Opinion
MOSK, J.This case presents a variation on a theme with which the Courts of Appeal have been struggling for over a decade without our direct guidance: under what circumstances does Prudhomme v. Superior Court (1970) 2 Cal.3d 320 [85 Cal.Rptr. 129, 466 P.2d 673], justify prosecutorial discovery of pretrial statements made by defense witnesses to defense investigators? Because Prudhomme’s suggestion (at p. 327) that such discovery may sometimes be allowed has caused little but confusion, and because we are not the proper body to create procedural rules that tend to impinge on the traditional and arguably constitutional rights of a criminal defendant, we disapprove of any compelled production of defense evidence absent explicit legislative authorization.
I.
On the evening of July 6, 1978, defendant Bertram Collie visited his estranged wife at her residence, as he had often done before. She and her daughter were in the bedroom watching television and, as defendant entered, the daughter retired to her own room, where she remained for the evening.
Defendant invited his wife to drink and to have sexual intercourse with him, but she refused. He subsequently bound her feet and hands and forcibly sodomized her. He then taped her mouth, ransacked the bedroom, left the room, and locked the door behind him. His wife heard him leave the house at about midnight.
Mrs. Collie then detected the odor of gas. She managed to free herself and unlock the door, and found that the stove burners were turned on, unlit. She turned them off. In the dining room she discovered and extinguished a lighted candle surrounded by combustible material. She then awakened her daughter, who was safe in her own bedroom and oblivious to all that had occurred.
*49Defendant testified that he and his wife had consensual intercourse early in the evening, after which he told her he was moving out permanently. He then proceeded to a neighborhood bar, where he telephoned a friend, Cynthia Morris, and asked to visit her. He testified that he arrived at her apartment about 11:30 p.m.
Ms. Morris was called as a defense witness. During cross-examination she revealed that she had spoken previously to a defense investigator. The prosecution immediately requested discovery of the notes prepared by the investigator. Defense counsel objected on the basis of the work-product doctrine and the attorney-client privilege. The court overruled the objections and ordered discovery.
On subsequent cross-examination, Ms. Morris admitted telling the investigator that defendant had visited her on July 8 or 9 rather than July 6. She was also impeached regarding relatively minor details of her testimony: for example, she told the investigator defendant called from a different bar than the one he had named in his testimony.
The jury found defendant guilty of attempted first degree murder of his wife, attempted second degree murder of his daughter, and forcible sodomy.
II.
Defendant first contends that the statements made by Cynthia Morris to the defense investigator were privileged communications not subject to prosecutorial discovery. Although his principal contention of error is that the order violated his privilege against self-incrimination under Prudhomme v. Superior Court, supra, 2 Cal. 3d 320, he made no such objection at trial and cannot raise it for the first time on appeal. (Evid. Code, § 353.)1 Defendant indirectly asserts the privilege, however, by contending that his attorney’s failure to object on the basis of Prudhomme and its progeny constituted ineffective assistance of counsel. If counsel failed to perform in a manner to be expected of a reasonably competent attorney acting as a diligent advocate, and if his failure deprived defendant of a potentially meritorious de*50fense, reversal is required. (People v. Pope (1979) 23 Cal.3d 412, 425 [152 Cal.Rptr. 732, 590 P.2d 859].) Preliminary to deciding this question, we reach the issue of whether the order violated defendant’s privilege against self-incrimination. A brief synopsis of the short history of prosecutorial discovery in California is helpful in framing the issue.
The first case to authorize discovery directed at the defense was Jones v. Superior Court (1962) 58 Cal.2d 56 [22 Cal.Rptr. 879, 372 P.2d 919, 96 A.L.R.2d 1213]. There, defendant sought a continuance on the day set for trial, and as a condition of granting his request the trial court ordered limited defense discovery. On appeal we held that the privilege against self-incrimination did not shield defendant from being forced to produce the names and identities of witnesses who would be used to prove defendant’s late proffered affirmative defense of impotency in a rape case. Language in the case declaring that discovery procedure “should not be a one-way street” (id., at p. 60) was taken as establishing a broad principle of reciprocity in criminal discovery, and led to routine and wide-ranging discovery orders directed at defendants.
In Prudhomme v. Superior Court, supra, 2 Cal.3d 320, we reviewed one such order requiring defendant to divulge the names, addresses, and anticipated testimony of all defense witnesses. In limiting Jones virtually to its facts, we struck down the order because we found that disclosure of the requested information “conceivably might lighten the prosecution’s burden of proving its case in chief.” (Prudhomme, supra, at p. 326.) We further held that “the privilege forbids compelled disclosures which could serve as a ‘link in a chain’ of evidence tending to establish guilt of a criminal offense; in ruling upon a claim of privilege, the trial court must find that it clearly appears from a consideration of all the circumstances in the case that an answer to the challenged question cannot possibly have a tendency to incriminate the witness.” (Ibid.) We left open, however, the possibility that prosecutorial discovery of some sort could be permitted: “A reasonable demand for factual information which, as in Jones, pertains to a particular defense or defenses, and seeks only that information which defendant intends to introduce at trial, may present no substantial hazards of self-incrimination and therefore justify the trial judge in determining that under the facts and circumstances in the case before him it clearly appears that disclosure cannot possibly tend to incriminate defendant.” (Id. at p. 327.)
