After a Massachusetts jury convicted petitioner-appellee Barbara Ouber on a drug-trafficking charge, she exhausted her state-court remedies and then sought habeas corpus relief in the federal district court. That court granted the writ. The Commonwealth's ensuing appeal raises nuanced questions concerning the interplay between the proper resolution of claims asserting ineffective assistance of counsel and the deferential standard of review imposed upon federal habeas courts by the Antiterrorism and Effective Death Penalty Act (AEDPA), Pub.L. No. 104-132, 110 Stat. 1214 (1996). Although our reasoning differs significantly from the district court's as to the prejudice component of the ineffective assistance test, we agree that habeas relief is appropriate in the unique circumstances of this case.

To put matters into perspective, we recount the background facts, the case's procedural history, the genesis of the petitioner's conviction (including a précis of the evidence adduced at trial), and what transpired thereafter.

The petitioner and her brother (Nick Tsoleridas) resided at 9 Beth Lane in Hyannis, Massachusetts. On January 25, 1992, Todd Shea, an undercover narcotics agent, accompanied by a confidential informant (CI), went to that address. Tsoleridas greeted them. He and the CI then went into the house. Shea was told to wait in the car.

The CI emerged alone. He and Shea waited for Tsoleridas (a suspected drug dealer). After some time had elapsed, the two men grew impatient and approached the front door. The petitioner was standing just inside the entrance and Tsoleridas was descending from upstairs. Shea said something to the effect that he wanted to look at "the package" before turning over any money. Tsoleridas escorted his visitors outside, saying that he did not want to "deal" in the house. After the three men drove a short distance, Tsoleridas sold Shea an ounce of cocaine for $1,100. He also gave Shea his cell phone number and told him that he could supply much larger quantities.

Tsoleridas delivered comparable amounts of cocaine to Shea on February 19 and March 2. At approximately 4:40 p.m. on March 8, Shea called Tsoleridas and indicated that he wished to purchase ten ounces of cocaine. Tsoleridas tried to persuade Shea to come to Boston to consummate the transaction. When Shea demurred, Tsoleridas offered to supply two ounces to tide him over, and told him to come to the parking lot of Bud's Country Lounge in Hyannis where Tsoleridas's sister would exchange the drugs for $2,000.

Shea testified that the transaction occurred as follows. He reached the parking lot at the appointed time. He saw the petitioner arrive, driving a Toyota. When he entered the passengers' side of the Toyota, the petitioner identified herself as Tsoleridas's sister and handed him two sealed envelopes. Shea asked if this was the same "coke" as before and if the envelopes aggregated the agreed quantity. After receiving an affirmative response, he gave the petitioner $2,000. She counted the money and dropped the bills on the floor of the Toyota. Meanwhile Shea broke the seals, withdrew a clear plastic bag from inside each envelope, and inspected the contents. He then debarked, entered his own vehicle, and departed with the contraband.

At the time of the transaction, the parking lot was deserted except for two law enforcement officers who were observing from a distance. They saw very little. One of them testified, however, that he watched the Toyota enter the parking lot and leave a few minutes after Shea exited the vehicle.

On March 13, Tsoleridas sold Shea the ten ounces of cocaine that Shea had "ordered." Shortly thereafter, the authorities searched the house at 9 Beth Lane and found drugs, large sums of cash, and drugrelated paraphernalia. The petitioner was present during the search. When she asked to see the warrant, however, the officers claimed to have lost it.

A Barnstable County grand jury subsequently indicted both Tsoleridas and the petitioner for trafficking in cocaine. See Mass. Gen. Laws ch. 94C, § 32E(b). The petitioner was tried on a single charge, based upon her alleged complicity in the March 8 transaction. She stood trial alone (Tsoleridas having fled the country). Shea and the petitioner were the main witnesses, and they gave sharply conflicting accounts as to what had occurred inside the Toyota.

Shea's testimony was along the lines described above. The petitioner, however, testified that she knew nothing of the drugs, but had been coerced by her brother into doing what she thought was a non-drug-related errand for him. Her version of what happened in the Toyota differed from Shea's in no fewer than four crucial respects. She denied having handed Shea the envelopes, saying that he removed them from the right front seat. She also denied that she and Shea had the conversation he described (or any conversation relating, directly or indirectly, to cocaine). She denied that she counted the money, instead saying that Shea threw it at her (with the result that the bills fluttered to the floor of the Toyota). And, finally, she denied that Shea opened the envelopes or inspected their contents in her presence.

To buttress this account, the defense presented the testimony of the petitioner's friend, Patricia Gisleson. Gisleson testified that she was at the petitioner's home on March 8 and overheard Tsoleridas and the petitioner arguing. Tsoleridas wanted her to deliver two envelopes for him. After the petitioner succumbed to Tsoleridas's bullying, Gisleson helped to move the petitioner's Toyota out of the garage. In the process, she noticed that Tsoleridas had placed two sealed envelopes on the front passenger's seat. The petitioner then drove away. Gisleson was still at 9 Beth Lane when the petitioner returned. The petitioner seemed very upset.

Due to the fact that the search party had been unable to display a warrant, a suppression order issued. Thus, the Commonwealth could not introduce the evidence seized in the house search during its case in chief. After the petitioner testified, however, the trial justice allowed the Commonwealth to introduce that evidence for impeachment purposes. Following arguments of counsel and the court's charge, the jurors could not reach agreement and the trial justice declared a mistrial.

