POVEDA--Tony-G.-Estimating-wrongful-convictions.-Justice-Quarterly--18-3--689-708--2001..md

This article was downloaded by: [University of Chicago Library] On: 11 December 2014, At: 04:21 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Justice Quarterly Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/rjqy20

Estimating wrongful convictions Tony G. Poveda a a State University of New York at Plattsburgh Published online: 20 Aug 2006.

To cite this article: Tony G. Poveda (2001) Estimating wrongful convictions, Justice Quarterly, 18:3, 689-708, DOI: 10.1080/07418820100095061

To link to this article: http://dx.doi.org/10.1080/07418820100095061

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/ page/terms-and-conditions RESEARCH NOTE

ESTIMATING WRONGFUL CONVICTIONS*

TONY G. POVEDA** State University of New York at Plattsburgh

Although numerous cases of wrongful convictions have been documented in the literature and in the media, criminologists have yet to devise a methodology for estimating the extent of such errors in the criminal justice system. I explore several methodologies with this purpose in mind, including the use of official data, inmates' self-reports, and case study approaches. Specifically, I use court-ordered discharges from imprisonment as a basis for measuring official error. In addition, I employ data from the RAND inmate surveys to estimate the extent of convicted offenders who deny their commitment offenses. Studies that attempt to catalogue individual wrongful convictions also serve as a basis for estimating false positive errors. Each methodology has its own limitations, but by employing multiple measures and approaches, I make possible an estimate of the "dark figure" of wrongful convictions.

The extent of wrongful convictions has been a topic of much speculation and remains a "dark figure" in the study of criminal justice. One of the few attempts to estimate the prevalence of wrongful convictions was a 1986 study by C. Ronald Huff and associates. In that research a variety of justice system officials (mostly in Ohio) were surveyed for their perceptions of this type of miscarriage of justice in felony cases (Huff, Rattner, and Sagarin 1986,

• An earlier version of this article was presented at the annual meetings of the American Society of Criminology, held in San Francisco on November 16, 2000. I am indebted to Paul H. Korotkin, assistant director, and to William Chapman of the DOCS office for Program Planning, Evaluation and Research for providing me with data on court-ordered discharges from imprisonment in New York State. I am also grateful to my colleague, Dr. Tie-ting Su, for writing a computer program to transform the data from the RAND Inmate Survey into a more manageable data set and for producing the resulting cross-tabulated tables. ** Tony G. Poveda is professor of sociology and criminal justice at the State University of New York at Plattsburgh. He received his doctorate in criminology in 1970 from the University of California at Berkeley. His research interests include white-collar crime (especially organizational crime), the FBI, and, most recently, wrongful convictions. His book publications are Lawlessness and Reform: The FBI in Transition (1990), Rethinking White-Collar Crime (1994), and a co-authored work, The FBI: A Comprehensive Reference Guide (1999). In addition, Professor Poveda has published numerous papers on these topics in scholarly journals.

JUSTICE QUARTERLY, Vol. 18 No. 3, September 2001 © 2001 Academy of Criminal Justice Sciences

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 690 WRONGFUL CONVICTIONS

1996). 1 More recently, a statistical study of errors in capital cases

conducted by James Liebman and colleagues has highlighted the

extent of reversible error in capital cases, some of which involve the

erroneous conviction of the innocent (Liebman, Fagan, and West

2000). 2 Research on the scope of the wrongful-conviction problem is

still in its infancy; it is plagued by methodological problems and by

the absence of any central entity for tracking official errors in the

justice system.

A 1998 scholarly exchange between Richard Leo, Richard Ofshe, and Paul Cassell underscores the difficulty of quantifying the

wrongful-conviction problem but also highlights its importance for public policy 3 (Cassell 1998; Leo and Ofshe 1998a, 1998b). 4 The present article is a preliminary effort at exploring several methodologies for estimating wrongful convictions. I emphasize two approaches to the problem: the use of official records on court-ordered discharges from imprisonment, and data from inmates' self-reports.

DEFINING WRONGFUL CONVICTIONS

Wrongful convictions can be understood along a continuum of

justice-system errors ranging from persons who are falsely accused

(arrested, prosecuted, and tried), to those who are wrongly convicted and imprisoned, to death row inmates who are erroneously

executed. In this article I focus on errors that result in the conviction and imprisonment of innocent persons. These are wrongperson errors, as distinct from procedural errors in the conviction of

a defendant. Although this distinction is sometimes difficult to

maintain in practice, wrong-person convictions are cases in which

1 The great majority of respondents (71.8%) in the Huff study estimated that fewer than 1 percent of felony convictions involved innocent persons (Huff et al. 1986:523). 2 In the Liebman study of capital cases, 7 percent of persons whose death sentences were overturned were found not guilty upon retrial, or charges against them were dismissed (Liebman et al. 2000:132). This is most evident in the movement for a moratorium on the death penalty. The. release of 95 innocent death row inmates nationwide since 1973 (as of May 7, 2001) has raised serious doubts about the administration of capital punishment in the United States (Death Penalty Information Center (DPIC) 2001; Dieter 1997). On January 31, 2000 Illinois became the first death-penalty state to issue such a moratorium. In that state, for each death row inmate executed over the last 20 years, another has been released because of innocence (Johnson 2000). 4 The exchange between Leo/Ofshe and Cassell centered on the role of policeinduced false confessions in producing wrongful convictions, the magnitude of the problem, and the relevance of the Miranda warning to this issue. Cassell (1998) maintained that the false-confession problem is minor compared with the problem of lost confessions from guilty persons (and therefore lost convictions); this problem, he argued, is due to restrictions on police interrogation. Leo and Ofshe (1998b) challenged Cassell's attempt to quantify the false-confession problem and rejected his claim that quantification is even necessary for public policy decisions. In addition, they reaffirmed the importance of Miranda and the need to develop other procedural safeguards to minimize false confessions and other sources of wrongihl convictions.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 POVEDA 691

individuals are exonerated because they are factually (not merely legally) innocent of the crime for which they were convicted--and are not released solely on the grounds that their due process rights were violated (e.g., by illegal search, Miranda warningnot given, trial-court error). This definition is in the tradition of Edwin Borchard's ([1932] 1970) classic study of wrongful convictions, the landmark work of Hugo Bedau and Michael Radelet (1987; Radelet, Bedau, and Putnam 1992) on erroneous convictions in capital cases, and the research of Huff and associates (1986, 1996) on convicted innocents.