Shortly after Prudhomme, we had occasion to consider a trial court’s attempt to restrict a discovery order by limiting it to advance notice of *51the identity of any person who would be called as an alibi witness. (Reynolds v. Superior Court (1974) 12 Cal.3d 834 [117 Cal.Rptr. 437, 528 P.2d 45].) The trial court ordered the defense to produce the information, but after a thorough assessment of relevant state and federal precedents, we unanimously concluded that the wisest course was to refrain from any attempt to create or adopt a prosecutorial discovery scheme through our inherent power to administer matters of criminal procedure. In an opinion that is a model of judicial restraint, Chief Justice Wright concluded, “We see little to recommend our attempting at once to consider the desirability of creating a notice-of-alibi procedure and to pass objectively on the constitutionality of any such procedure which might result. It is one thing for a court to prescribe judicial procedures necessary to protect some fundamental constitutional guarantee of individual liberty. [Citations.] It is quite another thing for a court to design judicial procedures which are in no way required by higher law but which may seem to some socially desirable and perhaps may be permitted—at least to some extent—by our state and federal Constitutions. In the former instance, constitutional principles guide the court’s hand; in the latter instance constitutional principles may well have to stay the court’s hand. Given the difficulty of the constitutional questions posed by notice-of-alibi procedures, it is far better for this court to pass judgment, if and when necessary, on an integrated legislative document than on our own conditional decree by which we might seek to smooth the constitutionally rough edges of the order of the court below.” (Id. at pp. 845-846.)
In our most recent pronouncement on the subject, Allen v. Superior Court (1976) 18 Cal.3d 520 [134 Cal.Rptr. 774, 557 P.2d 65], we held invalid an order made on the court’s own motion at the commencement of trial compelling disclosure of the names of prospective defense and prosecution witnesses for the purpose of voir dire. Chief Justice Wright again spoke for the court, but this time found no need to hesitate in reaching the constitutional issue: the order was found clearly violative of Prudhomme standards.2 Although the order was accompanied by an injunction against prosecutorial contact with any of the defense witnesses until their names were otherwise revealed, we found that knowledge of the names alone could provide the “essential link in a chain of evidence” of which Prudhomme had spoken.
*52In the present case, the discovery order is different from those we have discussed in the past: it issued after the prosecution rested and after the direct examination of the defense witness whose prior statements it forced the defense to produce. Several Courts of Appeal have considered similar orders and, failing either to presage or to heed the Reynolds counsel of caution in formulating rules of prosecutorial discovery,3 have instead struggled with the singular task of limiting the constitutional privilege to serve the practical goal of effective prosecution. The difficulty those courts have had in agreeing on the maximum amount of discovery consistent with the minimum rights of a defendant lends support to the conclusion in Reynolds that the courts are not the proper bodies to initially formulate prosecutorial discovery rules.
As at least one court has recognized, Prudhomme does not dictate an unequivocal result in this context. (People v. Thornton (1979) 88 Cal. App.3d 795, 799 [152 Cal.Rptr. 77].) Prudhomme expressly contemplates discovery of otherwise privileged information only if it will be produced by the defense at trial (Prudhomme, supra, 2 Cal. 3d at p. 327), yet it also would seem to imply that other information which cannot conceivably assist the prosecution in proving its case in chief is unprotected by the privilege, whether or not the defense intends to use it. (Id. at p. 326.) Three Courts of Appeal have attempted to resolve the issue by holding that the prosecution may discover prior statements of a defense witness at least for purposes of impeachment. (People v. Ayers (1975) 51 Cal.App.3d 370 [124 Cal.Rptr. 283]; People v. Chavez (1973) 33 Cal.App.3d 454 [109 Cal.Rptr. 157]; People v. Bais (1973) 31 Cal.App.3d 663 [107 Cal.Rptr. 519].) To that end, they have imposed or approved screening procedures in which the documents or statements are reviewed by the trial court, and the impeaching material is sifted from the rest.
These courts do not agree, however, on the justification for discovery. Ayers relied on federal precedent in reasoning that the privilege was *53personal to defendant and did not apply to third party statements. (Id. at p. 379.) But our reaffirmation of Prudhomme on state constitutional grounds in Allen (see fn. 2, ante) undercut this analysis. Chavez found its justification in the fact that the witness had already testified when discovery was requested, and thereby opened the door to discovery relevant to his testimony on direct examination. (Id. at p. 459.) Presumably, this third-party waiver approach would allow discovery of information that directly incriminates defendant so long as it is within the scope of the direct testimony of the witness. Bais did not find the timing of the request determinative, but suggested instead that statements which merely impeach a defense witness do not incriminate the defendant under the Prudhomme standards, because they do not assist in proving the prosecution’s case in chief. (Bais, supra, at p. 672; see also People v. Thornton, supra, 88 Cal.App.3d 795, 806-810 (conc. opn. of Scott, J.).) Under this approach, it appears that discovery must be limited to statements that impeach the witness but that otherwise provide no substantive assistance to the prosecution in proving defendant’s guilt.