The Commonwealth elected to retry the petitioner. Much the same proof scenario obtained at the second trial, except that Gisleson's testimony was much more detailed. She stated, inter alia, that Tsoleridas had slapped the petitioner when she initially refused to do his bidding. She also elaborated on the reason that Tsoleridas gave for wanting the petitioner to run the errand: the man she was to meet owed him money, and she was to give the man some drill bits and collect $2,000. Then, too, Gisleson volunteered that the petitioner had told her that, when she met Shea, he had thrown the money at her. Despite Gisleson's more expansive testimony, the jury deadlocked once again.

This brings us to the third trial. Because of their relationship to the issues on appeal, we describe the events that played out during this trial in greater detail.

As the third trial began,1 the petitioner's counsel — the selfsame lawyer who had represented her at the two earlier trial — selected to deliver his opening statement on the heels of the prosecutor's opening. In the course of this statement, the lawyer promised — not once, but four times — that the petitioner would testify. In the bargain, the lawyer emphasized the importance of this testimony. He pointed out that the case revolved around the petitioner's knowledge (or lack of knowledge) that the envelopes delivered to Shea contained cocaine, and that her version of the relevant events — particularly those that transpired in the car — was very different from Shea's. Counsel's peroration drove home these points. He told the jurors:

The case is going to come down to what happened in that car and what your findings are as you listen to the credibility and the testimony of Todd Shea versus what you[r] findings are as you listen to the testimony of Barbara Ouber.

. . . .

.... You're going to hear a difference of opinion as to whether [the envelopes] were handed to Mr. Shea, whether he opened them in front of her; and as to the conversation.

And you're going to have to decide the truth and veracity of those two witnesses; and that will be your ultimate decision in this case.

As in the earlier trials, the Commonwealth's case in chief hinged on Shea's testimony. His direct examination yielded the version of the transaction described above. On cross-examination defense counsel brought out a few inconsistencies (e.g., that Shea originally had claimed that the envelopes were unsealed when he received them whereas he now admitted that they were sealed). Defense counsel also attempted to show that Tsoleridas's actions on the occasion of Shea's first visit to 9 Beth Lane indicated that Tsoleridas was trying to conceal his drug trafficking from the petitioner.

Up to a point, the defense case seemed similar to that presented in the previous trials. The defense paraded a large number of character witnesses before the jury, including an Eastern Orthodox bishop and several priests from the petitioner's community. These witnesses were unanimous in attesting to the petitioner's good character and reputation for veracity. A number of them did double duty, declaring that Tsoleridas was abusive and domineering insofar as his sister was concerned. Gisleson also testified along the same lines as at the second trial — although she again added new details. These embellishments included testimony that Tsoleridas had threatened to kill the petitioner if she did not go to meet Shea; that the petitioner told Gisleson, after she returned, that the man she met had tried to get her to enter is vehicle; and that the petitioner never touched the envelopes.2

The trial then veered dramatically from the previous iterations. Although the petitioner had testified in both of the earlier trials, this time around the defense tested without calling her as a witness. Closing arguments followed. In his summation, the petitioner's attorney apologized for not presenting "more of a case" as he had promised, but opined that elements of Shea's and Gisleson's testimony supported a claim that the petitioner lacked knowledge of the envelopes' contents. The prosecutor responded that Shea's testimony, taken as a whole, showed that the petitioner was fully aware that the envelopes contained cocaine, and that there was no reason to doubt his credibility. The prosecutor contrasted this testimony with Gisleson's, which, he argued, had been tailored to protect the petitioner.

Jury deliberations began that afternoon, but court adjourned without a verdict. Deliberations resumed the next morning. Sounding a familiar refrain, the jurors soon reported that they were deadlocked. The trial justice urged them to deliberate further, giving them a supplemental instruction based on Commonwealth v. Rodriquez, 364 Mass. 87, 300 N.E.2d 192, 202-03 (Mass.1973) (suggesting suitable language for a "dynamite" charge). Later that day, the jury found the petitioner guilty as charged.

The petitioner moved for a new trial based on ineffectiveness of counsel. To understand the etiology of that claim, we must explore the genesis of the petitioner's decision not to testify. We glean the relevant facts, as did the state courts, primarily from affidavits submitted by the petitioner and her trial attorney in support of the ineffectiveness of counsel claim.

The third trial lasted only two days. According to the lawyer, he first focused on the possibility of withholding the petitioner's testimony on the evening of the first day (after the Commonwealth had rested). A discussion took place in which several priests and other friends of the petitioner participated. The petitioner apparently wanted to testify, but the lawyer persuaded her that it would be in her best interest not to do So.3 The following day, the lawyer conferred privately with the petitioner, but on the record (i.e., in the presence of a court reporter), so that the petitioner could confirm that she had decided not to testify. The affidavits and the record of that lobby conference make clear, however, that counsel's earlier promises to the jury were not discussed, and that the petitioner was never advised that her decision to refrain from testifying might be counterproductive in light of those promises. This confluence of factors — the decision to withhold the petitioner's testimony after having emphasized its importance and having repeatedly promised the jurors that they would hear it — constituted the essence of the petitioner's ineffective assistance claim.

The state courts were unreceptive to the petitioner's plea. The trial justice denied the motion for a new trial, and the Massachusetts Appeals Court affirmed the trial justice's order. See Commonwealth v. Ouber, 46 Mass.App.Ct. 1112, 707 N.E.2d 408 (Mass.App.Ct.1999) (table). The appellate court concluded that the petitioner's lawyer approached the question of whether she should testify "cautiously" and advised her to remain silent because she likely would suffer grievously in cross-examination. Because the attorney was "working with an intrinsically weak defense," the court, applying the test articulated in Commonwealth v. Saferian, 366 Mass. 89, 315 N.E.2d 878, 882-83 (Mass. 1974), found his performance constitutionally acceptable. As a fallback, the court observed that the attorney's advice did not prejudice the petitioner because the evidence against her was solid and the jury had been instructed not to draw a negative inference from her silence. The court made only a passing reference to the promises contained in counsel's opening statement, characterizing them as neither "dramatic" nor "memorable." The court added that, when the petitioner decided not to testify, she knew what the consequences would be because she had been through two trials and "[a]n inference about the jury's possible attitude would not be remote or difficult."

