dissenting in part, concurring in part.
I cannot agree that trial counsel’s pretrial investigation and preparation met the standard for effective assistance of counsel or that the issue is precluded. Unfortunately, I also cannot agree with the court’s *163conclusion that an attorney’s complete failure to investigate and prepare a capital case can be anything but prejudicial error.
To hold, as the majority does, that no matter what was done, the “jury would not have been persuaded to accept Carriger’s uncorroborated testimony concerning what he did at the time of the crime ... [so that] Carriger has not established that he was prejudiced” (at 106) is to conclude simply that this court believes Carriger is guilty and therefore did not need representation. One may justifiably wonder why we bother to try defendants before a jury when appellate courts assume the power to weigh the evidence and divine what the jury would or would not have believed. We are not given the gift of divination, only the duty of decision.
I prefer the more conservative approach of determining whether the evidence not offered was the type which might have influenced the jury and thus might have affected the verdict. If it is not, then the error is harmless; if it is, the error is prejudicial. Until today that seemed to have been the test in Arizona.
Not all extraneous information is so prejudicial as to require reversal. The standard ... is that the defendant is entitled to a new trial if it cannot be concluded beyond a reasonable doubt that the extrinsic evidence did not contribute to the verdict. [Citations] We adopt the reasonable possibility standard applied by the Ninth Circuit as an appropriate balancing test and in harmony with the standards applied in other circuits. See, e.g., ... United States v. Marx, 485 F.2d 1179 (10th Cir.1973), cert. denied 416 U.S. 986, 94 S.Ct. 2391, 40 L.Ed.2d 764 (1974), ‘slightest possibility that harm could have resulted’; ... Farese v. United States, 428 F.2d 178 (5th Cir.1970), ‘reasonable possibility that information affected the verdict.’
State v. Poland, 132 Ariz. 269, 283, 645 P.2d 784, 798 (1982). Thus, we are not to decide what the jury would have believed, but whether the evidence, if believed, possibly or probably would have had an effect on the verdict.
With these traditional tests in mind, we must look at the facts regarding the claimed ineffective assistance of counsel in the case at bench. First, the court correctly holds that many of the claims raised by defendant are precluded because they could have been raised on appeal. On a subsequent Rule 32 petition, however, defendant cannot, be precluded from raising those claims of ineffective assistance which are based upon information obtained after his direct appeal had been decided. It is precisely to allow contentions based upon after-acquired information that Rule 32 was adopted. See Standards Relating to Post-Conviction Remedies § 2.1 commentary at 33 (1967).
The case is not complicated. The facts left only two possibilities. The murderer was either Carriger or Dunbar (or, perhaps, both). Carriger was the defendant and Dunbar was the state’s only direct witness. To procure Dunbar’s testimony, the state agreed not to prosecute him. Such deals may be necessary at times, but when one is made, we — and the jury — are obliged to suspect the ensuing testimony. There was no direct evidence of Carriger’s guilt. The case was tried, therefore, on a simple basic issue: Did Dunbar tell the truth when he blamed the robbery and murder on Carriger, or was Carriger’s contention that the robbery and murder were committed by Dunbar to be believed? All other issues were peripheral. To corroborate his story and his credibility, and to rebut Carriger’s charges against him, Dunbar testified that he could not have committed the robbery and murder because by both history and nature he was neither a robber nor a killer; his long criminal history showed only participation in nonviolent crimes.
While Dunbar was impeached in many ways, none of them addressed these fundamental contentions. Impeachment on these issues was available. Carriger, who had known Dunbar in prison, specifically asked trial counsel to obtain Dunbar’s prison *164records and inspect them for impeachment information. Trial counsel failed to do so. The majority seems to excuse this rather glaring omission by indicating that the case permitting discovery of such files was not decided until three years after Carriger’s conviction (slip op. at 20). See State v. Morales, 129 Ariz. 283, 630 P.2d 1015 (1981). The majority fails to acknowledge, however, that four years before Carriger’s trial the United States Supreme Court had ruled that such information was subject to discovery. Davis v. Alaska, 415 U.S. 308, 319-20, 94 S.Ct. 1105, 1112, 39 L.Ed.2d 347 (1974).
We now find that the file which trial counsel was requested to obtain, which he had a right and duty to obtain, and which he did not attempt to obtain, contained information that directly impeached some of the key statements which Dunbar made on the witness stand. For instance, while Dunbar portrayed himself as a person who would not rob because he could not use violence or threat of force, there was evidence in the file that he had committed armed robberies in Phoenix and in California. While Dunbar claimed innocence because he could neither kill nor comprehend the act of killing, there was evidence in the file that he had twice attempted to carve his father and mother with a butcher knife. While Dunbar claimed that he would not lie and implicate a “buddy,” there was evidence in the file that over a period of years he had several times committed a crime and then called the police to turn in his “buddy.” While Dunbar was portrayed by the state as a peaceful person and a credible witness, there was evidence in the file that he was in fact a violent, habitual criminal, ready to lie for any reason or simply as a matter of habit. While the state portrayed Dunbar as an accurate observer and reporter of events, the file indicated that he had been committed twice to the state hospital for the insane.