On a conceptual level, wrongful convictions denote a distinct type of justice-system error: when a wrong person has been convicted and incarcerated. Operationally, however, it is difficult to determine when someone is factually innocent, The fact that an offender's conviction is overturned on appeal and that the offender is acquitted in a subsequent trial does not in itself establish innocence. It simply means that the state was not able to establish guilt beyond a reasonable doubt. Clearly, "not guilty" and "innocent" are not synonymous. For example, according to New York State's Court of Claims Act (Article II, Sec.8-b), if innocent persons wish to seek compensatory damages for unjust conviction and imprisonment, they must "demonstrate by clear and convincing evidence" that they did not commit the crime for which they were convicted and incarcerated (New York State Court of Claims 1998). More is required than merely showing that the conviction was vacated or that the indictment against them was dismissed.

The problem is illustrated further by the April 2000 verdict in the Sam Sheppard civil trial. The Sheppard jury had to determine "by the greater weight of the evidence" whether Sheppard was innocent and therefore wrongfully imprisoned by the State of Ohio for the 1954 murder of his wife. The civil jury ruled that Sheppard's family ~ had not met the burden of proof for innocence, even though a 1966 jury, in Sheppard's second criminal trial, had found him not guilty (Ewinger and Hagan 2000; Hagan and Ewinger 2000).

Thus, the operational definition of wrongful conviction for research is problematic. What constitutes evidence and criteria for innocence? How compelling should that evidence be? The tradition of studying wrongful convictions, beginning with Borchard in 1932, has erred on the side of a restrictive definition of wrongful conviction: evidence is required showing that someone else committed the offense or that the convicted person was uninvolved in the crime

5 Dr. Sam Sheppard died of liver failure in 1970 at age 46 ("History of Sheppard Case" 2000).

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 692 WRONGFUL CONVICTIONS

(Gross 1998:129; Radelet and Bedau 1998:106). 6 Bedau and Radelet (1987:47) maintained that "there is no quantity or quality of evidence that could be produced that would definitively prove innocence." At best, they argued, a "majority of neutral observers, given the evidence at our disposal, would judge the defendant in question to be innocent. ''7 Huff and his colleagues similarly included only cases in which there was evidence of innocence "beyond a reasonable doubt" (Huff et al. 1986:519) or when convicted persons had been "clearly exonerated" (Huff et al. 1996:10).

An additional issue is whether such wrong-person errors must be acknowledged by officials in the criminal justice system, particularly by the courts. Surely the more restrictive definition would consider only officially acknowledged errors. This criterion is followed in Huffs database of felony cases (Huff et al. 1986:519). Bedau and Radelet also relied on official judgment of error (e.g., appellate court decisions, executive pardons, indemnity claims) in 90 percent of their 416 capital cases. In the other 10 percent, however, their decision to include the case was based on unofficial judgments. These latter cases included research by scholars or journalists who had reexamined evidence in a particular case (and often had discovered new evidence), and had concluded that the defendant was not guilty, even though no court had determined this (Radelet et al. 1992:17-18). Insofar as we allow unofficial judgments of error, we are shifting toward a more expansive definition of wrongful conviction.

ALTERNATIVE METHODOLOGIES

A critical methodological problem in the study of wrongful convictions is how to determine the extent of such convictions and imprisonment in the criminal justice system. The dominant approach in the literature has been to catalogue individual cases of wrongful conviction. This approach has been pursued by legal scholars, journalists, sociologists, and other researchers (Borchard [1932] 1970; Connors et al. 1996; Dieter 1997; Huff et al. 1996; Radelet et al. 1992; Radin 1964; Rosenbaum 1990-1991; U.S. House 1994; Yant 1991); none claim to have identified all such cases. Indeed, many of

6 Radelet and Bedau (1998:106-07) also note that the concept of innocence can be broadened in a variety of ways including the counting of accidental killings, selfdefense, and killings by a mentally ill offender.

7 The most one can hope to obtain is a consensus of investigators after the case reaches its final disposition, Consensus can be measured in degrees, and the cases that we have included in our catalogue are those in which we believe a majority of neutral observers...would judge the defendant in question to be innocent" (Bedau and Radelet 1987:47).

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 POVEDA 693

these researchers point to the role of tuck and chance in the discovery of wrongful-conviction cases (U.S. House 1994:12). Bedau and Radelet (1987:29) emphasized "how accidental and unsystematic the discovery of relevant cases actually is." Gross (1998:150) concluded that "most miscarriages of justice in capital cases never come to light"; when they do so, it is typically because of heightened judicial attention to the case, because the real criminal confesses, or simply because of luck.

The fundamental flaw of the case study approach is that there is no way to determine how many cases go undiscovered--the "dark figure" of wrongful convictions--and whether these cases differ systematically from those which are identified. The documenting of individual cases has helped to sensitize the public and policy makers to the existence of erroneous convictions, but it falls short as a method for determining the prevalence of such convictions. Another approach, taken especially since the late 1970s, has been to employ laboratory or experimental methods in studying mistaken eyewitness identification, a significant factor in many wrongful convictions (Cutler and Penrod 1995; Loftus 1993, 1996; Wells 1993). This research has attempted to discover error rates of eyewitness identification under different experimental conditions, with emphasis on police identification procedures in lineups, event factors, and postevent information. Factors such as whether the culprit is present or absent in the lineup, police instructions given to the witness, the trauma of the crime, the use of a deadly weapon, and information obtained by the witness after the crime are all relevant to shaping the memory of an event and to eyewitnesses' accuracy (Devine 1995; Loftus and Ketcham 1991; Wells 1993). One problem with this line of research is that it is often based on simulated events (staged crimes and lineups) rather than on actual crimes; thus it raises questions of generalizability from experimental to forensic ("real") contexts (Yuille 1993). Moreover, the purpose of eyewitness identification experiments is to isolate the role of particular factors in producing error, not to estimate the overall accuracy of eyewitness identifications in actual cases (Cutler and Penrod 1995). Thus, again, this methodology is limited in providing an empirical basis for estimating the prevalence of wrongful convictions. In 1999, however, when the Innocence Project examined the 62 DNA exonerations to date in the United States, it found that mistaken eyewitness testimony was a factor in 84 percent of those wrongful convictions (Scheck, Neufeld, and Dwyer 2000:246). Another possible line of inquiry is the use of field research to determine the extent of error produced by a particular investigative

