Harris v. Pulley Motion for Leave to File Brief Amicus Curiae and Brief Amicus Curiae
Public Court Documents
July 1, 1985

Cite this item
-
Brief Collection, LDF Court Filings. Harris v. Pulley Motion for Leave to File Brief Amicus Curiae and Brief Amicus Curiae, 1985. 175f4971-b59a-ee11-be36-6045bdeb8873. LDF Archives, Thurgood Marshall Institute. https://ldfrecollection.org/archives/archives-search/archives-item/9c40d398-58dc-4ccb-ac51-ebd9b3a9fad4/harris-v-pulley-motion-for-leave-to-file-brief-amicus-curiae-and-brief-amicus-curiae. Accessed April 19, 2025.
Copied!
IN THE UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT No. 84-6433 ROBERT ALTON HARRIS, Petitioner-Appellant, -against- R. PULLEY, Warden Of The California State Prison At San Quentin, California, Respondent-Appellee. On Appeal From The United States District Court For The Southern District Of California MOTION FOR LEAVE TO FILE BRIEF AMICUS CURIAE AND BRIEF AMICUS CURIAE OF THE NAACP LEGAL DEFENSE AND EDUCATIONAL FUND, INC. JULIUS L. CHAMBERS JAMES M. NABRIT, III JOHN CHARLES BOGER DEVAL L. PATRICK 99 Hudson Street New York, New York 10013 (212) 219-1900 ANTHONY G. AMSTERDAM New York University School of Law 4-0 Washington Square South New York, New York 10012 ATTORNEYS FOR AMICUS CURIAE 1 2 2 7 13 14 14 15 15 15 19 25 28 TABLE OF CONTENTS MOTION FOR LEAVE TO FILE BRIEF AS AMICUS CURIAE STATEMENT OF THE ISSUES PRESENTED ...................... STATEMENT OF THE CASE ................................... A. The History Of The Issue ................ B. The Current Status Of The Issue ......... C . The Questions Presented By The Order Of The District Court .............. JURISDICTION STANDARD OF REVIEW ARGUMENT ......... I. An Extensive Body Of Scientific Research Provides Overwhelming Proof That Death-Qualified Juries Are 'Less Than Neutral' On The Issue Of Guilt Or Innocence ............. A. The Hovey Evidence .................. B. The Word & Sparks Evidence .......... II. Because Death-Qualified Juries Are Biased In Favor Of The Prosecution And Unduly Prone To Convict, The Use of Such Juries To Try The Guilt Or Innocence Of Capital Defendants Violates The Sixth And Fourteenth Amendments ............... CONCLUSION . STATEMENT OF RELATED CASES APPENDIX A - Description of Empirical Studies APPENDIX B - Article by Dr. Joseph Kadane, "Juries Hearing Death Penalty Cases: Statistical Analysis of a Legal Procedure," 78 J.Am.S.A. 544 (1983) l TABLE OF CASES Page Cases: Adams v. Texas, 4-4-8 U.S. 38 (1980)........................ 5 Ballew v. Georgia, 435 U.S. 223 (1978) .................. 28 Beck v. Alabama, 447 U.S. 625 (1980) ..................... 27 Commonwealth v. Martin, 465 Pa. 134, 348 A.2d 391 (1975) 6 Cuyler v. Sullivan, 446 U.S. 335 (1980)....................... 14 Davis v. Georgia, 429 U.S. 122 (1977)(per curiam) ....... 5 Duren v. Missouri, 439 U.S. 357 ( 1979) 11 Grigsby v. Mabry, 758 F.2d 226 (8th Cir. 1985)(en banc).. 9,25,27,28 Grigsby v. Mabry, 637 F . 2d 525 (8th Cir. 1980)................ 8 Grigsby v. Mabry, 569 F. Supp. 1273 (E.D. Ark. 1983).......... 9 Hovey v. Superior Court, 28 Cal.3d 1, 616 P .2d 1301 (1980)............................................... 7,8,15 Keeten v. Garrison, 742 F .2d 129 (4th Cir. 1984).............. 10 Keeten v. Garrison, 578 F. Supp. 1164 (W.D.N.C. 1984).... 10,13,25 McCleskey v. Kemp, 753 F .2d 877 (11th Cir. 1985) (en banc) ....................................... 11 People v. Rhinehart, 9 Cal.3d 139, 507 P.2d 642 (1973)............................................ 6 Smith v. Balkcom, 660 F.2d 573 (5th Cir. Unit B, 1981), modified on other grounds, 671 F .2d 858 (1982)............................................ 11 Spinkellink v. Wainwright, 578 F .2d 582 (5th Cir. 1978), cert, denied, 440 U.S. 976 (1979) ............ 10,25 Taylor v. Louisiana, 419 U.S. 522 (1975) ................ 11 TABLE OF CASES (cont'd .) Page United States v. Harper, 729 F .2d 1216 (9th Cir. 1984).. 12 r United States ex rel. Clark v. Fike, 538 F.2d 752 (7th Cir. 1976)................................. 6 >12 United States ex rel. Townsend v. Twomey, 452 F . 2d 350 (7th Cir. 1972)....................... 6,12 Wainwright v. Witt, __ U.S.__, 83 L.Ed.2d 841 (1985).... 5,28 Witherspoon v. Illinois, 391 U.S. 510 (1968)............ 2,26,27,28 Other Authorities: Hastie, R., Inside the Jury (Harvard University Press, Cambridge, Mass., 1983)............................. 18 • 8 Law and Human Behavior, Nos. 1/2 (June 1984)........... Federal Statutes: 28 U.S.C. §2241 ............................................... 14 28 U.S.C. §2253 ............................................... 14 V iii IN THE UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT No. 84-6433 ROBERT ALTON HARRIS, Petitioner-Appellant, -against- R. PULLEY, Warden Of The California State Prison At San Quentin, California, Respondent-Appellee. On Appeal From The United States District Court For The Southern District Of California MOTION FOR LEAVE TO FILE BRIEF AMICUS CURIAE The NAACP Legal Defense & Educational Fund, Inc. ("the Fund"), by its undersigned counsel, moves this Court, pursuant to Rule 29 of the Federal Rules of Appellate Procedure, for leave to file the annexed brief amicus curiae in support of petitioner-appellant Robert Alton Harris. STATEMENT OF INTEREST OF AMICUS CURIAE The Fund is a non-profit corporation established to assist black citizens in securing their constitutional rights. In 1967, it undertook to represent all condemned prisoners in the United States, regardless of race, for whom adequate representation could not otherwise be found. It has frequently represented IV condemned inmates before the Supreme Court of the United States. E . g . , Furman v. Georgia, 4-08 U.S. 238 (1972); Jurek v. Texas, 428 U.S. 262 (1976); Woodson v. North Carolina, 428 U.S. 280 (1976); Coker v. Georgia, 433 U.S. 584 (1977); Lockett v. Ohio, 438 U.S. 586 (1978); Enmund v. Florida, 458 U.S. 782 (1982); Francis v. Franklin, _U.S. __, 53 U.S.L.W. 4495 (U.S., April 30, 1985)(No. 83-1590). The Fund is serving as counsel for petitioners in two recent federal cases in which death-gualification claims — virtually identical to the claims asserted on this appeal by petitioner- appellant Robert Harris -- have been evaluated on a full evidentiary record. See Grigsby v. Mabry, 569 F. Supp. 1273 (E.D. Ark. 1983), aff'd , 758 F .2d 226 (8th Cir. 1985)(en banc); Keeten v. Garrison, 578 F. Supp. 1164 (W.D.N.C.), rev'd, 742 F.2d 129 (4th Cir. 1984), cert. pending, No. 84-5187 (filed February 2, 1985). The Fund rep resents numerous other inmates in a number of jurisdictions, includ ing this Circuit, who have asserted or intend to assert this consti tutional claim. In light of (i) the extensive experience of the Fund with the factual and legal issues presented by this claim, and (ii) the direct interest of clients of the Fund in the ultimate resolution of this claim, counsel respectfully request leave to file this brief amicus curiae. Dated: New York, New York July 1, 1985 Respectfully submitted, JULIUS L. CHAMBERS JAMES M. NABRIT, III JOHN CHARLES BOGER DEVAL L. PATRICK 99 Hudson Street New York, New York 10013 (212) 219-1900 v ANTHONY G. AMSTERDAM New York University School of Law AO Washington Square South New York, New York 10012 ATTORNEYS FOR AMICUS CURIAE BY: vi IN THE UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT No. 84-64-33 ROBERT ALTON HARRIS, Petitioner-Appellant, -against- R. PULLEY, Warden Of The California State Prison At San Quentin, California, Respondent-Appellee. On Appeal From The United States District Court For The Southern District Of California BRIEF AMICUS CURIAE OF THE NAACP LEGAL DEFENSE & EDUCATIONAL FUND, INC. STATEMENT OF THE ISSUES PRESENTED 1. Whether the jury trial guarantee of the Sixth Amendment and the Due Process Clause of the Fourteenth Amendment prohibit the systematic exclusion for cause from a capital defendant's guilt or innocence trial of all jurors who could fairly and impartially try that issue, solely because their opposition to the death penalty would make them ineligible for service in the event of a subsequent penalty trial -- despite extensive proof that this practice produces guilt-phase juries that are (i) biased in favor of the prosecution, (ii) unduly prone to convict, and (iii) unrepresentative of the communities from which they are drawn? 2. Whether the District Court could properly reject petitioner's claims on their factual merits without allowing petitioner an evidentiary hearing on the important factual issues it resolved adversely to petitioner? STATEMENT OF THE CASE A . The History Of The Issue The constitutionality of "death-qualified" juries as triers of guilt or innocence in capital cases was first addressed by the Supreme Court in 1968 in Witherspoon v. Illinois, 391 U.S. 510, 520 n.18 (1968), and was left unresolved. Since 1968 it has been the subject of a large and rapidly growing body of litigation in capital cases in the state courts and the lower federal courts. Although it has produced conflicting opinions from these courts, it has never been addressed by the Supreme Court again. The term "death-qualification" refers to the practice of identifying and excluding from capital juries those potential jurors whose views on the death penalty are considered incompatible with their duties as jury members. This practice is universal in jury trials of capital cases, and with rare exceptions those who are excluded oppose the death penalty. It is hardly surprising that many people have long believed that this process alters the composition and functioning of the resulting jury in a systematic and predictable way: that it weighs the jury against the defendant. But belief is one thing and proof another, and until the last few years there was no conclusive proof. In 1968 the Supreme Court was 2 asked in Witherspoon v. Illinois to hold that death-qualified juries are unconstitutionally conviction-prone. The Court declined to do so, finding that the empirical record before it was too weak to form the basis for such a decision. The Court did not rule that Witherspoon was factually wrong in his claim that death-qualified juries are conviction-prone; rather, it with held judgment. Addressing the preliminary drafts of some early social scientific studies on the effects of death-qualification which Witherspoon's attorneys had asked it to judicially notice, id., 391 U.S. at 517, n.10, the Court found: "The data adduced by the petitioner . . . are too tentative and fragmentary to establish that jurors not opposed to the death penalty tend to favor the prosecutor in the determination of guilt. We simply cannot conclude, either on the basis of the record now before us or as a matter of judicial notice, that the exclusion of jurors opposed to capital punishment results in an unrepresentative jury on the issue of guilt or substantially increases the risk of con viction. In light of the presently available information, we are not prepared to announce a per se constitutional rule requiring the reversal of every conviction returned by a jury selected as this one was." Id., at 517-518 (fn. omitted). The basis for this assessment was clearly and carefully stated in footnote 11 of the Court's opinion: "During the post-conviction proceedings here under review, the petitioner's counsel argued that the prosecution-prone character of 'death-qualified' juries presented 'purely a legal question,' the resolution of which required 'no additional proof' beyond 'the facts . . . disclosed by the transcript of the voir dire examination . . . .' Counsel sought an oppor tunity to submit evidence in support of several contentions unrelated to the issue involved here. On this issue, however, no similar request was made, and the studies relied upon by the petitioner 3 in this Court were not mentioned. 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. Under these circumstances, it is not surprising that the amicus curiae brief filed by the NAACP Legal Defense and Educational Fund finds it necessary to observe that 'with respect to bias in favor of the prosecution on the issue of guilt, the record in this case is almost totally lacking in the sort of factual information that would assist the Court.'" Ld., at 517-518, n.ll. The Witherspoon Court was quite explicit in leaving open both the factual and legal questions of conviction-proneness of death-qualified juries. It said in so many words that these questions were reserved for future consideration, and it invited further research and proof on the factual question: "Even so, a defendant convicted by such a [death-qualified] jury in some future case might still attempt to establish that the jury was less than neutral with respect to guilt. If he were to succeed in that effort, the question would then arise whether the State's interest in submitting the penalty issue to a jury capable of imposing capital punishment may be vindicated at the expense of the defendant's interest in a completely fair determination of guilt or innocence -- given the possibility of accommodating both interests by means of a bifurcated trial, using one jury to decide guilt and another to fix punishment. That problem is not presented here, however, and we intimate no view as to its proper resolu tion." Id., at 391 U.S. 520, n.18. On the record before the Court in that case, its decision appears to have been inescapable. Indeed, although the Witherspoon Court was not required to, and did not so hold, the entire body of scientific evidence available in 1968 may well have been insufficient <4 to support a factual conclusion concerning the conviction-proneness of death-qualified juries. Since 1968, however, the state of knowledge on this issue has changed greatly. While it did not reach the question of bias on the issue of guilt, the Witherspoon Court did, of course, restrict the permissible scope of death-qualification in order to insure fair jury determination of the issue of penalty. Under Witherspoon, potential jurors could be excused from service in capital cases because of their opposition to the death penalty only if they were unwilling to consider the death penalty in any case, or unable to be fair and impartial in deciding the defendant's guilt. 391 U.S. at 522-23 n.21. The Court held that a jury from which venirepersons were excused on any basis broader than these two is "a tribunal organized to return a verdict of death," id., at 521, and that "[n]o defendant can constitutionally be put to death at the hands of a tribunal so selected." Id., at 522-23. This rule does not affect the validity of any sentence other than death or the validity of a defendant's conviction, as opposed to his sentence. Recently, in Wainwright v.-Witt, __U.S.__, 83 L.Ed.2d 84-1 (1985) the Supreme Court modified the constitutional limitations on death- qualification. But while Witt changes the specific standards that govern this practice, it "adhere[s] to the essential balance struck by the Witherspoon decision." _Id, at n.5. Neither Witt nor the Supreme Court's earlier cases applying Witherspoon, see Davis v. Georgia, 429 U.S. 122 (1977)(per curiam); Adams v. Texas, 448 U.S. 38 (1980), address the present claim: that death-qualified juries are less than neutral with respect to guilt. 