In due course, the Massachusetts Supreme Judicial Court (SJC) denied further appellate review. Commonwealth v. Ouber, 429 Mass. 1104, 709 N.E.2d 1120 (Mass. 1999) (table). The petitioner then repaired to the federal district court and prosecuted an application for a writ of habeas corpus against the appropriate state correctional official. See 28 U.S.C. § 2254. The district court found the Appeals Court's decision to be an unreasonable application of the standard articulated in Strickland v. Washington, 466 U.S. 668, 687, 104 S. Ct. 2052, 80 L. Ed. 2d 674 (1984). See Ouber v. Guarino, 158 F. Supp. 2d 135, 149 (D.Mass. 2001). In the district court's view, the state court improperly focused on a peripheral matter — whether the petitioner was (or was not) fully informed about her right to testify when she decided to remain silent—and brushed aside the critical error in professional judgment: making a promise to the jury and then breaking it. Id. at 150. The court found it unreasonable that the Appeals Court did not evaluate the attorney's advice in light of the initial promises that had been communicated to the jury. Id. at 153. On this basis, the court concluded that the lawyer's actions fell below the Strickland benchmark and that the state court's application of Strickland's performance prong was unreasonable. Id. at 154. To cap matters, the court found that the state court had applied the wrong test as to prejudice and concluded that prejudice should be presumed in this case. See id. at 155. The court then went a step further and found, in the alternative, that the lawyer's error was outcome-determinative. Id. at 155-56.

Consistent with these findings, the district court ordered the petitioner relieved from her sentence unless the Commonwealth vacated her conviction and afforded her a new trial within a stipulated time frame. See id. at 156. This appeal ensued. The petitioner has remained free on bail pending the outcome of the habeas proceeding.

As said, this appeal turns on the interplay between the constitutional standard articulated in Strickland and the limited review permitted by the AEDPA in habeas cases. We comment on each of these elements.

The controlling principles for deciding ineffective assistance of counsel claims are limned in Strickland. Under these principles, a defendant alleging ineffective assistance of counsel must establish two elements in order to prevail:

First, the defendant must show that counsel's performance was deficient. This requires showing that counsel made errors so serious that counsel was not functioning as the "counsel" guaranteed the defendant by the Sixth Amendment. Second, the defendant must show that the deficient performance prejudiced the defense.

Strickland, 466 U.S. at 687, 104 S. Ct. 2052.

As to the first element, "[j]udicial scrutiny of counsel's performance must be highly deferential." Id. at 689, 104 S. Ct. 2052. The practice of law is not a mechanical exercise (like, say, kicking a foot press), and an inquiring court must leave ample room for variations in professional judgment. See id. By like token, a reviewing court must not lean too heavily on hindsight: a lawyer's acts and omissions must be judged on the basis of what he knew, or should have known, at the time his tactical choices were made and implemented. Bell v. Cone, 535 U.S. ___, 122 S. Ct. 1843, 1854, 152 L. Ed. 2d 914 (2002); United States v. Natanel, 938 F.2d 302, 309 (1st Cir.1991). Only if, "in light of all the circumstances, the [alleged] acts or omissions of counsel were outside the wide range of professionally competent assistance," can a finding of deficient performance ensue. Strickland, 466 U.S. at 690, 104 S. Ct. 2052.

The second Strickland element ensures that, even if a lawyer's performance is constitutionally unacceptable, relief will be withheld unless the quondam client has demonstrated that "there is a reasonable probability that, but for counsel's unprofessional errors, the result of the proceeding would have been different." Id. at 694, 104 S. Ct. 2052. While this level of prejudice may be presumed in a few settings, id. at 692, 104 S. Ct. 2052, that is the exception, not the rule. For the most part, the petitioner must carry the devoir of persuasion and prove that he was prejudiced, i.e., that his attorney's parlous conduct may have altered the outcome of the case. See Smith v. Robbins, 528 U.S. 259, 285-86, 120 S. Ct. 746, 145 L. Ed. 2d 756 (2000). In this regard, we caution that, although the possibility of a different out-come must be substantial in order to establish prejudice, it may be less than fifty percent. See Strickland, 466 U.S. at 693, 104 S. Ct. 2052 (explaining that "a defendant need not show that counsel's deficient conduct more likely than not altered the outcome in the case").

Under the AEDPA, a federal court may grant habeas relief to a state prisoner only if the state court adjudication

(1) resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States; or

(2) resulted in a decision that was based on an unreasonable determination of the facts in light of the evidence presented in the state court proceeding.

28 U.S.C. § 2254(d).

A state-court decision fits within the "contrary to" rubric if the state court either applies a legal rule that contradicts an established Supreme Court precedent or reaches a different result on facts materially indistinguishable from those of a controlling Supreme Court precedent. Williams v. Taylor, 529 U.S. 362, 405-06, 120 S. Ct. 1495, 146 L. Ed. 2d 389 (2000). Where a relevant but not factually congruent precedent exists, the state court need only apply a test consistent with the one announced by the Supreme Court in order to avoid the toils of section 2254(d)(1)'s "contrary to" clause.