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 694 WRONGFUL CONVICTIONS

practice. Leo and Ofshe's research on police interrogations reflects this approach (Leo 1996). Their fieldwork in a limited number of police departments has led to observations on how standard police interrogation practices can cause suspects to make false confessions (Leo and Ofshe 1998a; Ofshe and Leo 1997). These authors, however, make no claim that their methodology can be used to identify the universe of false confessions. On the contrary, they maintain that the "methodological problems inherent in arriving at a sound estimate [of the frequency of false confessions] are formidable and unsolved, and we have concluded that no welt-founded estimate has yet been published" (Leo and Ofshe 1998b:560). Thus they are quite skeptical about the possibility of quantification, given the absence of official records on the number of police interrogations and the frequency of false confessions resulting from those interrogations.

Finally, another approach to discovering errors in the criminal criminal justice system is the use of appellate court decisions in which convictions or sentences have been overturned. This methodology has been adopted by Liebman and associates in their research on errors in capital cases. Liebman relied on published judicial decisions at three levels of the capital review process: state direct appeal, state postconviction appeal, and federal habeas corpus review. This is a monumental task, as noted by Liebman, Fagan, and West (2000:24): it requires a "painstaking search" for cases relying on various sources of information. To complicate this task, there is no central repository for these judicial decisions.

Liebman's study of capital errors between 1973 and 1995 began in 1991 and continues today. Although appellate courts (at the three review levels) found reversible error in 68 percent of capital cases during the study period, Liebman's findings on the outcomes of these reversals are limited to the 301 cases in the state postconviction appeals. Upon retrial, 247 of the plaintiffs in the state postconviction cases were sentenced to less than death, 22 (7%) were found not guilty, and 54 were resentenced to death (Liebman et al. 2000:132). As Liebman et al. (2000:27) pointed out, state postconviction decisions, unlike state direct appeal and federal habeas corpus reviews, often are not published and therefore are difficult to locate. At the present, the list of such cases is incomplete. The Liebman study has provided substantial evidence demonstrating a high rate of error in capital judgments. So far, however (except for the state postconviction results), it has provided only limited information on the outcome of these reversals, notably in regard to the prevalence of wrongful convictions.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 POVEDA 695

As noted, criminal justice agencies do not keep statistics on errors made by officials at various stages of the justice process, including the number of persons wrongfully convicted and imprisoned. In the remainder of this paper, I consider two different approaches to this quantification problem. One approach employs official agency records or statistics as a basis for constructing a measure of wrongful convictions and imprisonment; the other relies on inmates' self-reports of their own criminal activity, including denial of committing the offense for which they were incarcerated. The development of multiple methods for estimating wrongful convictions has certain advantages. Insofar as each method has its own limitations, the use of multiple and independent measures to estimate the prevalence of wrongful convictions will increase our confidence in the findings of each approach, if we assume a consistency in results (Webb et al. 1966:3-5).

USE OF OFFICIAL STATISTICS ON COURT-ORDERED DISCHARGES

This approach develops an estimate of wrong~l convictions based on actions taken by justice officials, namely court-ordered discharge of inmates from imprisonment. "Court-ordered discharge," however, is a broader category than "wrongful conviction": it encompasses not only substantive errors (wrong-person convictions) but also procedural errors in the conviction and sentencing of the offender. Ultimately it is necessary to distinguish these two types of errors if our use of these official data is to have merit.

Data From the New York State Department of Correctional Services

A series of research reports and the availability of data from the New York State (NYS) Department of Correctional Services (DOCS) make New York a convenient jurisdiction for examining the value of court-ordered discharges, s Every year the DOCS publishes reports on inmates under its jurisdiction, but its annual report, Characteristics of Inmates Discharged, is most relevant to this study. The DOCS recognizes 10 categories of inmate release, including parole, maximum expiration of sentence, escape, death, court-ordered discharges, and others. A court-ordered discharge is defined by the DOCS as a case %vhere the individual is discharged from custody while further

s The DOCS disseminates official statistics through its Office of Program Planning, Evaluation, and Research.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 696 WRONGFUL CONVICTIONS

court proceedings take place (i.e., an appeal, new trial), or is discharged as a result of court proceedings in which the individual is found guilty of a lesser charge for which he may already have served the length of time legally required, or discharged because the original charge was dismissed" (DOCS 1997a:3). Over the 22- year period from 1976 through 1997, the DOCS has averaged 155 court-ordered discharges per year. There were 199 such discharges in 1995, the year that we examine here. Because of the relatively small number of inmates discharged under court order in any single year, the DOCS annual reports do not provide cross-tabulated data on inmate characteristics for those who are released in this manner. (Such data are available only for parole release.) The DOCS Office of Program Planning, Evaluation, and Research, however, provided me with more highly detailed information on the characteristics of court-ordered discharges for 1995.

