OPINION

BIRD, C.J.

This petition asks this court to decide whether a prospective juror who can be fair and impartial in determining the guilt or innocence of an individual accused of a capital offense may be removed for cause from serving at the guilt phase because the juror is unequivocally opposed to the imposition of the death penalty at the penalty phase?

I

By an information filed in the Alameda County Superior Court, petitioner, Richard Adams Hovey, stands accused of murder and kidnaping. (Pen. Code, §§ 187, 207.) Two "special circumstances" are alleged in connection with the murder charge. If petitioner is convicted of first degree murder and if the special circumstances allegations are found to be true, the jury will determine whether the sentence is "death or confinement in the state prison for life without possibility of parole. . . ." (Former § 190.2.)

Petitioner brought a pretrial motion to limit the exclusion for cause of prospective jurors to be called to try his case. The thrust of his motion was that the guarantee in the state and federal Constitutions to due process of law and an impartial jury prohibit the trial court from excluding at the guilt phase of the trial prospective jurors who would be fair and impartial, but who are unequivocally opposed to imposing the death penalty at the penalty phase (should this second phase prove necessary).

The evidentiary basis for petitioner's motion was developed at an extensive evidentiary hearing held in August and September 1979, in a separate proceeding in the same county. (People v. Kenneth Lynn Moore and David Lee Moore, Alameda Co. Super. Ct. No. 67113.) The Moore brothers' motion covered 17 court days and produced a reporter's transcript of more than 1,200 pages. Seven expert witnesses testified, five for the defense and two for the prosecution. In excess of 1,000 pages of exhibits — primarily sociological studies and graphs and charts — were admitted into evidence, as were several videotapes. By stipulation in the present case, these transcripts and exhibits were introduced at petitioner's motion. No new evidence was presented.

Petitioner's motion to limit exclusions for cause under section 1074, subdivision 8, was denied by the judge who ruled on the Moores' motion. Petitioner sought to invoke the original jurisdiction of this court by filing a petition for writ of mandamus. This court issued an alternative writ based on the importance of the issue presented.

II

LEGAL BACKGROUND

Section 1074, subdivision 8, appears to have been given an expansive interpretation in another respect. The statute authorizes the exclusion of jurors whose views "preclude" them from returning verdicts of guilt or death. Prior to 1968 this court held that jurors were properly excused for cause under the statute if they had any conscientious scruples against the infliction of the death penalty or if they stated that they "did not believe in capital punishment."

However, this court was not always consistent with respect to this aspect of the statute. For example, in People v. Bandhauer (1967) 66 Cal.2d 524, 531 [ 58 Cal.Rptr. 332, 426 P.2d 900], the court indicated that mere "doubts with respect to the death penalty. . . are not sufficient to disqualify a juror so long as he conscientiously believes that he could return a death penalty verdict in a proper case." (See also People v. Riser, supra, 47 Cal.2d at pp. 575-576 [suggesting that § 1074, subd. 8, compels the exclusion of jurors "incapable" of returning a death verdict]; People v. Rollins (1919) 179 Cal. 793, 796 [ 179 P. 209] [the statute excludes jurors who are "so bound by conscientious opinions as to capital punishment that they will be unable to act as the circumstances of the case demand in view of the law" (italics added)].)

This apparent conflict was conclusively resolved in June 1968 when the United States Supreme Court issued its landmark decision in Witherspoon v. Illinois (1968) 391 U.S. 510 [20 L.Ed.2d 776, 88 S.Ct. 1770]. In Witherspoon, a criminal defendant had been convicted of murder and sentenced to die. The prosecutor had removed for cause from the jury which had tried the case all individuals who "[had] conscientious scruples against capital punishment, or . . . [were] opposed to the same." On appeal, the defendant argued that the jury was unconstitutionally biased in favor of conviction and death.

The Supreme Court reversed Witherspoon's death sentence but upheld his underlying murder conviction. With respect to the penalty imposed, the court held it to be "self-evident" that if a state excuses prospective jurors for cause on the basis of "general objections to the death penalty or . . . conscientious or religious scruples against its infliction," then the resulting jury "cannot speak for the community" and is "uncommonly willing to condemn a man to die." This, the court suggested, violates one of the "basic requirements of procedural fairness . . . [i.e.,] that the decision whether a man deserves to live or die must be made on scales that are not deliberately tipped toward death." ( Id., at pp. 521-522, fn. 20 [20 L.Ed.2d at pp. 784-785].) The only prospective jurors who could constitutionally be excused for cause due to their opposition to or doubts about capital punishment were "those who made unmistakably clear (1) that they would automatically vote against the imposition of capital punishment without regard to any evidence that might be developed at the trial of the case before them, or (2) that their attitude toward the death penalty would prevent them from making an impartial decision as to the defendant's guilt." ( Id., at pp. 522-523, fn. 21 [20 L.Ed.2d at p. 785], original italics.)

The court did not reverse Witherspoon's substantive conviction for murder. It found the empirical studies tendered on his behalf on appeal "too tentative and fragmentary" to establish that the broad exclusion of jurors with scruples against capital punishment "results in an unrepresentative jury on the issue of guilt or substantially increases the risk of conviction." ( Id., at pp. 517, 518 [20 L.Ed.2d at p. 782].) Noting that Witherspoon had specifically declined to present evidence below in support of his contention, the court stated, "We can only speculate, therefore, as to the precise meaning of the terms used in those studies, the accuracy of the techniques employed, and the validity of the generalizations made." ( Id., at p. 517, fn. 11 [20 L.Ed.2d at p. 782].) On the issue of guilt, the court found "the record . . . 'almost totally lacking in the sort of factual information that would assist the Court.'" ( Id., at p. 518, fn. 11 [20 L.Ed.2d at p. 782].)

Nevertheless, the court indicated that the question was still an open one. Indeed, in footnote 18 of its opinion, page 520 [20 L.Ed.2d, page 784], the court noted that a defendant "in some future case might still attempt to establish that [a] jury [which had been 'death-qualified' in compliance with the newly announced Witherspoon standards] was less than neutral with respect to guilt."

In the 12 years since the Witherspoon decision, the Supreme Court has not addressed the constitutionality of using a " Witherspoon-qualified" jury to determine the issue of guilt or innocence in a capital case. However, the court has expounded upon the Sixth Amendment's guarantee of "trial by impartial jury" in two pertinent cases.

In Taylor v. Louisiana (1975) 419 U.S. 522 [42 L.Ed.2d 690, 95 S.Ct. 692] the Supreme Court reversed the state court criminal conviction of a male defendant, since women had been systematically excluded from the jury pool. In doing so, the court held for the first time that the states were bound by a Sixth Amendment fair-cross-section requirement. The court held that the gross underrepresentation of women in the jury panel deprived Taylor of "the kind of factfinder to which he was constitutionally entitled." (419 U.S. at p. 526 [42 L.Ed.2d at p. 696].) No burden was placed on the accused to establish anything concerning the attitudes, propensities, or juryroom voting behavior of the group excluded (women) or the group included (men). As has been recently noted, the Taylor court did not even consider "whether a jury composed of men only would be prejudiced or would have a 'tendency' to convict or acquit; it . . . simply [relied on] the requirement that the jury be selected from a panel representative of the community." ( Grigsby v. Mabry (E.D.Ark. 1980) 483 F. Supp. 1372, 1380, app. pending 8th Cir.)

The second important case dealing with the Sixth Amendment was Ballew v. Georgia (1978) 435 U.S. 223 [55 L.Ed.2d 234, 98 S.Ct. 1029]. There the court held that a criminal conviction rendered by a five-person jury violated the Sixth and Fourteenth Amendments. Ballew represented the culmination of a line of decisions, concerning jury size or unanimous verdict, which set forth a new analysis of the Sixth Amendment. Under this approach the court determines whether the "purpose and functioning of the jury in a criminal trial" is impaired to a significant degree by the challenged procedure (e.g., by the procedure of permitting a conviction to be based on the verdict of a five-member jury). ( Ballew v. Georgia, supra, 435 U.S. at p. 239 [55 L.Ed.2d at p. 246].) To ascertain the impact on the jury's purpose and functioning, the court in Ballew relied extensively on relevant scholarly and empirical studies. ( Id., at pp. 231-239, 244-245 [55 L.Ed.2d at pp. 241-246, 249-250].)

The high court has stated that trial by jury in a criminal case "safeguard[s] against the corrupt or overzealous prosecutor and against the compliant, biased, or eccentric judge" and "assure[s] a fair and equitable resolution of factual issues." These purposes are achieved "by the participation of the community in the determinations of guilt and by the application of the common sense of laymen who, as jurors, consider the case." ( Ballew v. Georgia, supra, 435 U.S. at p. 229 [55 L.Ed.2d at p. 240].)

In the context of a challenge to the size of a criminal jury, the court in Ballew articulated several concerns of constitutional dimension. It held that a jury should be "of sufficient size to promote group deliberation, to insulate members from outside intimidation, and to provide a representative cross-section of the community." ( Id., at p. 230 [55 L.Ed.2d at p. 240]; see also Williams v. Florida, supra, 399 U.S. at p. 100 [26 L.Ed.2d at p. 460].) A five-member jury was inadequate, the court reasoned, because "recent empirical data" suggested such a jury was "less likely to foster effective group deliberation," resulting in "inaccurate factfinding and incorrect application of the common sense of the community to the facts." (435 U.S. at p. 232 [55 L.Ed.2d at p. 242].)

In addition, smaller juries were "less likely . . . to overcome the biases of its members to obtain an accurate result. . . . [T]he counterbalancing of various biases is critical to the accurate application of the common sense of the community to the facts of any given case." ( Id., at pp. 233-234 [55 L.Ed.2d at pp. 242-243].) Further, the court enunciated an independent constitutional interest in protecting the representation on juries of minority viewpoints in the community and of minority groups. ( Id., at pp. 236-237 [55 L.Ed.2d at p. 244].) It concluded that decreasing jury size from six to five "foretells problems" for each of these interests. ( Ibid.)

Empirical research was also used by the court in its determination that smaller juries decreased the jury's ability to "remember all the important facts and arguments" (435 U.S. at p. 241 [55 L.Ed.2d at p. 247.]) and reduced the likelihood that juries would "compromise over the various views of their members, an important phenomenon for the fulfillment of the commonsense function." ( Id., at p. 235, fn. omitted [55 L.Ed.2d at pp. 243-244].) Smaller juries were also found to result in inconsistent verdicts and to decrease the incidence of hung juries. ( Id., at pp. 234, 236 [55 L.Ed.2d at pp. 242-243, 244].)

Thus, the court concluded, "the assembled data raise substantial doubt about the reliability and appropriate representation of panels smaller than six. Because of the fundamental importance of the jury trial to the American system of criminal justice, any further reduction that promotes inaccurate and possibly biased decisionmaking, that causes untoward differences in verdicts, and that prevents juries from truly representing their communities, attains constitutional significance." ( Id., at p. 239 [55 L.Ed.2d at p. 246].)

III

THE ISSUES PRESENTED

Two separate constitutional theories are advanced to reach this conclusion. One is bottomed on Witherspoon. (3) (See fn. 37.) Under it, petitioner contends there is evidence which can "establish" that a jury from which all "guilt phase includables" have been removed for cause is "less than neutral with respect to guilt." (391 U.S. at p. 520, fn. 18 [20 L.Ed.2d at p. 784], italics omitted.) The other constitutional theory relies on the "purpose and functioning" analysis developed in Ballew v. Georgia, supra, 435 U.S. 223, and the line of decisions it represents. This theory would require the court to determine — under a "substantial doubt" standard — whether the interests protected by the Sixth Amendment are significantly inhibited by the disqualification at the guilt phase of all the "guilt phase includable" jurors.