The "unreasonable application" component of section 2254(d)(1) comes into play when the state court identifies the correct legal principle, but unreasonably applies that principle to the facts of the prisoner's case. Williams, 529 U.S. at 407-08, 120 S. Ct. 1495. The "unreasonable application" clause also encompasses situations in which a state court either unreasonably extends a legal principle derived from Supreme Court precedent to an inappropriate context or unreasonably refuses to extend that principle to an appropriate context. Id. In all events, a state-court decision must be unreasonable, as opposed to merely incorrect, before a federal court can grant habeas relief. Id. at 410, 120 S. Ct. 1495.

The AEDPA also requires that the relevant legal rule be clearly established in a Supreme Court holding, rather than in dictum or in holdings of lower federal courts. Id. at 412, 120 S. Ct. 1495. This does not mean, however, that other federal court decisions are wholly irrelevant to the reasonableness determination. "To the extent that inferior federal courts have decided factually similar cases, reference to those decisions is appropriate in assessing the reasonableness vel non of the state court's treatment of the contested issue." O'Brien v. Dubois, 145 F.3d 16, 25 (1st Cir.1998). Reference to such cases may be especially helpful when the governing Supreme Court precedent articulates a broad principle that applies to a wide variety of factual patterns.

So it is here. The Strickland principles for deciding ineffective assistance of counsel claims are "clearly established" for purposes of the AEDPA. See Williams, 529 U.S. at 371-74, 120 S. Ct. 1495. Because the Supreme Court has yet to adopt more particularized guidelines for ineffectiveness of counsel claims, it is helpful to examine precedents from lower federal courts to determine how the general standard applies to a particular set of facts. Although such decisions are not themselves binding on a state court under the AEDPA framework, see id. at 412, 120 S. Ct. 1495, resort to them is appropriate for the purpose of discerning the requirements of Strickland in factually similar cases. See Mountjoy v. Warden, N.H. State Prison, 245 F.3d 31, 35-36 (1st Cir. 2001).

Another category of state-court errors that may be remedied on federal habeas review involves unreasonable determinations of fact. See 28 U.S.C. § 2254(d)(2). Under this standard, the state court's factual findings are entitled to a presumption of correctness that can be rebutted only by clear and convincing evidence to the contrary. Mastracchio v. Vose, 274 F.3d 590, 597-98 (1st Cir.2001). But the special prophylaxis of section 2254(d)(2) applies only to determinations of "basic, primary, or historical facts." Sanna v. Dipaolo, 265 F.3d 1, 7 (1st Cir.2001). Inferences, characterizations of the facts, and mixed fact/law conclusions are more appropriately analyzed under the "unreasonable application" prong of section 2254(d)(1). Cf. Townsend v. Sain, 372 U.S. 293, 309 n. 6, 83 S. Ct. 745, 9 L. Ed. 2d 770 (1963) (stating that mixed questions of fact and law do not fall within the purview of section 2254(d)(2)); Sanna, 265 F.3d at 7 (suggesting that only witness credibility and recitals of external events qualify as basic or primary facts for purposes of section 2254(d)(2)). Inasmuch as "both the performance and the prejudice components of the ineffectiveness inquiry are mixed questions of law and fact" for the purposes of federal habeas review, Strickland, 466 U.S. at 698, 104 S. Ct. 2052, section 2254(d)(2) is of limited utility in this case.

Consistent with the Strickland paradigm, we divide our analysis into two parts: performance and prejudice.

At the heart of this appeal lies a broken promise (or, more precisely put, a series of broken promises): defense counsel's repeated vow that the jurors would hear what happened from the petitioner herself. Thus, the error attributed to counsel consists of two inextricably intertwined events: the attorney's initial decision to present the petitioner's testimony as the centerpiece of the defense (and his serial announcement of that fact to the jury in his opening statement) in conjunction with his subsequent decision to advise the petitioner against testifying. Taken alone, each of these decisions may have fallen within the broad universe of acceptable professional judgments. Taken together, however, they are indefensible. Neither the state court nor the Commonwealth has managed to identify any benefit to be derived from such a decisional sequence, and we are unable to see the combination as part and parcel of a reasoned strategy. We therefore conclude that, in the absence of unforeseeable events forcing a change in strategy, the sequence constituted an error in professional judgment. Cf. Anderson v. Butler, 858 F.2d 16, 19 (1st Cir.1988) (finding a mistake, rather than a strategic choice, where nothing could be gained from counsel's approach).

This assessment does not end our inquiry. The complex dynamics of trial engender numerous missteps, but only the most inexcusable will support a finding that counsel's performance was so substandard as to compromise a defendant's Sixth Amendment right to proficient legal representation. See, e.g., Nix v. Whiteside, 475 U.S. 157, 164-65, 106 S. Ct. 988, 89 L. Ed. 2d 123 (1986) (quoting Strickland, 466 U.S. at 687, 104 S. Ct. 2052). To separate wheat from chaff — lapses of constitutional dimension from garden-variety bevues — we must assess the gravity of the error and then consider potential justifications for the attorney's actions, given what he knew or should have known at each relevant mo-ment in time. See Natanel, 938 F.2d at 309. And, finally, because this case comes to us on habeas review, we must examine the reasonableness of the state-court conclusion that counsel's performance was not constitutionally deficient. We turn to these interrelated tasks.