Measures of Wrongful Conviction

Unfortunately, the data mentioned above need further elaboration and interpretation before we can make any inferences about the extent of wrongful convictions. As stated earlier, official data on court-ordered discharges do not distinguish between substantive and procedural errors, and we are concerned with the former (wrong-person convictions and imprisonment). Nothing in the official data makes this distinction, except a 1989 DOCS report on vacated murder sentences in New York. That study, conducted by the New York State Department of Correctional Services, identified 33 offenders who had been convicted of murder and released from DOCS custody between 1980 and 1987 for new trials. Those cases "represent current instances where appellate courts found error so substantial as to warrant the vacation of the conviction for murder" (DOCS 1989:1). In the DOCS follow-up analysis on those 33 cases, it was found that on retrial the great majority (26, or 78.8%) were reconvicted of a lesser offense (17 for first-degree manslaughter). The remaining seven (21.2%) were acquitted after a new trial or after local prosecutors dismissed the original indictment. 9 If we regard these seven cases as wrong-person convictions, using acquittal after a new trial or dismissal of indictment (following a court-ordered discharge) as measures of wrongful conviction, this internal DOCS study offers some empirical basis for estimating the proportion of court-ordered

9 Five of the seven cases involved acquittals; the other two, dismissals. It is possible that the indictment dismissals were made on procedural grounds, but it is more likely that the conviction reversal in appellate court undermined the evidentiary basis for pursuing the case (e.g., recanted witness testimony, discovery of police report exonerating defendant).

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 POVEDA 697

discharges that represents substantive error (21.2%), at least in regard to murder convictions and commitm, ents. Samuel Gross (1996, 1998) argues persuasively that homicide cases in general, and capital cases in particular, are handled differently by officials in the criminal justice system, and for a variety of reasons are more likely to involve erroneous convictions. Therefore, in line with Gross's thesis (and the limits of DOCS data), we must limit our objective here to estimating the prevalence of wrongful murder convictions rather than wrongful felony convictions.

Findings

An examination of 1995 DOCS data on court-ordered discharges reveals that of the 199 inmates released from custody by the courts, 24 (12.1%) had been convicted of murder (DOCS 1997a). 1° If we assume that only 21.2 percent 11 of these court-ordered discharges represent substantive errors, then only five discharges represent wrongful murder convictions. However, the relevant population base for calculating prevalence of wrongful convictions is the number of murder commitments to the Department of Correctional Services. A study conducted by the New York State Defenders Association (NYSDA) 12 found that the median time from conviction to reversal (acquittal/dismissal) in wrongful homicide convictions in New York is approximately three years (NYSDA 1989; Rosenbaum 1990-1991). 13 Accordingly, murder inmates released by court-ordered discharge in 1995 were probably committed to the Department of Correctional Services in 1992, when 357 murder commitments were made to the DOCS (DOCS 1997b:10). If five of these were wrongful convictions, this translates to an error rate of 1.4 percent in murder convictions/commitments. 14

lo Data on the specific characteristics, including the commitment crime, of 1995 court-ordered discharges were provided by Paul Korotkin, assistant director of program planning, evaluation, and research of the DOCS. 11 The 21.2 percent figure is based on the findings of the 1989 DOCS study on vacated murder sentences. 12 The New York State Defenders Association is a nonprofit organization of public defenders and legal aid atton]eys who provide services to criminal defense lawyers throughout the state. 13 On the basis of the wrongful conviction cases cited in the NYSDA study, the median time between conviction and inmate's release was 38.5 months. The DOCS (1989:3) study placed the median at 29 months (from prison reception to reversal). Even if we allow for some variation in the reversal period (from one to four years), the average number of annual murder commitments in NYS between 1991 and 1994 was 370, only a slight departure from the 1992 figure of 357 that t used. 14 If we use the 7.3 percent figure (instead of the 21.2 percent from the DOCS study), derived from the Liebman study of state postconviction reversals, as the basis for estimating the proportion of errors involving wrong-person convictions in vacated murder cases, the error rate ~br murder convictions/commitments is only .5 percent. The Liebman study, however, is based on an incomplete sample of capital cases reviewed in state postconviction appeals (1973-1995) in 26 states (Liebman et al. 2000:53), not including New York (which did not restore the death penalty until

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 698 WRONGFUL CONVICTIONS

The NYSDA study, described above, also attempted to catalogue wrongful homicide convictions in New York State from 1965 to 1988. As part of its Wrongful Conviction Study Project, the study identified 59 wrongful homicide convictions during this period, 45 of which involved murder convictions (Rosenbaum 1990-1991:807-08). The NYSDA's definition of wrongful homicide conviction, however, was broader than that employed in the present study. It included not only cases in which the defendant, whose conviction had been overturned, was acquitted on retrial or the charges were dismissed, but also cases in which the defendant was reconvicted of a nonhomicide crime. If the NYSDA study cases are limited to murder sentences reversed between 1980 and 1987 and to cases in which the defendant was subsequently acquitted on retrial or the indictment was dismissed (and omits cases in which the defendant was reconvicted of a lesser offense), 15 we were left with 23 wrongful murder convictions in New York State between 1980 and 1987. During the period 1977 to 1984,16 2,276 murder commitments were made to the Department of Correctional Services (Chapman and Zausner 1985); this translates to a 1.0 percent error rate in murder convictions/commitments. The estimated prevalence of wrongful murder convictions reached by using the NYSDA study (1.0%) is comparable to the estimate of 1.4 percent calculated from DOCS data on court-ordered discharges. This is the case even though very different methodologies, each with its own assumptions and limitations, are used in each approach. Both, however, are based on officially acknowledged errors in which the convicted murder defendant has been released from custody and subsequently is acquitted on retrial, or in which the original indictment is dismissed.

USE OF INMATES' SELF-REPORTS

The use of inmates' self-reports is another methodological approach to estimating the prevalence of wrongful convictions. Unlike court-ordered discharges, which rely on the reliability and validity of agency records, inmates' self-reports depend on the reliability and validity of inmates' accounts of their own criminality. The use of these self-reports provides another approach, with a different set of assumptions and limitations, to the quantification problem.