The second basic challenge to the current system of selecting a jury for a capital case focuses on the voir dire procedure by which the jury is selected, rather than on the composition or functioning of the jury itself. The argument is advanced that whoever is chosen to serve on a jury in a capital case, he or she will be affected by the methods now employed to identify those persons whose views on capital punishment render them ineligible. The result, it is contended, is a jury likely to be "less than neutral" with respect to guilt and penalty. Petitioner suggests that certain procedural refinements would minimize this possibility.

IV

THE CONCEPT OF A NEUTRAL JURY

When the Supreme Court in Witherspoon held that "as arbiter of the punishment to be imposed, this jury fell woefully short of that impartiality to which the petitioner was entitled under the Sixth and Fourteenth Amendments," it was not ruling that any of the jurors who actually sat on Witherspoon's penalty trial were themselves biased in a constitutional sense. If this had been the court's concern, it would have ordered that such jurors be excluded from future capital cases.

Instead, the high court was relying on a different facet of the constitutional requirement of impartiality, an aspect that has been referred to as a guarantee of "diffused impartiality." It is denoted in this opinion (and in Witherspoon ) as "neutrality." (4) A neutral jury is one drawn from a pool which reasonably mirrors the diversity of experiences and relevant viewpoints of those persons in the community who can fairly and impartially try the case.

The concept of neutrality through diversity is demonstrated by the holding in Witherspoon. Assume that a jury must be empanelled to determine the question of punishment in a capital case. In the group of persons from the community who are statutorily competent to act as trial jurors, it can be expected that an entire spectrum of beliefs concerning the infliction of capital punishment — from persons who would invariably and automatically refuse to impose capital punishment to jurors who would automatically vote to impose the death penalty following a conviction for a capital offense — would be found. This spectrum of community viewpoints may be depicted as follows:

-------------------------------------------------------------------------- | "AUTOMATIC | "FAVOR | "INDIFFERENT" | "OPPOSE | "AUTOMATIC | | DEATH | DEATH | GROUP | DEATH | LIFE | | PENALTY" | PENALTY" | | PENALTY" | IMPRISONMENT" | | GROUP | GROUP | | GROUP | GROUP | |----------------|------------|---------------|-----------|----------------| | will | favors the | neither | opposes | will | | automatically | death | favors nor | or has | automatically | | vote for | penalty | opposes the | doubts | vote for life | | the death | but will | death | about the | imprisonment | | penalty | not vote | penalty | death | | | | to impose | | penalty | | | | it in | | but will | | | | every | | not auto- | | | | case. | | matically | | | | | | vote | | | | | | against | | | | | | it in | | | | | | every | | | | | | case. | | --------------------------------------------------------------------------

Under the state procedure in effect at the time Witherspoon's penalty was determined, all persons in the "oppose death penalty" group and the "automatic life imprisonment" group were excluded for cause. Thus, Witherspoon's jury was drawn from a pool of persons which did not reflect the range of community viewpoints on a critical aspect of the case. It also contained at least one group — the "favor death penalty" group — whose members came into trial with preconceived attitudes about the imposition of the death penalty which tended to favor one of the litigants, i.e., the prosecution.

In holding that a jury so constituted "crossed the line of neutrality," the court found that the jury was not impartial within the meaning of the Sixth and Fourteenth Amendments because it was too narrowly drawn. Members of the community who could "obey the oath" to fairly decide the case before them were excluded. As a result, a segment of the population that would tend to be favorable toward the accused was eliminated. The resulting jury was less than neutral with respect to penalty so a reversal of the penalty determination was required.

Clearly, the constitutional principle of achieving jury neutrality through diversity is relevant to a determination of guilt as well as penalty. Every juror brings to the guilt phase a number of personal characteristics which will "play an inevitable role" in assessing the accused's guilt or degree of guilt. (See Witherspoon v. Illinois, supra, 391 U.S. at p. 519 [20 L.Ed.2d at p. 783].) As members of this court have recently observed, "each juror brings to the deliberations [on guilt or innocence] his personal store of experience, knowledge, and judgment; these are the tools by which he tests 'the credibility, the probability of the testimony of witnesses, or of the inferences to be drawn from circumstances. . . .'" ( People v. Brigham (1979) 25 Cal.3d 283, 299 [ 157 Cal.Rptr. 905, 599 P.2d 100] (conc. opn. of Mosk, J.) quoting Trickett, Preponderance of Evidence, and Reasonable Doubt (1906) 10 The Forum 75, 82.) In addition, the juror must determine whether the evidence thus evaluated amounts to proof "beyond a reasonable doubt" of the truth of the charges. The difficulty in defining "this elusive and undefinable state of mind" has been chronicled by scholars. "Each of us, in effect, has his own subjective sense of when a chance of innocence can be disregarded as de minimis, but our respective senses are surely different."

Manifestly, fair and impartial jurors will bring to the determination of guilt a diversity of experience, knowledge, judgment, and viewpoints, as well as differences in their "thresholds of reasonable doubt." If some of these jurors are systematically removed from the guilt determination, this may result in a disproportionate elimination of persons with characteristics favorable to the accused. If so, the ensuing jury will be "less than neutral with respect to guilt," just as the jury at the penalty phase in Witherspoon was not neutral with respect to penalty.

Including diverse viewpoints on a jury serves several purposes. Diversity aids the accuracy of jury decision making by "counter-balancing . . . various biases" of the jury members. ( Ballew v. Georgia, supra, 435 U.S. at p. 234 [55 L.Ed.2d at p. 242].) As this court recently noted in a fair-cross-section case, "in our heterogeneous society jurors will inevitably belong to diverse and often overlapping groups defined by race, religion, ethnic or national origin, sex, age, education, occupation, economic condition, place of residence, and political affiliation; . . . it is unrealistic to expect jurors to be devoid of opinions, preconceptions, or even deep-rooted biases derived from their life experiences in such groups; and hence . . . the only practical way to achieve an overall impartiality is to encourage the representation of a variety of such groups on the jury so that the respective biases of their members, to the extent they are antagonistic, will tend to cancel each other out." ( People v. Wheeler (1978) 22 Cal.3d 258, 266-267, fn. omitted [ 148 Cal.Rptr. 890, 583 P.2d 748].)

Diversity serves to complement as well as neutralize viewpoints and attitudes. Diversity enhances the accuracy of a jury's decision making by improving its ability to recognize and appropriately evaluate evidence. Testimony from the hearing below, as well as studies in social psychology, help to explain why this is so. Human perception is selective, influenced by the very beliefs and attitudes which venirepersons bring into the courtroom. New data which tend to contradict one's beliefs may be quickly "forgotten" or may not even be perceived in the first place. The members of a homogeneously composed jury are more likely to perceive evidence in a similar fashion. Also, they are more likely to filter out any evidence inconsistent with their shared attitudes and values. Insofar as a jury is composed of members whose attitudes, preconceptions, and experiences are diverse, the jury is more likely to perceive and remember all the important evidence and arguments presented at trial.

In similar fashion, the human mind often tends to make any new information with which it is confronted logically consistent with its prior conscious beliefs. Thus, if a juror's beliefs do not correspond to the evidence presented at trial, the juror's "rational nature" may tend to impel him or her to distort or exclude the perception so as to protect the apparent reasonableness of the belief. Thus, diversity provides a corrective to the distortions which can occur when the evidence presented at trial is inconsistent with the preconceptions of some members of the jury.

Jury diversity helps to insure the full and accurate consideration of the evidence presented at trial in another way. A jury resolves conflicting propositions of fact and does so by drawing inferences from physical evidence and the testimony of witnesses. Yet "[f]acts are always elusive and often two-faced. What may appear to one to imply guilt may carry no such overtones to another." ( Johnson v. Louisiana, supra, 406 U.S. at p. 392 [32 L.Ed.2d at p. 177] (dis. opn. of Douglas, J.).) As this court noted more than a century ago, "the human mind is so constituted, that facts and circumstances do not always produce the same results; the judgment of two men upon the same set of facts may be diametrically opposite, particularly in the determination of a criminal case, when every doubt is carefully weighed and scrupulously balanced." ( People v. Stewart (1857) 7 Cal. 140, 144.)

If a jury is accurately to assess evidence, it should have some expertise both in generating the inferences which may reasonably be drawn from the evidence and in evaluating the relative plausibility of the competing inferences. The greater the diversity of individual viewpoints and experiences on the jury panel, the broader the range of appropriate inferences the jury can draw from the evidence at trial and the more knowledgeable their interpretation and weighing of these inferences one against the other. For example, jurors in criminal cases are often called upon to infer mental states from behavior. In a culturally pluralistic society, particular behavior can have dramatically different meanings to members of different subcultures. A jury with diverse membership will recognize a fuller range of possible meanings or explanations for particular behavior, and it will be able to evaluate those possible meanings in light of the diverse experiences of the panel members regarding values, norms, behavior, motivation, and psychology.

Finally, even if the evidence and inferences were agreed upon by all jurors, the legal effect may be subject to dispute. For example, jurors must determine whether the evidence amounts to proof beyond a reasonable doubt of the guilt of the accused. However, jurors' thresholds of reasonable doubt will necessarily differ. Diversity of viewpoint and experience on the jury tends to insure that the common sense of the community is accurately expressed in applying this standard to the facts.

Of course, a juror's attitudes, experiences, knowledge, judgment, and threshold of reasonable doubt are not the sole factors which decide how that juror votes. Rather, as the defense experts repeatedly testified below, the weight of the evidence is the primary determinant of the juror's decision. Thus, Kalven and Zeisel reported in their classic study, The American Jury, that roughly one-third of the juries to which criminal cases have been submitted reach a unanimous verdict for acquittal or conviction on the very first ballot. ( Supra, p. 488.) "That means," Dr. Zeisel testified with respect to these cases, "that whatever differences between personalities exist, [they] are overwhelmed by the clarity of the evidence, be it for acquittal or conviction."

The corollary to this finding is that in two-thirds of criminal trials, the jurors disagree in their initial ballot. Since the 12 jurors were exposed to the identical external events or "stimuli" — i.e., the same witnesses, evidence, instructions, courtroom personalities, preliminary deliberations (if any) — the inescapable conclusion is, as Dr. Zeisel testified, that "there must be something inside these people which makes them differ." It is petitioner's contention under the Witherspoon analysis that the process of "death-qualifying" a jury produces a jury whose composition is unbalanced. It tends disproportionately to remove from the jury pool those persons who have "something inside" which results in their voting in favor of the accused more often than do persons who remain. As a result, jurors drawn from the remaining pool will tend to vote in favor of the prosecution more often than would jurors drawn from a pool which includes a more complete spectrum of attitudes on capital punishment. (5) One final point deserves emphasis. In determining the constitutional "neutrality" of a "death-qualified" jury, it is unnecessary to make any judgment that one viewpoint, attitude, experience, etc., is preferable, morally or legally, to any other. What is essential under this analysis is that none of these viewpoints, attitudes, experiences, etc., are systematically excluded from the jury pool, either directly or by de facto operation of otherwise neutral laws or practices, to the detriment of the accused.

V

THE EVIDENCE RELATING TO THE WITHERSPOON AND BALLEW CONTENTIONS

Roughly two dozen studies, experiments, and surveys were introduced at or were the subject of expert testimony during the evidentiary hearing below. This material can logically be grouped for analytical purposes into four categories: (1) the research relating directly to conviction proneness or juror voting behavior; (2) the research relating to juror attitudes; (3) the research relating to demographic characteristics; and (4) the research relating to differences in juror evaluation of evidence.

CONVICTION PRONENESS

However, such an experiment is legally impossible. Thus, the hypothesis must be tested indirectly. One method of indirect analysis is to locate persons who have actually sat on juries and ask them what they did. The strength of this method — which was used in the Zeisel study that follows — is that its data are derived from a very realistic setting: an actual trial. The weakness of this method is the difficulty of controlling for a very important variable: the strength of the evidence. Since it is difficult to assure that the subjects of this type of research experienced the same kind of evidence, it is correspondingly difficult to determine whether any differences discovered reflect true differences in the subjects or merely differences in the stimuli to which they have been exposed.