The most recent case to address the subject, People v. Thornton, supra, 88 Cal.App.3d 795, held that the defendant’s privilege against self-incrimination extends even to statements that impeach defense witnesses without otherwise inculpating the defendant. In support of its holding, the court quoted a footnote in Allen v. Superior Court casting doubt on Prudhomme’s distinction between information tending to prove the prosecution’s case and materials tending to either discredit or prematurely expose a defense.4 As a result, Thornton found no basis for a screening process; under its analysis anything that would be of use to the prosecution in securing a conviction would for that reason be incriminatory, and thus privileged.
This tangle of Court of Appeal cases demonstrates only one certitude—prosecutorial discovery for the purpose of impeaching defense witnesses is no less constitutionally questionable than was the notice-of-alibi issue we declined to resolve in Reynolds. The theoretical disparities among the courts seem to arise from the dilemma of protecting *54the privilege, as broadly defined by the Prudhomme opinion, while attempting to actualize the Prudhomme dictum which appeared to leave some avenues of prosecutorial inquiry open. Unfortunately, the lower courts have given little or no consideration to Reynolds,5 our only unanimous case to touch on the subject. In so doing, they have joined in dubious battle over an issue which, like that in Reynolds, is more appropriately left to the Legislature for initial consideration.
In recognizing the original primacy of the Legislature in the field of creating rules of criminal procedure, we are not unmindful of the almost insurmountable hurdles likely to thwart any attempts to devise constitutionally permissible discovery rules applicable to defendant or defense material. As is true of notice-of-alibi rules, any discovery rule in this setting will inevitably raise serious state and federal constitutional questions. We have already reviewed the difficulty encountered in resolving the self-incrimination problem under the state Constitution. In addition, although the United States Supreme Court has laid to rest any serious fear that discovery of this type would offend the federal self-incrimination privilege (United States v. Nobles, supra, 422 U.S. at pp. 233-234 [45 L.Ed.2d at p. 151]), it is unclear whether the same can be said of the federal due process requirement of reciprocity in discovery established in Wardius v. Oregon (1973) 412 U.S. 470 [37 L.Ed.2d 82, 93 S.Ct. 2208]. As we noted in Reynolds, the Wardius opinion dictates that notice-of-alibi procedures provide an opportunity for the defense to learn both the prosecution’s most accurate account of the time and place of the alleged crimes (id. at pp. 478-479, fn. 12 [37 L.Ed.2d at pp. 89-90]), and the names and addresses of state witnesses who might refute defendant’s alibi (Id. at p. 475 [37 L.Ed.2d at p. 87]). If the prosecution is allowed to discover evidence from the defense to impeach a defense witness, the rationale of Wardius may require that the prosecution contemporaneously divulge any information that might reinstate the credibility of the defense witness or might otherwise help establish the truth of his direct testimony. Although Wardius was directly relevant to the issue in Reynolds, we declined to employ its guidance in formulating a general rule. Here, without direct guidance, we can be less dogmatic about the procedures, if any, federal due process permits; thus it is even less advisable to prescribe standards- that would be unavoidably shrouded in constitutional doubt.
*55Furthermore, defendant’s constitutional right to assistance of counsel is potentially threatened if, as the Chavez court concluded, the decision to present the witness justifies discovery to the extent of the witness’ direct testimony or if, as the Attorney General suggests, placing the defendant on the stand waives the privilege not only for purposes of cross -examination (see People v. Schader (1969) 71 Cal.2d 761, 770-771 [80 Cal.Rptr. 1, 457 P.2d 841]), but for purposes of discovery of all defense information conceivably relevant to defendant’s direct testimony. In either event defense counsel’s ability to freely investigate and effectively present the defense could be seriously compromised.6 A rule that would open the defense files if a witness or the defendant testified could penalize the defendant whose attorney was most vigilant in gathering, documenting, recording and studiously analyzing evidence to prepare the defense.
Finally, in addition to its constitutional implications, prosecutorial discovery risks conflict with the attorney-client privilege, the work-product doctrine, and the statutory prohibition of forced revelation of allegedly privileged material for the purpose of ruling on its protected status. (Evid. Code, § 915; see also Evid. Code, § 404.) Although these concerns are not of constitutional origin, the first two, at least, may have important constitutional implications, and all three undeniably contribute to the monumental complexity of the problem.
The reasoning of Reynolds dictates that we disapprove the trial court’s order and all other judicial attempts to frame prosecutorial discovery orders, and refrain from attempting ourselves to articulate a unitary principle on which discovery by the People can be based.7 In its *56uncritical assumption that courts should perform the rule-making function, Prudhomme failed to properly assess the difficult and awkward responsibility placed on the judiciary. In this connection rules that permit, and rules that deny, discovery are equally rules.