September 1995). The 21.2 percent figure is based on New York vacated murder cases during the 1980-1987 period. 15 These parameters then correspond to those of the 1989 DOCS study. 16 This allows for a three-year lag period from conviction to reversal.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 POVEDA 699

The RAND Inmate Survey

In late 1978 and early 1979, the RAND Corporation of Santa Monica, California conducted surveys of convicted male offenders (in 14 jails and 12 prisons) in California, Michigan, and Texas (Peterson et al. 1982). The surveys were administered in small groups (10 to 30 inmates) by temporary RAND employees who were hired because of their experience in working with incarcerated inmates; prison officials did not participate in these sessions. The sample was representative of adult male inmates admitted to prisons and jails in the above jurisdictions at the time of the survey. Participation was voluntary and required informed consentF Altogether self-report data were obtained from 1,380 prison inmates and 810 jail inmates.

The questionnaire, which took about 50 minutes to complete, asked for detailed information about crimes the inmates had committed in a one- to two-year period preceding their incarceration, including questions about arrests and convictions. The survey also questioned inmates about their current conviction offense, demographic characteristics, and other personal details. In addition, official criminal records were collected on 85 percent of the prison inmates who responded to the questionnaire (Peterson et al. 1982).

Measure of Wrongful Conviction

Although the purpose of the RAND Inmate Survey was to collect data on criminal careers and to develop policy implications from those data, two questionnaire items in the "1978 Jail/Prison Survey Booklet" (Chaiken and Chaiken 1982:App. E) are relevant to our research. These questions (Numbers 6 and 7 on p. 39 of the booklet) pertain to the inmate's current offense: "What charge(s) were you convicted of that you are serving time for now? (Check all that apply)," and "For these convictions, what crimes, if any, do you think you really did? (Check all that apply)." Each of these questions is followed by a list of 15 different offenses, plus an "other" category. In addition, the last response choice for the second question is "Did no crime." Inmates' denials of the conviction offense (i.e., self-reports that they "did no crime") provide another basis for measuring wrongful conviction. Clearly, inmates who claim that

17 The response rate for jail inmates was 70 percent; for prison inmates in California and Michigan, 50 percent; for prison inmates in Texas, 82 percent (Peterson et al. 1982:viii). The RAND researchers attempted to correct the problem of nonrespondents with a replacement procedure. Moreover, although the survey was confidential, it was not anonymous. The questionnaires were coded to make future identification of respondents possible for subsequent research (Peterson et al. 1982:x).

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 700 WRONGFUL CONVICTIONS

they are innocent of the offense for which they have been incarcerated meet the operational definition of wrongful conviction.

Certainly, inmates' self-reports raise serious questions of reliability and validity that must be addressed before we can take them at face value. Built into the RAND research design were several ways of checking on the internal consistency (reliability) as well as the validity of responses. These included asking respondents for essentially the same information in different parts of the questionnaire, retesting some of the respondents a week later, and comparing inmates' responses with official records. Overall, "83 percent of respondents tracked the questionnaire with a high degree of accuracy and completeness, and were very consistent in their answers" (Chaiken and Chaiken 1982:224-25).

When researchers compared questionnaire data with the prison inmates' official records, is they found a close correspondence. In fact, in California and Texas, respondents reported 6 percent more convictions than the records show; in Michigan, convictions were underreported by 6 percent (Marquis and Ebener 1981:34). Inmates were reluctant, however, to provide information on sex offenses, including rape; these are more likely than other offenses to be understated in self-reports (Peterson et al. 1982:21). There was also evidence that some errors in inmate-reported crimes were due to confusion about offense categories; thus the problem was one of misclassification rather than omission. For example, "robbery" and "theft" were sometimes confused (Marquis and Ebener 1981:38). In assessing the quality of prisoners' self-reports, RAND researchers began with the conventional wisdom that respondents are likely to conceal undesirable information about themselves. After reviewing the methods literature as well as results from the RAND Inmate Survey, however, they concluded just the opposite: "We found evidence that respondents usually reveal more arrests and convictions in questionnaires or interviews than can be found in official records" (Marquis and Ebener 1981:2). Moreover, the RAND evaluation of inmates' self-reports specifically examined the quality of reports on questionnaire items pertaining to the current offense. The investigators concluded that "on a general level, the data are close to unbiased" (Marquis and Ebener 1981:32). Although the RAND researchers observed item-by-item variation,

is This portion of the assessment of the quality of responses was limited to prisoner respondents because official records were available only for prisoners, not for jail inmates. RAND researchers used official records to check the correspondence between prisoners' self-reports and official arrests and convictions.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 POVEDA 701

they found in general that reliability was "moderately high" (Marquis and Ebener 1981:32). Overall the RAND findings refute the widely held belief that inmates are likely to underreport their criminal activity: "In general, the prisoner respondents do not appear to be systematically denying their conviction offenses in the questionnaire" (Marquis and Ebener 1981:47).

Method

The RAND Inmate Survey (National Archive of Criminal Justice Data 2000; Peterson et al. 1982) is available to the public at the National Archive of Criminal Justice Data website. 19 Because I was interested in only a few of the variables contained in the survey, I extracted those variables from the RAND data set to create a smaller data set for analysis. Any cases with missing data were eliminated. The specific variables in the newer, smaller data set included current conviction offense, crimes actually committed, state, and type of institution. In addition, because in the present study I am interested in wrongful convictions and imprisonment at the felony level, I examined only data on prison inmates: 1,282 cases in Michigan, California, and Texas.

Findings

The basic finding is that 197 of the 1,282 prison inmates questioned in the RAND Survey, or 15.4 percent, claimed that they did not commit the crime for which they had been convicted and imprisoned. (See Table 1) Those inmates also denied committing any of the other 15 crimes included on the "current conviction offense" checklist, including "other." This finding did not vary widely by state: 14.1 percent of Michigan prisoners denied having committed any crime, as did 14.6 percent in California and 16.7 percent in Texas. Denials varied more widely, however, by current conviction offense. Rapists and other sex offenders denied their conviction offense at the highest rates by far: 37.7 and 26.9 percent respectively. Those convicted of murder, weapons violations, assault, and robbery fell into the middle range: 17.5, 13.4, 12.8, and 11.5 percent respectively denied their offense. Drug and nonviolent property offenders were least likely to deny their conviction offense. It is possible that in this rather lengthy, complex questionnaire the inmate respondents misunderstood the two key questions that are of concern to us. Specifically, in marking "Did no crime," they