A second type of approach in researching this issue is a controlled study, i.e., a study done in a controlled setting, in which realistic stimuli are so presented that every subject/juror is exposed to the exact same stimuli. If this is done well, then any differences observed must be real differences between the people and not differences in the kinds of experiences to which they are responding. The remaining five studies in this portion of this opinion are of this "controlled study" variety.

The Zeisel Study

In the fall and winter of 1954-1955, Professor Hans Zeisel and two associates interviewed a number of jurors who had actually deliberated on felony cases in the Criminal Court of Chicago, Illinois, or the County Court of Kings County (Brooklyn), New York. Zeisel's purpose was to explore whether the jurors' attitudes toward capital punishment correlated with a tendency to vote for guilt or acquittal in criminal cases. He asked the jurors whether they had any conscientious scruples against the death penalty and how they had voted on the first ballot after the jury started to deliberate.

Asking only these two questions would have resulted in a wholly uncontrolled study, for in this posture, the factor of the strength of the evidence had not been taken into account. Zeisel devised a rather ingenious question to get at this factor. He concluded that the strength of the evidence in a given case could be roughly estimated and compared with other cases by determining how the jury as a whole voted on the first ballot after deliberations began. Thus, a first ballot vote of 11-to-1 in favor of acquittal suggested a weak case against the accused; a 10-to-2 vote for acquittal indicated a weak case but not quite so weak; a 9-to-3 vote for acquittal, an even less weak case; and so on, through all 11 possible jury splits, to 11-to-1 for conviction. By grouping each juror's vote into one of the eleven categories or "constellations" based on the strength of the evidence — from the weakest prosecution evidence to the strongest — Zeisel could roughly control for the weight of the evidence.

Zeisel conducted his interviews in the jury assembly room on the last day of the jurors' terms of service. He and his associates interviewed as many jurors who actually sat on a case as possible. Some had deliberated on two cases. Most of the jurors contacted agreed to be interviewed. Zeisel ended up with data on 464 first ballot votes. These were classified according to the strength of the evidence, using the first ballot split as an index of this strength. He compiled the results into table form as follows:

Zeisel found that in 10 of the 11 constellations of evidence strength, jurors with conscientious scruples against capital punishment voted to acquit more often than jurors without such scruples. In 9 of the 11 constellations, jurors without such conscientious scruples voted to convict more often than jurors with scruples.

Zeisel determined that these differences between the first ballot guilty votes of the jurors without conscientious scruples against capital punishment and the first ballot guilty votes of jurors with such scruples were statistically significant at the .04 level. ( Zeisel, supra, at p. 32.) At the hearing below, he elaborated on his findings. "Whatever the case is, if there is a split ballot, my statement is [that] jurors who have scruples against the death penalty as a group will have a lower percentage of guilty votes than jurors [who] have no scruples."

The Goldberg Study

Dr. Faye Goldberg (now Girsh) undertook a controlled study of the relationship between conviction proneness and attitudes toward capital punishment in 1966-1967. The subjects in this study were 200 undergraduate students enrolled in colleges in Atlanta, Georgia. One hundred of the students were white, one hundred were black. Eighty-four were female, one hundred sixteen were male.

The subject/jurors were given a written questionnaire, containing short (four or five sentence) descriptions of each of sixteen simulated criminal cases. The 16 cases depicted primarily murder charges, and all were "cases in which the death penalty could be given in most jurisdictions which [had] the death penalty." ( Goldberg, supra, 5 Harv.Civ. Rights-Civ.Lib.L.Rev. at p. 59.)

The described cases varied in severity, heinousness, justification, and amount of evidence available. The subject/jurors were instructed to decide the issue of guilt or innocence on the basis of the evidence summarized in the written description; no descriptions of the law were given. The subject/jurors were told to decide which of the following four categories of guilt or innocence they felt was appropriate: guilty of first degree murder, guilty of a lesser offense, not guilty by reason of insanity, and not guilty for lack of evidence. In addition, as this study occurred prior to Witherspoon, the subject/jurors were asked, "Do you have conscientious scruples against the use of the death penalty?" (Cf., post, fns. 72, 79.)

Dr. Girsh found that subject/jurors without conscientious scruples against the death penalty voted "guilty" on some crimes (i.e., first degree murder or a lesser offense) in 75 percent of their votes, whereas subject/jurors with such scruples voted guilty in 69 percent. The 6 percent difference between the groups was found to be marginally significant. (P = .08.)

Dr. Girsh also found that subject/jurors with conscientious scruples voted "not guilty by reason of insanity" in 14 percent of the votes, compared with 9 percent for the subject/jurors without such scruples. This finding was statistically significant. (P = .03.)

Dr. Girsh concluded that her results "do tend to corroborate the trends found" in the studies by Dr. Zeisel and Dr. Wilson (see post). ( Goldberg, supra, 5 Harv.Civ. Rights-Civ.Lib.L.Rev. at p. 69.) As Dr. Girsh testified, her study did not employ "sophisticated methodology," but the purpose of her study was to "try to find out if there was any reason to believe that this relationship existed at all. . . . [W]e were trying to first tap the most superficial level of the relationship to see if there was anything." Based on the data from her study, Dr. Girsh concluded "there may be a basis . . . to think that when people with conscientious scruples against the death penalty are removed from the jury . . . the resulting juries [may be] more likely to convict."

The Wilson Study

The subjects of Wilson's study were 187 junior and senior college students at the University of Texas. They were given one-half page written descriptions of "the facts presented to the jury in five criminal court cases." ( Wilson, supra, at p. 2; the written descriptions are set forth at pp. 10-12 of the study.) Four of these cases involved homicide charges, and one of the four (a robbery murder) involved two defendants. The fifth described case was a rape charge. Four of the case descriptions included brief jury instructions. The subject/jurors were asked to "assume that you are a member of the jury before whom each of these cases is being tried. . . . On the basis of these facts and your interpretation and evaluation of them, decide whether you feel the defendant is guilty or innocent and indicate your feelings on the answer sheet provided." ( Id., at p. 9.)

Wilson used the total number of guilty verdicts rendered by each subject/juror as "an index of [that juror's] tendency to say guilty." ( Id., at p. 2.) Thus, because one case involved two defendants, the maximum number of guilty or not guilty verdicts was six. Wilson then compared the percentage of guilty votes cast by the subject/jurors who indicated they had "conscientious scruples against the death penalty, or capital punishment, for crime" with the percentage of guilty votes cast by subject/jurors without such scruples. The results he reported are illustrated by the following graph.

Wilson found that a majority of both groups voted guilty on three or four occasions. However, 22 percent of the subject/jurors with conscientious scruples against the death penalty voted guilty on two or less occasions, as compared to only 9 percent of the subject/jurors without such scruples. At the other end of the spectrum, 30 percent of the jurors without conscientious scruples against capital punishment voted guilty in five or six cases, as compared to 17 percent of the jurors with such scruples. These results were statistically significant; the "p" value was less than .02.

As the testimony below indicated, the Wilson study, like Goldberg, was "certainly not a very sophisticated study." It had generally the same shortcomings as Goldberg. It was useful primarily to suggest there may be some "relationship between attitudes toward the death penalty and propensity to vote guilty in criminal cases."

The Jurow Study

Professor Jurow undertook the first controlled experiment in the post- Witherspoon era. His subjects were 211 employees at a Sperry-Rand Corporation plant on Long Island, New York. Virtually all of these individuals were eligible for jury duty. One-third of the subjects had had prior jury service. The group was overwhelmingly white (98.6 percent), fairly well educated, and had a relatively high median family income. Women constituted 26 percent of the subjects; Catholics nearly 50 percent; engineers 39 percent; and clerical workers and laborers 29 percent.

Jurow prepared two audiotapes of simulated criminal trials. A script was prepared for each case, which "attempt[ed] to weight the evidence as evenly as possible between acquittal and conviction." ( Jurow, supra, 84 Harv.L.Rev. at p. 581.) The script included opening statements by the attorneys, witnesses' testimony, direct and cross-examination, arguments of counsel, and the judge's instructions to the jury. The audiotapes were pretested to insure that the evidence was appropriately balanced and that the recordings were realistic and did enhance listener involvement. ( Ibid.)

The first audiotape lasted 33 minutes (case I). It involved "the murder of a liquor store proprietor during a holdup and the apprehension and trial of an ex-convict seen running from the vicinity of the store who denied any knowledge of the robbery and murder." ( Ibid.) The second tape (case II) lasted 50 minutes. It portrayed "a narcotics addict charged with robbing, raping, and killing a college girl in her apartment." ( Ibid.) At the conclusion of each tape, the subject/jurors were given time to think about their verdict and mark a ballot. There were no deliberations by the subject/jurors as a group.

The attitudes of the subject/jurors toward capital punishment were determined on a five-part spectrum, closely corresponding to the spectrum portrayed on page 20, ante. The percentage of guilty votes cast within each group can be depicted as follows:

In both cases, Jurow found some tendency for an attitude favorable to capital punishment to correlate with a tendency to vote for conviction. In case I (the liquor store robbery-murder), this tendency was highly significant (the "p" value was less than .01); in case II, the tendency was at best marginally significant (the "p" value is greater than .05).

Petitioner has reorganized the Jurow data, to compare directly the percentage of guilty votes cast by the "automatic life imprisonment" category with the percentage of guilty votes cast by the remaining categories lumped together. This reorganization has been graphed as follows:

Jurow represented a significant advance in the controlled studies in two respects. One, since his study followed Witherspoon, Jurow was able to ask his capital punishment questions in a manner precisely relevant to post- Witherspoon capital trials. Two, Jurow used a method for presenting evidence to his subject/jurors (i.e., audiotapes) that was much closer to a real trial than were the written questionnaires employed in Goldberg and Wilson. In that regard, it can be expected that the differences reported in Jurow are probably a closer approximation to reality than are those reported by the other two studies.

Jurow is not a definitive controlled study, however. Its presentation of evidence involved only oral, not visual, stimuli. The subjects were drawn from a limited population. There were no group deliberations and there may have been little felt responsibility on the part of the subjects as compared to actual jurors sitting in a real trial.

The Harris 1971 Study

In 1971 the Harris Poll organization conducted a nationwide survey regarding capital punishment. The subjects of the study were 2,068 adults, who comprised a stratified sample of the adult population of the United States. These subjects were given direct person-to-person interviews lasting about an hour. One portion of the interview involved a conviction-proneness study.

The subject/jurors were presented with a card containing certain instructions. The subject/jurors were told to assume that they were members of a jury in a criminal case and that they had to determine their verdicts from the evidence presented. They were instructed to keep three legal rules "in mind": the concept of proof beyond a reasonable doubt, the fact that the prosecution has the burden of proof, and the rule that the accused is not required to testify.

The subject/jurors were then presented detailed written descriptions of four criminal cases. These were referred to throughout the hearing below as the "typewriter robbery," the "manslaughter" case, the "assaulting an officer" case, and the "automobile larceny" case. The subject/jurors were provided with written legal definitions of the crimes for which each accused was on trial. They were told to use that definition in order to assess the accused's guilt or innocence.

The subject/jurors were also asked whether in a murder trial "there would be any situations in which you might vote for the death penalty, or do you think you could never vote for the death penalty, regardless of the circumstances?" This question differentiated the "automatic life imprisonment" group from the rest of the populace (i.e., the "Witherspoon qualified" group, which in the chart below petitioner calls the "death qualified" jurors). (Cf., ante, fn. 71.) It was found that in all four of the cases presented, the "automatic life imprisonment" jurors (petitioner calls these the "excludables") voted to convict less often than did the remaining jurors. In three of the four cases, the differences were highly significant ("p" value less than .01) and in the fourth case, the difference was marginally significant ("p" value less than .10). Combining each groups' votes across all four cases revealed an overall difference of 7 percent in the frequency with which these groups voted to convict. These results are portrayed below.

