The opinion of the court was delivered by: BERNARD WEISBERG
Petitioner James Free was convicted on June 22, 1979 in the Circuit Court of Du Page County for the April 24, 1978 murder and attempted rape of Bonnie Serpico and the attempted murder and attempted rape of Lori Rowe. The facts of the crime do not concern us here. The prosecutor requested a hearing to determine whether the death penalty should be imposed, and a two-stage hearing was held in accordance with Illinois law. The jury unanimously found beyond a reasonable doubt that the victim was murdered in the course of a burglary and rape, aggravating factors permitting the imposition of the death penalty. After hearing evidence in aggravation and mitigation, the jury found no mitigating factors that precluded the imposition of the death penalty and on August 7, 1979 Free was sentenced to death.
On January 24, 1983, the Illinois Supreme Court affirmed Free's conviction and sentence. People v. Free, 94 Ill.2d 378, 447 N.E.2d 218, 69 Ill. Dec. 1 (1983), cert. denied, 464 U.S. 865 (1983). Free filed two separate petitions for post-conviction relief in the Illinois state courts. Each petition was dismissed and the Illinois Supreme Court affirmed each dismissal. People v. Free, 112 Ill. 2d 154, 492 N.E.2d 1269, 97 Ill. Dec. 396 (1986), cert. denied, 479 U.S. 871 (1986); People v. Free, 122 Ill.2d 376, 522 N.E.2d 1184, 119 Ill. Dec. 325 (1988), cert. denied, 488 U.S. 872 (1988). Having exhausted all available state remedies, Free filed this habeas petition raising 21 separate grounds for relief. The state has stayed Free's execution pending the final disposition of this petition.
In an opinion dated November 5, 1991, Free v. Peters, 778 F.Supp. 431, Judge Aspen dismissed all but three grounds for relief. The remaining claims, Grounds 5, 10 and 14, attack the Illinois Death Penalty Act and jury instructions. Free contends that those instructions provide constitutionally inadequate guidance to capital sentencing jurors, specifically, that they confuse the jury on five important issues: (1) whether jurors must unanimously agree on the existence of mitigating factors, (2) whether jurors may consider mitigating factors not mentioned by the court as reasons not to impose death, (3) which party bears the burden of persuasion, (4) whether the existence of a mitigating factor mandates a non-death verdict and (5) whether the jury must unanimously agreed as to its ultimate verdict of life or death.
Judge Aspen noted that the constitutionality of the Illinois statute and jury instructions had been upheld in Silagy v. Peters, 905 F.2d 986 (7th Cir. 1990), cert. den. U.S. , 111 S. Ct. 1024) 112 L. Ed. 2d 1106 (1991), and Williams v. Chrans, 742 F.Supp. 472 (N.D. Ill. 1990)(Aspen, J.), aff'd 945 F.2d 926(7th Cir. 1991), cert. denied, 1992 U.S. LEXIS 3792, 60 U.S.L.W. 3858 (U.S. 1992). In Silagy, the Seventh Circuit rejected arguments that the Illinois death penalty statute and jury instructions unconstitutionally impose a presumption in favor of death (Free's Ground 5) and shift the burden of proof to the defendant to overcome that presumption (Free's Ground 6). Silagy, 905 F.2d at 997-1000. Williams rejected the related argument that the failure of the Illinois sentencing scheme to assign a specific standard of proof as to the ultimate issue renders the scheme unconstitutional (Free's Ground 14). Williams, 742 F. Supp. at 499-500. Williams additionally rejected the argument that the statute is unconstitutionally vague and fails to narrowly channel the sentencing authority's discretion, creating the impermissible risk that the death sentence will be imposed arbitrarily and capriciously (Free's Ground 10). Id. at 500.
Nevertheless, Judge Aspen reserved ruling on these claims because Free had offered the results of a juror study conducted in April 1990 by Professor Hans Zeisel which, according to Free, shows that jurors are not properly guided by the semantics of the statute. Judge Aspen believed that the study results,
if statistically reliable and unbiased . . . call into question the empirical assumptions as to juror comprehension which served as the predicate to the rulings in [Silagy and Williams ].