19 The NACJD is sponsored by the Bureau of Justice Statistics (U.S. Dept. of Justice), operated by the Inter-University Consortium for Political and Social Research, and headquartered at the University of Michigan.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 702 WRONGFUL CONVICTIONS

Table 1. Current Conviction Crime by Self-Reported Denial

Current Conviction Offense Did Not Commit n %

Rape 23 37.7 Sex Offense (not rape) 14 26.9 Murder 18 17.5 Weapons 20 13.4 Assault 22 12.8 Robbery 44 11.5 Forgery 8 9.9 Burglary 33 9.0 Drug Sate 7 8.1 Drug Possession 6 5.2

All Offenses 197 15.4

Source: National Archive of Criminal Justice Data (2000). NOTE: N = 1,272 (missing values in 10 of the 1,282 cases)

may not have understood their response. As a check on respondents' confusion, I examined whether inmates who reported that they "did no crime" (in response to question 7 in the booklet) also indicated, nonetheless, that they really had committed another crime. Such a response, of course, would be contradictory to the question "What crimes, if any, did you really commit?" This did not seem to be a problem, however: only 19 (1.5%) of the inmate respondents indicated that they had committed no crime and yet checked one of the other crime categories on the list.

Overall, inmates' reports of their conviction crime corresponded to crime they actually committed. For example, 103 inmate respondents reported that they had been convicted of murder, and 70 of those respondents admitted that they had committed murder. Many of the remaining 33 respondents admitted to crimes less than murder (e.g., manslaughter, robbery) and therefore questioned the state's classification of the crime they had committed, but only 18 claimed that they had committed no crime at all. Similarly, 382 inmate respondents reported a conviction of robbery, and 287 admitted that they had committed that crime. Only 44 of the remaining 95 respondents convicted of robbery claimed that they had committed no crime. This pattern holds for the great majority of crimes; the exceptions are rape and other sex offenses, for which the rate of denial was much higher. In the case of rape, 61 inmate respondents reported a conviction of rape, only 17 conceded that they had committed that crime, and 23 claimed to have committed no crime at all.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 POVEDA 703

DISCUSSION AND CONCLUSIONS

The prevalence of wrongful convictions as measured by inmates' self-reports is significantly greater than the estimate derived from using official data based on court-ordered discharges (15% versus 1%). 20 This finding perhaps is not surprising, but it does not support the idea that alternative methodologies might produce similar results. Although the RAND assessment of prisoners' self-reports challenged the conventional wisdom that inmates in general deny their criminal activities, some prisoners nonetheless have a variety of self-serving motives for underreporting their crimes. Even though they have been convicted and imprisoned, some inmates still may have forthcoming appeals, and therefore, in spite of the anonymity or confidentiality of a survey, a personal stake in minimizing their involvement in the conviction crime. It is also possible, as Loftus (1996) argued so persuasively with eyewitness testimony, that postevent information may reshape inmates' memories of their criminal conduct, including their current conviction offenses. 21 Perhaps defense arguments and accounts of the crime presented at trial or on appeal, by minimizing or denying defendants' involvement, may subtly influence memories of the actual crime. 22 It may also be that inmates' perceptions of their criminal conduct are at variance with the legal definition. For example, Koss, Gidycz, and Wisniewski (1987), who researched the prevalence of rape on college campuses, observed that male offenders tended not to recognize their conduct as rape; yet they admitted to having sex with a female against her will. 23 This disparity in definitions of

20 Technically, the comparison here should be between 17.5 percent (the proportion of convicted murderers denying the charges in the RAND survey) and the 1.4 percent error rate for murder convictions/commitments based on court-ordered discharges. 21 This is evident in the growing number of cases where postconviction DNA testing provides incriminating evidence rather than exonerating the inmate. A case in point is that of Texas death row inmate Ricky McGinn. In June 2000 Governor George Bush gave McGinn a 30-day reprieve to allow time for DNA testing before his execution. All of the DNA results subsequently showed a match between McGinn and evidence recovered from the victim. Even so, McGinn maintained his innocence (Texas Man" 2000). This also occurred in the Scott William Davi case in South Dakota: Davis's DNA results provided a match confirming the rape and murder of his ex-wife. Davi, too, maintained his innocence. The results of DNA testing at two major forensic laboratories (Forensic Science Associates and Cellmark) have confirmed prisoners' guilt in 60 percent of the cases (Cohen 2000:A1). ~2 Penningten and Hastie (1993) have written about the "story model" of juror's decision making, arguing that prosecutors and defense attorneys construct stories of the crime to make it more understandable to the jury. The story then is a way to organize the facts of the case in a manner consistent with the prosecution or the defense version of events. These "stories" might later influence the defendants' recollections of their crimes. 23 In particular, respondents disagree about what constitutes coercive or consensual sexual relations.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 704 WRONGFUL CONVICTIONS

rape (legal versus offender's definition) perhaps accounts for much of the underreporting of rape (and other sex offenses) in the RAND Inmate Survey. Another possibility, as noted by the RAND researchers, is that inmates misclassify their offense, perhaps confusing robbery with theft or murder with manslaughter. It is more difficult, however, to understand the complete denial of an inmate's conviction offense, short of lying or wrongful conviction.

An earlier RAND survey of 624 California male prisoners also concluded that "inmates did not deny committing crimes" (Peterson, Braiker, and Polich 1981:xxi), but found that 81 (13%) of those inmates denied committing any of the 11 crimes on the survey list. Peterson and his colleagues (1981:20) maintained that about half of the 81 had committed a crime not on their list; about 6 percent denied committing any crime, including the crime for which they had been imprisoned. Although this figure is less than the approximately 15 percent denial rate found in the 1978 RAND Inmate Survey, it is larger than the i percent estimate of wrongful convictions based on court-ordered discharges.