5 Neither petitioner Harris nor amicus are challenging the exclusion of veniremen who could not be fair and impartial in deciding their guilt or innocence. Nor do they dispute the right of the state to exclude venirepersons who could not fairly try the issue of penalty from participation in sentencing delib erations. Their claim is limited to the assertion that it is unconstitutional to preclude venirepersons who would be fair and impartial in determining their guilt or innocence from trying that issue, solely because they would not be eligible to participate in a possible subsequent penalty trial. In the decade between 1968 and 1978, the courts witnessed a steady incremental growth in the body of empirical evidence on death-qualification: the early studies that had been presented to the Supreme Court in preliminary form were completed, and a few new studies were conducted. This new evidence was presented in the lower courts in support of arguments against the constitutionality of death-qualification, but in each case the courts held that despite these developments, the available scientific evidence on this point remained "tentative and fragmentary." See, e .g ., People v. Rhinehart, 9 Cal.3d 139, 507 P.2d 642 (1973); Commonwealth v. Martin, 465 Pa. 134, 348 A.2d 391 (1975); United States ex rel. Townsend v. Twomey, 452 F .2d 350, 362-63 (7th Cir. 1972); United States ex rel. Clark v. Fike, 538 F.2d 752 (7th Cir. 1976). In 1978 and 1979, however, a large scale research project on death-qualification was conducted under the direction of some of the most eminent forensic social scientists in the country, and it produced a major new set of studies on the effects of death-qual ification on jury composition, attitudes and behaviors. These 6 studies conclusively demonstrate that death—Qualified juries are indeed biased in favor of the prosecution, unduly prone to convict, and unrepresentative of the communities from which they y are drawn. B . The Current State of the Issue In the past several years, lower court cases have followed one of two irreconcilable paths: (i) some have addressed the death- qualification issue as an empirical question and have held that death- qualification is constitutional despite its effects on capital juries; while (ii) some have held that this practice presents a legal issue only. (i) Major Cases Addressing the Empirical Issue Hovey v. Superior Court. The first appellate case that addressed this issue on the basis of a complete factual record including the more recent studies on death-qualification was Hovey v. Superior Court, 28 Cal.3d 1, 616 P .2d 1301 (1980). In Hovey, on the basis of a detailed record, the California Supreme Court found that a death-qualified jury selected by the procedures out lined in Witherspoon — a "Witherspoon-qualified" jury — "would not be neutral" on the question of guilt. Hovey, supra, 28 Cal.3d at 68, 616 P .2d at 1346. The court declined, however, to rule on the constitutionality of death-qualification as practiced in California, because under California law jurors may also be excluded from capital cases for an additional reason: because they would always vote for the death penalty in every capital case. The court reasoned that the exclusion of these potential jurors might, conceivably, offset the demonstrated effects of excluding those who would never consider the death penalty. I_d. , 28 Cal. 3d 1/ These studies have been published in a special journal issue on death-qualification, 8 Law and Human Behavior, Nos. 1/2 (June, 1984). 7 at 63-69, 616 P .2d at 1343-46. While the California Supreme Court recognized that this group, the "automatic death penalty" jurors, may only exist "in theory," id., 28 Cal.3d at 63, 616 P .2d at 1343, it held that the existence, the size and the impact of this group were factual questions on which the defendant had the burden of proof, and it found that the record in that case did not include any substantial evidence on this point. Therefore, the court concluded that "until further research is done . . . this court does not have a sufficient basis on which to bottom a consti tutional holding" on jury selection practices in capital cases in California. _Id. , 28 Cal.3d at 68 , 616 P.2d at 1346. In short, the Hovey case identified one narrow factual gap that remained to be filled in order to complete the proof of the claim that death-qualified juries are conviction-prone. Grigsby v Mabry. In 1980 the Eighth Circuit held that Grigsby, a state prisoner in Arkansas, was entitled to a federal habeas corpus hearing on his claim that the death-qualified jury that tried him was unconstitutional. Grigsby v. Mabry, 637 F .2d 525 (8th Cir. 1980). The central question that the Court of Appeals identified for consideration at that hearing was "whether death- qualified jurors are more likely to convict than jurors selected without regard for their views on the death penalty;" if that question is answered affirmatively "Grigsby has made a case that his constitutional rights have been violated and he would be entitled to a new trial." Id. at 527. 8 The hearing on remand in Grigsby was conducted in the United States District Court for the Eastern District of Arkansas, and was comparable in scope to the Hovey hearing. It included evidence on all the new studies first presented in Hovey, and it also included extensive new evidence demonstrating that the "automatic death penalty" group constituted only 1% to 2% of the population nationally, and that the occasional exclusion of an "automatic death penalty juror" from a capital case has an insignificant impact on the demonstrated biasing effects of death-qualification. In August of 1983, Judge G. Thomas Eisele of the Eastern District of Arkansas filed a detailed opinion based on this record, holding that the present form of death-qualification is unconstitutional on two grounds: (i) that it biases capital juries against the defendants in the determination of guilt or innocence and (ii) that it denies capital defendants their Sixth Amendment right to a jury on the question of guilt or innocence that is drawn from a fair cross- section of the community. Grigsby v. Mabry, 569 F. Supp. 1273 (E.D.Ark. 1983). Judge Eiselie's opinion has now been upheld on appeal by the Eighth Circuit en banc. Grigsby v. Mabry, 758 F .2d 226 (8th Cir. 1985)(en banc). Keeten v. Garrison (District Court). At the district court level, Keeten was similar to the Grigsby case both in form and in content. The petitioners were state prisoners who claimed in a federal habeas corpus proceeding that their constitutional rights were violated by the use of death-qualified juries to deter mine their guilt or innocence in the North Carolina state courts. The evidentiary record was comparable to that in Grigsby, and*the 9 judgment of trial court judge, the Honorable James B. McMillan of the Western District of North Carolina, was the same: the evidence in the record answers the question posed in Witherspoon and demonstrates that death-qualification produces juries that are "less than neutral with respect to guilt." Keeten v. Garrison, 578 F. Supp. 1164- (W. D. N.C. 1984). The judgment of the District Court was reversed by the Fourth Circuit in Keeten v. Garrison, 742 F.2d 129 (4th Cir. 1984), but the opinion of the Court of Appeals did not rest on the factual issues, but on its view of the law. Thus, since 1979, every court that has considered the entire body of empirical evidence now available has concluded that the evidence proves what remained unproven in 1968: that death- qualified juries are more likely to convict than ordinary criminal j uries. (ii) Major Cases That Reject The Empirical Issue Spinkellink v. Wainwright. Until 1978, every court that had considered the issue agreed that if proof could be marshalled that such juries are more likely to convict than ordinary, fully representative criminal juries, a constitutional violation would be established. In 1978, in the case of Spinkellink v. Wainwright, 578 F .2d 582 (5th Cir. 1978), cert, denied, 440 U.S. 976 (1979), the former Fifth Circuit held otherwise: "When petitioner asserts that a death-qualified jury is prosecution-prone, he means that a death- qualified jury is more likely to convict than a non-death-qualified jury . . . . Even if this is true the petitioner's contention must fail. That a death qualified jury is more likely to convict than a non-death-qualified jury does not demonstrate which jury is impartial." Id. at 594. 10 In essence, Spinkellink holds that there is no factual issue to be decided at all: if each juror who tried the defendant was individually "fair and impartial," nothing more can be asked. Spinkellink flies in the face of the Supreme Court's statement that factual proof that death-qualified juries are un commonly conviction-prone would implicate "the defendant's interest in a completely fair determination of guilt or innocence." Wither spoon v. Illinois, 391 U.S. at 520 n.18. Spinkellink ignores the basic constitutional rule that a jury is not just any group of jurors who are individually fair-minded. If that were so, juries could consist entirely of fair-minded whites or fair-minded Democrats. A jury must also fairly reflect the community from which it is drawn, both in its composition, see Taylor v. Louisiana, 419 U.S. 522 (1975); Duren v. Missouri, 439 U.S. 357 (1979), and in its predispositions. This rule was applied by the Court in Witherspoon when it prohibited juries that are "uncommonly willing to condemn a man to die." 391 U.S. at 521. Yet Spinkellink explicitly endorses the use of juries that are uncommonly willing to convict. Nonetheless, the holding in Spinkellink on this issue remains the law in the present Fifth and Eleventh Circuits, the two successor courts to the former Fifth Circuit. See Smith v. Balkcom, 660 F.2d 573 (5th Cir., Unit B, 1981), modified on other grounds, 671 F.2d 858 (1982). McCleskey v. Kemp, 753 F .2d 877, 901 (11th Cir. 1985)(en banc). Keeten v. Garrison(Circuit Court). As noted, the Fourth Circuit reversed the judgment of the District Court in Keeten without making any attempt to resolve the factual issues raised by the record and the opinion of the trial court judge. Instead, 11 the Circuit Court chose to follow Spinkellink, and concluded that the conviction-proneness of death-qualified juries is of no consti tutional consequence. (iii) Other Federal Cases Seventh Circuit. While the Seventh Circuit has not faced the task of analyzing all presently available empirical evidence on death-qualification, its position on the legal issue is clear and it is in direct conflict with Spinkellink and its progeny: "the decision on this issue rest[s] on empirical analysis . . . ." United States ex rel. Clark v. Fike, 538 F.2d 750, 762 (7th Cir. 1976); see also United States ex rel. Townsend v. Twomey, 542 F . 2d 350, 362-63 (7th Cir. 1972). Ninth Circuit. This issue has not been determined by this Court, although the Court has previously alluded to the question and its ramifications. See United States v. Harper, 729 F .2d 1216 (9th Cir. 1984)(Fletcher, J., concurring): "[W]hether a verdict returned such a 'death-qualified jury' can withstand constitutional scrutiny is a complex and difficult constitutional question" that should not be decided before a decision is necessary. In sum, the Fourth, Fifth and Eleventh Circuits have held that it is constitutional to use death-qualified juries to determine guilt, even if they are uncommonly conviction-prone. The Seventh Circuit has held that factual proof that death-qualified juries are conviction-prone would require a holding that the practice is unconstitutional. The Eighth Circuit has recently held that such a demonstration has been made,and has outlawed the practice. 12 The matter is before this Circuit on petitioner Harris' appeal; the remaining circuits have not yet been obligated to face this question. C . The Questions Presented By The Order of the District Court The District Court addressed and analyzed petitioner Harris' death-qualification claim as an empirical issue. In other words, it accepted the overall analytical framework set forth in Hovey, and determined that Harris was obligated to prove two things: first, that Hovey itself correctly evaluated the death—qualification studies and their results; and second, that Dr. Kadane's additional 2/ study presented in Word & Sparks filled the one evidentiary gap — concerning the size and significance of the so-called "automatic death penalty" group of prospective jurors ("ADPs") _ identified by the Supreme Court of California in Hovey. The District Court proceeded to bypass the first question, the accuracy of Hovey, and decided as a matter of fact, without any evidentiary hearing, that Dr. Kadane's evidence was insufficient to make Hovey whole. By contrast, the District Court that received the Kadane evidence in Keeten v. Garrison, 578 F. Supp. 1164-, 1175-77 (W.D.N.C. 1984-) , found it "credible, consistent, and essentially uncontradicted. Id. at 1177. Yet the District Court here rejected that same evidence without any adversary testing of Dr. Kadane's testimony. 2/ See discussion of the Word & Sparks evidence at pages 19- 25, infra. 13 JURISDICTION The District Court had subject matter jurisdiction of this case pursuant to 28 U.S.C. §224-1. This Court has juris diction on appeal pursuant to 28 U.S.C. §2253. The District Court's memorandum decision and order was filed on October 17, 1984; it is final and appealable to this Court. A timely notice of appeal was filed on November 14, 1984. STANDARD OF REVIEW The death-qualification claim asserted by petitioner Harris and addressed in this brief by amicus curiae raises questions of federal statutory and constitutional law and mixed questions of law and fact requiring independent, de novo review by this Court. See, e . g ., Cuyler v. Sullivan, 446 U.S. 335, 341-42 (1980). ARGUMENT I AN EXTENSIVE BODY OF SCIENTIFIC RESEARCH PROVIDES OVERWHELMING PROOF THAT DEATH- QUALIFIED JURIES ARE 'LESS THAN NEUTRAL' ON THE ISSUE OF GUILT OR INNOCENCE______ A . The Hovey Evidence The District Court did not explicitly pass upon the validity of the California Supreme Court's findings in Hovey v. Superior Court, 28 Cal.3d 1, 616 P.2d 1301 (1980), but even a cursory review of the scientific evidence reveals that those findings were compelled by an overwhelming and one-sided record. Attitudinal and Demographic Surveys The Hovey record includes five surveys that study the attitudinal and demographic characteristics of the jurors who are _3/ excluded from service by the process of death-qualification. These surveys uniformly find that jurors who are now permitted to serve in capital cases hold attitudes that are more hostile to the defendant and more favorable to the prosecution than the attitudes of those who are excluded. The surveys also uniformly find that women and blacks are disproportionately excluded by death-qualification. Conviction-Proneness Studies Six studies in the Hovey record examine the voting behavior of death-qualified jurors and of those who are excluded by death- ±J qualification, in actual and in simulated criminal trials. These _3/ A more complete description of each attitudinal study appears in Appendix A at 1-6. V See Appendix A at 6-11. - 15 - range from a study by Professor Hans Zeisel of the votes of actual jurors in felony trials in Brooklyn and Chicago, to studies of responses of college students in Georgia and of industrial workers in New York, to a sophisticated trial simulation using jury-eligible adults in California. In each instance the results are the same: death-qualified jurors are more likely to convict than those who are excluded from capital juries because of their opposition to the death penalty. Studies Of The Mechanisms That Produce __________ The Biasing Effects___________ _ Several of the studies in the record examine how the removal of opponents of the death penalty pursuant to Witherspoon _5/ changes the functioning of a jury. These studies show that death-qualified jurors are more likely than those who are excluded to believe prosecution witnesses and to disbelieve defense witnesses, that differences in voting behavior between these two types of jurors persist after jury deliberations, and that the exclusion of jurors who would not consider voting for the death penalty adversely affects the quality of jury delibera tions. In addition, one study (the Haney Study) demonstrates that the process of death-qualification — the questioning of potential jurors at the outset of a capital trial on their attitudes toward the death penalty, and the removal of those who are unwilling to consider imposing the death penalty — predisposes even those jurors who are permitted to serve to believe that the defendant is guilty. 