The Ellsworth Conviction-Proneness Study

The most recent controlled study of conviction-proneness was undertaken in 1979 by Drs. Phoebe Ellsworth, William Thompson, and Claudia Cowan. They recruited as subjects 288 adults eligible for jury service in California. Of the group of 288, most (218) had responded to a local newspaper advertisement asking for volunteers for a study of "how juries make decisions;" 37 were recruited from the jury lists for Santa Clara County Superior Court after being discharged from further jury duty; and 33 were friends of other persons who participated as subjects in the experiment. Forty-five percent of the subjects had had prior jury service.

The subject/jurors participated in the study in groups of 12 to 36. They were shown a two-and-one-half-hour videotaped reenactment of an actual criminal trial. The videotape had been developed by Harvard Professor Reid Hastie for use in his studies of nonunanimous jury verdicts. Hastie based the tape on an actual trial in Massachusetts, although it was not a scripted reenactment of that trial but a spontaneous one. The actors in the tape had been provided with a transcript of the actual trial and other relevant materials. The "prosecutor" and "defense counsel" were told to conduct the case as they saw fit, and the "witnesses" themselves were to testify as closely as possible to the "real" testimony in the transcript. The role of the district attorney was played by a local prosecutor. The defense counsel was a defense attorney in private practice. The judge presiding over the trial was a Massachusetts judge and the role of the police officer witness was played by a recently retired policeman. Ellsworth altered the Hastie tape somewhat. Since the judge in the Hastie tape gave jury instructions based on Massachusetts law, that portion of the videotape was recreated with a law school dean acting as judge and giving CALJIC instructions.

After this videotape had been played, each subject/juror was asked to indicate how he or she would vote, based upon his or her "own personal individual decision." There were four possible verdicts: guilty of first degree murder, guilty of second degree murder, guilty of voluntary manslaughter, and not guilty by reason of self-defense or excusable homicide.

The subjects' views on capital punishment had been previously determined. Of the 288 persons participating in the experiment, 30 were "guilt phase includables" and 258 were "Witherspoon-qualified" (i.e., persons who could impose the death penalty in at least some circumstances).

The results of the Ellsworth Conviction-Proneness Study are tabulated below:

Ellsworth Conviction-Proneness Study: Juror Voting Behavior, Comparison of "Witherspoon-Qualified" and "Guilt Phase Includable" Jurors

"Guilt Phase "Witherspoon- Verdict Includable" Jurors Qualified" Jurors

First Degree Murder 3.3% (1) 7.8% (20)

Second Degree Murder 23.3% (7) 21.3% (55)

Manslaughter 26.7% (8) 48.9% (126)

Acquittal 46.7% (14) 22.1% (57) __________ ___________ 100% (30) 100% (258)

These results, Ellsworth concluded, "provide strong support for the hypothesis that death qualified jurors are more likely to convict than are jurors excludable under the Witherspoon criteria. The most direct test of this hypothesis is a comparison of the relative proportion of guilty and not guilty verdicts among the two groups of jurors. Among the [ Witherspoon] qualified jurors, 22.1% voted not guilty while 77.9% found the defendant guilty of some level of homicide. Among the [guilt phase excludable] jurors, 46.7% voted not guilty, and 53.3% voted guilty of some offense. This difference is highly significant ['p' value of less than .01] and indicates that the departure from representativeness created by the process of restricting juries in capital cases to [ Witherspoon] qualified jurors only may have important negative consequences for defendants in death penalty trials." ( Ellsworth Conviction-Proneness Study, supra, at p. 7.)

Using a statistical process known as multiple regression analysis, Ellsworth also analyzed the data to determine whether the differences in the jurors' voting behavior could be attributable to factors other than differences in the jurors' attitudes toward capital punishment. She found that none of the other factors examined — prior jury service, age, sex, and source from which the subjects were recruited (i.e., newspaper advertisement or venire list) — correlated with voting behavior.

Of the Ellsworth conviction-proneness study, Professor Zeisel testified, "this study, with respect to its stimulus, comes as close to the ideal experiment as one can ever come. . . . There's no way of doing it better."

OVERVIEW OF CONVICTION-PRONENESS STUDIES

The expert witnesses called on behalf of the defense testified that the studies convincingly established a strong correlation between the tendencies of jurors to vote for conviction and juror attitudes toward capital punishment. Dr. Zeisel noted "the almost monotony of the results [which are] obviously the same whether you take the experiment at Sperry-Rand in New York or students in Atlanta or jurors in Chicago or Brooklyn or eligible jurors here in [California]; it comes always out the same way. . . . And since all of the studies show the same result, no matter with whom, no matter with what stimulus, no matter with what closeness of simulation, there is really only one conclusion that we can come to. The relationship is so robust — and this is a term of art among scientists — that no matter how strongly or how weakly you try to discover it in terms of your experimental design, it will come through."

Drs. Ellsworth and Girsh reached similar conclusions. Ellsworth amplified her testimony by using a statistical technique that yields a single overall significance level for the combined results of the five conviction-proneness studies which had preceded hers. Combining the results from Jurow's case I with the other four studies produced an overall significance level of less than .00001 (i.e., a probability of less than 1 in 100,000 that the combined results of those studies could have been achieved through chance alone.) Using the results from Jurow's case II, the overall significance level was less than .0005 (i.e., less than 1 in 2,000). Indeed, when Ellsworth combined only the Jurow and Harris 1971 studies, which differentiated their subjects' views on capital punishment by use of questions based on the Witherspoon decision, she found that the overall significance level was less than .04 using Jurow case I and less than .009 using Jurow case II.

The prosecution has not sought either at the hearing below or in its briefs to adduce evidence that there is no correlation between conviction-proneness and attitudes toward capital punishment. Rather, the main thrust of its position has been to attack the methodology of the studies relied upon by petitioner and to question the propriety of generalizing from the results of those studies to the constitutional conclusions which petitioner must prove. As the most significant critiques apply to virtually all of the studies, they will be discussed in the next section of this opinion.

For present purposes, it should be noted that the prosecution did call two expert witnesses at the hearing below to raise methodological criticisms of some of the studies. The primary expert for the prosecution, Dr. Thomas Haines, discussed the Goldberg, Jurow, Harris 1971, and Ellsworth Conviction-Proneness studies. He concluded his direct examination by testifying that "there is a tendency or direction indicated here [i.e., in the studies]" but that "if this is in fact the sum total of the data to present the argument, I would have to say that it is definitely very tentative and fragmentary. . . ." He later explained that "basically my criticism is that [the studies in their written form] do not give me enough information" as to the procedures used. He had not been present for nor given a transcript of the testimony of the defense experts at the evidentiary hearing, and he conceded that this testimony could have answered the methodological concerns he raised.

Haines did note that the various studies "all point in the same direction. . . . There is some relationship, apparently." He agreed that "as the stud[ies] get methodologically better the effect that they are finding doesn't go away. . . . [T]he better study [e.g., the Ellsworth Conviction-Proneness Study] is not producing a weaker finding. . . ."

The second expert witness called by the prosecution was Dr. Gerald Shure, who discussed only one of the conviction-proneness studies, Ellsworth's. As to this study, he stated that "the evidence presented suggests that in fact a death-qualified juror is likely to be more biased in certain respects [particularly] [a]t the level of attitudes. . . ." He felt that the study could not "predict . . . the actual behavior of jurors in specific cases. . . ." (Italics added.) That, of course, is not the issue. (Cf., ante, fn. 57.)

THE ATTITUDE SURVEYS

The Wilson Survey

In addition to directly studying conviction proneness ( ante, at pp. 32-33), Wilson also presented his subject/jurors with seven statements regarding the insanity defense and asked them to indicate whether they agreed or disagreed. The number of "agree" responses was used to measure "the tendency to be biased against insanity as a defense plea." ( Wilson, supra, at p. 3.) Wilson reported that 65 percent of the subjects with scruples against capital punishment agreed with 0, 1, or 2 of the statements, compared with 53 percent of the subjects without such scruples. These results were marginally significant.

The Bronson-Colorado Survey

About 1967, Professor Edward Bronson surveyed 718 persons drawn from jury lists in 8 Colorado counties. Half of these persons were interviewed in person, the balance over the phone. These subject/jurors were asked to "agree" or "disagree" with five statements related to crime and the criminal justice system. The number of "agree"-responses was used as a measure of the tendency to favor the prosecution. (Note the "response set' problem. Cf., ante, fn. 82.)

Bronson also asked each respondent to indicate whether he or she strongly favored, favored, opposed, or strongly opposed capital punishment. He found that the number of "agree"-responses differed for each of the four capital punishment groups and that the amount of agreement tended to increase as support for the death penalty increased. This pattern of responses was statistically significant at the .01 level or less. In an attempt to determine what effect this had on the representation of attitudes on capital juries, Bronson compared the responses of the "strongly opposed" group — which he postulated was a Witherspoon-excludable group — with the combined responses of the remaining three groups. On all of the five questions, the "strongly opposed" group gave fewer "prosecution prone" responses than did the others. The differences between the groups ranged from 6 percent on question 2 to 30 percent on question 3. The results are graphed below.

The Harris 1971 Survey

In addition to measuring conviction proneness by obtaining votes in four specified cases, the Harris 1971 study also asked its subject/jurors to respond to a large number of questions or statements about crime and the criminal justice system. The responses of the "automatic life imprisonment" subjects were then compared to the remaining (i.e., "Witherspoon-qualified") subjects.

A majority of both groups of subjects agreed that "in most cases the jury should ignore a defense of insanity because it is a loophole that allows too many guilty people to go free." However, as in Goldberg, Wilson, and Bronson-Colorado, more of the "Witherspoon-qualified" subjects agreed than did the "automatic life imprisonment" subjects (67 percent to 57 percent). This difference was highly significant, with a "p" value of less than .01.

Similarly, while a majority of both groups believed there was less "law and order in this country" compared with five years earlier, the "Witherspoon-qualified" subjects were more likely to hold that belief (74 percent to 61 percent). They were also more likely to view blacks (39 percent to 30 percent) and the courts (34 percent to 19 percent) as "major cause[s]" of the "breakdown in law and order in this country." Again, these results were all significant at the .01 level.

The two groups differed little in their evaluations of judges in criminal cases, but there were consistent and statistically significant differences in their respective attitudes towards the lawyers. Prosecutors tended to fare better in the eyes of both groups than did defense counsel. However, the "automatic life imprisonment" group saw the gap as considerably smaller than did the "Witherspoon-qualified" group.

Finally, the subjects were asked about the attitudes concerning nine trial occurrences. These inquiries were prefaced with the following instructions: "Suppose you were a juror in a criminal case, and the judge told you that the prosecution must prove the defendant guilty beyond a reasonable doubt, otherwise the jury should find the defendant innocent. If the case involved each of the situations I am going to read to you how difficult would it be for you to vote 'not guilty?'" Examples of the "situations" to which responses were elicited are: "If the defendant does not take the witness stand in his own defense"; "If an aide to the judge claimed he knew the defendant had committed the crime although the judge said the jury should ignore such statements;" and "If a policeman testified that he saw the defendant commit the crime." The subjects were then asked to indicate whether they would "certainly find [defendant] innocent," "probably find [defendant] innocent," "not find [defendant] innocent" or were "not sure."

In each of the nine situations presented to the subjects, the "Witherspoon-qualified" group was more likely to give a "not innocent"-response than was the "automatic life imprisonment" group. Conversely, in seven of the nine situations, the "automatic life imprisonment" subjects were more likely to give "innocent"-responses. If the percentage of innocent responses is subtracted from the percentage of not innocent responses in order to calculate a "net pro-guilty" response, it is found that the "Witherspoon-qualified" subjects had higher "net pro-guilty" responses on eight of the nine "situations." In all eight, the "p" values were .05 or lower.