It is apodictic that a defendant cannot be compelled to testify in a criminal case, see U.S. Const. amend. V, and criminal juries routinely are admonished — as was the jury here — not to draw an adverse inference from a defendant's failure to testify. But the defendant has the right to testify in her own defense, and, when such testimony is proffered, the impact on the jury can hardly be overestimated. See Green v. United States, 365 U.S. 301, 304, 81 S. Ct. 653, 5 L. Ed. 2d 670 (1961) ("The most persuasive counsel may not be able to speak for a defendant as the defendant might, with halting eloquence, speak for himself."). When a jury is promised that it will hear the defendant's story from the defendant's own lips, and the defendant then reneges, common sense suggests that the course of trial may be profoundly altered. A broken promise of this magnitude taints both the lawyer who vouchsafed it and the client on whose behalf it was made.

The Commonwealth argues that a defendant's decision about whether to invoke the right to remain silent is a strategic choice, requiring a balancing of risks and benefits. Under ordinary circumstances, that is true. It is easy to imagine that, on the eve of trial, a thoughtful lawyer may remain unsure as to whether to call the defendant as a witness. If such uncertainty exists, however, it is an abecedarian principle that the lawyer must exercise some degree of circumspection. Had the petitioner's counsel temporized — he was under no obligation to make an opening statement at all, much less to open before the prosecution presented its case, and, even if he chose to open, he most assuredly did not have to commit to calling his client as a witness — this would be a different case. See Phoenix v. Matesanz, 233 F.3d 77, 85 (1st Cir.2000) (finding no ineffectiveness where, in the absence of an express promise, counsel chose not to call a potentially important witness).

Here, however, the circumstances were far from ordinary. The petitioner's counsel elected to make his opening statement at the earliest possible time. He did not hedge his bets, but, rather, acted as if he had no doubt about whether his client should testify. In the course of his opening statement, he promised, over and over, that the petitioner would testify and exhorted the jurors to draw their ultimate conclusions based on her credibility. In fine, the lawyer structured the entire defense around the prospect of the petitioner's testimony.4

In the end, however, the petitioner's testimony was not forthcoming. Despite the fact that the lawyer had called the petitioner to the stand in both prior trials, he did a complete about-face. The lawyer states in his affidavit that he only realized that keeping his client off the witness stand was an option after the first day of trial. This realization came much too late. Indeed, the attorney's delayed reaction is sharply reminiscent of the situation in Anderson, in which we observed that even "if it was ... wise [not to have the witness testify] because of the damaging collateral evidence, it was inexcusable to have given the matter so little thought at the outset as to have made the opening promise." 858 F.2d at 18.

The Commonwealth argues that defense counsel's mid-trial decision should be excused as a justified reaction to unfolding events. The theoretical underpinnings for this argument are sound: unexpected developments sometimes may warrant changes in previously announced trial strategies. See, e.g., Dutton v. Brown, 812 F.2d 593, 598 (10th Cir.1987). But although we cannot fault counsel for not guarding against the unforeseeable the case at hand does not fit that description. Here, everything went according to schedule; nothing occurred during the third trial that could have blindsided a reasonably competent attorney or justified a retreat from a promise previously made. After all, the petitioner's lawyer had represented her during two previous trials for the same offense; the prosecution's case in chief did not differ significantly at the third trial; and the situation that confronted the attorney when he changed his mind about the desirability of presenting the petitioner's testimony was no different from the situation that existed at a comparable stage of the earlier trials.

The Commonwealth suggests that the tenor of Shea's testimony justified counsel's last-minute change of heart. Shea's testimony, it says, was stronger and more consistent this time around. The record belies this claim; it shows beyond hope of contradiction that the new wrinkles in Shea's testimony were of marginal significance. Some uncertainties were clarified on direct examination in preparation for the defense's cross-questioning, but this slight tightening-up of the prosecution's case should readily have been anticipated. What is more, even if Shea's testimony was less vulnerable than originally predicted, it remains a mystery why, in response to adverse evidence that proves stronger than expected, a lawyer should decide to abandon the only available avenue of controverting it.

The Commonwealth has another arrow in its quiver: it asserts that, had the petitioner testified, she would have been heavily impeached (and, thus, the decision not to testify was a legitimate one). Because of the damaging evidence that was available for impeachment had the petitioner testified — the drugs and cash found in the search — this argument has a patina of plausibility. The difficulty, however, is that counsel knew of this sword of Damocles — the threat that the impeaching evidence would be introduced — when he made his opening statement.5 Indeed, that evidence was used to cross-examine the petitioner during the two prior trials, and counsel appeared ready, willing, and able to handle that contingency.

The Commonwealth next argues that enough of the petitioner's story was presented through Gisleson that counsel reasonably could have advised the petitioner not to testify. This is little more than whistling past the graveyard. Gisleson was not present when Shea and the petitioner met on March 8, and so could only relate what she saw and heard before the petitioner left the house and after the petitioner returned. Thus, Gisleson's testimony, on its own, neither provided an adequate defense for the petitioner nor fulfilled the explicit promises made to the jury in the lawyer's opening statement.6

In all events, Gisleson had testified at each of the earlier trials, and defense counsel knew the substance of her testimony when he promised the jury that the petitioner would testify at the third trial. We add that, to the extent that Gisleson's testimony at the third trial contained variations from her two previous appearances as a witness, those variations do not change the calculus. Some of them — such as the more detailed account of her conversation with the petitioner after she (the petitioner) returned from the parking lot — were probably helpful to the defense, while others — such as her failure to explain that Tsoleridas often made his sister run errands related to his carpentry business — were perhaps detrimental. The inescapable fact, however, is that a witness's testimony is rarely identical two times running. Cf. Beachum v. Tansy, 903 F.2d 1321, 1326 (10th Cir.1990) (noting that "uncertainties and minor variations [are] normal to the recollection of honest witnesses after lapse of time"). Thus, the dispositive question must be whether, viewed as a whole, the testimony may be characterized as materially different. We think not: comparing Gisleson's testimony at the second and third trials, the differences are minor and amounted to neither a qualitative change nor an unexpected event justifying an abrupt switch in strategy.7

If more were needed—and we doubt that it is — the lawyer's about-face regarding the need for his client's testimony took place between the first and second day of trial. In other words, he changed his mind before Gisleson even testified. This chronology erases any suspicion that differences in Gisleson's testimony may have prompted the reversal of strategy.