Firmly embedded in our law is the doctrine that the courts are the final interpreters and guardians of the Constitution. As the heirs of that institutional obligation, we must be reluctant to step out of our traditional role by undertaking to define in the first instance procedures that upon reflection may appear to undermine the foundational principles it is our primary responsibility to protect.
We can do justice neither to the legitimate needs of the prosecution nor to the rights of the defendant in this context if we undertake judicial rule-making in an attempt to accommodate both ends simultaneously. Any effort to further the truth-seeking function bears considerable risk of encroaching on constitutional and other protections: as we have noted, the problem is complicated by an interrelated composite of state and federal constitutional concerns, statutory rules and common law privileges. We realize the same problems would confront the Legislature if it were to consider the issue, and we have grave doubts that a valid discovery rule affecting criminal defendants can be devised. But if the Legislature undertakes to formulate a comprehensive solution that purports to be practical in application and consistent with the public interest, any legislative error would be subject to judicial review. Ours is likely to be the last word on the subject; for that reason, it should not also be the first.8 As Reynolds observed, the most appropriate course of action is to refrain from acting in the absence of word from the Legislature. We therefore hold that the trial court erroneously ordered discovery of the witness’s prior statement.
*57Because the present issue is the competence of counsel, however, our inquiry does not end with that observation. It would be unrealistic to conclude that counsel should have anticipated our extension of Reynolds to an area to which it had not previously been applied by the Courts of Appeal. It is more likely that counsel erred in failing to invoke the existing analysis of the Courts of Appeal. Although the applicable rules were far from clear, even the cases least solicitous of defendant’s rights validated discovery only after the judge had screened the documents to exclude nonimpeaching evidence. (People v. Ayers, supra, 51 Cal.App.3d at p. 378; People v. Chavez, supra, 33 Cal.App.3d at pp. 459-460.) Defense counsel did not request such screening.9 But assuming arguendo counsel did not act diligently on defendant’s behalf, he did not thereby deprive defendant of a potentially meritorious defense.
In the first place, the trial judge reviewed the documents sua sponte10 and stated that he found nothing in them that would uncover other witnesses, thereby implying that he had considered the incriminating effect of the statements under the Prudhomme standard.11 Defendant offers no proof that the statements were in fact used for any purpose other than impeachment. Thus it appears defendant was not prejudiced by his attorney’s failure to request that the statements be screened in a manner suggested by existing cases.
*58Moreover, even assuming the court would have excluded the statements entirely, their use did little to harm defendant’s case, because Ms. Morris’ direct testimony did little to help it. The matters about which Ms. Morris testified were collateral to the underlying question of what transpired at the Collie residence on the night of the alleged crimes. Furthermore, her testimony regarding those collateral matters was consistent with the prosecution’s case: her statement that defendant arrived at her apartment at 12:30 did not controvert Mrs. Collie’s assertion that defendant departed her residence at about midnight. In fact, Ms. Morris’ recollection of the chronology cast doubt on defendant’s assertion that he arrived at her apartment at 11:30. She further called into question defendant’s veracity by testifying that he told her he habitually wore dark glasses because he liked them, and not because of light sensitivity, as he had earlier testified.
Ms. Morris’ testimony regarding defendant’s demeanor after the incident was no more helpful to his cause. Although she testified that defendant seemed incredulous when confronted with a newspaper article reporting the incident, she also stated, over defense counsel’s objection, that he appeared “nervous and jumpy” after the date of the alleged crimes and before seeing the article. Finally, she testified that Mrs. Collie had reacted with extreme jealousy to defendant’s association with other women. Although this information could have raised questions in the jurors’ minds about Mrs. Collie’s truthfulness and motive in assisting the prosecution, it also provided a possible explanation for defendant’s alleged hostility to his wife.
Considering the contrast between the minimal utility of Ms. Morris’ testimony and the persuasive eyewitness testimony of Mrs. Collie, as corroborated by her neighbors and the investigating police officers, the impeachment of Ms. Morris, even if it substantially impaired her credibility, did not deprive defendant of a potentially meritorious defense.
Because we find discovery harmless with regard to the privilege against self-incrimination, we next discuss defendant’s contention that discovery violated the work-product privilege and prejudiced his case by exposing his defense investigator’s theories and thought processes to the prosecution.
*59We have never explicitly held the work-product doctrine applicable to criminal cases and neither has the Legislature, although it has codified the rule as to civil trials. (Code Civ. Proc., § 2016, subd. (b).) There is little reason, however, to withhold its protection from the criminally accused. Many other jurisdictions, including the federal courts, have applied the rule to criminal cases. (See, e.g., United States v. Nobles, supra, 422 U.S. 225, 238 [45 L.Ed.2d 141, 153-154], and cases cited.) As the United States Supreme Court noted in Nobles: “Although the work-product doctrine most frequently is asserted as a bar to discovery in civil litigation, its role in assuring the proper functioning of the criminal justice system is even more vital. The interests of society and the accused in obtaining a fair and accurate resolution of the question of guilt or innocence demand that adequate safeguards assure the thorough preparation and presentation of each side of the case.” (Ibid.)12 Nobles also persuasively reasons that the privilege should extend not just to the attorney’s work product, but to the efforts of those who work with him to prepare the defense: “At its core, the work-product doctrine shelters the mental processes of the attorney, providing a privileged area within which he can analyze and prepare his client’s case. But the doctrine is an intensely practical one, grounded in the realities of litigation in our adversary system. One of those realities is that attorneys must often rely on the assistance of investigators and other agents in the compilation of materials in preparation for trial. It is therefore necessary that the doctrine protect material prepared by agents for the attorney as well as those prepared by the attorney himself.” (Id. at pp. 238-239 [45 L.Ed.2d at p. 154].) We conclude that the work-product doctrine applies to criminal cases and protects the work product of defense investigators.