The Bronson-California Surveys

Professor Bronson, who had earlier conducted a survey of jurors in Colorado (cf., ante, pp. 43-44), was interested in seeing whether the results of that survey could be replicated in California. He, therefore, conducted two similar surveys in this state, the first in Butte County in 1969-1970 and the second in Los Angeles, Sacramento, and Stockton in 1974-1975. These surveys were administered by students over the telephone except in Los Angeles. There, the jury commissioner handed out Bronson's questionnaire.

The Butte County survey involved 755 persons on the jury list of a small, relatively homogeneous, rural county. ( Bronson-California, supra, at p. 12.) The jurors were given substantially the same five questions as in the Bronson-Colorado survey. In addition, Bronson added two new questions to the survey. As in Bronson-Colorado, an "agree" answer was used as a measure of the tendency to favor the prosecution.

Bronson found that on each question the percentage of "agree" responses increased as the jurors' approval of the death penalty increased. As in Bronson-Colorado, Professor Bronson again compared the responses of the group "strongly opposed" to the death penalty with the combined responses of the remaining groups. On all seven of the questions, the "strongly opposed" group gave fewer "agree" responses than did the others. The differences ranged from 16 percent on question 6 to 52 percent on question 3. The results are graphed below.

The Los Angeles-Sacramento-Stockton portion of the survey involved 707 jurors and 7 questions. Four of the questions were identical to those asked in the Butte County survey; two questions were reworded in an attempt to overcome problems of "response set" (questions 2 and 6); and a new question was substituted for question 7 of the Butte County survey.

The responses to four of the seven questions showed a statistically significant relationship between support for capital punishment and the number of pro-prosecution responses. ("P" values of less than .001.) The differences among the groups on the remaining three questions (numbers 2, 6, and 7) were not statistically significant. The chart below reflects Bronson's comparison of the pro-prosecution responses of the "strongly opposed" group with the combined pro-prosecution responses of the remaining groups:

The Ellsworth Attitude Survey

The final attitude survey was conducted in April 1979 in Alameda County. The subjects were 811 persons randomly selected from the adult population of the county. All participants were registered to vote in the county or had a California driver's license; all lived in a household with an operational telephone.

The survey questionnaire was prepared by Drs. Ellsworth and Fitzgerald. The questions were pretested to insure that "the questions were comprehensible and the responses were sufficiently variable so that meaningful between-group comparisons would be possible." ( Id., at p. 3.) The survey was administered by the Field Research Corporation (FRC). The subjects were selected by FRC using a process known as random digit dialing, and the interviews were conducted by experienced professional interviewers at FRC. The methodology of this survey was universally praised by the experts below. The prosecution experts called it "very well done," "an excellent job," and "far better in quality" than earlier surveys such as Bronson's. The results of this study were considered representative of the above-delineated adult population of Alameda County.

The survey asked questions concerning the subjects' attitudes toward capital punishment. ( Ellsworth Attitude Survey, supra, appen. I, p. 2.) Of the 811 subjects participating, 73 indicated that their opposition to the death penalty would preclude them from being fair and impartial at the guilt phase (i.e., these 73 were "guilt phase nullifiers"); 21 more subjects gave no answer on the nullification question. These two sets of subjects apparently did complete the interview procedure. However, since petitioner does not quarrel with the exclusion of these persons from the guilt phase of a capital trial, their responses are excluded in the results reported in this survey. Thus, the total sample size was 717 subjects, and Ellsworth was able to compare directly the "guilt phase includable" jurors with the "Witherspoon-qualified" jurors.

The subjects were asked 13 attitudinal questions. There were statistically significant differences between the "guilt phase includable" subjects and the "Witherspoon-qualified" subjects on 11 of them.

The subjects were asked initially whether unemployment or violent crime was the more serious problem for Alameda County residents. A significant majority of the "Witherspoon-qualified" subjects (62.5 percent) chose "violent crime," while a slight majority of the "guilt phase includables" (50.4 percent) selected unemployment. (P = .01.)

As in many of the previous surveys, there was a majority of both groups who viewed the insanity defense as "a loophole allowing too many guilty people to go free," but the majority was much larger for the "Witherspoon-qualified" jurors (78 percent) than for the "guilt phase includable" group (59.2 percent). (P .001.)

The subjects were asked to agree or disagree with the statement "It is better for society to let some guilty people go free than to risk convicting an innocent person." A majority of the "guilt phase includable" group (62.5 percent) agreed, whereas a majority of the "Witherspoon-qualified" group disagreed (56.1 percent). The table below sets forth these results, which are significant at a level less than .001:

Differences in attitudes toward the privileges against self-incrimination were measured in two questions. About one-third of the "Witherspoon-qualified" subjects (32.3 percent) indicated agreement with the statement "A person on trial who doesn't take the witness stand and deny the crime is probably guilty." Only 23.5 percent of the "guilt phase includables" agreed. (P .006.) In a related item, subjects were told that "a principle of law that you would have to follow if you were a juror in a criminal trial" is the rule that "if the defendant does not testify, this should not be treated as showing guilt." They were asked to indicate whether they agreed or disagreed with this principle. Twenty-four percent of the "Witherspoon-qualified" subjects disagreed with it, compared to 14.0 percent of the "guilt phase includable" subjects. (P .02.)

The two groups also indicated sharp divergences in their attitudes toward the statement that "Even the worst criminal should be considered for mercy." The "Witherspoon-qualified" subjects indicated overall disagreement with the statement (56.0 percent), whereas a large majority of "guilt phase includables" agreed (77.8 percent). (P .001.) On a somewhat related proposition — i.e., "Harsher treatment of criminals is not the solution to the crime problem" — 20 percent of the "guilt phase includable" subjects disagreed, compared with 41.0 percent of the "Witherspoon-qualified" group. The pattern of attitudinal differences was significant at better than a .001 level.

The subjects were also asked to "suppose you're a juror in a criminal trial. . . . From the newspaper and television stories you have seen you know that the defendant made a confession to the crime. But the confession isn't presented during the trial. The judge instructs you that you must make your decision about guilt or innocence only on the evidence you heard during the trial. Without the confession the prosecution's case is weak; it would not convince you beyond a reasonable doubt of the defendant's guilt. In reaching your verdict what would you do?" A majority of the "guilt phase includable" subjects (60.2 percent) indicated they would not consider the confession, whereas a slight majority of the "Witherspoon-qualified" subjects indicated they "would take the confession into consideration." (P .04.)

Finally, the survey dealt with the attitudes toward counsel at a criminal trial. The results were similar to those in Harris 1971. In separate questions, the subjects were asked to agree or disagree with the statement that "defense attorneys [or district attorneys] have to be watched carefully, since they will use any means they can to get their clients off [or get convictions]." The results are tabulated below. A majority of "guilt phase includable" subjects agreed that both the district attorney (53.1 percent) and the defense counsel (64.7 percent) must "be watched carefully." However, whereas only 48.9 percent of the "Witherspoon-qualified" subjects agreed as to the district attorney, nearly three-quarters (73.5 percent) believed that defense counsel must be watched carefully.

Ellsworth sought to combine the results from all 13 questions into one "general index of prosecution-proneness." In doing so, she computed that there was a difference of about 12 percent in the proportion of pro-prosecution attitudes between the "Witherspoon-qualified" group and the "guilt phase includable" group. This overall difference was statistically significant (p .0001), and indicated that the former group was "significantly more prosecution-prone" than the latter.

DEMOGRAPHIC CHARACTERISTICS

(7) There has long been survey research done which related attitudes towards capital punishment to such demographic characteristics as race and sex. This research is relevant to the Ballew issue in the present case insofar as it may show that excluding persons based upon their opposition to capital punishment "foretells problems . . . for the representation [on juries] of minority groups in the community." ( Ballew v. Georgia, supra, 435 U.S. at p. 236 [55 L.Ed.2d at p. 244].)

Sex

Surveys conducted over the past 30 years by Gallup, Harris, and the National Opinion Research Center (N.O.R.C.) have uniformly shown that women are opposed to capital punishment more frequently than are men. From 1953 through 1978, Gallup and N.O.R.C. conducted nationwide polls asking the survey respondents, "Are you in favor of the death penalty for persons convicted of murder?" As the graph below indicates, there have been consistent differences between the responses of men and those of women, with women expressing opposition to the death penalty more often (by an average of 11 percent).

Harris polls conducted in 1973 and 1977 found similar disparities in persons' responses to the question "Do you believe in capital punishment (death penalty) or are you opposed to it?" The 1973 poll found that women expressed opposition 7 percent more often than men (35 percent to 28 percent); in 1977, the gap was 9 percent (29 percent to 20 percent).

These differences between the sexes do not disappear when the question posed is tailored to the Witherspoon criteria. The Harris 1971 survey asked, "Suppose you were a juror in a murder trial and it was completely up to the jury to choose whether the penalty would be death or imprisonment, do you think there would be any situations in which you might vote for the death penalty or do you think you could never vote for the death penalty, regardless of the circumstances?" In response to this question, 37 percent of the women indicated they could never vote for the death penalty, compared to 24 percent of the men.

Similarly, the Ellsworth Attitude Survey asked, "Is your attitude toward the death penalty such that as a juror you would never be willing to impose it in any case, no matter what the evidence was, or would you consider voting to impose it in at least some cases?" Twenty-one percent of the women and only 13 percent of the men responded they would be unwilling to impose the death penalty in any case. (P = .01.)

Race

The disparities between whites and blacks is even greater than between males and females. As the chart below shows, Gallup and N.O.R.C. polls during 1953 through 1978 have shown that blacks are much more opposed to capital punishment than are whites. The average difference in opposition to capital punishment over the entire 25-year period is 23 percent, but the gap has been increasing (it averaged 28 percent in the 1970s).

Harris polls in 1973 and 1977 are consistent with the Gallup and N.O.R.C. data, finding disparities of 21 percent (50 percent to 29 percent) and 26 percent (48 percent to 22 percent) respectively.

As was true for the sexual differences, the significant disparities remain when a Witherspoon question is posed. The Harris 1971 survey found that 46 percent of blacks indicated they "could never vote for the death penalty, regardless of the circumstances," whereas only 29 percent of whites so responded. (P .01.) The Ellsworth Attitude Survey reported that 17 percent of the whites in Alameda County, but 26 percent of the blacks, stated they "would never be willing to impose [the death penalty] in any case, no matter what the evidence was." (P = .05.)

As the evidence adduced below established, these correlations between opposition to capital punishment and racial and sexual characteristics "have tended to appear with boring regularity ever since these topics have been researched. . . ." "No one who has ever done [such] a survey . . . has failed to find" these differences.

JUROR EVALUATION OF EVIDENCE

While the conviction-proneness studies ( ante, pp. 27-42) may indicate that "guilt phase includable" jurors will tend to differ in their decisions on guilt or innocence from "Witherspoon-qualified" jurors, they do not themselves directly explain how these differing conclusions come about. Some inferences as to how this occurs may be drawn from the attitude and demographic surveys. In the evidentiary hearing below, three studies by Ellsworth were used to suggest some "possible mechanism[s] by which these attitude[ ] [differences] might translate into the relevant behavior."

One study, referred to below as the Ellsworth Post-Deliberation Follow-up Study, sought to measure the differences in jurors' thresholds of reasonable doubt by measuring the regret they might feel if they returned an erroneous verdict (either conviction or acquittal) in a criminal case. (Cf., ante, fn. 101.) The subjects in this study were all persons eligible for jury service in California. Fourteen were "guilt phase includable" and fifteen were "Witherspoon-qualified." The subjects were asked to imagine that they had actually served on a jury in a homicide case, that the jury had reached a unanimous verdict, and that they later learned that the verdict was wrong. They were then asked to rate on a scale from 1 to 100 the amount of regret they would feel for each type of erroneous verdict in such a case.