5/ See Appendix A at 12-14. 16 Three overall points about this record deserve brief mention. First, it is noteworthy that so many studies, conducted over a long period of time by different researchers in different locations, using various methodologies and varying subject pools, have all found the same thing: death-qualified juries are more likely to convict than ordinary criminal juries. As Professor Hans Zeisel explained in his testimony in this record: "The reason I have put these six studies together is the following, namely, I'm sure that it couldn't escape anybody who has listened to this testimony . . . the almost monotony of the results. It is 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 Stanford; it comes always out the same way. "And Your Honor, I should add that it happens seldom in the social sciences that the problem is being studied even twice, not to speak of six times . . . . "So this is an unusual fact. 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 one con clusion that we can come to. The re lationship is too 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." (Hovey RT 8A-85.) Second, the studies are of high quality. The body of empirical research on death-qualification includes studies by such eminent social scientists as Professor Hans Zeisel and Professor Phoebe Ellsworth, and surveys by such nationally 17 known organizations as Louis Harris Polls and the Field Research Corporation. This research has been enthusiastically reviewed by, among others, Professor Reid Hastie, author of the most ex tensive work on jury functioning in recent years, Inside the Jury (Harvard University Press, Cambridge, Mass., 1983). The most recent studies -- those conducted after 1978 -- are of particular note. They are exemplary in their methodology and they are tailored to address the exact legal issue at stake. The subjects in each of these studies were classified on the basis of their attitudes toward the death penalty using questions based on Witherspoon, and the two Witherspoon-excludable groups -- those who would never consider voting for the death penalty and those who would not be fair and impartial on guilt or innocence in a capital case -- were identified separately. In each study, the subjects who could not be fair and impartial on guilt were excluded from consideration. As a result, these studies directly demonstrate the biasing effects of excluding potential jurors who would be fair on guilt but who would never consider voting for the death penalty, after those who would not be fair and impartial on guilt have already been removed. Third, there are no studies whatever that reach a contrary conclusion. The views on both sides were thoroughly aired in the testimony in the record, and the criticisms failed to sway the courts that evaluated this body of research on its merits. One of the major reasons for the conclusions of the Hovey court is the fact that after sixteen years and a dozen or more studies, nothing has ever been shown to contradict the uniform finding that death- qualification biases juries against the defendant. To quote Professor Zeisel's testimony once more: 18 "But I just want to say, given the difficulties of coming to conclusions about human nature, I would say that there are few things about which I am so certain than this relationship between death-qualification and the tendency to vote guilty. And it is supported by the attitude studies. I don't see how one can sensibly come to doubt it. You see, these cross-examinations, if you will forgive me, have gone on now for 15 years, and nobody has ever produced a study which shows that this is not true." (Hovey, RT, 163-64.) B . The Word and Sparks Evidence The Word & Sparks record consists, essentially, of the expert testimony of Dr. Joseph B. Kadane, an eminent professor of statistics at Carnegie-Mellon University in Pittsburgh, Pennsyl vania, and exhibits offered in support of that testimony. Dr. Kadane reanalyzed two of the key Hovey studies -- the Ellsworth Conviction-Proneness Study and the Ellsworth Attitude Survey -- in light of two more recent surveys: (i) a 1981 Field Research Corpora tion Survey that found that 12.6% of the fair and impartial jury- eligible population of California would never consider imposing the death penalty in any capital case; and (ii) a 1981 Harris Poll that found that 1% of the national adult population who could be fair and impartial would always vote for the death penalty in any capital case. 0 J (These surveys are described in Exhibit I to the Harris petition.)— Essentially, Dr. Kadane's analysis answers the following question: given the size of the "guilt phase nullifier" group (couldn't be fair and impartial on guilt-or-innocence in a capital case), the "guilt phase includable" group (could be fair and im partial on guilt-or-innocence, but could never consider imposing death), and the "Witherspoon-qualified" group (could consider 6/ An article written by Dr. Kadane describing his studies is annexed as Appendix B to this brief, for the convenience of the Court. 19 imposing death) -- all found by the 1981 Field Survey -- and given the size of the "automatic death penalty" group (would always give death to a convicted capital defendant) found by the 1981 Harris Survey, what effect does the exclusion of "automatic death penalty" jurors from capital juries have on the findings of the Ellsworth Attitude Survey and the Ellsworth Conviction-Proneness Survey, as applied to the population of California venirepersons? Dr. Kadane "had to assume something about the behavior [and attitudes] of the automatic death penalty group" (Word & Sparks RT 56) since they "were not directly measured" by the two prior Ellsworth studies. (Kadane Study, Exh. B, No. 6, at 19). Kadane took "the most conservative stance on this issue" (id.), and assumed that the automatic death penalty group "would be as strongly opposed to the defense as it could be on each issue, [and as] strongly pro prosecution as it could be on each issue." (Word &■ Sparks RT 56) . Specifically, he assumed that they would take the most pro-prosecution position on each question in the Ellsworth Attitude Survey, and "that they would vote, all of them, to convict in the behavior experiment [the Ellsworth Conviction-Proneness Study]." (Id.) This assumption therefore weights Dr. Kadane's results against the position urged by petitioner: "By assuming that the group who would ALWAYS impose the death penalty have the most extreme views in favor of the prosecution, I am making it as difficult as I can to show that the amal gamated ALWAYS and NEVER group is nonetheless more pro defense, less pro prosecution than the remainder. So what this means is that my estimate of the bias will be too low, that is, if I had accurate data on the views of the group that would ALWAYS impose the death penalty, they will not be as extreme as what I have assumed. I don't 20 know by how much. Consequently, my estimate of the odds of bias will be lower, lower in the calculations that I do here than they would be if I had better data by an amount . . . [that] I don't know. So I am making the assumption that is most conservative, most in accord with [the prosecution's] interest rather than [the defense's] in the sense that the numbers that I am calculating will . . . tend to show less bias than they would if I had data on this question." (Id. at 57; emphasis added.) Using these assumptions, Dr. Kadane calculated the effect of excluding "automatic death penalty" jurors on the findings of the Ellsworth Attitude Survey and the Ellsworth Conviction-Proneness Z fStudy. The results, with respect to the Ellsworth Attitude Survey are set forth in Table 5 of the Kadane Study (Exhibit B, No. 6,at 22; see Figure A). (The form of this table is identical to that of Table 2 of the Kadane Study (see Figure 2).) Dr. Kadane states: "I conclude from Table 5 a showing of substantial bias against the defense from the current procedure." (Kadane Study, Exhibit B, No. 6, p . 21.) With respect to the Ellsworth Conviction-Proneness Study, Dr. Kadane reports: "The showing of substantial bias in the Ellsworth and Fitzgerald study is con firmed by the reanalysis of the Ellsworth, Thompson and Cowan experiment. Here $ is 1.519 (so the estimated odds of NEVER or ALWAYS juror being more favorable to the defense than a SOMETIMES AND SOMETIMES NOT juror is more than 3 to 2), with a standard deviation of .228, which means that ((f) -1)/SD = 2.27. Hence, the probability of neutrality or bias against the prosecution ( $ ^ 1) is 1.3%. Pgain we have a finding of substantial bias against the defense." (Kadane Study, Exhibit B , No. 6, at 21) . 7/ The mathematical calculations involved are discussed in the Kadane Study (Exhibit B, No. 6) at pp. 18-19 and 26-40; a minor technical issue ("effective sample size") is discussed at 20- 21 (see RT 58). - 21 - Comparing these data to the original findings of the Ellsworth Conviction-Proneness Study, Dr. Kadane stated: "They are very similar. The estimated odds drop from 1.65 to 1.52, which is a very slight drop." (RT 66.) The underlying reason for these findings is described at the conclusion of the Kadane Study: "What makes all the results true is that the ALWAYS group is so small (1% of the population) that the ALWAYS or NEVER group is dominated by the NEVER part, and the AT LEAST SOMETIMES group is dominated by the SOMETIMES AND SOMETIMES NOT majority. Even attributing the least favorable views to the ALWAYS group does not disturb the finding of substantial bias against the defense." (Exhibit B, No. 6, at 23.) Dr. Kadane calculated the effect of excluding the the "automatic death penalty" group using data from the Ellsworth Attitude Survey and the Ellsworth Conviction-Proneness Study be cause "they are the most recent and most thorough of their respec tive types." (Kadane Study, Exhibit B, No. 6, at 4.) He testified, however, that the same general conclusions would apply to the findings of the other studies on death-qualification. (Words 6 Sparks RT 68.) The linchpin of the District Court's order is its rejection of Dr. Kadane's finding. This rejection is based on three grounds, all untenable. First, the District Court complains that Dr. Kadane 'performed no original research.' (Order, p. 7). The significance of-this criticism is not apparent. Dr. Kadane may not have interviewed research subjects personally, but he testified about a major new study of death-qualification that he conducted. His study does reanalyze the data of earlier studies, but it does so on the basis of two new surveys that provide new data addressed to the issue left open in Hovey. 22 Second, the District Court speculates that the proportion of automatic death penalty jurors (ADPs) may in fact be higher than the 1% that Harris found. There is no evidentiary basis for this speculation. The Court cites a reference in Hovey to a possible ADP figure over 25%; in fact that figure comes from a study that does not measure ADPs. In any event, the proportion of ADPs is a factual question — put in dispute by the pleadings ̂/ It cannot be decided without evidence. 8/ The prosecutor at the Word 8 Sparks hearing faulted Dr. Kadane for not using the Jurow Study, or Smith, A Trend Analysis of Attitudes Toward Capital Punishment, 1936-1974 (1975) for their data on the size of the "automatic death penalty" group. (Word & Sparks RT 89-95.) This is an empty criticism. Dr. Kadane explained why he used Harris 1981 rather than Jurow's finding of 2% automatic death penalty jurors: "The Harris study, being as it was done by a national polling organization, done under very strict standards . . . is a very reliable thing for me to use. I had confidence in that study, and that's why I used it." (Id. at 93-94.) In comparison to the Jurow Study, "the Harris study, it seemed to me, was the sounder work." (_Id. at 94.) The Hovey court concurred. Jurow's finding that "2 percent of his 211 subjects fell into the 'automatic death penalty' category . . . did not afford a reliable basis for generalizing to the percentage of such jurors in the entire population, since . . . [it] was [not] based on a random sample of the population." (28 Cal.3d at 64 n.lll; see also id.., at 63 n.109 and accompanying text.) (Jurow's finding was based on interviews with 211 workers at one plant.) Dr. Kadane stated that he had not read the Smith study. (Word & Sparks RT 89-95.) However, the Smith study -- which was before the court in Hovey -- contains only "tentative indications" as to the number of automatic death penalty jurors (28 Cal.3d at 64); it has no data on the issue, as the Hovey opinion reflects. The "suggestion" in Smith relating to this question (see id., at 64, n. Ill) is contained in data from a 1973 Harris poll (see Smith, supra, Appendix 2 for the text of the relevant questions) which merely indicate that 28% of the 1973 national sample favored the death penalty for all persons convicted of first degree murder. The 1973 Harris survey contains no indication of how many of these people, as jurors, would have personally voted for the death penalty, automatically, in every capital case, much less how many of them would have done so in the face of explicit legal 23 [cont'd .] Third, the District Court then argues that Dr. Kadane's conclusions are untrustworthy because he has no data on the "be havioral characteristics that an always-death group might possess," and the Court speculates that this group might have a disproportionate impact if it "tended to vote in a unified fashion." (Order at 8). This is plain error. Dr. Kadane assumed that this group would vote in a unified fashion, and that it would be as pro-prosecution as possible. His findings of bias are reliable precisely because he makes such a conservative assumption, and therefore necessarily understated the bias that the process of death-qualification produces. The District Court completely missed this important point. The basic problem with the District Court's evaluation of Dr. Kadane's research is that the Court purports to resolve contested factual issues without a hearing. The errors the Court makes — its remarkable oversights and extreme misinterpretations of the evidence — are classic examples of the most basic reason why such in camera fact finding is prohibited: because it leads almost inexorably to error. In this case the errors are all the more glaring because they contrast harshly with the actions of other federal courts faced with the identical issue. Evidence that was rejected here without a instructions to them, as jurors, to consider all possible penalties. Faced with this evidence, the court in Hovey concluded that "there is no reliable data" on the size of the automatic death penalty group. (I_d. , at 6k.) 2k hearing was found to be valid and reliable after a hearing by another federal court, Keeten v. Garrison, supra, 578 F. Supp. at 1175-77; and while the District Court here speculates that there may be more ADPs than the 1% Dr. Kadane estimated, other federal courts — examining Dr. Kadane's underlying data in light both of competing evidence and of testimony on direct and cross-examination — found that the number of ADPs is negligible. Grigsby v. Mabry, supra, 558 F. Supp. at 1308, aff'd , 758 F.2d at 238. The Court's key factual assumptions are thus unwarranted. II BECAUSE DEATH-QUALIFIED JURIES ARE BIASED IN FAVOR OF THE PROSECUTION AND UNDULY PRONE TO CONVICT, THE USE OF SUCH JURIES TO TRY THE GUILT OR INNOCENCE OF CAPITAL DEFENDANTS VIOLATES THE SIXTH AND FOURTEENTH AMENDMENTS Although the District Court does not refer to this issue, other circuits have held that the fact that death-qualified juries may be biased in favor of the prosecution and unduly prone to con vict is of no constitutional significance. The leading case in this line is Spinkellink v. Wainwright, 578 F.2d 582, 594- (5th Cir. 1978). If it disposes of the factual and procedural questions and ultimately reaches the legal merits on this appeal, we urge this Court to reject such an erroneous view of the Sixth Amendment. As noted earlier, the Fifth Circuit held in Spinkellink that "[w]hen petitioner asserts that a death- qualified jury is prosecution-prone, he means that a death-qualified jury is more likely to convict than a non-death-qualified jury . . . . Even if this is true the peti tioner's contentions must fail. That a death-qualified jury is more likely to convict than a non-death-qualified jury does not demonstrate which jury is impartial." (Id. at 594.) 