The short of it is that, without exception, the events that occurred at the third trial should have been easily foreseeable to competent counsel at the time he made his opening statement. There were no surprises — and, thus, the lawyer's tergiversation could not be excused by changed circumstances. Compare, e.g., Magill v. Dugger, 824 F.2d 879, 887-88 (11th Cir. 1987) (finding ineffective assistance because counsel's strategy failed to account for foreseeable testimony), with Drake v. Clark, 14 F.3d 351, 356 (7th Cir.1994) (reaching the opposite conclusion when counsel's strategy was frustrated by an unforeseeable development).

Were we sitting in direct review, the foregoing analysis would lead us to find counsel's performance constitutionally unacceptable. In the exercise of habeas jurisdiction, however, we must take another step and evaluate the reasonableness of the Appeals Court's contrary conclusion. See 28 U.S.C. § 2254(d)(1). Strickland constitutes the established Supreme Court precedent, and the state court purported to apply the functional equivalent of Strickland's performance prong.8 Because it did so — and because the facts of this case differ significantly from those of Strickland — this case does not fit within the confines of section 2254(d)(1)'s "contrary to" clause insofar as counsel's performance is concerned. Rather, the crux of the matter is whether the state court applied Strickland's performance standard in an objectively reasonable manner when it determined that the lawyer's performance did not fall below the constitutional minimum.

We start this phase of our analysis with the text of the state-court decision. The state court first absolved the attorney from responsibility for failing to present the petitioner's testimony because the petitioner herself possessed enough sophistication to make such a decision. This determination misses the point of the petitioner's constitutional claim. While a decision about whether to testify ultimately rests with the defendant, see Rock v. Arkansas, 483 U.S. 44, 49-53, 107 S. Ct. 2704, 97 L. Ed. 2d 37 (1987), a defendant's waiver of the right to testify must be knowing, informed, and intelligent. This implies an understanding of the consequences of the decision. See United States v. Manjarrez, 258 F.3d 618, 623-24 (7th Cir.2001). Yet, the affidavits of both the petitioner and her trial counsel make it clear that she was not informed about the potential impact that the broken promises might have on the jury should she decide not to testify.9

A second problem with the state-court decision lies in its characterization of defense counsel's approach in his opening statement. Despite the unambiguous, emphatic, and oft-repeated comments regarding both the imminence and the salience of the petitioner's testimony, the state court asserted that "counsel approached cautiously the question of [the petitioner] testifying." Here, however, the record makes manifest that trial counsel's approach to the question of calling the petitioner as a witness — making an unconditional promise, repeating it four times over, and then breaking it without justification — was the antithesis of caution. Because the state-court's characterization is not borne out by any plausible reading of the record, we deem it unreasonable. See O'Brien, 145 F.3d at 25 (stating that if a state-court determination is devoid of record support, it fits within the "unreasonable application" prong of section 2254(d)(1)).

Finally, the state court offered only a single reason why counsel might legitimately have changed his mind about calling the petitioner to the witness stand, namely, that the barbed cross-examination of Gisleson intimated that the petitioner would undergo an even fiercer attack. That hypothesis does not withstand scrutiny. For one thing, counsel reversed course before Gisleson testified. For another thing, the Commonwealth's strongest attack on Gisleson relied on her knowledge of the suppressed evidence. Because Gisleson's cross-examination put some of that same information before the jury, see supra note 2, her testimony actually removed part of the rationale for not putting the petitioner on the witness stand. In other respects, Gisleson's testimony was no more severely impeached than in the previous trials — trials in which both Gisleson and the petitioner had testified and had held their own under withering cross-examination. At the very least, the petitioner's counsel should have anticipated the ferocity of potential cross-examination when he was deciding what to tell the jury in his opening statement. The Appeals Court's attempted justification is, therefore, plainly insupportable.

To sum up, counsel committed an obvious error, without any semblance of a colorable excuse. There is simply no record support for the state court's finding that the attorney's conduct constituted a reasonable strategic choice. To the contrary, the only sensible conclusion that can be drawn from this record is that the attorney's performance was constitutionally deficient under Strickland — and severely so. We hold, therefore, that the state-court finding on this point constituted an unreasonable application of the Strickland performance prong.

The remaining issue involves the state court's determination that counsel's performance, even if constitutionally deficient, did not prejudice the petitioner. The district court found fault with the state court's application of Commonwealth v. Saferian, supra, positing that, insofar as prejudice is concerned, Saferian articulates a standard contrary to Strickland. Ouber, 158 F.Supp.2d at 154 (arguing that Saferian concentrates on whether counsel's mistake deprived the accused of a substantial ground of defense rather than whether the mistake altered the outcome of the trial). We disagree with this assessment.