The two approaches employed in this study to estimate the prevalence of wrongful convictions not only resulted in different outcomes, but also are based on different methodologies. One draws on officially acknowledged error (court-ordered discharges); the other, on the uncorroborated claims of prisoners (inmates' selfreports), and thus on a more expansive definition of wrongful conviction. Because court-ordered discharges are the result of inmates who pursue their claims of justice-system error (whether substantive or procedural), it is not surprising that the outcome of that process will yield a smaller estimate, given the attrition of cases. Moreover, future research on court-ordered discharges must continue to distinguish between persons released for procedural and for substantive reasons, and must reexamine the proportion of cases involving wrong-person convictions. 24 Although the extent of wrongful convictions remains a "dark figure," the divergent fmdings of the two approaches used here may delineate the parameters of the problem: prevalence ranges from 1 to 15 percent, depending on assumptions and operational definitions. Future research also must take into account variation in wrongful convictions over time. Drawing on Packer's (1968) distinction between "crime control" and "due process," Huff and his colleagues (1996:143-44) note a possible link between wrongful

24 The 21.2 percent estimate used in this study is based on one study, the 1989 DOCS research on vacated murder convictions, which needs to be explored more fully.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 POVEDA 705

convictions and a crime-control model of the criminal justice process. This suggests that in historical periods which emphasize crime control over due process (as in "get tough" crime policies), errors of justice are also more likely, including wrongful convictions. Until now, the absence of a methodology for measuring justice-system error has made this proposition untestable.

Estimating the prevalence of wrongful convictions is not merely a matter of academic interest. IZ~owing the magnitude and scope of the problem has important public policy implications. Although we may never know exactly how many innocent persons have been convicted and imprisoned, the extent of this problem indicates how well (or how poorly) our criminal justice system is sorting the guilty from the not guilty; certainly this is a major goal of the system. The extent of error provides a quality control mechanism and alerts us to systemic flaws, whether they pertain to mistaken eyewitness identification, police interrogation practices, ineffective counsel, or the use ofjailhouse informers.

In a 1999 report by the Institute of Medicine of the National Academy of Sciences (NAS), it was estimated that medical mistakes annually kill from 44,000 to 98,000 hospital patients (Corrigan, Kohn, and Donaldson 1999). The report emphasized the systemic basis for most errors in the health care system, rather than individual recklessness, and identified some of the organizational measures that can be taken to prevent death and injury from medical mistakes (NAS 1999). Although officials and providers in the health care system have been slow 25 to recognize the extent of medical mistakes and their implications for public safety, this report represents a reversal.

Officials in the criminal justice system can learn from the experience of other institutions by publicly acknowledging justicesystem errors, conducting inquiries when they occur, and beginning to quantify the extent of the problem. 26 Only then can the problem of wrongful convictions be addressed in terms of systemic flaws, not merely as individual wrongdoing nor as honest mistakes that are quickly forgotten.

25 Action has been slow in relation to some other sectors of the economy in which public safety issues are well established, such as the airline and automobile industries and workplace safety in general. Medical malpractice suits, of course, have contributed to the health care system's reluctance to publicly acknowledge medical mistakes. 26 In May 2000, New York Governor George Pataki proposed legislation to cre~ ate a DNA Review Committee. This committee would have the authority to review cases in which DNA has exonerated offenders, and to recommend legal or procedural changes to prevent future recurrences ("Gov. Pataki" 2000).

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 706 WRONGFUL CONVICTIONS

REFERENCES

Bedau, H.A. and M. Radelet. 1987. "Miscarriages of Justice in Potentially Capital Cases." Stanford Law Review 40:21-179. Borchard, E.M. [1932] 1970. Convicting the Innocent: Errors of Criminal Justice. New York: Da Capo. Cassell, P. 1998. "Protecting the Innocent From False Confessions and Lost ConfessionsAnd From Miranda." Journal of Criminal Law and Criminology 88:497- 556. Chaiken, J.M. and M.R. Chaiken. 1982. Varieties of Criminal Behavior. Santa Monica, CA: RAND. Chapman, W.R. and S.A. Zausner. 1985. Characteristics of Inmates Under Custody: A Ten-Year Trend Study. Albany, NY: Department of Correctional Services. Cohen, L.P. 2000. "Someone Who's Guilty Would Never Want DNA Testing, Right?" Walt Street Journal, July 12, pp. A1, A12. Connors, E., T. Lundregan, N. Miller, and T. McEwen. 1996. Convicted by Juries, Exonerated by Science: Case Studies in the Use of DNA Evidence to Establish Innocence After Trial. Washington, DC: U.S, Department of Justice. Corrigan, J., L. Kohn, and M. Donaldson, eds. 1999. To Err Is Human: Building a Safer Health System. Washington, DC: Institute of Medicine, National Academy of Sciences. Cutler, B.L. and S.D. Penrod. 1995. Mistaken Identification: The Eyewitness, Psychology, and the Law. New York: Cambridge University Press. Death Penalty Information Center (DPIC). 2001. "Innocence: Freed From Death Row." Death Penalty Information Center Website. Retrieved May 7, 2001 (http:// www.deathpenaltyinfo.org). Department of Correctional Services (DOCS). 1989. Murder Commitments With Subsequent Sentence Vacations. Albany, NY: Division of Program Planning, Research, and Evaluation. --.1997a. Characteristics of Inmates Discharged, 1995. Albany, NY: Division of Program Planning, Research, and Evaluation. --.1997b. Characteristics of New Commitments, 1995. Albany, NY: Division of Program Planning, Research, and Evaluation. Devine, P.G. 1995. "Getting Hooked on Research in Social Psychology: Examples From Eyewitness Identification and Prejudice." Pp. 161-84 in The Social Psychologists, edited by G. Brannigan and M. Merrens. New York: McGraw-Hill. Dieter, R.C. 1997. "Innocence and the Death Penalty: The Increasing Danger of Executing the Innocent." Death Penalty Infbrmation Center Website. Retrieved April 13, 2000 (http://www.deathpenaltyinfo.org). Ewinger, J. and J.F. Hagan. 2000. "Jurors Issue Rapid Verdict." Cleveland Plain Dealer, April 13. Retrieved April 13, 2000 (http://www.cleveland.com/news). "Governor Pataki to Establish DNA Review Committee." 2000. Press release, May 8. Retrieved Aug. 10, 2000 (http://www.state.ny.us/governor/press/year00/may8 _00.htm). Gross, S.R. 1996. "The Risks of Death: Why Erroneous Convictions Are Common in Capital Cases." Buffalo Law Review 44:1-21. Retrieved Dec. 22, 1998 (http'J/ web.lexis-nexis.com/universe). --.1998. "Lost Lives: Miscarriages of Justice in Capital Cases." Law and Contemporary Ioblems 61:125-52. Hagan, J.F. and J. Ewinger. 2000. 'Third Trial Fails to Overcome Burden of Proving Sheppard Was Innocent." Cleveland Plain Dealer, April 13. Retrieved April 13, 2000 (http://www.cleveland.com/news). "History of the Sheppard Case." 2000. Cleveland Plain Dealer, Feb. 3. Retrieved April 13, 2000 (http://www.cleveland.. indepth/sam/stories). Huff, C.R., A. Rattner, and E. Sagarin. 1986. "Guilty Until Proved Innocent: Wrongful Conviction and Public Policy." Crime and Delinquency 32:518-44. --.1996. Convicted but Innocent: Wrongful Conviction and Public Policy. Thousand Oaks, CA: Sage. Johnson, D. 2000. "Illinois, Citing Verdict Errors, Bars Executions." New York Times, February 1, pp. A1, A16.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 POVEDA 707