Thus, the question we face is whether it is appropriate to consider such evidence as a means of supplanting those prior rulings. Although Free cites no case law on the question, he makes the compelling argument that a finding, which turns on how jurors comprehend the law they are told to apply, should be based if possible on substantial evidence rather than judicial speculation.
Free, 778 F.Supp. at 434-35.
Judge Aspen directed this court to hold an evidentiary hearing to consider the validity of the Zeisel study and its applicability to these three grounds for relief. A hearing was held on January 13, 14 and 15, and February 4, 5 and 7, 1992. The court heard the testimony of experts in empirical jury studies, statistics, survey methodology and linguistics. During the recess Zeisel carried out a second jury survey which was admitted in evidence. The parties filed a post-trial briefs and proposed findings of fact and conclusions of law.
This Report will address the issues before us in the following manner. In Part 1 we will review Illinois' procedure for imposing the death penalty. Part 2 summarizes Free's claims which have been referred here and the constitutional requirements invoked in this case. Part 3 describes the Zeisel studies (sometimes called the Illinois Capital Jury Surveys). Part 4 assesses the challenges to their validity and reliability. In Part 5 we discuss whether several of Free's claims have been procedurally defaulted on the ground urged by respondents that the Zeisel studies were available to Free during the period of his state court post-conviction proceedings. Finally, Part 6 analyzes the impact of the empirical evidence on the pertinent claims in Free's petition.
1. The Illinois Death Penalty Act
Under the Illinois Death Penalty Act, the prosecution may request a capital sentencing hearing after the defendant pleads guilty to or is convicted of murder. Ill. Rev. Stat. (1979) ch. 38 P9-1(d). In the first phase of the hearing the state must prove beyond a reasonable doubt that the defendant is eligible for the death penalty: that the defendant was at least eighteen years old at the time of the offense and that one or more of seven statutory aggravating factors exist.
If the jury unanimously finds that the defendant is eligible for the death penalty, the second phase commences. The state presents evidence of any aggravating factors, statutory or otherwise, and the defense presents mitigating circumstances. The statute lists five potentially relevant mitigating factors, but only by way of example.
The defendant may present any aspects of the defendant's character or record, and any of the circumstances of the offense that militate against the imposition of the death sentence. The criminal rules of evidence are relaxed in this phase. The following statutory language provides the standard by which the jury is to assess the mitigating and aggravating factors:
If the jury determines unanimously that there are no mitigating factors sufficient to preclude the imposition of the death sentence, the court shall sentence the defendant to death.
Unless the jury unanimously finds that there are no mitigating factors sufficient to preclude the imposition of the death sentence the court shall sentence the defendant to a term of imprisonment under Chapter V of the Unified Code of Corrections.
Ill. Rev. Stat. (1979) ch. 38, P9-1(g). The death sentence is imposed only if all jurors agree that the mitigating factors are insufficient to preclude a death sentence. If the jurors unanimously agree that no such mitigating factors exist, then the court must impose a death sentence. However, a juror may find a mitigating factor even if the defendant presents no evidence at the sentencing hearing.
The Illinois Supreme Court has held that the statute requires the jury to weigh aggravating and mitigating factors. People v. Brownell, 79 Ill. 2d 508, 38 Ill. Dec. 757, 770, 772, 404 N.E.2d 181, 194, 196 (1980). That court has also said that the defendant has the burden of persuasion on the question of whether sufficient mitigating circumstances exist to preclude the imposition of the death penalty. People v. Bean, 137 Ill. 2d 65, 560 N.E.2d 258, 291, 147 Ill. Dec. 891, 924 (1990). On the other hand, in the same opinion, the court said, "we note that the defendant does not alone have a burden of persuasion at this balancing stage, for the State is the movant, the party seeking the death penalty, and so bears the primary burden of persuading the jury that, as the statute states, there are no mitigating factors sufficient to preclude the sentencer from imposing the sentence of death for which the defendant is eligible." Id., 147 Ill. Dec. at 925, 560 N.E.2d at 292 (emphasis added). The court then added:
In Silagy, the Seventh Circuit agreed with the Illinois Supreme Court's reading and held that this aspect of the statute satisfies the Eighth and Fourteenth amendments. Silagy, 905 F.2d at 998-1000.