25 With due respect to the Fifth Circuit, the argument does not make sense. If ordinary, non-death-qualified juries are acquittal- prone and unfair, why are they used in all criminal trials except capital cases? Could a State seriously contend that it would not have received a fair trial if petitioner had been tried for non capital murder, because his jury would have been acquittal- prone? The issue here is whether a State can increase a defendant s chances of conviction -- tip the balance on guilt or innocence by placing him on trial for a capital crime, rather than a non capital one, and then death-qualifying the guilt-or-innocence jury. Such an argument, moreover, is squarely refuted by Wither- spoon. The Supreme Court in Witherspoon condemned the systematic exclusion of opponents of the death penalty from capital juries be cause it "stacked the deck against the petitioner" on the issue of penalty. 391 U.S. at 523. If an inordinate tendency to prefer a particular outcome were constitutionally acceptable, the Supreme Court would not have condemned this practice. However, the Court recognized that a jury must express the "conscience of the community," id. at 519, and that its performance must be measured against the yardstick of that community. Pre-Witherspoon juries failed that test because they were "uncommonly willing to condemn a man to die." Id., at 521. The Spinkel1 ink position necessarily implies that a lesser standard of reliability governs determinations of guilt in capital cases than do determinations of penalty -- that a jury "uncommonly willing" to convict on capital charges is 26 constitutional despite the holding in Witherspoon. There is no justification for this distinction, and it is directly refuted by Witherspoon itself. The Court there recognized that a de fendant might someday prove that death-qualified juries are guilt-prone, and stated: "If he were to succeed in that effort, the question would then arise whether the State1s interest in submitting] the penalty issue to a jury capable of imposing capital punishment may be vindicated at the expense of the defendant's interest in a completely fair determination of guilt or innocence -- given the possibility of accommodating both interests by means of a bifurcated trial, using one jury to decide guilt and another to fix punishment." 391 at 520 n .18 (emphasis added). The Supreme Court's question, we submit, is rhetorical: the State's interest in using a single jury in capital cases cannot possibly outweigh the right to a fair trial. The Eighth Circuit addressed and disposed of the Spinkellink argument in Grigsby v. Mabry: "We feel the reasoning of Spinkellink . . . fails to analyze or decide the fundamental issue . . . The issue is not whether a jury would be biased one way or the other, but whether an impartial jury can exist when a distinct group in the community is excluded by systematically challenging them for cause. As the district court noted, the petitioners were not seeking _9_/ 9/ In Beck v. Alabama, 4.27 U.S. 625 (1980), the Supreme Court has expressly held that the higher standard of procedural fairness required at the penalty phase of the capital trials extends to the guilt-or-innocence phase as well: "To ensure that the death penalty is indeed imposed on the basis of 'reason rather than caprice or emotions,' we have invalidated procedural rules that tended to diminish the reliability of the sentencing determination. The same reasoning must apply to rules that diminish the reliability of the guilt determination." Id. at 638 (emphasis added). 27 a jury composed entirely of [Witherspoon excludables]; what petitioners sought was a 'jury drawn from the entire cross section of the community as a whole, including both those who strongly favored the death penalty and those who strongly opposed it.'" Grigsby v. Mabry, supra, 758 F.2d at 24-0-4-2. CONCLUSION When Witherspoon was decided in 1968, the evidence on the possible bias of death-qualified juries was "too tentative and fragmentary. " Witherspoon v. Illinois, supra, 391 U.S. at 517 n.10. Since that time, however, the scientific evidence has be come definite and whole, and totally one-sided. It has convinced two federal district courts; it is the basis for a major en banc decision by the Eighth Circuit; and it will likely be soon re viewed by the Supreme Court. See Witt v. Wainwright, __U.S. __, 8k L .Ed.2d 801 (Marshall & Brennan, JJ., dissenting from denial of certiorari)("This Court is certain to grant certiorari in the immediate future to resolve this issue.") It would be appropriate and correct for this Court to accept the factual findings of the Eighth Circuit and follow Grigsby v. Mabry. The Supreme Court expressed some concern in Witherspoon that the studies adduced there had not been presented initially at the trial court level, see 391 U.S. at 517 n.ll. Similarly, several justices in Ballew v. Georgia, 4-35 U.S. 223 (1978) were troubled that studies relied upon there by the majority had never been "subjected to the traditional testing mechanisms of the adversary process." 4-35 U.S. at 24-6. But the 28 overwhelming evidence here has been thoroughly tested in a number of lower courts, and it may not be necessary for every court that faces this claim to hear it again at the trial level. It would be inappropriate, however, to rely upon the present record to reject petitioner's factual contentions. The District Court's adverse findings are the product solely of surmise and speculation, not adversary testing. If the Court concludes that the Grigsby and Word & Sparks records do not yet constitute a sufficient factual foundation upon which to base its legal conclusion, the case should be remanded so that full evidence may be presented on the factual issues that remain in dispute. Dated: New York, New York July 1, 1985 Respectfully submitted, JULIUS L. CHAMBERS JAMES M. NABRIT, III JOHN CHARLES BOGER DEVAL L. PATRICK 99 Hudson Street New York, New York 10013 (212) 219-1900 ANTHONY G. AMSTERDAM New York University School of Law 4-0 Washington Square South New York, New York 10012 (212) 598-2638 ATTORNEYS FOR AMICUS CURIAE STATEMENT OF RELATED CASES Amicus curiae is not aware of any cases currently pending in this Court which related to the issues it has addressed in this brief. 30 CERTIFICATE OF SERVICE I hereby certify that I am counsel for amicus curiae NAACP Legal Defense & Educational Fund, Inc., and that I served the annexed motion for leave to file brief amicus curiae and brief amicus curiae on the parties to this action, by placing copies in the United States mail, first class mail, postage prepaid, addressed as follows: Michael J. McCabe, Esq. 108 Ivy Street San Diego, California 92101 Charles M. Sevilla, Esq. 1010 Second Avenue, Suite 1001 San Diego, California 92101 Michael Wellington, Esq. Office of the Attorney General 110 West "A" Street, Suite 700 San Diego, California 92101 All parties required to be served have been served. Done this 1st day of July, 1985. P /; /I / ,6' APPENDIX A APPENDIX A This Appendix is provided for the convenience of the Court to summarize the major studies on death-qualification that have been presented to the courts. The studies are divided into four major categories: (i) studies on the relation between sub jects' death penalty attitudes and their demographic characteristics and other criminal justice attitudes (pages 1-6, infra); (ii) studies on the relation between subjects' death penalty attitudes and their behavior as jurors in actual or simulated trials (pages 6-14,infra); (iii) other related studies on the relationship between death penalty attitudes and behavior in capital cases (pages 11-14,infra); and (iv) studies on the incidence of "automatic death penalty jurors" in the population (pages 14-15, infra). I. ATTITUPINAL AMD DEMOGRAPHIC SURVEYS 1. BRONSON - COLORADO, 1970 AUTHOR: BRONSON, Edward C. TITLE: "On the Conviction Proneness and Representativeness of the Death-Qualified Jury: An Empirical Study of Colorado Veniremen," 42 Colo. L. Rev. 1 (1970). CITATIONS: Hovey v. Superior Court, 28 Cal.3d 1, 43-44,_616^P^ 2d 1306, 1327-28 (1980); Grigsby v. Mabry, 569 F. Supp. 1273, 1293- 94 (E.D. Ark. 1983); 758 F .2d 226, 232 (8th Cir. 1985)(en banc); Keeten v. Garrison, 578 F. Supp. 1164, 1172 (W.D.N.C. 1984). SUMMARY: This was the first study to examine the relationship between death-penalty attitudes and other attitudes relating to the administration of criminal justice. The respondents in this survey were 718 Colorado venirepersons. Each respondent was asked whether he or she "strongly favor[ed]," "favor[ed]," "oppose[d]" or "strongly oppose[d]" the death penalty. Each respondent was also asked five questions about his or her attitudes on criminal justice issues. Interviews were carried out, in person and by telephone, by trained students from the University of Colorado, in 1968 and 1969. The Bronson-Colorado, 1970 survey shows a consistent correlation between attitudes on the death penalty and attitudes on other criminal justice issues. On each of the five questions, the stronger the respondent's support for the death penalty (as measured on Bronson’s four- point scale), the stronger their support for positions most favorable to the prosecution. 2. BRONSON - CALIFORNIA, 1980 AUTHOR: BRONSON, Edward C. TITLE: "Does the Exclusion of Scrupled Jurors in Capital Cases Make the Jury More Likely to Convict? Some Evidence from California, 3 Woodrow Wilson L. Rev. 11 (1980). CITATIONS: Hovey, 28 Cal.3d at 47-4-9, 616 P . 2d at 1330-33; Grigsby, 569 F. Supp. at 1293-94; 758 F .2d at 232-33; Keeten, 578 F. Supp. at 1172-73. SUMMARY: The studies grouped together in Bronson - California, 1980 are similar in methodology and results to the Bronson - Colorado, 1970 survey. The first of these studies, Bronson/Butte County, 1980, was conducted in 1969-1970. Seven hundred and fifty-five people from Butte County, California jury venires were interviewed over the telephone by students at the California State University at Chico. As in Bronson - Colorado, 1970, respondents were asked to indicate their position regarding the death penalty on a scale from "strongly 2 They were also asked whether theyfavor" to "strongly oppose." agreed or disagreed with seven statements: five that were nearly identical to the one used in Bronson - Colorado, 1970, and two additional criminal justice items. The findings of the Butte County survey closely parallel those in Bronson - Colorado, 1970: the stronger the endorsement of the death penalty, the higher the level of agreement with pro-prosecution statements. Following the Butte County study, Professor Bronson admin istered a slightly modified questionnaire to a sample of 707 venire- persons from Los Angeles, Sacramento and Stockton, California. (Bronson - Los Angeles, 1980). These interviews were carried out in late 197k and early 1975. Once again, the data showed a consistent pattern: the more strongly the respondents favored the death penalty, the more likely they were to endorse pro-prosecution positions, and there were marked attitudinal differences between the "sbrcngly oppose" group and the other three groups combined. In a followup survey on some kOO Butte County prospective venirepersons in June 1971, Bronson found that 93% of those who "strongly opposed" the death penalty would be legally excludable under Witherspoon. 3. HARRIS, 1971 AUTHOR: LOUIS HARRIS & ASSOCIATES, INC. TITLE: Study N.o. 2016 (1971). CITATIONS: Hovey, 28 Cal.3d at k5-k7, 616 P .2d at 1328-30; Grigsby, 569 F. Supp. at 1293-9k; 758 F .2d at 233; Keeten, 578 F. Supp. at 1173-7k. SUMMARY: Harris, 1971 is a detailed national opinion survey on attitudes toward the death penalty, and the first study in which 3 a direct comparison can be made between respondents who are death- qualified and those who are excluded by Witherspoon criteria. It was administered in person to a representative sample of 2,068 respondents drawn from the adult population of the United States in 1971. The findings of the Harris, 1971 survey parallel those of the Bronson surveys, and greatly extend them. In response to dozens of questions on their attitudes toward various aspects of the crim inal justice system, death-qualified respondents were consistently more likely to favor the prosecution's position than Witherspoon- excludable respondents. Harris, 1971 also found that more blacks than whites would be excluded from jury service by death qualifica tion (4-6% vs. 29%), and more women than men (37% vs. 24%). (Harris, 1971 also collected data on the voting behavior of the respondents as jurors in criminal trials; see infra, at g-io). 4. NATIONAL POLL DATA AUTHOR: LOUIS HARRIS & ASSOCIATES, INC.: AMERICAN INSTITUTE FOR PUBLIC OPINION (Gallup); AND NATIONAL OPINION RESEARCH CENTER TITLE: Various national polls from 1953 through 1978 partially summarized in: Smith, Tom W. "A Trend Analysis of Attitudes Toward Capital Punishment, 1936-1974," in James A Davis (ed.) Studies of Social Change Since 1948, National Opinion Research Center, Report 127B, Chicago (1976). CITATIONS: Hovey, 28 Cal. 3d at 54-57, 616 P .2d at 1337-39. SUMMARY: Numerous surveys of the national population have established two major demographic facts about attitudes toward the death penalty: (1) Since 1953, women have consistently opposed the death penalty in greater proportions than men. (2) Since 1953, blacks have consistently opposed the death penalty in greater proportions than whites and that racial gap has grown steadily, from a difference of 8% in 1953 to 27% in 1978. 