The discovered documents in this case were reports to defense counsel from the defense investigator, in which the investigator paraphrased the statements of Ms. Morris, prefacing each with the expression, “subject reports.” The court asserted that in its view preparation of the documents in this manner was a blatant attempt to cloak the statements of the witness in the work-product privilege, and refused to be deceived by the “subterfuge.” Although the court may have been jus*60tified in its suspicion that the form of the report was influenced by a desire to shield the statements, the normal test is not intent but content. If the only “work” for which protection is claimed consists of prefacing the witness’s statements with a phrase such as “subject reports,” it is difficult to conceive how the policies served by the work-product doctrine would be offended by its disclosure, at least if it is the attorney’s agent rather than the attorney who records the statements.13 On the other hand, courts should not hastily conclude that summarized or edited remarks, or statements in response to questions formulated by the attorney or his aides, are without potential for invading the protected provinces of thought and strategy.
Assuming the court correctly concluded that the paraphrased statements were not a work product, nonetheless other parts of the discovered documents clearly were. At the top of the reports, the investigator made a “statement resume” in which he assessed the usefulness of Ms. Morris as a witness, and listed particular issues about which she could or would testify.14 The reports also revealed that Ms. Morris had been subpoenaed. These features of the reports, at the very least, should have been protected, by the work-product doctrine, from discovery.
Nevertheless, defendant fails to demonstrate any prejudice from the revelation. The reports reflected solely on the potential utility of a witness who had already been utilized. Thus even if the defense plan to use Ms. Morris was improperly revealed, no harm resulted because it had already been revealed when she was called to the stand by defense counsel. Consequently, it is not “reasonably probable that a result more *61favorable to the appealing party would have been reached in the absence of error.” (People v. Watson (1956) 46 Cal.2d 818, 836 [299 P.2d 243].)15
III.
Defendant correctly contends that the trial court erred in instructing the jury that it could convict of attempted second degree murder despite the absence of a specific intent to kill. The court instructed the jury in the language of CALJIC No. 8.31 (1974 rev.) which, as modified by the court in the language we have italicized, reads as follows; “Murder of the second degree is [also] the unlawful killing of a human being as the direct causal result of an act involving a high degree of probability that it will result in death, which act is done for a base, antisocial purpose and with wanton disregard for human life by which is meant an awareness of a duty imposed by law not to commit such acts followed by the commission of the forbidden act despite that awareness. [¶] When the killing, or the attempted killing in this case, is the direct result of such an act, it is not necessary to establish that the defendant intended that his act would result in the death of a human being.”
In People v. Murtishaw (1981) 29 Cal.3d 733 [175 Cal.Rptr. 738, 631 P.2d 446], we held that an assault with intent to commit murder required express malice, and could not be founded on the malice implied from reckless conduct with wanton disregard for human life. Relying on a line of authority originating in People v. Mize (1889) 80 Cal. 41 [22 P. 80], we observed, “once a defendant intends to kill, any malice he may harbor is necessarily express malice. Implied malice, as defined in CALJIC No. 8.11 [and CALJIC No. 8.31], cannot coexist with a specific intent to kill. To instruct on implied malice in that setting, therefore, may confuse the jury by suggesting that they can convict without finding a specific intent to kill.” (Murtishaw, supra, at pp. 764-765.)
*62The same reasoning applies to attempted murder. “Specific intent to kill is a necessary element of attempted murder. It must be proved, and it cannot be inferred merely from the commission of another dangerous crime.” (People v. Belton (1980) 105 Cal.App.3d 376, 380 [164 Cal. Rptr. 340].)16 Hence the trial court erred in instructing the jury that it need not find a specific intent to kill in order to convict of attempted murder.
The instructional error was harmless regarding the conviction for the attempted first degree murder of Mrs. Collie, because the jury was properly instructed that the verdict required findings of premeditation and deliberation, which entail a specific intent to kill. But the verdict of guilt on the attempted second degree murder charge was not insulated from error. Although the jury was properly instructed that a specific intent to kill would satisfy the intent requirement of an attempted second degree murder charge, it is impossible to determine whether the verdict rested on that ground, for which there was little evidence, or on the impermissible basis of defendant’s wanton conduct, which was more clearly supported by the record. Because we cannot know on which instruction the jury relied, the conviction for attempted second degree murder of defendant’s daughter must be reversed. (See, People v. Rhoden (1972) 6 Cal.3d 519, 526 [99 Cal.Rptr. 751, 492 P.2d 1143]; People v. Bail (1943) 22 Cal.2d 642, 653 [140 P.2d 828].)