Free's Ground Five, Petition PP62-65, claims that the Illinois Death Penalty Act and jury instructions create a presumption in favor of death. He maintains that although Blystone v. Pennsylvania, 494 U.S. 299 (1990), and Silagy held that such a presumption is not unconstitutional, the instructions are defective because they leave the jury without guidance as to which side has the burden of proof and what standards to apply in considering the aggravating and mitigating evidence. Petitioner's Pretrial Br. at 12-15.
The Tenth Ground for relief, Petition PP89-93, claims that the statute and the instructions in this case are unconstitutionally vague and fail to provide clear and objective standards to guide the jury in its sentencing determination. Petitioner's Reply filed September 7, 1990 at 33-34 contends that the key instruction is vague and incomprehensible and did not provide the jury with sufficient guidance concerning 1) whether they must unanimously find the existence of a mitigating factor before any one juror can consider it; 2) whether a mitigating factor must be substantially the same or similar to ones listed by the judge; 3) whether one mitigating factor must singly outweigh all aggravating factors in order to preclude the imposition of the death penalty; and 4) whether the defendant's failure to prove that mitigating factors exist mandates imposition of the death penalty.
Free's Fourteenth Ground claims that the statutory scheme is unconstitutional because it is devoid of any sentencing guidance concerning the burden of proof as to the ultimate issue of life or death. Petition PP106-10.
Respondents contend that, with two exceptions, Free's claims are procedurally defaulted because they were not fairly presented to the Illinois courts. That contention was rejected by Judge Aspen, Free, 778 F. Supp. at 435 n. 3, and it is inappropriate to reconsider that ruling here. However, respondents also attack Judge Aspen's alternative ruling that even if Free's claims in Grounds Ten and Fourteen are procedurally defaulted, that bar must be avoided because Free has shown cause, i.e., that the Zeisel survey evidence was unavailable to Free at the time of his prior state court petitions. Since respondents' attack is based on evidence developed at the hearing, we defer discussion of that issue until after our description of the Zeisel survey evidence.
We summarize briefly the constitutional requirements invoked by Free. Imposition of the death penalty may not be capricious or arbitrary and to that end the discretion of the sentencer must be channeled and controlled. At the same time the unique nature of the death penalty requires that the sentencer be able to consider anything that might plead in favor of clemency. Thus, the sentencer's discretion must be channeled by clear and specific and detailed guidance which make the process for imposing a sentence of death rationally reviewable, Godfrey v. Georgia, 446 U.S. 420, 428 (1980). The sentencer must be free to consider any potentially mitigating circumstances and therefore cannot be limited to mitigating factors enumerated in a statute or judicial instructions. Lockett v. Ohio, 438 U.S. 586, 604 (1978). A sentencing jury must be free to consider any mitigating circumstance; thus, requiring the jury to agree unanimously on the existence of a mitigating factor before it may be considered by any juror violates the Eighth Amendment. Mills v. Maryland, 486 U.S. 367 (1988); McKoy v. North Carolina, 494 U.S. 433 (1990).
3. The Illonois Capital Jury Surveys
In 1990 and 1992 Hans Zeisel, a pioneer and leading expert in empirical studies of jury behavior, carried out surveys of comprehension of jury instructions based on the Illinois capital sentencing statute. The surveys were given to a total of 238 persons who had been summoned for jury duty at the Daley Center in Chicago. With the approval of the Chief Judge of the Circuit Court of Cook County, jury officials assigned panels of prospective jurors already assembled for courtroom jury service to courtrooms in which the surveys were conducted.
