have changed the result, that effort buys nothing, because a finding that the lawyer erred would not make the judgment any less trustworthy. Skipping the performance prong therefore conserves judicial effort without costing the defendant a claim that could have succeeded, since a claim that fails on prejudice fails however the performance question comes out.

In practice, courts often assume deficient performance for the sake of argument and test prejudice on that assumption.⁵⁵ If no prejudice is found, the defendant loses and there is no reason to inquire into the lawyer's performance at all.⁵⁶ In Berghuis v. Thompkins, counsel failed to request a limiting instruction about an accomplice's earlier trial, and the Court assumed that omission was deficient representation. It then held that Thompkins could not show prejudice, because the surviving victim identified him as the shooter with support from a surveillance photograph, a friend testified that he confessed, and the details of the confession were corroborated by evidence that he stripped and abandoned the van.⁵⁷ An assumption of this kind is not a finding. The Court never said the lawyer erred, only that the error, if it was one, made no difference. The same path is open whenever the evidence of guilt is overwhelming and the alleged mistakes are minor.⁵⁸ Courts may also decide both prongs, as in Strickland itself, where the Court found counsel's conduct could not be held unreasonable and added that, even if it could, the prejudice was insufficient.⁵⁹

The shortcut runs in one direction only. A court may skip deficiency to deny a claim, but it may not skip deficiency to grant one, because relief requires both showings.⁶⁰ Because the Sixth Amendment guarantees effective assistance and a lost trial proves nothing about that by itself, a convicted defendant can nearly always point to some different tactic that might have produced a better result. The performance prong screens those claims by asking whether counsel's choice was reasonable when it was made, with the distorting effects of hindsight set aside.⁶¹ If prejudice alone could carry a claim, that screen would disappear, and the guarantee of competent counsel would become a guarantee of whatever strategy later seems best. Keeping both prongs mandatory is what stops the efficiency rule from becoming a route to relief without proof of deficient performance.

The contrast with Berghuis v. Thompkins shows which prong was doing the work. There prejudice failed, so performance never had to be decided and the claim ended. In the hypothetical, prejudice was shown, so performance had to be decided and the claim ended the other way. Students most often lose points by treating prejudice alone as enough, reasoning that if a better lawyer would have won, a new trial must follow. The mirror-image error is to read the permission to skip deficiency as a command to begin with prejudice. A court may start with either prong, and the choice follows whichever is easier to resolve.

The first step is the only place the shortcut operates, and on an exam it is often the most efficient route to a correct answer when the facts describe a lopsided case for the state.