(dissenting, with whom Liacos, C.J., joins). I cannot agree with this court’s ruling that defense counsel’s decision not to interview Wiggins and Kelly and not to call them as impeachment witnesses was neither “manifestly unreasonable” nor prejudicial to the defendant’s case.1 In a thorough and tightly reasoned opinion, the Appeals Court held that counsel’s decision was not supported by any sound trial strategy and that in fact the defendant had demonstrated that his case was prejudiced thereby. Commonwealth v. White, 28 Mass. App. Ct. 417 (1990). I find nothing in this court’s opinion that justifies a different conclusion. I dissent.
I would adopt the Appeals Court’s opinion in its entirety. See 28 Mass. App. Ct. 417 (1990). It is not necessary to reproduce the Appeals Court’s analysis here. Instead, I confine my remarks to those elements of this court’s opinion with which I do not agree.
1. Conduct of defense counsel. The court dismisses counsel’s failure to interview Collette Wiggins and to call her to testify with a neat syllogism: (a) it was a “legitimate goal” of defense counsel to minimize the number of times the jury heard the complainant’s allegations of rape; (b) Wiggins would have repeated the complainant’s allegations; therefore (c) counsel’s decision not to call Wiggins to the stand was not “manifestly unreasonable.” Ante at 273. This analysis, though logically precise, fails to take into account the probative value of different aspects of Wiggins’s possible testimony.
While Wiggins may have repeated the complainant’s allegation, such testimony would merely have been cumulative of the fresh complaint testimony of others and would be only slightly significant as corroboration. On the other hand, Wiggins could also have testified that the complainant initially *279alleged only sexual harassment and stated that she had escaped the defendant’s advances. This testimony was potentially devastating to the prosecution’s case, which was built largely on the credibility of the complainant. Nevertheless, the defendant’s lawyer chose to ignore this evidence (not even telephoning Wiggins to confirm her version of the event), and relied instead on much less compelling impeachment evidence, such as the complainant’s failure to scream.
In essence, the asserted rationalization for this decision is that counsel feared that the complainant’s allegation of rape, even though its repetition was redundant, would ring in the jurors’ ears longer than the complainant’s initial denial that the rape had occurred, which the jury never heard.2 This calculation of the risks and rewards inherent in calling Wiggins to the stand is indefensible, the deference owed to tactical decisions of defense counsel notwithstanding. In light of the potential impact of Wiggins’s powerful impeachment testimony in a case where the prosecution relied almost entirely on the credibility of the complainant (producing no compelling physical evidence and no corroborating testimony apart from the fresh complaint witnesses), counsel’s omission was manifestly unreasonable. That counsel had a “legitimate goal” in deciding not to call Wiggins to testify (to wit, minimizing the number of times the jury heard the complainant’s allegation of rape) does not, without more, justify his decision.
The defense attorney’s decision not to offer Kelly’s testimony, though it involves a greater number of variables, is also unjustifiable. The court again rides to the lawyer’s rescue, however, by failing to weigh the potential costs and benefits of the omitted testimony.
*280The court states that counsel reasonably decided not to interview Kelly or call him to testify because he did not want the jury to learn that the defendant “had told Kelly what amounted to two different stories.” In reaching this decision, the court concludes, counsel properly relied on the defendant’s statement to counsel that, in two telephone conversations with Kelly, the defendant first denied any sexual contact with the complainant and then admitted to consensual sexual intercourse with her. Ante at 274.
Concededly, a defense attorney is entitled to rely on his client’s description of the events underlying a criminal charge. Strickland v. Washington, 466 U.S. 668, 691 (1984). When a lawyer subsequently gains information that supplements his client’s version of events, however, common sense dictates that the reasonableness of his tactical decisions be judged based upon all the evidence in his possession.3 In this case, counsel learned through discovery that the complainant, too, had changed her story, at first telling Kelly that she had escaped the defendant’s advances and alleging only sexual harassment by the defendant. Thus, the issue is not simply whether it was reasonable for counsel to rely on his client’s description of the conversations with Kelly.
Rather, the issue is the reasonableness of counsel’s calculation that, on balance, the damage to the defendant’s case would outweigh the damage to the prosecution’s case if it were revealed that both the defendant and the complainant had changed their stories. Given the considerable potential of Kelly’s testimony to create reasonable doubt by calling into question the complainant’s credibility, I conclude that counsel’s determination was manifestly unreasonable. Counsel’s decision not to interview Kelly, therefore, was also manifestly unreasonable.
Furthermore, as the court recognizes, had counsel interviewed Kelly, he would have discovered that Kelly’s testi*281mony would not have suggested that the defendant had contradicted himself. Instead Kelly would have testified that the defendant first denied only a charge of sexual harassment and later admitted only having consensual sexual intercourse with the complainant. See ante at 269 n.4. Both statements are entirely consistent with the defense’s theory of consent.