IV.
Defendant next contends he was improperly and excessively sentenced to six years in prison for the attempted murder of his wife. He relies on the decisions of two Courts of Appeal (People v. Montano (1979) 96 Cal.App.3d 221 [158 Cal.Rptr. 47]; People v. Gray (1979) 91 Cal.App.3d 545 [154 Cal.Rptr. 555]) each court holding that the attempted murder it reviewed was indistinguishable from an assault with intent to commit murder and therefore was governed by the particular sentencing provisions relating to the latter offense (former Pen. Code, § 217)17 rather than the residual sentencing provisions of the general attempt statute. (Pen. Code, § 664.)
*63We find the reasoning of Montano and Gray inapplicable to the present case. Defendant was convicted of attempted murder in the first degree, for which premeditation and deliberation are prerequisites. These elements are unnecessary to establish an assault with intent to commit murder. (People v. Sartain (1968) 268 Cal.App.2d 486, 489 [73 Cal.Rptr. 799].) Hence the more serious nature of defendant’s crime justifies imposition of the more severe penalty. (People v. Singleton, supra, 112 Cal.App.3d 418, 426-433 (conc. opn. of Staniforth, J.).)
V.
Defendant complains that the trial court should have instructed the jury sua sponte on the limited admissibility of evidence of previous assaults he allegedly committed on his wife.18 Although the trial court may in an appropriate case instruct sua sponte on the limited admissibility of evidence of past criminal conduct, we have consistently held that it is under no duty to do so. (People v. Holbrook (1955) 45 Cal.2d 228, 233 [288 P.2d 1]; People v. Weitz (1954) 42 Cal.2d 338, 347 [267 P.2d 295]; People v. Simeone (1945) 26 Cal.2d 795, 808 [261 P.2d 369]; People v. Findley (1901) 132 Cal. 301, 306 [64 P. 472]; see also People v. Morrisson (1979) 92 Cal.App.3d 787, 790-791 [155 Cal. Rptr. 152]; People v. Harris (1977) 71 Cal.App.3d 959, 966 [139 Cal.Rptr. 778]; People v. Jackson (1975) 45 Cal.App.3d 67, 70 [119 Cal.Rptr. 71].) We have more recently decided that in many cases sua sponte instructions regarding relevant defenses (People v. Sedeno (1974) 10 Cal.3d 703, 716 [112 Cal.Rptr. 1, 518 P.2d 913]; People v. Mosher (1969) 1 Cal.3d 379, 390-391 [82 Cal.Rptr. 379, 461 P.2d 659]) and lesser included offenses (People v. St. Martin (1970) 1 Cal.3d 524, 531 [83 Cal.Rptr. 166, 463 P.2d 390]; People v. Hood (1969) 1 Cal.3d 444, 449-450 [82 Cal.Rptr. 618, 462 P.2d 370]) are required because those matters are “closely and openly connected” with the evidence and the fate of the defendant in cases to which they apply. (Id. at p. 449, quoting from People v. Wade (1959) 53 Cal.2d 322, 334 [1 Cal.Rptr. 683, 348 P.2d 116].) Defendant fails to show that the *64limited admissibility of particular bits of evidence of past criminal conduct normally deserves similar unsolicited recognition and instruction by the trial court.
Evidence of past offenses may not improperly affect the jury’s deliberations if the facts are equivocal, the charged offense is dissimilar, or the evidence is obviously used to effect one or more of the many legitimate purposes for which it can be introduced. (See Evid. Code, § 1101, subd. (b); People v. Haylock (1980) 113 Cal.App.3d 146, 150 [169 Cal.Rptr. 658].) Neither precedent nor policy favors a rule that would saddle the trial court with the duty either to interrupt the testimony sua sponte to admonish the jury whenever a witness implicates the defendant in another offense, or to review the entire record at trial’s end in search of such testimony. There may be an occasional extraordinary case in which unprotested evidence of past offenses is a dominant part of the evidence against the accused, and is both highly prejudicial and minimally relevant to any legitimate purpose. In such a setting, the evidence might be so obviously important to the case that sua sponte instruction would be needed to protect the defendant from his counsel’s inadvertence. But we hold that in this case, and in general, the trial court is under no duty to instruct sua sponte on the limited admissibility of evidence of past criminal conduct.19
VI.
We have reviewed defendant’s remaining contentions. They present no new and important issue of law, and were correctly resolved by the Court of Appeal when the case was before that court. For the reasons stated by the Court of Appeal, we hold each of these contentions to be without merit. (People v. Scott (1978) 21 Cal.3d 284, 297 [145 Cal. Rptr. 876, 578 P.2d 123]; People v. James (1977) 19 Cal.3d 99, 118 [137 Cal.Rptr. 447, 561 P.2d 1135].)
The judgment is reversed as to the conviction of attempted second degree murder. In all other respects the judgment is affirmed.