The surveys (1) identified and excluded survey respondents who would always vote for the death penalty or would be subject to challenge under Witherspoon v. Illinois, 391 U.S. 510 (1968), because of personal opposition to the death penalty, (2) summarized the facts in the Free case with the names changed, (3) set out in the 1990 survey the 1987 Illinois Pattern Instructions (IPI) for the second phase of capital sentencing proceedings and in the 1992 survey the actual instructions given to the jury in the Free case, (4) asked sixteen questions designed to test the survey respondents' comprehension of those instructions by asking whether hypothetical jurors making certain judgments had followed the instructions and (5) asked whether the survey respondents would lean for or against the death penalty in the case described to them.
The surveys were designed to test juror comprehension of five issues.
1. Whether a jury must unanimously agree on the existence of a mitigating factor before that factor can be considered by an individual juror (questions 4 and 5).
2. Whether the, jury is free to consider mitigating factors not enumerated in the judge's instructions (questions 1, 2, 6, 7, 8, 9, 10 and 11).
3. Whether the existence of a mitigating factor bars a sentence of death (question 3).
4. Who has the burden of proof on the appropriate sentence (questions 12, 13, 14 and 15).
5. Whether a jury which is divided on the penalty issue should return a verdict or tell the judge they cannot reach a unanimous verdict (question 16).
The survey, questions are attached as Appendix A to this Report.
Free contends that the survey results show high levels of misunderstanding of the IPI Instructions and the substantially similar instructions given to the jury in this case and that those instructions therefore did not provide constitutionally adequate guidance to channel the jury's discretion. He points to the following results:
1. Between 15% and 36% (in this Report all fractions above .5 are rounded to the next highest integer) of the respondents answered questions 4 and 5 incorrectly, indicating their belief that all members of the jury must agree on the existence of a mitigating factor before any juror may consider that factor, contrary to Mills v. Maryland and McKoy v. North Carolina.
2. Between 39% and 68% of the respondents answered questions 2, 7, 8, 9, 10 and 11 incorrectly, indicating their belief that the jury is not free to consider mitigating factors not enumerated in the judge's instructions, contrary to Lockett v. Ohio. (Although question 1 was also designed to test comprehension on this issue, the answers to that question have been disregarded here and, where so indicated, elsewhere in this Report since Free's witnesses conceded that it does not have a clearly correct answer. Transcript (T.) 83-84, 135, 261-63.)
3. Forty-four percent in the 1990 survey and 63% in the 1992 survey answered question 3 incorrectly, indicating their belief that the existence of a mitigating factor bars the penalty of death, reflecting a failure to provide clear and objective standards to channel the jury's discretion and provide specific and detailed guidance which permits rational review, as required by Godfrey v. Georgia.
5. Fifty-one percent in the 1990 survey and 56% in the 1992 survey answered question 16 incorrectly, indicating their belief that if a minority of the jurors vote against the penalty of death the jury should tell the judge that they cannot reach a unanimous verdict instead of signing the verdict form for a sentence other than death, reflecting a failure to provide clear and objective standards to channel the jury's discretion and provide specific and detailed guidance which permits rational review, as required by Godfrey v. Georgia.
The hearings held in January 1992 focused on the initial task specified in Judge Aspen's referral order, assessing the validity of the 1990 survey. Much of the testimony at those hearings dealt with similarities and differences between the 1987 IPI instructions used in that survey and the instructions actually used at Free's sentencing hearing. The court had earlier granted respondents' request for a two week recess in the hearing after the close of Free's case to enable respondents to prepare expert testimony. During the recess Zeisel conducted the 1992 survey substituting the actual Free instructions for the IPI instructions and correcting a typographical error in question 6.
The principal witnesses who testified were these:
-- Hans Zeisel, professor emeritus, Law School of the University of Chicago. Law graduate of the University of Vienna, Fellow of the American Statistical Association, co-author with Harry Kalven, Jr., of The American Jury (1966), author of several empirical studies of jury behavior and numerous other publications listed in PX 1.
-- Valerie P. Hans, professor of criminal justice and psychology, University of Delaware, Ph.D. in social psychology, author with N. Vidmar, of Judging The Jury (1986) and several articles on empirical jury studies and other publications listed in PX 6.