4 5. ELLSWORTH ATTITUDE SURVEY, 1979 AUTHORS: ELLSWORTH, Phoebe C.; and FITZGERALD, Robert TITLE: "Due Process vs. Crime Control: Death Qualification and Jury Attitudes," published in 8 Law and Human Behavior, Issue 1-2, pp. 31-53 (198A). CITATIONS: Hovey, 28 Cal.3d at 50-54, 616 P .2d at 1333-37; Grigsby, 569 F. Supp. at 1293-94; 758 F .2d at 233; Keeten, 578 F. Supp. at 1171-72. SUMMARY: The Ellsworth Attitude Survey, 1979 is the most sophisticated of the surveys that have examined the relationship between death-qual ification and juror attitudes. The respondents in the Ellsworth Attitude Survey, 1979 were a probability sample of 811 jury-eligible adult residents of Alameda County, California, in 1979. The sample was drawn, and the subjects interviewed, by the Field Research Corpora tion of San Francisco, an independent professional polling organization. The respondents in the Ellsworth Attitude Survey, 1979 were asked carefully tailored questions that embody the two prongs of the Witherspoon standard: whether they would consider voting to impose the death penalty, and whether they could be fair and impartial in determining guilt or innocence in a capital case. Respondents who could not be fair and impartial ("nullifiers") were excluded from the analysis; of those who could be fair and impartial (717 out of the total of 811), 17.2% were Witherspoon excludable. Respondents were asked 13 attitudinal questions on criminal justice issues; on each, death-qualified respondents were more favorable to the prosecution, more crime-control oriented, and less concerned with constitutional protections for suspects than were excludable respondents. Most differences were sizeable and highly statistically significant. The 5 survey ctXso found that more blacks than whites are excluded by death-qualification (25.5% vs. 16.5%), and more women than men (21% vs. 13%). 6. PRECISION RESEARCH, 1981 AUTHOR: PRECISION RESEARCH, INC. TITLE: Precision Research Survey CITATIONS: Grigsby, 569 F. Supp. at 1294; 758 F .2d at 233. SUMMARY: In June, 1981, Precision Research, Inc., a polling organization in Little Rock, Arkansas, conducted a state-wide survey of death penalty attitudes using a representative sample of 407 respondents drawn from the adult population of the State of Arkansas. This survey used the same death penalty questions that had been used in the Ellsworth Attitude Survey, 1979. It found that (i) approximately 11% of Arkansas adults who could be fair and impartial in determining guilt or innocence in a capital case are excludable under Witherspoon because they would never consider voting for the death penalty; (ii) among those who would be fair and impartial, more blacks than whites are excludable in Arkansas (29% vs. 9%); and (iii) more women than men (13% vs. 8%). II. CONVICTION-PRONENESS STUDIES 1. ZEISEL, 1968 AUTHOR: ZEISEL, Hans TITLE: "Some Data on Juror Attitudes Toward Capital Punishment," Monograph, Center for Studies in Criminal Justice, University of Chicago Law School (1969). CITATIONS: Hovey, 28 Cal.3d at 27-30, 616 P .2d at 1315-17; Grigsby, 569 F. Supp. at 1295-96; 758 F.2d at 233; Keeten, 578 F. Supp. at 1174. 6 SUMMARY: This is the earliest study on the conviction proneness of death-qualified jurors. The data for the study were collected by Professor Zeisel and his late colleague Professor Harry Kalven, Jr. in 1954- and 1955, although the present monograph was not published until 1968. (In Witherspoon, the Supreme Court had before it some fragments of an early draft of this study; see 391 U.S. at 517 n.10). One distinctive feature of this study is that it examined the behavior of actual criminal trial jurors. The researchers interviewed jurors who had just completed service on felony trial juries in the Brooklyn Criminal Court in New York and in the Chicago Criminal Court in Illinois, and asked them three questions: (i) What was the first ballot vote of the jury as a whole? (ii) What was your own first ballot vote? (iii) Do you have conscientious scruples against the death penalty? In all, the researchers collected data on 464 such votes. Professor Zeisel analyzed these data, controlling for the strength of the evidence of the defendant's guilt, and deter mined what subjects with scruples against the death penalty voted to acquit significantly more often then those without scruples against the death penalty. 2. WILSON, 1964 AUTHOR: WILSON, W. Cody TITLE: "Belief in Capital Punishment and Jury Performance," unpublished (1964). CITATIONS: Hovey, 28 Cal.3d at 32-33, 43; 616 P .2d at 1318-19, 1327; Grigsby, 569 F. Supp. at 1295; 758 F .2d at 233-34; Keeten, 578 F. Supp. at 1174. SUMMARY: Wilson, 1964 was the first experimental study on the conviction-proneness of death-qualified jurors. The subjects -- 7 187 college students -- were presented in 1964- with written descriptions of five capital cases (four with a single defendant, one with two codefendants), asked to assume that they were members of the juries trying the cases, and requested to reach a decision on each defendant's guilt or innocence. Each subject was also asked "Do you have conscientious scruples against the death penalty, or capital punishment for a crime?" Wilson found that subjects without scruples against the death penalty voted for conviction more often than those who had scruples against the death penalty (difference significant at the p<.02 level). 3. GOLDBERG, 1970 AUTHOR: GOLDBERG, Faye (Faye Girsh) TITLE: "Toward Expansion of Witherspoon: Capital Scruples Jury Bias, and the Use of Psychological Data to Raise Presumptions in the Law," 5 Harv. C.R.-C.L.L. Rev. 53 (1970). CITATIONS: Hovey, 28 Cal.3d at 30-31; 516 P .2d at 1317-18; Grigsby, 569 F. Supp. at 1295; 758 F .2d at 233; Keeten, 578 F. Supp. at 1174. SUMMARY: The subjects in this 1966 study — 200 students in private liberal arts colleges in Georgia, 100 white and 100 black — were given 16 written descriptions of criminal cases involving various crimes, and were asked to assume that they were jurors and to indicate their vote on the case. They were also asked: "Do you have conscientious scruples against the use of the death penalty?" Subjects without scruples against the death penalty voted to convict in 75% of the cases, while those with scruples voted to convict in 69% (difference significant at the p<.08 level). 8 JUROW, 1971 AUTHOR: JUROW, George L. TITLE: "New Data on the Effects of a 'Death-Qualified' Jury on the Guilt Determination Process," 84 Harv. L. Rev. 567 (1971). CITATIONS: Hovey, 28 Cal.3d at 33-36; 616 P .2d at 1319-21; Grigsby, 569 F. Supp. at 1296-97; 758 F .2d at 234; Keeten, 578 F. Supp. at 1174. SUMMARY: Jurow's subjects — 211 employees of the Sperry Rand Corporation in New York — listened to two tape recordings of simulated murder trials including, in abbreviated form, opening statements, examination of witnesses, closing arguments, and the judge's instruction to the jury, and voted on the guilt or innocence of the defendant by marking a ballot. In addition, Jurow asked his subjects to complete a long questionnaire that contained several sets of questions relating to the death penalty, one of which (a five-point scale designated "CPAQ(B)") included a statement embodying the first prong of the Witherspoon criteria for exclusion: "I could never vote for the death penalty regardless of the facts and circumstances of the case." When the subjects are divided into groups on the basis of their positions on that five-point CPAQ(B) scale, the pattern that emerges resembles the patterns of responses to Bronson's attitudinal surveys: the subjects who more strongly favor the death penalty are more likely to convict. These differences are statistically significant at the .01 level in Jurow's first case, but not statistically significant in the second. 5. HARRIS, 1971 AUTHOR: LOUIS HARRIS & ASSOCIATES, INC. TITLE: Study No. 2016 9 CITATIONS: Hovey, 28 Cal.3d at 36-37, 616 P .2d at 1321-23; Grigsby, 569 F. Supp. at 1297-98; Keeten, 578 F. Supp. at 1173-74. SUMMARY: The Harris, 1971 study, in addition to its attitudinal and demographic data (see supra pp. 3-4) gathered behavioral data on conviction-proneness. Each of the 2,068 subjects in the national sample was instructed about three legal principles which apply to all criminal cases (the prosecutor's burden of proof, the requirement that a defendant be convicted by proof beyond a reasonable doubt, and the rule that the defendant is not required to testify) and was given written descriptions of the evidence in four criminal cases. After each description, the subject was given the legal definition of the crime involved, and asked whether, as a juror in the case, he or she would find the defendant guilty or not guilty. In each case the Witherspoon-excludable subjects voted to convict less often than death-qualified subjects. In three of the cases the differences between the two groups were significant at the .01 level; in the fourth case the difference was significant at the .10 level. 6. ELLSWORTH CONVICTION-PRONENESS STUDY, 1979 AUTHOR: ELLSWORTH, Phoebe C.; THOMPSON, William; COWAN,Claudia TITLE: "The Effects of Death Qualification on Jurors’ Predisposition to Convict and on the Quality of Deliberation," published in 8 Law and Human Behavior, Issue 1-2, pp. 53-79 (1984). CITATIONS: Hovey, 28 Cal.3d at 38-40, 616 P .2d at 1323-25; Grigsby, 569 F. Supp. at 1298-1301; 758 F .2d at 234; Keeten, 578 F. Supp. at 1174-75. SUMMARY: Ellsworth Conviction-Proneness Study, 1979 is the most sophisticated of the experimental studies on the conviction-prone- ness of death-qualified jurors. Its subjects -- 288 jury eligible 1 0 adult residents of San Mateo and Santa Clara Counties, California were screened by telephone when they were initially contacted in 1979 to determine: (i) whether they would be excluded from capital juries under Witherspoon because they would never consider voting to impose the death penalty; and (ii) whether they could be fair and impartial in judging guilt or innocence in a capital case. Those who could not be fair and impartial ("nullifiers ) were excluded from participation. The remaining subjects were shown a detailed and highly realistic 2\ hour videotape of a homicide trial _ including extensive witness testimony, arguments by prosecutor and defense counsel, and instructions by the trial judge — and were asked to give a written verdict. Nearly 78% of the death-qualified subjects voted to convict the defendant of some degree of homicide, compared with only 53.3% of the Witherspoon-excludable subjects (difference highly statistically significant, at the .01 level). III. OTHER EXPERIMENTAL STUDIES 1. ELLSWORTH POST-DELIBERATION DATA,1979 AUTHORS: ELLSWORTH, Phoebe C.; COWAN, Claudia; and THOMPSON, William TITLE: "The Effects of Death Qualification on Jurors' Predisposition to Convict and on the Quality of Deliberation," published in 8 Law and Human Behavior, Issue 1-2, pp. 53-79 (1982). CITATIONS: Hovey, 28 Cal.3d at 39-20, 57-58, 60; 616 P .2d at 1322- 25, 1339-21; Grigsby, 569 F. Supp. at 1302. SUMMARY: After filling out their written ballots, most of the subjects in the Ellsworth Conviction-Proneness Study, 1979 (see supra pp. 10-11 — 228 out of 288 -- were divided into juries of 12 members and deliberated for one hour on a verdict in the trial they had just seen. Half of the jury panels had only death-qual ified subjects, and half included two, three, or four Witherspoon- 1 1 excludable subjects ("mixed juries"). After the one-hour delib eration, the jurors were asked to fill out a second ballot form. While none of the juries reached a unanimous verdict in the time allotted, individual jurors did change their votes. The basic difference between death-qualified and Witherspoon-excludable jurors did not change: the death-qualified jurors remained far more likely to vote for conviction than the Witherspoon-excludable jurors (difference significant at the .01 level). The subjects were also asked to fill out a post-deliberation questionnaire about the trial. The questionnaire data reveal that: (i) the death-qual ified subjects were more likely to believe the prosecution witnesses and to disbelieve the defense witnesses than Witherspoon-excludable subjects; (ii) subjects on mixed juries viewed all witnesses (prose cution and defense) more critically than did subjects on death-qual ified juries; (iii) subjects on mixed juries remembered the facts of the case better than those on death-qualified juries. 2. ELLSWORTH WITNESS CREDIBILITY STUDY, 1979 AUTHORS: ELLSWORTH, Phoebe C.; HARRINGTON, Joan C.; THOMPSON, William and COWAN, Claudia. TITLE: "Death Penalty Attitudes and Conviction Proneness: The Translation of Attitudes Into Verdicts" published in 8 Law and Human Behavior, Issue 1-2, pp. 95-113 (1984.). CITATIONS: Hovey, 28 Cal.3d at 59-60, 616 P .2d at 1340-41; Grigsby, 569 F. Supp. at 1302; 758 F .2d at 234; Keeten, 578 F. Supp. at 1175. SUMMARY: Twenty death-qualified and 16 Witherspoon-excludable subjects who had participated in the Ellsworth Conviction-Proneness Study, 1979 (see supra, pp. 10-11) were shown a videotape of a prose cution witness (a police officer) and a defense witness (a possible defendant) giving two versions of an incident; they then answered 1 2 questions about the witnesses' testimony. On each of 16 questions, the death-qualified subjects gave answers that were more favorable to the prosecution than those of the Witherspoon-excludable subjects (on 10 of the 16 questions the differences were significant at or beyond the .05 level). 3. HANEY, 1979 AUTHOR: HANEY, Craig TITLE: "On the Selection of Capital Juries: The Biasing Effects of the Death-Qualification Process," published in 8 Law and Human Behavior, Issue 1-2, pp. 121-132 (1984.) . CITATIONS: Hovey, 28 Cal.3d at 75-79, 616 P .2d at 1350-53; Grigsby, 569 F. Supp. at 1302-05; 758 F .2d at 234; Keeten, 578 F. Supp. at 1175-76. SUMMARY: Haney, 1979 is the first study to investigate the effects of the process of death-qualification on the jurors who undergo it. The subjects — 67 jury-eligible adults from Santa Cruz County, California -- were screened by telephone following the procedures used in the Ellsworth Conviction-Proneness Study, 1979 (see supra, pp. 10-11) to identify Witherspoon-excludables and nullifiers; only death-qualified subjects who could be fair and impartial in deciding guilt or innocence in a capital case were used. Those subjects were randomly assigned to two groups: one group saw a realistic two-hour videotape of a voir dire in a capital murder case, including a half-hour of voir dire during which prospective jurors were death-qualified; the other group saw the same videotape with the death-qualification segment edited out. Following the viewing, both groups filled out a questionnaire on their beliefs about the case. The questionnaire data show that the randomly 13 assigned subjects who saw the death-qualifying voir dire were more likely to think that the defendant was guilty than those who did not, more likely to believe that he would be convicted, and more likely to think that the judge and the defense attorney believed that the defendant was guilty (all differences significant beyond the .05 level). IV. STUDIES ON "AUTOMATIC DEATH PENALTY" JURORS (ADP'S) 1. Harris, 1981 AUTHOR: LOUIS HARRIS & ASSOCIATES, INC. TITLE: Study No. 814002 (1981) CITATIONS: Grigsby, 569 F. Supp. at 1307; 758 F .2d at 236; Keeten, 578 F. Supp. at 1170. SUMMARY: This national Harris survey of a sample of 14-98 respondents drawn from the adult civilian population of the United States in January 1981 found that only 1% of the respondents who could be fair and impartial in determining guilt or innocence in capital cases would automatically vote for the death penalty upon conviction for any capital murder. 2. ARKANSAS ARCHIVAL STUDY, 1981 AUTHOR: YOUNG, Andrea TITLE: Arkansas Archival Study (1981) CITATIONS: Grigsby, 569 F'. Supp. at 1307; 758 F . 