As the preceding discussion suggests, I disagree with the court’s estimation of the impeachment value of Kelly’s and Wiggins’s possible testimony concerning the complainant’s changing her story. The court concludes that this testimony would be problematic for two reasons. First, according to the court, the complainant’s inconsistent allegations (first that the defendant forcibly fondled her, and then that he raped her) would both rebut the defendant’s consent theory. Ante at 274. Since the jury had already heard the complainant’s version of a violent rape, the incremental value of her initial, lesser allegation as rebuttal would be negligible. Second, the court concludes that defense counsel could reasonably predict that a jury would not forget that the complainant “never abandoned her charge that the defendant dealt with her violently.” Ante at 274-275. This conclusion overlooks an essential difference between the complainant’s first and second allegations: the defendant was only on trial for one of them. That is, even if the jury remembered that the complainant consistently alleged that the defendant “dealt with her violently,” that allegation does not amount to rape. Therefore the omitted testimony would have raised doubts about the complainant’s credibility as to the ultimate question — whether the defendant raped her.
In short, the testimony that defense counsel chose to omit was less damaging to the defendant’s case than the court suggests and was the most compelling impeachment evidence available. Of course a reviewing court should examine the tactical or strategic decisions of the defendant’s lawyer with some deference. Commonwealth v. Rondeau, 378 Mass. 408, 413 (1979). Under the facts of this case, however, where the balance of risks and rewards so favored calling Kelly and *282Wiggins to testify, I conclude that the decision not to use their testimony was “manifestly unreasonable.”
2. Prejudice. The court concludes that the omission of Kelly’s and Wiggins’s testimony did not prejudice the defendant’s case because the drawbacks of such testimony would have offset the advantages. Again, I disagree.
The omitted testimony was the most potent impeachment evidence available to the defendant. Like most evidence, it was imperfect (from the defendant’s perspective): by calling Kelly and Wiggins to the stand, defense counsel risked admitting redundant fresh complaint testimony and testimony that the complainant consistently alleged at least that the defendant forcibly assaulted her. The first point, however, would have added nothing of substance to the evidence the jury heard, and the second would not have restored the complainant’s credibility. The evidence thát the complainant had changed her story, by contrast, would directly and substantially affect the complainant’s credibility, which was at the heart of the prosecution’s case. Therefore, on balance Kelly’s and Wiggins’s testimony would clearly have favored the defendant.
The court states that Kelly’s testimony in particular would have been damaging to the defense because Kelly’s recollection of the conversations with the defendant “fundamentally” paralleled the defendant’s recollection. The record does not support this conclusion. The defense attorney testified, “[The defendant] originally . . . told me that when he spoke to James Kelly . . ., he initially and vehemently denied that he had anything to do with [the complainant] on that occasion, that night.” However, Kelly testified that in their initial telephone conversations the defendant never denied that something “of a sexual nature” occurred between the defendant and the complainant. Instead, Kelly stated, the defendant denied only a charge of sexual harassment and subsequently admitted only to consensual sexual intercourse with the complainant. See ante at 269 n.4. These statements are entirely consistent with the consent defense. The court’s characterization of Kelly’s testimony as describing simply a denial fol*283lowed by an admission is at best superficial and at worst misleading. In fact, Kelly’s testimony does not support the inferences of guilt or dishonesty that the court draws from it. See ante at 276.
It is impossible, of course, to conclude with absolute certainty how the testimony of Wiggins and Kelly would have influenced the jury’s decision, if at all. However, the case against the defendant rested squarely on the testimony — and thus the credibility — of the complainant. The defense lawyer chose nevertheless to omit potent testimony impeaching the complainant’s credibility, and the defendant was subsequently convicted essentially on the complainant’s word. Under the unique circumstances of this case, I find it impossible to conclude that the defendant has not met his burden of showing that “better work might have accomplished something material for the defense.” Commonwealth v. Satterfield, 373 Mass. 109, 115 (1977).
In sum, I conclude that defense counsel’s decision neither to interview Kelly and Wiggins nor to call them to testify was “manifestly unreasonable” and was prejudicial to the defendant’s case. Therefore I do not agree with the court’s conclusion that the defendant has failed to establish the elements of ineffective assistance of counsel. I respectfully dissent.
I agree with the court’s conclusion that counsel’s’ failure to inform the judge of the complainant’s initial conversations with her husband, Wiggins, and Kelly when the judge was considering the admissibility of fresh complaint evidence was not prejudicial, and therefore did not constitute ineffective assistance. See section 2, ante at 277.
Counsel also asserted that he did not call Wiggins to testify because he was afraid she would reveal that Motorola had fired the defendant but not the complainant, suggesting that the company believed her and not him. As the Appeals Court recognized, however, this evidence was entirely irrelevant and could have been excluded by a motion in limine. 28 Mass. App. Ct. 417, 423 (1990).
Obviously, this statement does not imply that defense lawyers may not rely on their client’s statements, so that they must “investigate and confirm virtually every statement that a client makes, no matter how apparently or circumstantially accurate,” as the majority suggests. Ante at 274.