-- Shari Seidman Diamond, professor of psychology and criminal justice, University of Illinois, Chicago, and Senior Research Fellow, American Bar Association, Ph.D. in social psychology and J.D., publications on jury studies and other subjects listed in PX 9.
-- Judith N. Levi, professor, Department of Linguistics, Northwestern University, Ph.D. in linguistics, publications on language in the law and other subjects listed in PX 15.
-- Peter E. Rossi, associate professor of econometrics and statistics, University of Chicago, Graduate School of Business, Ph.D. in econometrics, publications listed in PX 19.
-- Paul J. Lavrakas, professor, School of Journalism, Northwestern University, Director of Northwestern University Survey Laboratory, Ph.D. in research psychology, publications listed in RX 8.
An agreed list of exhibits admitted in evidence was filed on March 10, 1992.
Summary of Principal Testimony
The 1990 Illinois Capital Jury Survey was initiated at the suggestion of David Bradford, an attorney with the McArthur Justice Center with a special interest in capital cases. The jury panels used in the survey were a random selection obtained from the Circuit Court assignment room just as they would be provided to judges and represented as good a probability sample of the jurors there as he could design. T. 23, 57. The confidence intervals (margins of error) in PX 3 are based on a 95% standard calculation and show the range of incorrect answers expected in 95 out of 100 repetitions of the test. T.26-29. The percentages of incorrect answers summarized in PX 3, Table 1, are "massive". T.34. The reliability of the results based on the sample of 96 jurors is supported by his split-half reliability analysis, PX 4, showing that when the sample is divided into two halves the answering patterns are nearly identical, thereby demonstrating the consistency of the survey results. T.36-37. Based on his experience and research, the jury deliberation process would not do much to eliminate the very high levels of erroneous impressions shown in the survey results. T.37-41.
In her opinion, based on her experience and review of the 1990 survey, that study is generally sound, the method adopted by Zeisel is one commonly used and generally accepted as an appropriate scientific method for examining juror comprehension and the sample was a reasonable one. The results show significant misunderstandings and fairly high levels of miscomprehension of the judicial instructions provided in the survey. Based on the 1990 survey results and her general knowledge about jury decision making, she believes that the Free jury more likely than not misunderstood the instructions given to them. T. 129-34, 168, 221-23.
In her opinion the methodology employed in the 1990 survey meets generally accepted criteria for valid jury research. T.247-49. Based on the confidence intervals, the split-half reliability analysis, the pattern of results and the level of precision appropriate to the subject, the survey results are highly reliable. T.257-58. She found no differences in the 1990 survey based on age or gender. Over 86% of the Witherspoon qualified jurors in that survey were white and the data show a high level of miscomprehension among those white jurors. T. 1079-81, 1083.
She also testified about the Diamond-Casper Study, PX 11, a preliminary report of a jury study conducted in the Cook County Circuit Court branch in Maywood, Illinois.
This study exposed prospective jurors to a videotape simulation of a murder trial, including IPI jury instructions for capital sentencing. The jurors then gave their decisions and answered multiple-choice questions intended to test, among other things, their recall and understanding of the instructions. One set of jurors was divided in two; half of the jurors were divided into jury panels and deliberated together for an hour and fifteen minutes before giving their decisions, while the other half did not deliberate before giving their decisions and answering the questionnaire.
Professor Levi, a linguistics scholar with a special interest in the interaction between language and the law, testified about similarities and differences between the instructions used in the 1990 survey and the actual instructions in Free's case. Her comparisons of the two sets of instructions are set out in PX 16 and 17. She concluded, based on a linguistic analysis, that the IPI instructions and the instructions in the Free case were confusing on several key issues. Other portions of her testimony will be referred to in later discussion of specific issues.