Bird, C. J., Tobriner, J., and Abbe, J.,* concurred.
At least one case asserts that the statute would allow review of the unchallenged admission of evidence if defendant could establish a “miscarriage of justice.” (People v. DeVaney (1973) 33 Cal.App.3d 630, 635 [109 Cal.Rptr. 276].) But that reading of the statute is clearly incorrect: the statute allows reversal on the miscarriage of justice standard only if a timely and proper objection has been made. (Evid. Code, § 353.)
Prudhomme relied largely on federal constitutional principles. (Id. at pp. 323-325 of 2 Cal.3d.) Subsequently, however, two United States Supreme Court cases made clear that the defendant’s Fifth Amendment rights do not extend to statements made by *52third parties to the defense. (United States v. Nobles (1975) 422 U.S. 225 [45 L.Ed.2d 141, 95 S.Ct. 2160]; Williams v. Florida (1970) 399 U.S. 78 [26 L.Ed.2d 446, 90 S.Ct. 1893, 1914].) Recognizing Prudhomme’s establishment of a more solicitous attitude toward the privilege in this state, Allen reaffirmed Prudhomme’s holding on state constitutional principles. (Allen, supra, 18 Cal.3d at p. 525; see Cal. Const., art. 1, § 15.)
With regard to notice-of-alibi rules, however, Reynolds was foreshadowed by Rodriguez v. Superior Court (1970) 9 CaI.App.3d 493 [88 Cal.Rptr. 154], which determined that judicial abstention was the wisest course after noting the Legislature’s repeated albeit unsuccessful attempts to enact a statute on the subject.
In discussing the danger of revealing names of defense witnesses to the prosecution, Allen reasoned “For example, if the defense witnesses identified are friends or relatives of an accused, the prosecution can anticipate an alibi defense; if police officers the defense of entrapment can be projected.... investigation may reveal the details of the alibi or other defenses, or may yield other evidence useful to the prosecution including impeachment witnesses, inconsistent statements, and admissible evidence of specific instances of misconduct by the prospective witnesses.” (Id., 18 Cal.3d at p. 526, fn. 4.)
Chavez and Bais were decided before Reynolds, but Ayers and Thornton were later. The Ayers court acknowledged the existence of Reynolds, but because it wrongly assumed an answer was dictated by federal cases, it approved the trial court’s procedure.
In United States v. Nobles, supra, 422 U.S. 225 at page 240, footnote 15 [45 L.Ed. 2d 141 at pages 154-155], the high court summarily rejected a similar contention relating to discovery of a defense investigator’s documents after the investigator testified. But there, the discovered reports formed the basis of the witness’ direct testimony; thus the forced production was justified not by the decision to present the witness, but by the plan to use the documents at trial. The court has elsewhere recognized that procedures restricting counsel’s ability to investigate and present the case can deprive defendant of the assistance to which he is entitled. (See, e.g., Brooks v. Tennessee (1972) 406 U.S. 605 [32 L.Ed.2d 358, 92 S.Ct. 1891].) Furthermore, the people of California have independently safeguarded in the state Constitution the defendant’s right to competent and vigorous legal representation. (Cal. Const., art. I, § 15.)
Although we refrain from establishing discovery rules related to testimonial evidence, we leave intact the firmly established precedents that hold the self-incrimination privilege inapplicable to, and allow mandatory production of, nontestimonial evidence such as fingerprints, blood samples, breath samples, appearances in lineups, and handwriting and voice exemplars. (United States v. Wade (1967) 388 U.S. 218, 222-223 [18 *56L.Ed.2d 1149, 1154-1155]; Gilbert v. California (1967) 388 U.S. 263, 266-267 [18 L.Ed.2d 1178, 1182-1183]; Schmerber v. California (1966) 384 U.S. 757, 761 [16 L.Ed.2d 908, 914, 86 S.Ct. 1826]; Holt v. United States (1910) 218 U.S. 245, 252-253 [54 L.Ed. 1021, 1030, 31 S.Ct. 2]; Cramer v. Tyars (1979) 23 Cal.3d 131, 139 [151 Cal.Rptr. 653, 588 P.2d 793]; People v. Williams (1969) 71 Cal.2d 614, 625 [79 Cal.Rptr. 65, 456 P.2d 633]; People v. Ellis (1966) 65 Cal.2d 529, 533-535 [55 Cal.Rptr. 385, 421 P.2d 393]; People v. Sudduth (1966) 65 Cal.2d 543, 545-546 [55 Cal.Rptr. 393, 421 P.2d 401]; People v. Haeussler (1953) 41 Cal.2d 252, 257 [260 P.2d 8]; Finley v. Orr (1968) 262 Cal.App.2d 656, 660-663 [69 Cal.Rptr. 137].)
If physical evidence is altered or removed by defense counsel or his investigator, the prosecution may be able to obtain testimony describing the original condition or location of the evidence. (See People v. Meredith (1981) 29 Cal.3d 682, 694-695 [175 Cal.Rptr. 612, 631 P.2d 46].)