The majority argues that none of this would have made a difference because most, if not all, of the information in the file was not in admissible form. (At 1000-1002) The majority mistakes the issue. The issue is not whether the information would have been independently admissible; it is what use might have been made of such information by a skillful cross-examiner. It is impossible to believe that experienced trial counsel, cross-examining a man whom the state described at oral argument as a “cross-examiner’s dream,” would not have made considerable progress with this information. He might, for instance, have asked Mr. Dunbar to repeat his direct testimony about being a peaceful, nonviolent man who could not comprehend murder and then have asked him to explain the manner in which he had used the butcher knife on his mother. No purpose is served by further speculation. Experience with juries teaches us that skillful cross-examination of such a target with the ammunition in the file might have destroyed the witness. The majority, not impressed, holds that the jury still would have believed Dunbar. I have no way of knowing this, nor does the majority; most lawyers would conclude that such cross-examination is the type directly calculated to affect a juror’s belief of the witness’ testimony.
In arguing that the information in the file was “inadmissible,” the court also ignores the precept that one of the prime purposes of discovery is to obtain information that will lead to the discovery of admissible facts. It will not do to argue that the hearsay statements in the file were not independently admissible without acknowledging that the information would have provided a lead by which counsel could have obtained certified copies of records and other documents which could have been offered in evidence. The majority argues that they may have been remote and therefore irrelevant and would not have been admitted. That decision, I believe, is one for the trial judge under Rule 403, Ariz.R.Evid., 17A A.R.S. It involves a weighing of the circumstances. Given the serious nature of the crimes charged, the direct conflict between Dunbar and Cairiger, and whatever mileage counsel might have made by impeachment with the information contained in the file, it is impossible *165for us to say at this date what the trial judge would or would not have done in making the discretionary decision required of him under Rule 403.
This court has already found the same trial counsel to have been ineffective at the sentencing stage of this proceeding. State v. Carriger (Carriger II), 132 Ariz. 301, 645 P.2d 860 (1982). The standard applied at that time was “sham, mockery or farce.” The record of the trial proceeding with which we now deal reveals no less a mockery of effective trial representation. No Brady motion was filed, although the client sent numerous notes telling the lawyer what should be done and what information should be requested. Twenty witnesses, including Dunbar, testified for the state, but defense counsel interviewed only Dunbar and that for only a couple of hours. Counsel’s investigator interviewed only two or three of the other witnesses. Counsel made no motion under Rule 15, Ariz.R. Crim.P., 17 A.R.S., to obtain information on Dunbar’s past.
The state candidly concedes that a competent lawyer would have realized that the key to the case lay in Dunbar’s credibility, so that everything possible should have been discovered about Dunbar’s childhood, schooling, military record and criminal record. Almost nothing was done. To conclude, as the majority does, that it wouldn’t have made any difference is to render the guarantee of effective assistance of counsel meaningless. This court has no way of knowing whom the jury would have believed. It can only judge, consistent with the standard in the past (State v. Poland, supra), whether the evidence not presented was of the type calculated to have influenced jurors. With respect to the information contained in the prison file, that answer must be affirmative.
In State v. Lee, 142 Ariz. 210, 214, 689 P.2d 153, 157, we adopted the Strickland test (Strickland v. Washington, — U.S. -, 104 S.Ct. 2052 (1984)) and held that because the defendant is in a better position than the state to determine the prejudicial effect of errors of trial counsel, we would not reverse for ineffective assistance of counsel unless the defendant carried the burden of showing “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” (142 Ariz. at 214, 689 P.2d at 157.) We may have changed the burden of proof, and properly so, but surely this court did not impose upon a litigant the burden of showing that which it is never possible to show. One cannot offer proof of what would have happened had the lawyer done something which he failed to do. One can show only that the type of omission was one which had a reasonable probability of affecting the jury’s perception of the credibility of the witness. This is the rule we set in Poland. I gather it is now overruled sub silentio.
The new rule established today is one which allows this court, and those who follow us on the bench, to reach any result which they may desire in a particular case by simply deciding, on an unarticulated and subjective basis, that the jury would not have believed the evidence or would not have disbelieved the witness, regardless of the nature of the evidence or the character of the witness. The rule of law is essentially subverted when judges assume such power. The case should be remanded for retrial.