In his opinion, the 1990 survey very clearly shows that the potential jurors surveyed were very divided and confused in their responses. He replicated the calculations of confidence intervals in PX 3, found no numerical errors and does not believe there are any conceptual mistakes in the application of statistics. His own analysis of the survey responses, set out in PX 20, utilized a 99% rather than a 95% confidence interval because of the seriousness of the issue. The result was to widen the confidence interval slightly. T.632. The majority of the resulting confidence intervals do not range as high as 50% correct. Even as to the question with the highest confidence interval, question 4 (55-79%), "we are pretty sure that that population will not get that question right 80% of the time -- as much as 80% of the time." T.609-16.
Rossi also compared the distribution of correct responses actually observed with what would be expected under a random or "coin-tossing" view of respondent behavior. This calculation indicates that "the actual respondent distribution is very different from the expected distribution if the respondents were randomly selecting answers," PX 20, with "large numbers of people scoring systematically lower than one would expect if they were engaged in a casual sort of random response behavior." T.622. While a sample size of fifty is not accurate enough for judging the results of a presidential campaign where margins much less than 10% are important, the sample size here was entirely adequate for purposes of the Zeisel study. In his opinion the Zeisel survey is statistically sound and generalizable to the population as a whole. T.624-25.
Rossi does not see any systematic sub-groups of the survey respondents that either did well or did not do well. He sees rather a uniform pattern of consistent failure to get correct answers and fairly even dispersion across the types of answers. T. 644. To the extent there is systematic behavior the distribution responses are tilted toward incorrect responses. Cook County is a heterogenous population that would be expected to contain any subgroups that might be found elsewhere in Illinois. If there were a subgroup that did consistently better or worse, it would counsel caution in extrapolating the results to a different population that might not contain this subgroup in a similar proportion. Because no such subgroups appeared, Rossi was confident that the survey results could be extrapolated to other juror populations in Illinois. T.645-48.
The Phase II (1992) survey results are virtually identical to the 1990 results. T. 722. The 1992 survey satisfies critical requirements for a proper survey. T.723. The confidence intervals Rossi calculated on a 99% basis for the 1992 survey, with one exception, overlapped with the similar confidence intervals he computed for the 1990 survey. While in most instances the percentages of correct answers to individual questions in the 1992 survey fell outside the 1990 survey confidence intervals, that may be accounted for by the fact that different instructions were used. The 1992 survey differed from the 1990 survey in that an error in question 6 had been corrected, making it usable in the 1992 survey. In the 1992 survey there was a noticeable subgroup of 13 persons with 8 out of 11 answers correct, creating a "spike" contrasting with the 6 persons who scored 7 correct and the 7 persons who scored nine correct. This could have been a result of chance variation. PX 25. Rossi testified that since there was "no large group of people that perform extremely well on this test" he believed that the survey could be applied to other juror populations. T. 727.
According to Rossi, what is important is not the comparisons for individual questions, but the aggregate number of correct answers as to which the findings of the two surveys are roughly consistent, T. 752-65, 770-73. A clear majority of the respondents score less than 50% correct based on an analysis of answers to the eleven questions listed in PX 25. Although on some questions the respondents did slightly better and PX 20 and 25 show an increase in the mean percentage of correct answers from 40 to 48, both numbers are far away from any high percentage.
In his opinion, based on his experience as a specialist in research methodology and his review of the 1990 and 1992 surveys, both surveys were not rigorous from a methodological standpoint, there is insufficient evidence to conclude that the surveys measured comprehension and they do not justify any causal linkage between incorrect answers and a bias favoring the death penalty. There are too many uncertainties about the similarity of other counties to project the survey results beyond Cook County and the surveys have poor statistical validity. These opinions were based on Professor Lavrakas' analysis of four criteria for evaluating research which he referred to as construct validity, external validity, internal validity and statistical validity. Because his testimony is the basis for respondents' attack on the validity of the Zeisel surveys, that testimony will be described in more detail in the next section of this Report.