Two types of erroneous verdicts were possible: (1) the conviction of an innocent person, or the conviction of a guilty person for a degree of homicide more severe than the "true" crime warranted (Ellsworth called this "harsh error"), and (2) the acquittal of a guilty person, or the conviction of a guilty person for a degree of homicide less severe than the "true" crime warranted (denoted "lenient error").

Ellsworth found that the "Witherspoon-qualified" subjects expressed equal regret for "harsh" and "lenient" errors, whereas the "guilt phase includable" subjects felt a considerably higher amount of regret for "harsh" errors than for "lenient" errors. The differences were statistically significant. (P .05.) This suggests, Ellsworth testified, that the two groups differed in their thresholds of reasonable doubt and hence their propensity to convict. (Cf., ante, fn. 101.)

A second study was conducted to determine whether the two groups also differed in their "perception of the credibility of defense and prosecution witnesses." The subjects in this study were 36 persons who had previously participated in the Ellsworth Conviction-Proneness Study. Sixteen were "guilt phase includable" and twenty were "Witherspoon-qualified."

Ellsworth had prepared a 20-minute videotaped simulation of the conflicting testimony of a white police officer and a black defendant on trial for assaulting the officer. The simulated case involved a physical confrontation between the officer and the defendant during what the officer described as a routine crowd-control incident and the defendant described as police harassment. The experimenters' intention was to create two equally likely versions of the event. The script had been reviewed for realism by a group of lawyers and psychologists, and the videotape had been pretested for plausibility.

The videotape was shown to 36 subject/jurors, who were then asked 16 questions concerning (1) the accuracy and truthfulness of the 2 witnesses; (2) the extent to which specific facts mentioned by each witness seemed to be true; and (3) the appropriateness of each witness' behavior. On each of the 16 questions the "Witherspoon-qualified" subject/jurors gave (as a group) answers that were more favorable to the police officer than did the "guilt phase includable" group. The differences were statistically significant, with a "p" value of .05 or less, on 10 of the 16 questions. Summing up the responses to all questions revealed an overall difference of about 20 percent in the groups' assessments of credibility. This disparity was statistically significant at less than a .0002 level.

Ellsworth testified that this study demonstrated "a way in which the different attitudes that these people come in with do translate into different perceptions. And those perceptions are . . . very relevant to the actual behavior that jurors would engage in [when] deciding on guilt or innocence, that is, your perception of how credible various witnesses are."

The results of the Ellsworth Witness-Credibility Study were replicated to some extent in the Ellsworth Conviction-Proneness Study. As previously indicated, after the subject/jurors of the latter study had indicated their personal beliefs as to the accused's guilt or innocence, most of them were divided into 12-person juries and allowed to deliberate. (Cf., ante, fn. 78.) Some of the juries were composed entirely of "Witherspoon-qualified" jurors and some had a mixture of "Witherspoon-qualified" and "guilt phase includable" jurors.

After one hour, the deliberations were terminated and the subjects given a written questionnaire to fill out. Among the items on the questionnaire were six questions asking "How believable was the testimony?" of each of the six witnesses who testified. As shown by the chart below, the "Witherspoon-qualified" jurors (the "D.Q." jurors in the chart) believed each of the prosecution witnesses more than the "guilt phase includable" jurors ("Excl.") did. As for the two defense witnesses, their credibility was rated higher by the "guilt phase includable" jurors than by the "Witherspoon-qualified" jurors. As Ellsworth testified, these results "can be considered simply a replication across a whole series of witnesses in a different [situation] of what we got in the witness credibility study."

VI

CRITICISMS OF THE STUDIES RELATING TO THE WITHERSPOON AND BALLEW CONTENTIONS

(9) Several of the criticisms do not require extended discussion. For example, it is contended that none of the studies can be considered reliable indices of how real jurors would act, since the subjects of the studies did not have the same sense of "felt responsibility" as would a juror in an actual criminal trial.

There are several responses to the criticism. First, the criticism is inapplicable to the Zeisel study, since he drew his data from jurors who had actually sat and deliberated on real criminal cases. The fact that Zeisel's study is fully consistent with the results of the other, controlled experiments strongly indicates that the absence (if any) of felt responsibility did not produce spurious results. Second, it appears that the subjects in the Ellsworth Conviction-Proneness Study did "g[e]t very, very into deliberations." Those subjects in this study who had had prior jury experience reported that the experimental situation corresponded well with their jury service. The defense experts also testified that it was not at all unusual for subjects in controlled jury studies to become involved in their cases to the degree that Ellsworth's subjects did.

Significantly, a comparison among the controlled conviction-proneness studies shows that the differences found in voting behavior tend to get wider as the studies become more realistic and involving. Thus, the studies using written stimuli found differences of 6 and 7 percent ( Goldberg, Harris 1971). However, this rose to 13 percent when an audiotape was used ( Jurow) and to 25 percent when a videotape was employed ( Ellsworth Conviction-Proneness). This trend also indicates that any absence of felt responsibility would not alter the conclusions to be drawn from the studies.

Finally, it must be noted that in reaching its constitutional conclusions in Ballew v. Georgia, supra, 435 U.S. 223, the United States Supreme Court relied heavily on controlled studies, none of which involved "real" jurors and hence "real" felt responsibility. It would appear that the use of such studies has been already approved.

Similar considerations lead this court to reject the Attorney General's next criticism of the studies, i.e., that they did not involve juror interaction and deliberation. Again, this critique is inapplicable to the Zeisel study. Further, the subject/jurors in the Ellsworth Conviction-Proneness Study deliberated for an hour, and the significant differences between the "guilt phase includable" jurors and the "Witherspoon-qualified" jurors remained even after that time. Moreover, much of the force of this criticism is further overcome by the findings of Kalven and Zeisel, supra, and others that in the overwhelming number of cases, the jury will eventually reach a unanimous verdict consistent with the position of the first ballot majority. (Cf., ante, fn. 56.) This indicates that the critical factor in jury decision making is not the process of juror interaction but the jurors' original views of the case, which are formed at or near the time the case is submitted to them for decision.

Next, it is asserted that the studies (except the various polls and the Harris 1971 and Ellsworth Attitude surveys) must be disregarded because the subjects involved were not representative of the general population. This contention reveals a fundamental misunderstanding of the conclusions sought to be drawn from these studies, a misunderstanding which was repeatedly addressed by the experts at the hearing below. The studies are being used to show differences in behavior and attitudes between groups; they are not being used to show the relative size of one group to another. Under these circumstances, a researcher need not insure that the groups are included in the study in the same proportions as they appear in the general population. Rather, it must merely be insured that there is a sufficiently large number of persons in each group to enable the researcher to draw conclusions about the characteristics of that group.

In any event, the criticism concerning representativeness all but disappears when the studies are viewed as a whole, not in isolation. As Dr. Zeisel noted, the results of these studies are consistent irrespective of whether the subjects are southern college students, northern industrial workers, Illinois or New York jurors, persons eligible for jury service in California, or persons selected randomly from throughout the nation. It defies reason to attribute the studies' results to "unrepresentative" or nonrandom samples.

(8b) However, there is a more telling criticism of the conclusions which petitioner seeks to draw from the evidence relating to the Witherspoon and Ballew issues. The pool of jurors eligible to serve in a capital trial in California consists of those persons eligible to serve in a noncapital case whose attitudes toward capital punishment would place them in either the "favor death penalty," "indifferent," or "oppose death penalty" group. (Cf., ante, fns. 34, 48.) Thus, in order to establish that a capital case jury in this state is drawn from a pool that is less than neutral with respect to guilt or is inadequate to effectuate the purposes and functions of a jury, petitioner must establish deficiencies in a pool consisting of those three groups, i.e., the "California death-qualified" groups.

None of petitioner's studies expressly focused on a pool comprised in this manner. They do focus on the deficiencies in a pool of " Witherspoon-qualified" jurors. However; at least in theory, a pool of " Witherspoon-qualified" jurors differs from a pool of "California death-qualified" jurors. The former contains a fourth group in addition to the three "California death-qualified" groups, i.e., a "Witherspoon-qualified" jury pool also contains the "automatic death penalty" group. Thus, in order to draw conclusions about a pool of "California death-qualified" jurors on the basis of the evidence introduced below, petitioner must be able to account for the "automatic death penalty" jurors. It is asserted by the Attorney General that petitioner is unable to do this in a scientifically adequate fashion. In short, petitioner is said to have proved the wrong proposition.

Petitioner seeks to overcome the critique in several ways. First, he contends that the "automatic death penalty" group is a tiny group compared to the "Witherspoon-qualified" group as a whole. If this were true, it is urged, then the results of petitioner's studies would not be appreciably altered by subtracting the few "automatic death penalty" jurors from the large class of "Witherspoon-qualified" jurors.

However, there is no reliable evidence in the record to support petitioner's assumption as to the minute size of the "automatic death penalty" group. The defense experts below repeatedly admitted that "nobody knows" the size of this group. There are tentative indications in the record that this group may be as small as 1 percent (or less) of the adult population or as large as 28 percent.

Petitioner notes that there is no reliable data indicating that this group does appear "in substantial numbers" in California jury pools, but this observation does not advance his cause. Witherspoon placed the burden on the accused to establish the nonneutrality of a death-qualified jury. (Cf., ante, fns. 37 and 38 and accompanying text.) Until petitioner makes out at least a prima facie case of unconstitutionality, the state is not required under Witherspoon to make any factual showing. The lack of reliable evidence on the effect of California's exclusion of "automatic death penalty" jurors from capital cases means petitioner has not made any reliable showing that a pool of "California death-qualified" jurors differs from the pool of jurors who are eligible to serve at noncapital trials.

Following oral argument in the present proceeding, the United States Supreme Court decided Adams v. Texas (1980) 448 U.S. 38 [65 L.Ed.2d 581, 100 S.Ct. 2521]. There, the court struck down a Texas statute which disqualified a prospective juror from serving in a capital case "unless he states under oath that the mandatory penalty of death or imprisonment for life will not affect his deliberations on any issue of fact." (Tex. Pen. Code Ann., § 12.31(b) (Supp. 1980), quoted at 448 U.S. at p. 42 [65 L.Ed.2d at p. 587].) The Supreme Court reasoned that the statute violated the principles of Witherspoon, since it authorized the exclusion of jurors on "broader grounds based on their opinions concerning the death penalty" than is permitted by Witherspoon. ( Id., at p. 49 [65 L.Ed.2d at p. 592].)

In the course of seeking to uphold Adams' death sentence, the State of Texas argued the statute was "neutral" in that it permitted the disqualification of persons favoring the death penalty on the same basis as persons opposed. ( Id., at p. 49 [65 L.Ed.2d at p. 592].) The high court responded by stating that this feature of the statute created only an "appearance of neutrality." ( Ibid.) The court noted the "hypothetical existence of the juror who believes literally in the Biblical admonition 'an eye for an eye,'" but, it stated, "it is undeniable, and the State does not seriously dispute, that such jurors will be few indeed as compared with those excluded because of scruples against capital punishment." ( Ibid. [65 L.Ed.2d at p. 592].)

These comments by the United States Supreme Court do not persuade this court to alter its conclusions with respect to petitioner's failure to account for the "automatic death penalty" group in his challenge to the neutrality of a "California death-qualified" jury. Even if this court were to accept the Supreme Court's suggestion as to the relative sizes of the two groups in question, petitioner's problem would not be solved. In order to extrapolate from petitioner's showing about a " Witherspoon-qualified" jury to his intended conclusion as to a "California death-qualified" jury, it is necessary to know more than just the relative size of the "automatic death penalty" group. The impact of this group on petitioner's statistical calculations would depend on both its size and its voting behavior in the jury room. (Cf., post, fn. 115 and accompanying text.) A small group would have an impact beyond its sheer numbers if it tended to be more unified in its voting behavior (or other relevant characteristic) than a larger but less cohesive group. Petitioner's evidence contains no indication of the relevant characteristics of the "automatic death penalty" group as compared to the other groups.