Although Strickland and Saferian do not employ identical phraseology, we have described those variations as "minor" and have concluded that, for habeas purposes, Saferian is a functional equivalent of Strickland. Scarpa, 38 F.3d at 7-8. That is the law of the circuit. Moreover, that interpretation squares with the relevant Massachusetts case law. While Saferian predated Strickland, the SJC since has concluded that Saferian is at least as solicitous of Sixth Amendment rights as Strickland. See Commonwealth v. Urena, 417 Mass. 692, 632 N.E.2d 1200, 1202 (Mass.1994); Commonwealth v. Fuller, 394 Mass. 251, 475 N.E.2d 381, 385 n. 3 (Mass.1985). In light of these precedents, we are unable to sustain the district court's conclusion that Saferian runs contrary to Strickland (and, thus, that the Appeals Court's decision is contrary to settled Supreme Court case law).

The district court committed another error when it ruled that the state court's "no prejudice" decision was unreasonable because prejudice must be presumed when an attorney inexcusably fails to carry out an announced promise to present an important witness. Ouber, 158 F.Supp.2d at 155. To the extent that the district court meant that the prejudice inquiry demanded by Strickland is superfluous in such a case, that holding is not grounded in any established Supreme Court precedent. To the contrary, the Court repeatedly has emphasized the limited nature of any exceptions to the general rule that a defendant must demonstrate actual prejudice. See Mickens v. Taylor, ___ U.S. ___, 122 S. Ct. 1237, 1246, 152 L. Ed. 2d 291 (2002); Smith, 528 U.S. at 287, 120 S. Ct. 746; Strickland, 466 U.S. at 692, 104 S. Ct. 2052.

As recently as May 28, 2002, the Court reiterated that prejudice may be presumed only in three narrowly circumscribed situations. Bell, 122 S.Ct. at 1850. First, a trial is presumptively unfair if the accused is completely denied the presence of counsel at a critical stage of the proceedings. Id. (citing, inter alia, Hamilton v. Alabama, 368 U.S. 52, 54, 82 S. Ct. 157, 7 L. Ed. 2d 114 (1961)). Second, such a presumption is warranted if "counsel entirely fails to subject the prosecution's case to meaningful adversarial testing." Id. at 1851 (quoting United States v. Cronic, 466 U.S. 648, 659, 104 S. Ct. 2039, 80 L. Ed. 2d 657 (1984)). Third, prejudice may be presumed in the presence of circumstances under which a competent lawyer would likely not be able to render effective assistance. Id. (citing Powell v. Alabama, 287 U.S. 45, 53 S. Ct. 55, 77 L. Ed. 158 (1932)).

In this case, the district court did not assert that any of these exceptions apply, and, in all events, the record would not support such an assertion. Instead, the district court appeared to read Anderson as carving out a new exception. Ouber, 158 F.Supp.2d at 154. Whether or not Anderson intended to do so is beside the point, since the weight of recent Supreme Court precedent is to the contrary. See, e.g., Bell, 122 S.Ct. at 1850-51; Mickens, ___ U.S. at ___, 122 S.Ct. at 1246. We have heeded the Court's clear message in the past, e.g., Scarpa, 38 F.3d at 11, and we are constrained to heed it here. Because the circumstances of this case do not fall within the contours of any of the three recognized exceptions to the Strickland

formulation, a presumption of prejudice cannot be condoned.

Setting the misplaced presumption to one side, we turn to the task of determining whether, on the facts of this case, the error was prejudicial. For this purpose, an error generally is considered prejudicial if there is a strong possibility that it affected the outcome of the trial. See Strickland, 466 U.S. at 693-94, 104 S. Ct. 2052. Consequently, we must consider, on whole-record review, whether the trial might have ended differently absent the lawyer's blunder. This is normally a difficult endeavor, but we are aided here by a unique circumstance: this was the petitioner's third trial, and the only substantial difference among those trials relates to the omission of her testimony at the third trial. Thus, unlike in the vast majority of cases, we have actual rather than hypothetical reference points to guide our inquiry.10

When the petitioner testified, two different juries found the prosecution's case so evanescent that they were unable to reach a verdict. Even without the petitioner's testimony, the jury in the third trial was deadlocked for a time. Given these facts, we are bound to conclude that the case was exceedingly close.

In a borderline case, even a relatively small error is likely to tilt the decisional scales. See, e.g., Frey v. Fulcomer, 974 F.2d 348, 369 (3d Cir.1992). The error here — failing to present the promised testimony of an important witness — was not small, but monumental. See, e.g., Anderson, 858 F.2d at 18-19; cf. United States v. Gonzalez-Maldonado, 115 F.3d 9, 15 (1st Cir.1997) (finding reversible error when the trial judge initially agreed that an important witness could testify, but later barred that witness from testifying). The net result of the failure to call the petitioner to the witness stand was that the jury heard only Shea's version of what transpired in the car. Yet, the petitioner's version would have been materially different with respect to certain critical aspects, such as whether Shea opened the envelopes in front of her and whether any verbal exchange regarding the contents (e.g., the weight and quality of the cocaine) occurred. Because these contradictions were not introduced into evidence, the jury never had an opportunity to assess the conflicting testimony or to weigh the petitioner's credibility against Shea's. What is worse, counsel's belated decision not to present the petitioner's testimony sabotaged the bulk of his efforts prior to that time (and, in the process, undermined his own standing with the jury, thereby further diminishing the petitioner's chances of success). Because the error was egregious, we are fully persuaded that, but for its commission, a different outcome might well have eventuated. Accordingly, the case satisfies the prejudice prong of the Strickland framework.