4. Validity and Reliability of the Zeisel Surveys Construct Validity
The first issue raised by Professor Lavrakas was based on his analysis of construct validity, a criterion which asks whether a research instrument measures what it is claimed to measure, in this case, juror comprehension. T. 825-26. Lavrakas' conclusion that there is not sufficient evidence to conclude that the Zeisel surveys measure comprehension, T. 871, was based on two statistical tests he performed on the 1990 survey data. The first was a factor analysis which depicts patterns of consistency in correct and incorrect answers to the survey questions. The second was an internal consistency test known as Cronbach's Alpha. His factor analysis is explained at T. 843-61 and depicted in RX 5(a) and (b). His Cronbach Alpha calculation is explained at T. 862-65.
We do not believe these analyses significantly undermine the validity of the 1970 survey or, by inference, the 1992 survey. First, Lavrakas essentially conceded that, as Rossi and Diamond testified, it is not appropriate to apply factor analysis and a Cronbach's Alpha calculation to the Zeisel survey data because they are binary or dichotomous, i.e., answers to yes - no questions, as opposed to numbers on a continuum. T. 845-65, 926, 1221-24 (Lavrakas) T. 774-76, 1176-86 (Rossi), T. 1085-89 (Diamond). Lavrakas admitted that multi-dimensional questions -- which we understand to mean questions that could turn on the understanding of more than one piece of information, or more than one step of possibly fallible reasoning -- would reduce factorial consistency, T. 929, and further admitted that his factorial analysis did show consistency in the responses to some of the questions. T.930-31, 937-38, 941. Lavrakas could not point to any specific flaws in the questions that would account for the lack of consistency he claimed to find, and acknowledged that the inconsistency could be a result of the subjects' lack of understanding. T. 944-49, 977.
Apart from technical reasons, Rossi explained that the use of factor analysis is logically inappropriate because any interpretation based on patterns of consistency is necessarily based on unverifiable assumptions about how heterogeneous the survey respondents are in their ability to comprehend the instructions. Lavrakas, who had the last word, offered no response to that observation.
Lavrakas' construct validity analysis is therefore rejected. The Zeisel surveys on their face appear to be designed to measure jury comprehension of capital sentencing instructions. The testimony of Zeisel, Hans, Diamond and Rossi supporting that conclusion is persuasive while Lavrakas' criticism is based on two types of statistical analysis which are inappropriate here.
While RX 6 comparing persons more than 24 years [probably, of age] in the two counties shows higher percentages of high school graduates in DuPage County, no evidence was presented comparing the educational levels of eligible jurors in the two counties or those of the Free jury compared to the Zeisel survey subjects. What is more significant is that the survey data show high levels of miscomprehension and Rossi testified without dispute that the data show a uniform pattern of consistent failure to answer questions correctly and fairly even dispersion across the types of answers with no systematic sub-group doing well or poorly. On this evidence, we believe that the educational differences depicted in RX 6 do not represent a significant demographic variance for this purpose. As to racial demographics, Lavrakas ultimately agreed that race as a variable was not important in determining comprehension level. T. 1225-27 (Lavrakas) 1190-92 (Rossi) and 1080-83 (Diamond).
Internal Validity and Statistical Validity
Lavrakas testified that an inquiry into internal validity asks whether a study was designed and implemented to test a cause and effect relationship, T. 889-98, in this case, whether the survey results demonstrate a correlation between incorrect answers and the likelihood of voting in favor of the death penalty. PX 28. This criterion is not discussed further because, in our opinion, while the pattern of correlation reflected in PX 28 is suggestive, it does not provide a firm basis for any cause and effect conclusion. See T. 1198-99 (Rossi). For the same reason it is not necessary to discuss Lavrakas' testimony at T. 894-99, illustrated in RX 7, that the Zeisel study has weak statistical validity (compare Diamond's contrary testimony at T. 1093-1118) since Lavrakas' focus in that connection was also on whether the survey results support a causal inference relating incorrect answers and the likelihood of voting in favor of the death penalty.
Lavrakas stated that the results of the 1992 survey suggested "volatility, variation in the numbers", T. 869, because the answers to most questions fell outside the confidence intervals predicted on the basis of the 1990 survey results. Beyond stating that it could be interpreted as a mark of inconsistency, Lavrakas did not offer any explanation of volatility or the application of that concept to the survey data.