2d at 234. SUMMARY: A review of 41 transcripts of voir dires of capital cases on file at the Arkansas Supreme Court, taken from trials in 1973 through 1981, reveals that of all the jurors who were questioned (a total of 1,914) 14.4% were excluded on Witherspoon grounds but only 0.5% were excluded because they would always vote to impose 14 3. KADANE, 1981 AUTHOR; KADANE, Joseph B. TITLE: "Juries Hearing Death Penalty Cases: Statistical Analysis of a Legal Procedure," published in 78 Journal of the American Statistical Association, No. 383, pp. 54-4-552 (1983) (see also 8 Law and Human behavior, Issue 1-2, pp. 115-120 (1984-) for a more concise account). CITATIONS: Keeten, 578 F. Supp. at 1176-77. SUMMARY: In this study Dr. Kadane reanalyzed the results of the Ellsworth Attitude Survey, 1979 (see supra, pp. 5-6) and the Ellsworth Conviction-Proneness Study, 1979 (see supra,pp. 10-11) using data from the Harris, 1981 (see supra, p. 14) survey for the proportion of ADP's in the population, and data from a 1981 state-wide Field Research Corporation survey of California for the proportion of fair and impartial Witherspoon-excludables in the population. The reanalysis reveals that even when the most conservative possible assumptions are made about the attitudes and behavior of ADP jurors, their exclusion from capital juries does not alter the fact that death-qualification biases capital juries against the defendant on the issue of guilt or innocence. the death penalty upon conviction, in every capital case. 15 APPENDIX B Juries Hearing Death Penalty Cases: Statistical Analysis of a Legal Procedure JOSEPH B. KADANE* Potential jurors in capital cases are often queried on their attitudes toward capital punishment. The extreme groups say they would never or they would always approve cap ital punishment, given a guilty verdict. In many jurisdic tions, these two groups are routinely excluded from juries deciding whether the defendant is guilty in capital cases. This exclusion persists even when the potential jurors say they could be fair and impartial in deciding guilt or in nocence. The current study shows that this exclusion cre ates a bias that almost certainly works against the de fendant. KEY WORDS: Death penalty; Law; Missing datapBaye- sian analysis; Survey research. This article considers a question of current legal inter est: whether the present jury selection procedure used in cases in which capital punishment is a possibility is biased against the defense. Section 1 gives the legal background of the question addressed here. Section 2 describes two major studies on jurors in capital cases that were recently presented to the California Supreme Court in Hovey v. Superior Court (1980): the Fitzgerald and Ellsworth (1984) work on the attitudes of jurors favoring either the defense or the prosecution, and the Cowan, Thompson, and Ellsworth (1984) experiment on juror behavior. The response of the California Supreme Court in Hovey is then discussed. Section 3 introduces the basic measures used in this article, and applies them to the Fitzgerald and Ellsworth, and the Cowan, Thompson, and Ellsworth data. Section 4 reviews two surveys on attitudes toward the death penally conducted since Hovey. a Field Re search Corporation survey of Californians and a Harris national survey. Then Section 5 gives the assumptions needed to justify the likelihood function. Section 6 gives the numerical results and discusses their implications. Substantial evidence of bias against the defense is found. 1. LEGAL BACKGROUND Some 90 percent of the criminal cases prosecuted in America are settled by plea bargaining without a trial. Of * Joseph B. Kadane is Professor of Statistics and Social Science, Carnegie-Mellon University, Pittsburgh, PA 15213. The author thanks Don Rubin for continuing to act as Editor for this article beyond the expiration of his term He also thanks Morris DeGroot, Pheobe Ellsworth, Eugene Ericksen, Steve Fienberg, Robert Fitzgerald, Robert Kass, Jill Larkin, Kathy Blaekmond Laskey, Timothy McGuire, and Michael Shamos for their comments and help. The referees and As sociate Editor gave unusually helpful comments. This article is a re vision of the paper that was the basis for the author's testimony in People v. Word and Sparks, California Superior Court, 31 August and I September 1981. the remainder, fewer than half are heard by juries. And of those heard by juries, in only a portion is the outcome truly in doubt. Hence in terms of frequency of invocation, a trial by jury is the exception rather than the rule. None theless, juries play a vital role in American criminal law. Because every defendant has the right to a trial by jury, the likely outcome were the case to go to a jury forms the background to a plea bargain or an agreement to let a judge decide the case without a jury. It is precisely the unusual case in which the jury has to work hard that sets the tone for the system as a whole. As but one example, the refusal of a jury in New York City to convict John Peter Zenger of sedition for printing criticism of British Colonial authorities established practical freedom of the press in that era and context even before the American Revolution (Alexander 1963). In each state and in each federal district, law and court rules establish the basic lists from which juries are to be chosen (see Van Dyke 1977, Appendix A). Generally the list of those registered to vote is used, although recently proposals have been made to supplement the voters list with others (Kadanq_and Lehoczky 1976; Kairys, Ka- dane, and Lehoczky 1977). These people are then sum moned to appear on a particular day, to serve for a set length of time. (To avoid large numbers of excusals, some jurisdictions have reduced the duration of service to one day or one trial.) When a case is ready to be heard by a jury, a group of summoned potential jurors are examined— usually by the judge, often by the attorneys as well— to see if they should be excused from the case for cause. The reasons a person may be excused for cause are restricted to those that would directly impair their ability to hear the case impartially, for example being a close relative of the de fendant, the victim, or one of the attorneys, or having a fixed opinion on the defendant’s guilt. In addition, the law provides for peremptory challenges for each side (dif ferent numbers in,.each state, see Van Dyke 1977, Ap pendix D), which can be used by the defense or the pros ecution to eliminate potential jurors without having to explain their reasons. In capital cases jurors are also questioned on their views about the death penalty, and some are excused because of their attitudes toward the death penalty. At one time all jurors w'ho opposed the death penalty were excluded from capital cases. In Witherspoon v. Illinois © Journal of the American Statistical Association September 1983, Volume 78, Number 383 Applications Section K a d a n e : D e a th P en a lty Juries 5 4 5 (1968) the Supreme Court limited this practice. Under Witherspoon the only jurors who may be constitutionally excused from capital cases for cause because of their op position to the death penalty are those who say either (a) that they would automatically vote against the death pen alty in every capital case, regardless of the evidence, or (b) that they could not be fair and impartial in deciding a capital defendant's guilt or innocence. In addition, in some states— notably California— jurors are also ex cused for cause from capital cases if they state that they would automatically vote for the death penalty in every capital case in which they made a finding of first degree murder and aggravated circumstances, regardless of the evidence. This article analyzes the Cafifornia procedure. Commonly in capital cases a single jury decides both (a) the guilt or the innocence of the defendent and (b) if the defendent is convicted of a capital offense, the pen alty. Undercurrent practice jurors who would never vote for the death penalty and jurors who would always vote for the death penalty are excluded from the guilt or in nocence phase of a capital trial even though they say they could fairly and impartially try that issue. The Supreme Court recognized in Witherspoon that this practice might be prejudicial to the capital defendant in the determina tion of guilt or innocence (Witherspoon v. Illinois, p. 520, fn. 18). This article examines whether the practice is in fact prejudicial. 2. RECENT STUDIES AND THE HOVEY DECISION For our analysis, consider each potential juror to be in one and only one of the following four groups: A|t those who say they would not decide the question of guilt or innocence in a fair and impartial manner; A2: those who say they are fair and impartial, and say they would ALWAYS vote for the death penalty, regardless of the facts, if the defendant is found guilty; A3: those who say they are fair and impartial, and say they would NEVER vote for the death penalty, regardless of the facts, if the defendant is found guilty; A4: those who say they are fair and impartial, who say they would consider the death penalty and would SOMETIMES AND SOMETIMES NOT vote for it, depending on the facts, if the defendant is found guilty. There are two other groups for which it is useful to have notation: A5: those who say they are fair and impartial, who say they would consider the death penalty and would AT LEAST SOMETIMES vote for it, if the de fendant is found guilty (A5 "= A2 U A4); A6: those who say they are fair and impartial, and say they would ALWAYS OR NEVER vote for the death penalty (A6 = A2 U A3), if the defendant is found guilty. Under the procedure in effect in California and various other states, the question that the Supreme Court left open in Witherspoon can be restated as follows: does the exclusion from the pool of jurors who could fairly and impartially try a capital defendant’s guilt or innocence of two groups of jurors— the automatic death penalty (or ALWAYS death penalty) group, and the automatic life imprisonment (and hence NEVER death penalty) group— bias the jury pool against the defendant? In 1980, a major presentation citing many studies was made to the California Supreme Court on this question. Two of these studies will be reviewed here in detail be cause they are the most recent and most thorough of their respective types. The first, by Fitzgerald and Ellsworth (1984), studies the relationship between attitudes toward capital punishment and attitudes on other relevant crim inal justice issues. A questionnaire was administered to 811 persons eligible for jury duty in Alameda County, California at random by random digit dialing by the Field Research Corporation of San Francisco, CA. The ques tionnaire asked whether they could decide the question of guilt or innocence fairly and impartially; 717 respond ents indicated that they could. (The remaining analyses are limited to those self-designated fair and impartial re spondents.) They were also asked whether their views on capital punishment were such that they could never vote to impose it, or whether they would vote to impose it in some cases. The results are shown in Table 1. Fitzgerald and Ellsworth summarize these data by find ing that those who would AT LEAST SOMETIMES im pose the death penalty “were consistently more prone to favor the point of view of the prosecution, to mistrust the criminal defendant and his counsel, to take a punitive attitude toward criminals, and to be more concerned with crime control than with due process.” By contrast, those who would NEVER impose the death penalty “tended to be more concerned with mercy, more oriented toward due process, and less mistrustful of the defendant and his legal representative.” These findings confirm those of Bronson (1970) and of Vidmar and Ellsworth (1974) that jurors’ attitudes with respect to whether they would ever impose the death penalty are related to many other juror attitudes of concern to a defendant seeking a fair trial. The second study of major interest here, by Cowan, Thompson, and Ellsworth (1984), reports an experiment relating attitudes toward capital punishment to simulated juror behavior. In this study, 288 adults eligible for jury service in California were shown a videotaped reenact ment of an actual murder trial that occurred in Boston. Thirty-seven subjects were recruited from venire lists of the Santa Clara Superior Court after completing their terms, 218 responded to a newspaper advertisement for volunteers for a study of “how juries make decisions,” and 33 were referred by friends who had seen the ad vertisement. The tape included appropriate instructions to the jury on California law. All the subjects in this study said they would be fair and impartial in deciding guilt or innocence; 30 would NEVER impose capital punishment, the other 546 Jou rn a l o f the A m e r ic a n S tatis tical A ssoc ia tio n , S e p te m b e r 1983 Table 1. The Relationship Between Attitudes of Jurors on the Death Penalty and Defense Bias Items (from Fitzgerald and Ellsworth). Percent of Death Penalty Groups NEVER (A3) and AT LEAST SOMETIMES (A5) Giving Each Answer and the Number Answering Agree Group of Respondent Strongly Agree Somewhat Disagree Somewhat Disagree Strongly Number Answering 1. Better some guilty go free NEVER 32.5 30.0 20.8 16.7 120 ALS 16.1 27.9 27.2 28.9 585 2. Failure to testify indicates NEVER 10.9 12.6 31.9 44.5 119 guilt ALS 16.0 16.3 39.5 28.2 582 3. Consider worst criminal for NEVER 40.2 37.6 10.3 12.0 117 mercy ALS 15.0 29.0 15.5 40.5 575 4. District attorneys must be NEVER 21.2 31.9 32.7 14.2 113 watched ALS 24.0 25.0 26.4 24.6 568 5. Enforce all laws strictly NEVER 22.3 24.0 25.6 28.1 121 ALS 38.1 19.0 25.3 17.6 585 6. Guilty if brought to trial NEVER 14.9 11.6 17.4 56.2 121 ALS 17.2 15.1 17.7 49.9 581 7. Exclude illeaaly obtained NEVER 50.0 13.9 17.2 18.9 122 evidence ALS 38.4 18.1 24.0 19.6 576 8. Insanity plea is a loophole NEVER 27.5 31.7 22.5 18.3 120 ALS 51.5 26.5 13.7 8.3 581 9. Harsher treatment not NEVER 55.0 25.0 14.2 5.8 120 solution to crime ALS 32.7 26.3 17.9 23.1 571 10. Defense attorneys must be NEVER 21.0 43.7 23.5 11.8 119 watched ALS 38.9 34.6 17.4 9.1 581 Unemployment Crime Number Answering 11. More serious problem: NEVER 50.4 49.6 117 Unemployment, crime ALS 37.5 62.5 581 Would Not Consider Would Consider 12. Consider confession NEVER 60.2 39.8 118 reported by news media ALS 49.1 50.9 581 Should not infer Should infer 13. Infer guilt from silence NEVER 86.0 14.0 118 ALS 76.0 24.0 588 NOTE: NEVER = NEVER vote for Death Penalty. ALS = AT LEAST SOMETIMES vote for death penalty. 