Under the AEDPA, an erroneous determination is not necessarily an unreasonable determination. Williams, 529 U.S. at 410, 120 S. Ct. 1495. Thus, it remains for us to address whether the state court's finding of no prejudice was not only incorrect but also unreasonable. The test is an objective one. See id. It focuses on the state court's ultimate conclusion rather than on the strength of the court's announced rationale. See Bui v. DiPaolo, 170 F.3d 232, 243-44 (1st Cir.1999) (stating that "state courts are not required to supply the specific reasons that a federal court thinks are most persuasive for upholding the judgment"), cert. denied, 529 U.S. 1086, 120 S. Ct. 1717, 146 L. Ed. 2d 640 (2000); accord Hurtado v. Tucker, 245 F.3d 7, 19 (1st Cir.), cert. denied, ___, 122 S. Ct. 282, 151 L. Ed. 2d 208 (2001); O'Brien, 145 F.3d at 25. In other words, the hallmark of a reasonable determination is the result reached by the state court, not the ratiocination leading to that result.

Here, the Massachusetts Appeals Court's "no prejudice" determination is not a credible outcome. That tribunal dealt with the question of prejudice vel non in a single paragraph, which reads:

To omit to call a witness who has been promised can, of course, be a serious mistake, but whether it is such in any given case is dependent on the circumstances, as the law recognizes. [string citations omitted] The promise here was not made dramatically or memorably, as it was in [Anderson ]. Counsel's apology in closing was brief and subdued. That the jury were not overcome by the unfulfilled promise is indicated by the fact that it took a [dynamite] charge to inspire the verdict. The verdict itself found solid support in the evidence. The judge charged against any invidious implication from the defendant's silence.

As can be seen, this paragraph contains several assertions — but these assertions are either irrelevant or devoid of record support. We explain briefly.

The state court opined that defense counsel's opening promises were not "dramatic" or "memorable." We find it hard to imagine, however, how the court could have reached that conclusion. The attorney made the promises explicitly and repeatedly. He also exhorted the jurors to base their ultimate decision on their collective assessment of the contrasting accounts that would be given by Shea and the petitioner, respectively. This call for a credibility judgment was the crowning element of the lawyer's opening statement and could not have failed to make an impression on the jury. The single promise in Anderson, 858 F.2d at 17 — a case in which counsel did not urge the jurors to rest their decision on the credibility of the witness who was promised but not produced — was certainly far less dramatic and memorable.

The state court also implied that the alleged error was inconsequential because the jury initially deadlocked and therefore was not immediately overborne by the detrimental effect of the broken promises. This argument effectively assumes that because a blunder did not lead to a summary conviction, it was of negligible effect. We believe that such an assumption is unreasonable; the fact that the jury convicted the petitioner only after prolonged deliberations and a supplemental "dynamite" charge necessarily underscores the closeness of the case (and, therefore, the gravity of any error).

The Appeals Court also posited that the petitioner's case was "intrinsically weak," and that the jury's verdict rested on solid evidence. The court, however, did not buttress these conclusory statements with any specific findings, and they are belied by the record. Indeed, the very fact that the first two trials ended in hung juries is powerful proof that those statements are insupportable. At each of those trials, the evidence marshaled against the petitioner was so underwhelming that the jurors were unable to reach a decision. The state court failed to consider this fact, or to suggest why doing so might be unhelpful.

The remaining factors mentioned by the Appeals Court bear little relevance to the prejudice inquiry. The fact that counsel's apology to the jury was "subdued" neither establishes the insignificance of the original promises nor palliates the effect of the mistake. The fact that the jury was advised not to draw a negative inference from the petitioner's failure to testify is likewise irrelevant; the attorney's mistake was not in invoking the petitioner's right to remain silent, but in "the totality of the opening and the failure to follow through." Anderson, 858 F.2d at 17.

To sum up, this was the petitioner's third trial and the only salient difference between it and the two prior trials was the absence of her testimony. This time around, defense counsel made a promise, hammered it home, and then broke it. The first two trials, at which the petitioner testified, offer a prime example of how this trial likely would have ended in the absence of this stunning error. We believe that it was unreasonable for the state court not to have taken such obvious reference points into account. Had it done so, it would have been bound to conclude that the case was a close one in which counsel's egregious error was likely to have made a dispositive difference.

That ends the matter. Since neither the state court's opinion nor our own careful perscrutation of the record reveals an objectively reasonable ground for the state court's "no prejudice" determination, we are constrained to set it aside.

We need go no further. Had the state court applied Strickland in an objectively reasonable manner, it would have been bound to conclude that defense counsel's abandonment of the oft-repeated promise that the petitioner would testify, enunciated in his opening statement, amounted to ineffective assistance of counsel in violation of the Sixth Amendment. The lawyer was intimately familiar with the case before he made this promise (having represented the petitioner in two prior trials on the same charges). Yet, he staked his client's defense on the strength of her testimony and then, with no discernible justification, changed his mind and decided that she should not testify. No significant change in circumstances occurred between the time of the lawyer's opening statement and the time of his about-face. This was a serious error in professional judgment, and the state court's contrary determination represented an unreasonable application of Strickland's performance prong.

Here, moreover, the prior trials serve as a meaningful benchmark for determining the likelihood that the outcome of the third trial was affected by the lawyer's mistake. Those trials, neither of which was marred by the same error, produced results materially different from the one reached in the third trial. Yet, the state court inexplicably failed to undertake this comparative analysis. We conclude, therefore, that the state court's harmlessness determination represented an unreasonable application of Strickland's prejudice prong.

For these reasons, we affirm the judgment of the district court. The petitioner shall be entitled to a writ of habeas corpus unless the Commonwealth affords her a new trial within the period prescribed.

Affirmed.