Zeisel and Rossi calculated confidence intervals differently. Zeisel calculated 95% confidence intervals based on percentages of incorrect answers, while Rossi calculated 99% confidence intervals based on percentages of correct answers. Eliminating questions 1, 13, 14, 15 and 16 as questions with no correct answers (see p. 9 supra, PX 3, Table I and Part 6 infra) and question 6, which was not included in 1990 results because of a typographical error, leaves 10 questions. Using Zeisel's confidence intervals, the 1992 results fell within predicted confidence intervals for 3 questions, fell slightly outside (1% to 2%) for 2 questions, outside by 4% to 6% for 3 questions and outside by 9% to 12% for 2 questions. Using Rossi's confidence intervals, the 1992 results fell within predicted confidence intervals for 3 questions, slightly outside for 1 question, outside by 5% to 8% for 5 questions, and outside by 12% for 1 question.
We do not agree with Rossi's premise that the primary importance of the Zeisel surveys is as evidence of comprehension of the instructions "in the aggregate". While the likelihood of misunderstanding on more than one issue is significant, our analysis necessarily focuses on the likelihood of misunderstanding with respect to specific issues on which the Constitution requires that the jury be given clear guidance. Nevertheless, the facts relating to confidence intervals do not, in our opinion, detract from the validity of either survey.
First, we put aside questions 12 to 15 dealing with the burden of persuasion. Except possibly for question 12, which we explain in Part 6 has a potential ambiguity, these questions do not have correct answers. As discussed more fully in Part 6 of this Report, the significance of the survey results for those questions lies in the extent of the disagreement among the respondents reflecting substantial confusion about the burden of persuasion.
As Rossi observed, the two surveys were not identical and the use of different instructions may account for some of the variations under discussion. What is more important is that the levels of misunderstanding shown in both surveys, with few exceptions, are very high and the levels of deviation in the 1992 survey from the results predicted are modest. Finally, there is no suggestion that confidence intervals are a critical test of validity. Lavrakas said only that the deviations "suggest" and "could be interpreted" as a sign of variation or inconsistency. For these reasons, and considering all of the evidence on validity, we conclude that the facts relating to confidence intervals do not detract from the validity of either survey.
Other Possible Sources of Error
The court has considered other possible sources of error suggested by Lavrakas but finds them unsupported. Inadvertence, fatigue and literacy problems, T. 839, are possibilities. Diamond testified that if fatigue had been a factor, one would expect the subjects to score better on questions coming earlier on the list, but she found no evidence of that. T. 1153-54. Lavrakas did not dispute Rossi's analysis showing that the actual distribution of responses was very different from the distribution expected if the respondents were randomly selecting answers. PX 20. Lavrakas criticized the surveys for the absence of internal consistency checks. However, he did not address the inclusion of multiple questions on three of the five topics, a more meaningful consistency check than those he suggested. T. 138, 206, 638. Finally, in criticising the surveys for inadequate documentation, he only specified the question whether the surveyed population was a representative sample of all persons eligible to serve as jurors in Cook County. T.873. But he nowhere indicated what type of documentation on that point is needed in view of 1) the uncontradicted testimony of Zeisel and attorney James Bailinson, T. 1041-76, that the survey subjects were panels of persons summoned for jury duty at the Daley Center and previously grouped by jury officials in the same manner as groups sent to courtrooms for actual jury service and 2) the evidence summarized above that the demographic differences he stressed are not significant for this purpose. Reserved for later discussion is one other issue brought up by Lavrakas, the significance of the difference between taking a written test and the real life situation of jurors making an actual decision on life or death.
The results of the Zeisel studies are consistent with other studies of juror comprehension. Where jury instructions have been based upon statutes and appellate court decisions with no attempt to rewrite them for better comprehension, every empirical study has shown a startlingly low level of comprehension. The literature on such studies is summarized in W. Steele and E. Thornburg, Jury ...