Next, petitioner contends that the Ellsworth Conviction-Proneness Study enables him to draw conclusions about the appropriate "California death-qualified" groups. At the time this study was conducted, none of the subject/jurors was asked a question which would have identified him or her as an "automatic death penalty" juror. However, when it appeared from the testimony of earlier witnesses at the hearing on the Moore brothers' motion that this may have been an "imperfection" in her study, Ellsworth recontacted by telephone those subjects who had previously indicated they "strongly favored" the death penalty. Of the 23 subject/jurors in the "strongly favor" category, she was able to contact 22. All 22 indicated they "would be willing to consider voting for life imprisonment or death depending on the evidence . . . in the case."

While the Ellsworth Conviction-Proneness Study was in other respects an outstanding piece of research, the propriety of relying on such a delayed call-back procedure as a reliable means of retroactively identifying the attitudes of one's subjects is doubtful. Ellsworth herself indicated it was "conceivable" (albeit "doubt[ful]") that different answers would have been obtained had the "automatic death penalty" question been asked several months earlier along with the other questions about capital punishment attitudes.

Petitioner's assertion that the Ellsworth Conviction-Proneness Study demonstrates the nonneutrality of a "California death-qualified" jury presents another problem as well. Even if this study reliably showed that "California death-qualified" jurors tend to be more conviction-prone than "guilt phase includable" jurors, it does not necessarily follow that the former would also tend to be more conviction-prone than are jurors in a noncapital case. It is this latter showing which is the crux of the Witherspoon issue. However, the study's omission of the "automatic death penalty" group — which comprises an unknown proportion of the pool of jurors in noncapital cases — renders unsound petitioner's leap from the results of this study (that "California death-qualified" jurors are more conviction-prone than "guilt phase includable" jurors) to the conclusion that "California death-qualified" jurors are also more conviction-prone than a "neutral" jury.

In view of these gaps in the Ellsworth Conviction-Proneness Study, this court cannot ground a constitutional decision on its results alone.

Finally, petitioner urges, in response to the Attorney General's contention, that he (petitioner) "is entitled to the representation [in] his jury [pool] of all viewpoints and outlooks that are represented in the population; he is entitled to any fair-minded juror who may properly be randomly selected to serve." The first half of the statement is correct, but his evidence has not established that any "viewpoints or outlooks" have been excluded from the jury pool. All he has shown is that if a state used all four " Witherspoon-qualified" groups in a capital trial, the jury would not be neutral. California is not such a state.

The second half of petitioner's statement may be a correct statement of law under a cross-sectional analysis, but petitioner has repeatedly declined to invoke such an analysis. However, petitioner's assertion is incomplete under the Witherspoon analysis. The exclusion of "fair-minded" jurors is improper under Witherspoon only if it can be established that such exclusions tend to result in a nonneutral jury, i.e., in the elimination or underrepresentation of those fair and impartial jurors who would tend to draw conclusions favorable to the accused about issues in the trial, and in the overrepresentation of such jurors favorable to the prosecution. Petitioner has failed to make such a showing as to "death-qualified" juries in California.

Therefore, until further research is done which makes it possible to draw reliable conclusions about the nonneutrality of "California death-qualified" juries in California, this court does not have a sufficient evidentiary basis on which to bottom a constitutional holding under Witherspoon and Ballew. It is, indeed, unfortunate that so much research, time, and energy has been expended in this area with the result that no one stopped to consider the differences between a " Witherspoon-qualified" jury and a "California death-qualified" jury.

The trial court properly rejected petitioner's motion under Witherspoon and Ballew to limit the exclusion for cause of jurors unalterably opposed to capital punishment.

VII

THE USE OF SEQUESTRATION IN DEATH-QUALIFYING A JURY

(10a) Petitioner asks this court to consider whether the procedures currently used in this state to identify death-qualified jurors alter the jury to the detriment of an accused. The specific question is whether the voir dire process in death-qualifying a jury inclines jurors to become more favorable to the prosecution than if a more refined technique were used. At the hearing below, petitioner presented testimony which indicated that current procedures create in the minds of the jurors certain expectations unfavorable to the accused and predispose the jurors to receive and interpret evidence in ways favorable to the prosecution. Various refinements in the current procedure are suggested in order to minimize, if not eliminate, the most prejudicial of these effects.

Presently, the defense counsel and the prosecutor have the right to question potential jurors about their attitudes toward the death penalty in order to lay a foundation for possible challenges for cause. The voir dire is usually conducted in open court with the entire jury panel present. Although each juror is questioned personally, he or she has the opportunity to observe the examination of many other venirepersons.

A typical death-qualifying voir dire includes repeated explanations of the procedural route the jury will travel if it is to reach the penalty phase. Potential jurors are examined in detail about their attitudes toward capital punishment. Frequently, this questioning is couched in hypothetical terms. Potential jurors are asked to assume that the accused has been found guilty of first degree murder and to imagine that the special circumstances allegations have been found true. They are then asked to indicate whether they could fulfill their legal responsibility to choose the appropriate penalty based on the evidence presented. The court has the task of ruling on whatever challenges for cause may be lodged against them.

There are several reasons why these methods for death-qualifying a jury might be expected to have an impact on the expectations, perceptions, attitudes, and behavior at trial of the jurors exposed to them. These affect both the guilt and penalty determinations.

The process presently used focuses attention on penalty before the accused has been found guilty. As a result, some jurors may be more likely to believe the accused is guilty as charged. Modern psychological theory suggests several reasons to anticipate such a result.

Few human impulses are more fundamental than the need to make sense of one's surroundings. People who find themselves in novel, complicated, or confusing situations instinctively seek guidance in interpreting those situations. Accordingly, they become sensitive to the way other people around them seem to be behaving. When authority figures who appear to be familiar with the situation are present, they take on added importance and influence. In such a fashion, venirepersons who are in the unfamiliar and imposing surroundings of a courtroom, undergoing the oftentimes elaborate and sometimes baffling ritual of voir dire, will typically seek cues about appropriate ways of thinking, feeling, and believing. Such venirepersons are likely to look to the behavior of the most knowledgeable and respected figures in the courtroom, i.e., the judge and counsel.

In a typical death-qualifying voir dire, the judge and the attorneys repeatedly instruct the jurors about the steps leading to the penalty trial and question each prospective juror, oftentimes at considerable length, concerning his or her attitudes about capital punishment. These repeated displays of concern about the death penalty before any evidence of guilt has been presented may prompt the jurors to infer that the court and counsel assume the penalty trial will occur.

A penalty trial is contingent on a guilty verdict and a finding of special circumstances. Jurors undergoing death-qualification would have reason to infer that the judge and the attorneys personally believe the accused to be guilty or expect the jury to come to that conclusion. Only such an inference could serve to explain to the jurors why so much time and energy are devoted to an extensive discussion of penalty before trial. Provided with these cues from people who are not only experts in the courtroom but are also presumably acquainted with all the evidence in the case, the relevant law, and the "correct" application of the one to the other, death-qualified jurors may themselves become more inclined to believe that the accused is guilty as charged.

The effects of such a predisposition on the jury's eventual verdict would be expected to be magnified by the particular ways in which a jury functions. Diversity of experience and viewpoint enables a jury to compensate for the perceptual and evaluative limitations inherent in any one particular point of view. (Cf., ante, at pp. 23-25.) However, current methods used in voir dire may impede these corrective mechanisms by the operation of what is known as "perceptual set." The concept of perceptual set, explained one expert witness at the evidentiary hearing below, "suggests that when people begin to form a framework for understanding, particularly with respect to complicated events with which they are somewhat unfamiliar, they begin early on in the process to form a perspective, to form a way of interpreting that information. . . . Perceptual set affects not only what they do with information, but indeed it also affects what they view as information. There is a notion of selective attention, the idea that people only attend to certain kinds of things. When they are presented with complicated information they tend to select out of that information for the most part information which conforms with their belief structure, which conforms with their attitudes, which conforms with their expectations. That is what they look for, that is what they see, and they interpret it differently based on attitudes, beliefs, and expectations."

If a juror is predisposed by the very process of the death-qualifying voir dire to believe the accused is guilty, the juror will tend, in consequence of that perspective: (1) to selectively perceive the evidence — e.g., "forgetting," distorting, or actually failing to perceive evidence which conflicts with his or her preconceptions; (2) to draw only those inferences from the evidence which support those preconceptions and perhaps to "create" data to reinforce those beliefs; and (3) to evaluate the evidence perceived — e.g., assess credibility of witnesses, weigh the inferences drawn — in a manner which tends to fulfill his or her expectations. (See ante, at pp. 23-24.) Moreover, a juror who is predisposed to expect a guilty verdict will tend to arrive more readily at the conclusion that the evidence presented amounts to proof beyond a reasonable doubt.

These tendencies would not be nullified by the evidence actually presented in the courtroom because what the jurors perceive to be evidence will itself be a function of the attitudes, beliefs, and expectations with which they view the proceedings. Of course, the jurors' "perceptual sets" are unlikely to determine the outcome in those cases in which the evidence points overwhelmingly in one direction or another. But the influence of the jurors' "perceptual set" increases — along with the dangers of miscarried justice — as the evidence becomes more closely balanced between guilt and innocence. It becomes especially important to scrupulously safeguard the accused's constitutional rights to a fair trial in such cases, because what might in a more clear-cut case have only an insignificant effect on the way the jurors view the evidence could "provide the slight impetus which [swings] the scales toward guilt." ( Glasser v. United States (1942) 315 U.S. 60, 67 [86 L.Ed. 680, 698, 62 S.Ct. 457].) "[I]t is with respect to those cases that the jury trial right has its greatest value. When the case is close, and the guilt or innocence of the defendant is not readily apparent, a properly functioning jury system will insure evaluation by the sense of the community and will also tend to insure accurate factfinding." ( Ballew v. Georgia, supra, 435 U.S. at pp. 237-238 [55 L.Ed.2d at p. 245].)

A capital jury, which has been predisposed by virtue of the very process by which it has been selected to think the accused guilty in advance of trial, is unlikely to function properly or maintain its neutrality. As a jury's neutrality decreases, the quantum of evidence necessary to prove guilt also decreases.

In addition to making a capital jury prone to convict, the current method of death-qualification may alter the states of mind of the jurors exposed to it in ways which make them more likely to impose a death sentence. It is important to recognize that in the penalty phase, no less than in the guilt phase, the jury serves as a representative of community values. "[O]ne of the most important functions any jury can perform in making . . . a selection [between life and death] is to maintain a link between contemporary community values and the penal system — a link without which the determination of punishment could hardly reflect 'the evolving standards of decency that mark the progress of a maturing society.'" ( Witherspoon v. Illinois, supra, 391 U.S. at p. 519, fn. 15 [20 L.Ed.2d at p. 783], citation omitted.) A jury in a penalty trial is thus called upon to "express the conscience of the community on the ultimate question of life or death." ( Id., at p. 519 [20 L.Ed.2d at p. 783], fn. omitted.)

(11) A penalty jury can speak for the community only insofar as the pool of jurors from which it is drawn represents the full range of relevant community attitudes. In Witherspoon, the Supreme Court held that those who hold scruples against capital punishment must be included within this pool so long as they are willing to consider the choice of penalty provided by law and are not irrevocably committed to vote against the imposition of the death penalty regardless of the evidence. ( Id., at p. 522, fn. 21 [20 L.Ed.2d at p. 785].) The court acknowledged that the caution, reluctance, and even aversion with which many members of the community would pronounce a sentence of death were appropriately represented on a jury which purported to express the conscience of the community: a jury "[c]ulled of all who harbor doubts about the wisdom of capital punishment — of all who would be reluctant to pronounce the extreme penalty," the high court found, "cannot speak for the community." ( Id., at p. 520 [20 L.Ed.2d at pp. 783-784].)