258 would impose it AT LEAST SOMETIMES. After having seen the videotape, the jurors were asked whether they would vote for conviction of first-degree murder, second-degree murder, or manslaughter, or for acquittal. The data collected suggested that the only real issue in the case was the choice between manslaughter and ac quittal. Over 70 percent of the subjects in each category voted for one of these verdicts on the initial ballot; in a second ballot taken after an hour’s deliberations this fig ure rose to over 80 percent for each group. Accordingly, the focus of the analysis by Cowan, Thompson, and Ellsworth was on the dichotomy in the initial ballot be tween conviction and acquittal, rather than on the pos sible levels of conviction. The results of the study are given in Table 2. Cowan, Thompson, and Ellsworth conclude from the data in Table 2 that jurors in the AT LEAST SOME TIMES group are more likely to convict than are jurors in the NEVER group. Similar experiments are reported by Goldberg (1970), Jurow (1971), and Wilson (1964). The California Supreme Court in its Hovey decision reviewed both the Fitzgerald and Ellsworth attitude study and the Cowan, Thompson, and Ellsworth study of juror behavior in a simulated trial. The court found the evi dence relevant and persuasive except for one fact: both studies included the automatic death penalty group (AL WAYS) with those who would SOMETIMES AND SOMETIMES NOT impose the death penalty, rather than with the automatic life imprisonment group Table 2. Juror Attitudes Toward Capital Punishment and Behavior in a Simulated Trial (from Cowan, Thompson, and Ellsworth). Percent of Death Penalty Groups NEVER (A3) and AT LEAST SOMETIMES (A5) giving Each Verdict, and their Number Acquit Convict Number NEVER 46.7 53.3 30 ALS 22.1 77.9 258 NOTE. NEVER = NEVER vote for death penalty. ALS = AT LEAST SOMETIMES vote for death penalty. K a d a n e : D e a th P en a lty Juries 547 (NEVER). Thus the court remained unconvinced that the exclusion of both the ALWAYS and the NEVER groups results in a bias against the defense, and consequently it did not find the form of death qualification in use in Cal ifornia to be legally flawed. In the remainder of this article the Fitzgerald and Ellsworth study and the Cowan, Thompson, and Ellsworth study are extended to take into account the effect of moving the automatic death penalty group from the AT LEAST SOMETIMES to join the NEVER group in a new group called NEVER OR AL WAYS (A6). 3. STATISTICAL MEASURES'f That groups of people that differ in one respect (here views on the death penalty) differ also in others (attitude on the Fitzgerald-Ellsworth survey items and behavior in the Cowan-Thompson-Ellsworth experiment) is unsur prising. It is apparent in the data from Tables 1 and 2 that those who would consider voting for the death penalty are consistently more favorable to the prosecution than those who would never do so. It is not obvious, however, how strongly each item reveals a more favorable attitude toward the prosecution among the AT LEAST SOME TIMES (A5) group than among the NEVER (A,) group. To be most helpful to the court, I looked tor measures to quantify this difference, so that the court could judge whether the differences shown are sufficiently large to require the imposition of a new procedure for jury selec tion. > There are several sources in the statistical literature for measures of association: the measures proposed and studied by Goodman and Kruskal (1954,1959,1963,1972), the literature on log-linear models, and the nonparametric literature. In this section, I first reorient the data in Tables 1 and 2 so that answers favorable to the defense would be early, and answers favorable to the prosecution late, in the numbering system. Then I introduce the measure I chose, and explain its interpretation and relation to oth ers not chosen. Finally I disuss the statistical model ap propriate to the data of Tables 1 and 2 and apply the chosen measure to these data. Suppose items are indexed with an integer subscript g. Thus in considering the Fitzgerald and Ellsworth data, g - 1, 2, . . . , 13. Also, suppose that item g has ng re sponses. Then for the Fitzgerald and Ellsworth data, nt = n2 = ■■■ = nio = 4, and «n = ni2 = flu = 2. The ng responses can then be reordered if necessary so that re sponse 1 is most favorable to the defense, and response ng most favorable to the prosecution. This requires re versal of items 2, 5, 6, 8, and 10 in Table 1. The data in Table 2 do not require this reordering. Recall that groups holding the views A3 (NEVER) and A5 (AT LEAST SOMETIMES) are of interest here. Sup pose for the moment that the numbers p,-./ were known, where p,-./ is the proportion of jurors holding views A / on capital punishment (/ = 3 or 5), taking position i (1 ii < ng) on an item g. How might these numbers be sum marized in a single number representing how much more favorble to the defense is a typical juror in group A3 by comparison with a juror in group As on item g? One convenient notion, borrowed from nonparametric statistics, is the probability that a random juror from group A3 has a view less favorable to the defense than does a random juror from group As (with ties being re garded as equally likely to go in either direction). The Mann-Whitney Statistic (1947) is an unbiased estimate of this probability (Fraser 1957). Expressed symbolically, Wg = 2 P‘-S (2 Pi'-3 + -P'*3)* (!)I* I 1*36/ As an example, we do the calculations for question 12 or Table 1, interpreting the proportions from the data as if they were probabilities in the entire population. W 12 = Pl.5(P2.3 + P 1.3/2) + P2.5 (P2.3/2) = .491 (.398 + .602/2) + .509(.398/2) = .4445. Thus, for this example, the probability is less than 45 percent that a random NEVER juror is less favorable to the defense than is a random AT LEAST SOMETIMES juror on question 1 2, which relates to consideration of a confession reported in the news media. It is convenient to reexpress Wg in odds form as <t>, = (1 - Wg)/\Vg. In the example <(>12 = .5555/.4445 = 1.25. Thus the odds are 1.25 to 1, or 5 to 4, that a random juror in group A3 (NEVER) is more favorable to the defense on item 12 than is a random juror in group A 5 (AT LEAST SOME TIMES). If 4>k is greater than 1, this would indicate bias against the defense in the AT LEAST SOMETIMES group compared with the NEVER group; if 4>K is less than 1, this would indicate bias against the prosecution; finally, if <bs is exactly 1, this would indicate no bias on item g. (This is why the reorientation of the data is convenient.) The measures studied by Goodman and Kruskal (1963), particularly their measures y and Somers’s (1962) A are possible alternatives. Somers’s measure and the Mann- Whitney W are related by A = 1 - 2\Vg so that d> = (1 + A)/(l - A). Another alternative source of inspiration is the log-linear model literature, in which the work of Haberman (1974) specifically addresses ta bles with ordered classifications. Somers pointed out that his and the Goodman-Kruskal measures differ in their denominators. I found the Somers measure with the Mann-Whitney explanation convenient because its urn interpretation seemed natural to me for this problem. The log-linear approach would have imposed a particular par ametric form on the data, and then would have interpreted the extent to which one group is more favorable to the defense than is the other in terms of a parameter in this 548 Jou rn a l o f th e A m e ric a n Statistical A ssociatio n , S e p te m b e r 1983 distribution. This seemed to me to require extra assump tions that are really extraneous to the essential problem. For these reasons, I am satisfied with the Mann-Whitney- Somers Statistic transformed to odds (<}>). The next issue is an appropriate statistical model for the data reported in Tables 1 and 2, now taking into ac count that the numbers pu are not known with certainty. Two important alternatives are to consider the data as jointly multinomial, or to consider them independently multinomial, row by row. If one believed that the Fitz- gerald-Ellsworth sample of Alameda County residents were typical of California potential jurors in the propor tion of NEVER, and AT LEAST SOMETIMES views among them, the former would be appropriate. Since Ala meda County, containing Berkeley and Oakland, has the reputation of being more liberal than California as a whole, this assumption would strain credibility. The same assumption for the Cowan-Thomspon-Ellsworth experi ment, involving volunteers, would be even harder to jus tify. However, assuming a row-by-row multinomial dis tribution means assuming only that the views of the NEVER (AT LEAST SOMETIMES) jurors in the Fitz- gerald-Ellsworth study and their behavior in the Cowan- Thompson-Ellsworth experiment are typical of those groups in California. This is a more reasonable assump tion, and consequently 1 adopt it. Under this model, maximum likelihood estimates for 4> and an asymptotic standard error for it can be calcu lated. The former consists simply of substituting sample proportions for p,-.5 and p,-,3 in (1). Since the latter is a special case of a more complicated computation de scribed later, I defer discussion of it. Table 3 reports maximum likelihood estimates and asymptotic standard errors for each of the Fitzgerald- Ellsworth questions reported in Table 1. It shows that on each item, the odds are that a juror who would consider imposing the death penalty AT LEAST SOMETIMES (ALS) will be more favorable to the prosecution than one who would NEVER consider imposing the death penalty. The magnitudes of the odds of bias range from a modest finding of 1.131 on item 4, to very sizable odds on several other items. This analysis confirms and strengthens the Fitzgerald-Ellsworth conclusions about Table 1. Table 4. (Cowan, Thompson, and Ellsworth 1983). How Much More Likely to Vote for the Defense is the NEVER Group Than is the AT LEAST SOMETIMES 6 = 1.652 SD(6) = .265 (<t> - 1)/SD(<5) = 2.45 Pr{<t> « 1} = 7.1 x 1 0 '1 2 3 4 A discussion of the philosophy behind the computation of the fourth column is necessary. In this problem, the parameters are the p,/s. There is a function of them, <f>, of particular interest and a special value of 4>, namely 1, of substantive concern. One classical method would be to test cj> = 1 against 6 > 1 at, say, the .05 level. If (4> - 1)/5D(4>) is greater than 1.645, significance at the .05 level is achieved. To say that <5 is significantly dif ferent from 1 at the .05 level says that if the system were exactly nondiscriminatory, the probability is less than 5 percent that <t> would be as large, or larger, than the value observed. A Bayesian approach can be based on a theorem of Walker (1962), showing that the asymptotic posterior dis tribution of c}> is normal, with mean <t> and standard de viation SD{4>). This result does not depend on the par ticular prior probability distribution in the parameter space chosen for the analysis, as long as it is smooth and has positive probability everywhere. The probability that <{> is greater than 1 can now be calculated in a straight forward way. The fourth column of Table 3 can thus be interpreted either as the significance level of a one-tailed test of <J> = 1 against the alternative 4> > 1, or as the probability that 4) < 1. I find the Bayesian interpretation more responsive to the legal question of the probability that the currently used juror selection procedure discriminates against the defense: the Bayesian analysis attempts to answer the question at hand, while the classical testing approach does not. Table 4 gives a similar analysis of the Cowan-Thomp- son-Ellsworth experiment. Again the results indicate those who would NEVER impose capital punishment to Table 3. (Fitzgerald and Ellsworth 1983). How Much More Likely to Favor the Defense is the NEVER Group Than is the AT LEAST SOMETIMES Grobp? Item j> ' SD(4>) (i> - lyS D ft) Pr(i> s 1) 1. Better some guilty go free 2. Failure to testify indicates guilt 3. Consider worst criminal for mercy 4. District attorneys must be watched 5. Enforce all laws strictly 6. Guilty if brought to trial 7. Exclude illegally obtained evidence 8. Insanity plea is a loophole 9. Harsher treatment not solution to crime 10. Defense attorneys must be watched 11. More serious problem: unemployment, crime 12. Consider confession reported by news media 13. Infer guilt from silence 1.642 .124 5.16 1.427 .081 5.26 2.456 .160 9.12 1.131 .085 1.55 1.448 .091 4.93 1.149 .063 2.38 1.224 .085 2.62 1.796 .134 5.94 1.836 .122 6.84 1.451 .115 3.93 1.296 .093 3.20 1.250 .078 3.19 1.220 .047 4.67 1.3 X 10“ 7 8 X 10“8 7.5 X 10-20 6.1 X 10” 2 4.2 X 1 0 -7 8.8 X 1 0 -3 4.4 X 1 0 -3 1.5 X 1 0 -9 10 11 4.0 X i o 12 13 4.3 X 1 0 -5 6.9 X 10^" 7.1 X i o - * 1.5 X 10 6 K a d a n e : D e a th P e n a lty Juries 5 4 9 be more lenient to the defense than arc those who would impose it AT LEAST SOMETIMES, confirming the Cowan-Thompson-Ellsworth conclusions. 4. SURVEYS MEASURING THE SIZE OF THE DEATH PENALTY GROUPS In order to conduct a further analysis of the Fitzgerald and Ellsworth attitude survey and the Cowan-Thompson and Ellsworth experiment, taking into account the eftect of the exclusion of the automatic death penalty group, it is necessary to have data on the proportion of people in the population holding various attitudes toward the death penalty. In particular, it is necessary tb have data on the size of the automatic death penalty group (ALWAYS). Two recent surveys, which contain the needed data, form the basis for the analysis presented in this article. The first survey was conducted by the Field Research Corporation in March 1981 for the National Council on Crime and Delinquency. In this survey some 1,014 face- to-face interviews were conducted in a cluster sample of households. The respondents were asked about their gen eral views on the death penalty, whether they would fairly and impartially decide the question of guilt or innocence, and whether in the guilt phrase they would NEVER favor the death penalty, or whether in AT LEAST SOME cases they would consider voting for it. Field describes its sample as follows: “The sample presently consists of 200 primary sampling-point clusters. These primary sampling units enter the sample with a probability of selection in proportion to the population of California counties. Specific cluster locations are deter mined by random selection of key addresses, using cur rent telephone directories as the initial sampling frame within counties. Households in a given cluster are con secutively listed with a procedure to assure that the in terviewers exert no influence on the selection of house holds. This procedure also draws non-telephone homes into the sample and permits telephone-density bias to be removed.” The results of the interviews are then weighted by giv ing each cluster of interviews a weight inversely propor tional to the density of listed telephones found in that cluster. A second-stage weighting by age and sex within geographic area is done using Census data. The weighted results are given in Table 5. The data indicate that 11.3 percent of the total adult population of California— which amounts to 12.6 percent of those in the adult population who could fairly and impartially try the guilt or innocence Table 5. Attitudes of Californians on the Death Penalty (Field Survey) Would be fair and impartial in deciding guilt or 89.8% innocence NEVER 11-3% AT LEAST SOMETIMES 78.2% Don't know/no answer/refused -4% Would not be fair and impartial 7.9% Don't know/no answer/refused 2.2% of a capital defendant— would be excluded from jury ser vice in capital cases because they would never consider voting for the death penalty. The Field survey does not permit disaggregation of the AT LEAST SOMETIMES imposers of the death penalty into the ALWAYS group and the SOMETIMES AND SOMETIMES NOT group. For this, I turn to a Harris survey conducted in January 1981 by a national telephone poll, with 100 clusters (telephone exchanges) stratified by geographic region and metropolitan/nomnetropolitan res idence. The respondents were asked whether they strongly favored capital punishment, or whether there were some (or many, or most, or all) cases in which it is legally possible but not appropriate. Since only those who had said they were strongly in favor of the death penalty had given answers consistent with the ALWAYS posi tion, only they were asked whether they would be fair and impartial in determining guilt or innocence in a capital case. Those who said they would be fair and impartial were asked whether they would ALWAYS vote to im pose the death penalty, or whether they would consider the evidence, and hence be in the SOMETIMES AND SOMETIMES NOT category. Of the 1,499 people ques tioned, 13 were in the ALWAYS group. After reweighting for age, sex, and race, the data support