Although the lower courts do not share our final authority, their responsibility for protecting constitutional rights is no less central to their judicial role. We agree with *57Reynolds that the piecemeal development of discovery procedures by the lower courts “—with the inevitable inconsistencies of early cases as the procedure evolved [witness Thornton, Ayers, Chavez, and Bais, supra]—would be even less desirable than our Jovian creation of such procedure by comprehensive judicial fiat in the context of a single case.” (Id., 12 Cal.3d at p. 845, fn. 16.)
In stating his work product objection to discovery, defense counsel mentioned that he was aware some courts screen documents to remove information that is not “an actual statement of the subject witness.” But not only was this comment unrelated to the privilege against self-incrimination, counsel expressly disclaimed any intent to offer such a screening procedure as a compromise in lieu of total exclusion.
Shortly before oral argument, the Attorney General attempted to augment the record on appeal by submitting for our consideration a signed declaration of the trial judge describing the circumstances surrounding his inspection of the documents. The Attorney General offers no justification for this attempted departure from the sound and well-established principles that limit the scope of review to matters either preserved in the record or properly subject to judicial notice. (See In re Hochberg (1970) 2 Cal.3d 870, 875 [87 Cal.Rptr. 681, 471 P.2d 1]; 5 Cal.Jur.3d, § 478.) We therefore have given no consideration or effect to the preferred declaration for purposes of this decision, and grant defendant’s motion to strike the purported declaration from the record.
In ruling on the work product and attorney-client privilege objections, the court stated, “I want the record to be perfectly clear. The only thing on that particular form *58is that statement. There are no other notes. There is no reference to any other witness. There are no references from your client or anyone else.”
Even in civil cases the work-product doctrine is not an absolute bar to discovery (Code Civ. Proc., § 2016, subd. (b)); manifestly, it cannot be invoked by the prosecution to preclude discovery by the defense of material evidence, or to lessen the state’s obligaton to reveal material evidence even in the absence of a request therefor. (See Brady v. Maryland (1963) 373 U.S. 83, 87 [10 L.Ed.2d 215, 218, 83 S.Ct. 1194]; In re Ferguson (1971) 5 Cal.3d 525, 532 [96 Cal.Rptr. 594, 487 P.2d 1234].)
The most commonly recognized danger in ordering discovery of work product is that the attorney will be inhibited by the possibility that the strategies and ideas he develops to present his client’s case could be put in the hands of the opposing party to the detriment of his client. (Hickman v. Taylor (1947) 329 U.S. 495, 510-511 [91 L.Ed. 451, 462-463].) But also underlying the doctrine is the strong policy against putting the attorney’s credibility in issue by introducing material he has prepared, and thus creating the possibility that he will be forced to testify. (Id. at pp. 512-513; see also id. at p. 517 [91 L.Ed. at pp. 463, 465-466], conc. opn. of Jackson, J.; Nobles, supra, 422 U.S. at pp. 252-254 [45 L.Ed.2d at pp. 161-163], conc. opn. of White, J.)
In the first report, the resume stated, “Subj. can testify in a limited way about def and vic’s relationship, subj. can give a possible account of defs whereabouts on the evening of the A/O, subj. can give a limited account of the vic’s behavior, and finally subj. can testify to defs reaction when the subj. informed the def of the alleged incident.” In the second report, prepared eight months later, the resume stated “The subj. will testify to the extent of the defs surprise when learning of the news article charging def. with the alleged offense."
Although defense counsel at trial objected to discovery on the additional basis of the attorney-client privilege, that ground has rightly been abandoned by counsel on appeal. Because the communication with the defense investigator involved not the client but a third party witness, and because there has been no suggestion that the interview was the product of a confidential communication between attorney and client, the privilege is clearly inapposite.
To the extent it is inconsistent with this opinion, People v. Logan (1966) 244 Cal.App.2d 795 [53 Cal.Rptr. 549], is hereby disapproved.
Former section 217 authorized penalties of two to four years for assaults with intent to commit murder. The Legislature swiftly and unequivocally expressed its disapproval of the application of the section to attempted murder convictions in Gray *63and Montano by repealing the statute. (Stats. 1980, ch. 300, § 2, see People v. Murtishaw, supra, 29 Cal.3d 733, 762-763, fn. 24; People v. Singleton (1980) 112 Cal.App.3d 418, 424 [169 Cal.Rptr. 333].)
The alleged incidents never resulted in prosecution because after calling the police, Mrs. Collie was unwilling or unable to explain what had happened and did not sign a complaint. Defense counsel first mentioned the incidents with the apparent purpose of proving that Mrs. Collie was prone to summon the police without just cause. The prosecutor then relied on them ostensibly to establish defendant’s retaliatory motive and his knowledge that his wife would not press charges if he attacked her again.
People v. Williams (1970) 11 Cal.App.3d 970, 976-978 [90 Cal.Rptr. 292], and People v. Demes (1963) 220 Cal.App.2d 423, 439 [33 Cal.Rptr. 896], are disapproved to the extent they are inconsistent with this opinion.
Assigned by the Chairperson of the Judicial Council.