Thus, a process which systematically reduces whatever "doubts about the wisdom of capital punishment" or "reluctan[ce] to pronounce the extreme penalty" is as constitutionally infirm as a jury from which individuals who hold such views are systematically "culled." Neither jury can "speak for the community." ( Ibid.) Both juries are "less than neutral" with respect to the choice of penalty. ( Id., at p. 520, fn. 18 [20 L.Ed.2d at p. 784].)

Several theories discussed at the evidentiary hearing below explain why the current modes of death-qualification might increase jurors' willingness to impose a sentence of death. During a death-qualifying voir dire, venirepersons are questioned in open court about their attitudes toward the death penalty. The fact that the court dismisses those venirepersons who express unequivocable opposition to the death penalty is likely to be interpreted by the remaining jurors as an indication that the judge in particular and the law in general disapprove of such attitudes. Jurors whose scruples against capital punishment are not so irrevocable as to disqualify them under Witherspoon may feel that in the eyes of the law, their attitudes are improper, or at least suspect. Those jurors may in consequence feel less willing to express or rely on such attitudes in their consideration of penalty.

Another relevant concern is drawn from research into the process of desensitization. For many people, even those who are in favor of the death penalty, the prospect of having to make a personal decision about whether another human is to live or die poses an understandably intimidating duty. While the state has a legitimate interest in a death penalty jury which is willing to consider the imposition of capital punishment in an appropriate case ( Witherspoon v. Illinois, supra, 391 U.S. at p. 522, fn. 21 [20 L.Ed.2d at p. 785]), the instinctive caution and reluctance with which many jurors approach this duty are representative aspects of the community conscience which the jury is called upon to express.

When people are continually exposed to a stimulus which is intimidating or frightening to them, they become desensitized to what they earlier found to be threatening. In a capital voir dire, prospective jurors are repeatedly prompted to think about the penalty decision they may later be called upon to make. What was initially regarded as an onerous choice, inspiring caution and hesitation, may be more readily undertaken simply because of the repeated exposure to the idea of taking a life. Some jurors initially face the penalty decision with reluctance and aversion. This may represent a significant viewpoint in the community. A process which systematically erodes these attitudes would make the jury less representative of the community and more inclined to impose death.

In 1979, Dr. Craig Haney, an assistant professor of psychology at the University of California at Santa Cruz, devised a controlled study to determine whether the process of death-qualification actually alters jurors' states of mind so that it affects their evaluation of guilt or their choice of penalty. The subjects of the experiment were selected from adults eligible for jury duty in Santa Cruz County who had responded to a local newspaper advertisement. The researchers screened the respondents by telephone, and eliminated from the study those whose attitudes about capital punishment disqualified them under the Witherspoon criteria.

Haney had prepared a two-hour videotape of a simulated voir dire in a capital trial. The voir dire was not scripted; the attorneys depicted on the tape based their questioning on the facts of an actual capital case. The roles of judge, prosecutor, and defense counsel were played by experienced private defense attorneys. The roles of most of the venirepersons were played by adult residents of Alameda County whose responses to an earlier survey had indicated that they would be qualified to serve under the Witherspoon guidelines. Two of the venirepersons who were questioned on the videotape were confederates placed by Haney. These two were instructed to respond to the voir dire questions concerning capital punishment as if they had views which would render them ineligible under Witherspoon.

The 67 subjects of Haney's study were randomly divided into two groups for purposes of viewing the videotaped voir dire. Both groups were asked to imagine themselves to be "prospective jurors in this very case." One group — the "experimental" group — was shown the full two-hour videotape which included half an hour of death-qualification. The "control" group saw the same videotape with the death-qualifying segment deleted. Each version of the videotape contained various introductory remarks and questions from the judge, including a nullification question patterned on Penal Code section 1074, subdivision 8.

After viewing the videotapes, both groups completed a questionnaire. Their responses were consistent with the predictions that the present procedures for death-qualifying a jury alter the jurors' perspectives.

Haney found that the "death-qualified" subject/jurors in his experimental group were more likely than those in the control group to believe the accused was guilty of first degree murder, a finding that was statistically significant. Moreover, the experimental group was more likely to think that the prosecutor and the defense attorney personally believed the accused guilty as charged. The subject/jurors in the experimental group were also more inclined to think that the judge believed the accused to be guilty as charged.

Haney testified that these findings were consistent with the prediction that "the fairly extensive concern with the penalty phase on the part of the attorneys as well as the judge may well indicate to persons in a somewhat unfamiliar situation, prospective jurors, that . . . those participants in this process of death-qualification believe this person is guilty or believe that it is reasonably likely that the jury will find him guilty, such that the amount of time which is expended discussing penalty is justified."

The subject/jurors were also asked to assess the likelihood that the accused would later be found guilty of first degree murder and sentenced to death. The experimental group was more likely to predict that the accused would be convicted and sentenced to death. In their responses to another question, those in the experimental group also revealed that they were more likely to expect that the accused would be convicted of some lesser included offense if not of first degree murder.

These results lend support to the hypothesis that "people who are asked to imagine the occurrence of the penalty phase of the trial should be more likely to expect that that penalty phase will take place. . . . [S]o one would . . . predict that people who have been asked to imagine the occurrence of this penalty phase would predict that it was more likely that the defendant will be convicted and more likely that special circumstances will be found in this particular case."

The two groups were asked to select what they thought would be an appropriate penalty, assuming that the accused had been convicted and a special circumstance allegation of a prior conviction of first degree murder had been found true. Fifty-seven percent of the experimental group indicated that they would vote to impose the death penalty, compared to only 21.9 percent of the control group.

The subject/jurors were also asked whether they thought "the law generally disapproves of people who are opposed to the death penalty." The subject/jurors in the experimental group were substantially more likely than those in the control group to believe that attitudes in opposition to the death penalty are legally disapproved. Haney explained this result as follows: "What happens in the process of death-qualification is not only that those persons who were found to be acceptable from the point of view of Witherspoon are seated on the jury, but also of course people who are . . . not able to consider the imposition of the death penalty . . . are excluded from sitting on the jury. . . . It happens in the presence of other jurors. In fact it is quite likely, I think, to be interpreted by people who witness this event as a sign of disapproval, perhaps, on the part of the court, the judge who makes such a dismissal, perhaps more generally on the part of the law, the notion being that for the legal reason the law disapproves of people who have attitudes in opposition to the death penalty."

The responses to another question lend support to the surmise that the present procedures for death-qualification are likely to increase the juror's beliefs that the judge personally favors the death penalty. The subject/jurors were asked how they thought the judge "personally feels about the death penalty." Those in the experimental group rated the judge's support for the death penalty considerably higher than did those in the control group.

Haney testified that the altered perceptions of the judge's attitude by the subject/jurors in the experimental group may have resulted from "the judgments which the judge was asked to make when forced to decide whether or not someone should be [excused] for cause." These modified perceptions of the attitudes of the most highly respected figure in the courtroom would, he suggested, "lead [jurors who undergo the current procedures for death-qualification] to reconsider [their] attitude[s] and it may well lead [them] to favor the death penalty somewhat more than [they] had. . . ."

Asked to describe the overall conclusion he drew from his study, Haney testified that "people who have been through the process of death-qualification are in a different state of mind when compared to people who have not been through that process. They are in a different state of mind with relation to the ease with which they select the death penalty as an appropriate penalty or punishment. They are in a different state of mind with respect to the degree to which they believe the prosecutor, defense attorney and judge believe in the death penalty. They are in a different state of mind with respect to the degree to which they believe the prosecutor, defense attorney and judge believe in the guilt of the defendant. And they are also in a different state of mind with respect to the extent to which they expect the defendant would be convicted. And also different in respect to the extent to which they believe the defendant to be guilty.

"My conclusion is that they are in a very different state of mind, and that in each and every case they are in a state of mind which is prejudicial to the defendant in the case."

Haney's findings indicate that the current process for selecting capital jurors creates certain side effects which shape the jury's attitudes toward a death sentence. The courts are appropriately concerned if procedures encourage "[t]endencies, no matter how slight, toward the selection of jurors by any method other than a process which will insure a trial by a representative group." ( Glasser v. United States, supra, 315 U.S. at p. 86 [86 L.Ed. at p. 707].) It has always been the judiciary's duty to counteract processes which generate in jurors "a bias in favor of the prosecution." ( Ibid.) The high court has been vigilant in its review of procedures which "undermin[e]" and "weaken[ ] the institution of jury trial." ( Ibid.) These "undermining processes . . . should be sturdily resisted. . . . Steps innocently taken may, one by one, lead to the irretrievable impairment of substantial liberties." ( Ibid.)

Haney's study has served to alert this court to some of the pernicious consequences of our current voir dire procedures in capital cases. This court must be concerned about the threat these procedures present to an accused's constitutionally protected interests in a fair trial.

Haney testified that the prejudicial alteration in attitudes which resulted from a juror's observations of the death-qualification of his or her fellow venirepersons is "a function of exactly how extensive the questioning becomes. The more extensive the questioning, the more you would expect to find important differences between the state of mind of jurors who have been through the one process as compared with those who have been through the other." This proposition implies a corollary: "the extent to which [these effects] are minimal will be a function of the extent to which the questioning is minimized."

The most practical and effective procedure available to minimize the untoward effects of death-qualification is individualized sequestered voir dire. Because jurors would then witness only a single death-qualifying voir dire — their own — each individual juror would be exposed to considerably less discussion and questioning about the various aspects of the penalty phase before hearing any evidence of guilt. Such a reduction in the pretrial emphasis on penalty should minimize the tendency of a death-qualified jury to presume guilt and expect conviction.

While disputing petitioner's Witherspoon and Ballew contentions, the Attorney General indicated in oral argument that he had "no objection" to this court adopting a rule of sequestered voir dire in capital cases as a judicially declared rule of criminal procedure. He agreed with expert testimony that "We could get rid of a lot of [the concerns raised by the Haney study] by sequestered voir dire." He also reminded this court that "there's precedent for it in this state during the trial of Ruchell Magee and during the trial of Edmund Kemper in Santa Cruz County where the issue was venue. . . . So I can't see how I could have any objection to it."

(10b) In order to minimize the potentially prejudicial effects identified by the Haney study, this court declares, pursuant to its supervisory authority over California criminal procedure, that in future capital cases that portion of the voir dire of each prospective juror which deals with issues which involve death-qualifying the jury should be done individually and in sequestration. This rule will not in any way affect the open nature of a trial. Although trial counsel or the court may pose general questions to the panel, the venirepersons should not respond to any questions beyond those routinely asked in any criminal trial until they are outside the presence of their fellow venirepersons.

Of course, this court cannot insure that a rule of sequestered voir dire in capital cases will alleviate all the untoward effects of the current procedures. Unless a juror is able to understand and respond with certainty to the Witherspoon questions, the juror may be subjected at the sequestered proceeding to considerable voir dire on his or her attitude toward capital punishment. It is unknown at this point whether such personal voir dire would entail the same dangers of inducing bias as do the current procedures for voir dire.

Nonetheless, sequestered voir dire will minimize each juror's exposure to the death-qualifying voir dire of others. It will thereby minimize the deleterious effects of such exposure. Given the frailty of human institutions and the enormity of the jury's decision to take or spare a life, trial courts must be especially vigilant to safeguard the neutrality, diversity and integrity of the jury to which society has entrusted the ultimate responsibility for life or death.

VIII

Let a peremptory writ of mandate issue, directing the trial court to conduct the voir dire in accordance with the views set forth in part VII hereof. In all other respects, the requested relief is denied.

Tobriner, J., Mosk, J., and Newman, J., concurred.