an estimate that 1 percent of the adult American population falls into the automatic death penalty (ALWAYS) group. The Harris results are analyzed by region (East, Midwest, South, West), and area (city, suburbs, towns, rural), as well as by age, education, sex, race, income, type of work, re ligion, and union membership. No substantial regional differences in general attitudes toward the death penalty are revealed by the data. There are two other sources cited in the Hovey decision (fn. Ill) for indications as to the size of the automatic death penalty (ALWAYS) group. Jurow (1971) studied the views of 211 volunteer subjects who were employees of Sperry Rand. His subjects were more conservative than the population of New York City at the time. Of these, five answered that in a jury vote to determine the penalty for a serious crime, “I would always vote for the death penalty in a case where the law allows me to.” This finding of 2.37 percent is not inconsistent with the 1981 Harris survey finding of 1 percent ALWAYS jurors since (a) those not fair and impartial were not eliminated and (b) a more conservative than average, volunteer popu lation at one company cannot be taken to be represent ative of the national population. The second study cited by the California Supreme Court in Hovey is Smith (1976), which relies on a 1973 Harris survey. Here 28 percent of the respondents an swered that all persons convicted of first-degree murder should receive the death penalty. However, this question does not distinguish between people’s views on what the law should be and how they would vote as jurors. Jurow (1971), for one, found 21 percent of his population strongly favored the death penalty, while only 2.37 per cent would impose it in every case. Additionally, the Har ris 1973 question— unlike the 1981 question— does not 550 Jou rn a l of the A m e r ic a n S tatis tical A ssoc ia tio n , S e p te m b e r 1983 tell the respondents that it would be their obligation as jurors to follow the law, and that the judge would instruct them that the law requires them to consider all penalties and to weigh all the evidence before making up their minds. Thus the 1973 Harris survey cited by Smith did not ask the relevant question to determine the size of the ALWAYS group. Consequently, we use the 1981 Harris survey number of 1 percent because, unlike Jurow, it was based on a national probability sample, and, unlike the 1973 Harris survey, it asked the relevant set of questions. 5. ASSUMPTIONS AND A MODEL It is entirely possible that an analysis will reveal that on some questions the ALWAYS OR NEVER group is more favorable to the defense than the SOMETIMES AND SOMETIMES NOT group, and that on other ques tions the reverse might be true. In order to learn whether this is the case, the findings of the Cowan, Thompson, and Ellsworth experiment, and of each of the 13 attitude questions in the Fitzgerald and Ellsworth study are con sidered separately. The first assumption I make is that the Field survey and the Harris survey are random samples from the pop ulation of jurors in question. Thus if pi = Pr(A,), the Field survey contributes a likelihood function pJ'pJ'pJ', (2) where // is the weighted number of persons found in the Field survey to be in category A/ (/ = 1,3, 5). Similarly the Harris survey contributes a likelihood Pi'12 ~ Pi)h\ (3) * where h2 is the weighted number of persons in the Harris survey to be in category A2, and h7 are the others in the Harris survey. In making this assumption, I am relying on the geographical homogeneity found in both surveys to indicate that Californians are very similar to other Americans in their views, and on the expertise of the Field and Harris organizations in doing their surveying properly. My second assumption is that the attitude survey and the experiment constitute random samples from the pop ulations studied (A3 and A5). Thus ifj\ persons in group A3 take position i on item g, and A,- persons in group A5 take the position /(I < / < ;iA.), the contribution to the likelihood is nit n P i j p i . s ki (4)i= 1 One need not assume that the number of subjects in groups A3 and A 5 in the Cowan, Thompson, and Ells worth and the Fitzgerald and Ellsworth studies are typ ical of the population, but only that the views of the sub jects in groups A3 and A5 are typical of the views of those groups in the population at large. Finally, some assumption is necessary about the views ot the group A2, since these were not directly measured. The harsher these views are to the defense, the more difficult the defense’s task of showing that exclusion of the whole group A6 = A 2 U A 3 is disadvantageous to them. Accordingly Ltake the most conservative stance on this issue. Those in group A2 are assumed, on each item g, to take the position rif, most opposed to the de fense’s interests. Inevitably, this assumption is, to some extent, false; to the extent that it is false, my estimates (J> will be too low. To the extent that bias against the defense is shown using this assumption, the real extent of the bias against the defense is greater by some amount. In summary, the likelihood function for the data is nK % = p \ f< p i f i P 2h2 (i - P i ) h j n p j ‘p i .s*'. wi-1 where /?,.5 = (p2 Pi.i + Pa Pi.A)/{p2 + Pa), and where p,.2 = 0 (iV ng) and pnil.2 = 1. The maximum likelihood estimates are found in closed form by straightforward maximization of (5). The com putation of the asymptotic standard deviation is equally straightforward, but considerably more tedious. The prin cipal danger is that one might forget the Goodman-Krus- kal (1972) admonition not to change parameterization in midcomputation. An Appendix giving the details of both computations is available on request from the author. 6. IMPLEMENTATION AND RESULTS Before implementing the model in the last section, sev eral details must be discussed about how to relate the model of Section 5 to the surveys of Section 4. First, the “don’t know/no answer/refused” groups were distributed among the other categories in proportion to their size. After doing this, the Field survey numbers fu /3, and / 5 are taken to be proportional to .0808, .1162, and .8040, respectively. The Harris numbers h2 and h7 are taken from the weighted analysis to be proportional to .01 and .99, respectively. The question of effective sample size arises in both studies, because both used cluster sampling (see Kish 1965). Hence the actual sample sizes, 1,014 for the Field survey and 1,499 for the Harris survey, may overstate the amount of information available. The Field survey reports the results of six replicates of approximately egual size, weighted as were the results reported in Table 4. For NEVER, the replicate standard error is .0 1 1, compared with the standard error computed as if simple random sampling had occurred, which is [(.113)(.887)/1014]l/2 = .0099. Thus the ratio is.011/.0099 s 1.1. Similarly from the AT LEAST SOMETIMES group, the reported standard error is .0 1 0, which gives a ratio of 1.3. Finally the NOT FAIR AND IMPARTIAL question has a standard error of .088, yielding a ratio of 1.06. I take the median of these numbers, 1.1, as repre sentative of them, and hence use the effective sample size for the Field survey of 1,014/(1.1)2 = 838. In the Harris survey, each of the 13 in the ALWAYS group was found in a different cluster. This cannot be K a d a n e : D e a th P en a lty Juries 551 taken as evidence of a correlation due to being in the same cluster, so I use the unadjusted sample size, 1,499, as the effective sample size for the Harris survey. 6.1 The F itzgera ld a n d Ellsworth Study Using these effective sample sizes, values for 4>, SD(4>), (<{> - 1)/5D(4>), and Pr{6 < 1} were calculated for each item in the Fitzgerald and Ellsworth study, and are recorded in Table 6. In analyzing the results of Table 6, recall the assumption that the ALWAYS group take the worst position on each question from the defense viewpoint. To the extent that this extreme assumption is wrong, the numbers of <f> recorded in "fable 6 are too low. I propose a working assumption that estimated odds (4>) closer to 1 than .1 reveal an item close enough to nondiscriminatory that it need not be worried about. These items are those closer to even than 11 to 10 on the high side, and 9 to 10 on the low side. This eliminates four items, numbers 4, 6, 7, and 13. All nine of the others display bias against the defense, rising to estimated odds of 2 to 1 for item 3. I conclude from Table 6 a showing of substantial bias against the defense from the current procedure. 6.2 The C ow an , Thompson, a n d Ellsworth Study The showing of substantial bias in the Fitzgerald and Ellsworth study is confirmed by the reanalysis ot the Cowan, Thompson, and Ellsworth experiment. As shown in Table 7, <f> is 1.519 so the estimated odds of a NEVER or ALWAYS juror being more favorable to the defense than a SOMETIMES AND SOMETIMES NOT juror is more than 3 to 2. Hence the probability of neutrality or bias against the prosecution (4> ̂ 1) is 1.3 percent. Again we have a finding of substantial bias against the defense. What makes all the results true is that the ALWAYS group is so small (1% of the population) that the AL WAYS or NEVER group is dominated by the NEVER part, and the AT LEAST SOMETIMES group is domi nated by the SOMETIMES AND SOMETIMES NOT majority. Even attributing the least favorable views to the ALWAYS group does not disturb the finding of sub- Table 7. (Cowan, Thompson, and Ellsworth 1983). How Much More Favorable to the Defense are the Votes of the ALWAYS OR NEVER Group (A6) Compared to the SOMETIMES AND SOMETIMES NOT (Aa) Group?____________ <j> = 1.519 SD(6) = 229 (<j> - 1)/SD(<j>) = 2.27 Pr(<t> « 1) = 1.3% stantial bias against the detense. Ot course, this extreme assumption does reduce the estimated odds in Table 6 compared with those in 1 able 3, and in Table 7 compared with those in Table 4, as arithmetically it must. 7. CONCLUSION The Cowan-Thompson-Ellsworth experiment, because it deals with a simulated juror vote, is the evidence most closely related to actual juror behavior. Consequently, I would pay most attention to the odds (about 3 to 2) of a more proprosecution vote in that experiment among the SOMETIMES AND SOMETIMES NOT group than among the ALWAYS OR NEVER group. This conclu sion is strengthened by the finding that 9 of the 13 attitude questions from the Fitzgerald-EHsworth study show the same favoritism to the prosecution among the SOME TIMES AND SOMETIMES NOT group compared with the ALWAYS or NEVER group. (The other four attitude questions show essentially no difference.) This conclu sion is further strengthened by the tact these calculations are conducted under the extreme assumption that those who would ALWAYS impose the death penalty, but would reach the question of guilt fairly and impartially, would unanimously vote to convict in the experiment, and would unanimously take the most proprosecution view of each attitude question. Therefore there appears to be a distinct and substantial antidefense bias in the exclusion for cause of the ALWAYS OR NEVER group from the jury deciding guilt or innocence. [Received November 1981. Revised April 1983.] Table 6 (Fitzgerald and Ellsworth 1983). How Much More Favorable to the Defense is the ALWAYS OR NEVER Group (A&) Than is the SOMETIMES AND SOMETIMES NOT (A*) Group?________ Item 1. Better some guilty go free 2. Failure to testify indicates guilt 3. Consider worst criminal for mercy 4. District Attorneys must be watched 5. Enforce all laws strictly 6. Guilty if brought to trial 7. Exclude illegally obtained evidence 8. Insanity plea is a loophole 9. Harsher treatment not crime solution 10. Defense Attorneys must be watched 11. More serious problem: unemployment, crime 12. Consider confession reported by news media 13. Infer guilty from silence SD<8) (6 - 1)/SD(i>) Pr(i> « 1) .132 2.89 1.9 x 10~1 2 3 4 .108 1.75 4.0 x 10~2 .174 5.78 4.0 x 1 0 '9 10 11 12 13 .101 -.2 6 6.0 x 1 0 '1 .106 2.38 8.7 x 1 0 '3 .089 -.2 6 6.0 x 10"’ .100 .40 3.4 x 1 0 '' .133 4.18 1.6 x 1 0 '5 6 7 8 .143 3.54 2.0 x 10“ “ .120 2.13 1.7 x 1 0 '2 .091 2.02 2.2 x 1 0 '2 .086 1.42 7.8 x 10"2 .085 .52 3.0 x 10“ ' 1.380 1.189 2.006 .974 1.252 .977 1.040 1.558 1.506 1.256 1.184 1.121 1.044 Journal of the American Statistical Association, September 1983 REFERENCES ALEXANDER, JAMES (ed.) (1963), A Brief Narration o f the Case and Trial o f John Peter Zenger, Cambridge, Mass.: Harvard University Press. BRONSON, E.J. (1970), “On the Conviction-Proneness and Repre sentativeness of the Death-Qualified Jury: An Empirical Study of Col orado Veriremen,” 42 U. Colo. L Rev. 1. COWAN, CLAUDIA, THOMPSON, WILLIAM, and ELLSWORTH PHOEBE C. (1984), “The Effects of Death Qualification on Jurors’ Predisposition to Convict and on the Quality of Deliberation,” Law and Human Behavior, to appear. FITZGERALD, ROBERT, and ELLSWORTH, PHOEBE C. (1984). “ Due Process vs. Crime Control: Death Qualification and Jury At titudes," Law and Human Behavior, to appear. FRASER, D.S. (1957), Nonparametric Methods in Statistics, New York: John Wiley, 268 , 269. GOLDBERG, F. (1970), “Toward Expansion of Witherspoon: Capital Scruples, Jury Bias, and the Use of Psychological Data to Raise Pre sumption in the Law," Harvard Civil Rights and Civil Liberties Law Review, 5, 53. GOODMAN, L.A., and KRUSKAL. W.H. (1954), “ Measures of As sociation for Cross Classifications,” Journal o f the American Statis tical Association, 49, 732-764. -------- (1959), “ Measures of Association for Cross Classifications II,” Further Discussion and References, Journal o f the American Statis tical Association, 54, 123-163. -------- (1963), “ Measures of Association for Cross Classifications III. Approximate Sampling Theory,” Journal o f the American Statistical Association, 58, 310-364. -------- (1972), “ Measures of Association for Cross Classifications IV. Simplification of Asymptotic Variances,” Journal of the American Statistical Association, 67, 415-521. HABERMAN, S.J. (1974), “ Log-Linear Models for Frecjuency Tables with Ordered Classifications,” Biometrics, 30, 589-600 HOVEY V. SUPERIOR COURT (1980), 28 Cal. 3d 1, 616 P.2d 1301. JUROW, G.L. (1971), “ New Data on the Effect of a Death-Qualified Jury on the Guilt Determination Process," Harvard Law Review 84 567. KADANE, J.B., and LEHOCZKY, J.P. (1976), “ Random Juror Se lection from Multiple Lists,” Operations Research, 24 207-219 KAIRYS, D„ KADANE, J.B., and LEHOCZKY, J.P. (1977), “ Jury Representativeness: A Mandate for Multiple Source Lists,” Califor nia Law Review, 65, 776-827. KISH, L. (1965), Survey Sampling, New York: John Wiley. MANN, H.B., and WHITNEY, D.R. (1947), “ On a Test of Whether One of Two Random Variables Is Stochastically Larger than the Other,” Annals o f Mathematical Statistics, 18, 50-60. SMITH, T.W. (1976), "A Trend Analysis of Attitudes Toward Capital Punishment, 1936-1974," in Studies in Social Change Since 1948, ed. James A. Davis, NORC Report 127B Chicago: National Opinion Re search Center. SOMERS, R.H. (1962), “A New Asymmetric Measure of Association for Ordinal Variables,” American Sociological Review, 27, 799-811. VAN DYKE, J.M. (1977), Jury Selection Procedures, Cambridge, Mass.: Ballinger. VIDMAR, N„ and ELLSWORTH, PHOEBE (1974), “ Public Opinion and the Death Penalty,” Stanford Law Review, 26. WALKER, A.M. (1962), “On the Asymptotic Behavior of Posterior Distributions,” Journal o f the Royal Statistical Society, Ser. B 31 80-89. WILSON, W.C. (1964), “ Belief in Capital Punishment and Jury Per formance,” University of Texas, unpublished. WITHERSPOON V. ILLINOIS (1968), 391 U.S. 510. V i * f