Koss, M.P., C.A. Gidycz, and N. Wisniewski. 1987. "The Scope of Rape: Incidence and Prevalence of Sexual Aggression and Victimization in a National Sample of Higher Education Students." Journal of Consulting and Clinical Psychology 55:162-70. Leo, R.A. 1996. "Inside the Interrogation Room." Journal of Criminal Law and Criminology 86:266-303. Leo, R~ and R.J. Ofshe. 1998a. "The Consequences of False Confessions: Deprivations of Liberty and Miscarriages of Justice in the Age of Psychological Interrogation." Journal of Criminal Law and Criminology 88:429-96. .1998b. "Using the Innocent to Scapegoat Miranda." Journal of Criminal Law and Criminology 88:55%78. Liebman, J.S., J. Fagan, and V. West. 2000. "A Broken System: Error Rates in Capital Cases, 1973-1995." The Justice Project Website. Retrieved June 30, 2000 (http://207.153.244.129/liebman1.pdf). Loius, E.F. 1993. "Psychologists in the Eyewitness World." American Psychologist 48:550-52. --.1996. Eyewitness Testimony. Cambridge, MA: Harvard University Press. Loftus, E.F. and K. Ketcham. 1991. Witness ['or the Defense: The Accused, the Witness, and the Expert Who Puts Memory on Trial. New York: St. Martin's. Marquis, K.H. and P.A. Ebener. 1981. Quality of Prisoner Self-Reports: Arrest and Conviction Response Errors. Santa Monica, CA: RAND, National Academy of Sciences (NAS), Institute of Medicine. 1999. "Preventing Death and Injury From Medical Errors Requires Dramatic, System-Wide Changes." National Academies News, November 29. Retrieved December 30, 1999 (http:// www4.nationalacademies.org/news). National Archive of Criminal Justice Data. 2000. "Survey of Jail and Prison Inmates, 1978: California, Michigan, and Texas." ICPSR Study 8169. Ann Arbor, MI: ICPSR. Retrieved May 2000 (http://www.icpsr.umich.edu/NACJD). New York State Court of Claims. 1998. "Court of Claims Act." Retrieved June 10, 1998 (http://www.nyscourtofclaims.state.ny.us). New York State Defenders Association (NYSDA). 1989. Convicting the Innocent: Wrongful New York State Homicide Convictions, 1965-1988. A Preliminary Report. Albany, NY: Public Defense Backup Center. Ofshe, R.J. and R.A. Leo. 1997. "The Decision to Confess Falsely: Rational Choice and Irrational Action." Denver University Law Review 74:979-1122. Packer, H. 1968. The Limits of the Criminal Sanction. Stanford, CA: Stanford University Press. Pennington, N. and R. Hastie. 1993. "The Story Model for Juror Decision Making." Pp. 192-221 in Inside the Juror: The Psychology of Juror Decision Making, edited by R. Hastie. Cambridge, UK: Cambridge University Press. Peterson, M.A., H.B. Braiker, and S.M. Polich. 1981. Who Commits Crimes: A Survey of Prison Inmates. Cambridge, MA: Oelgeschtager, Gunn, and Haln. Petersen, M., J. Chaiken, P. Ebener, and P. Honig. 1982. Survey of Prison and Jail Inmates: Background and Method. Santa Monica, CA: RAND. Radelet, M.L. and H.A Bedau. 1998. "The Execution of the Innocent." Law and Contemporary Problems 61:105-24. Radetet, M.L., H.A. Bedau, and C.E. Putnam. 1992. In Spite of Innocence: Erroneous Convictions in Capital Cases. Boston, MA: Northeastern University Press. Radin, E.D. 1964. The Innocents. New York: Morrow. Rosenbaum, M.I. 1990-1991. "Inevitable Error: Wrongful New York State Homicide Convictions." Review of Law and Social Change 18:807-30. Scheck, B., P. Neufeld, and J. Dwyer. 2000. Actual Innocence. New York: Doubleday. "Texas Man Gets New Execution Date." 2000. New York Times, August 15. Retrieved August 15, 2000 (http://www.nytimes.com). U.S. House of Representatives. 1994. Innocence and the Death Penalty: Assessing the Danger of Mistaken Executions. Staff Report by Subcommittee on Civil and Constitutianal Rights of the Committee on the Judiciary. 103rd Congress, 2nd Session. Washington, DC: U.S. Government Printing Office. Webb, E.J., D.T. Campbell, R.D. Schwartz, and L. Sechrest. 1966. Unobtrusive Measures: Nonreactive Research in the Social Sciences. Chicago, IL: Rand McNally. Wells, G.L. 1993. "What Do We Know About Eyewitness Identification?" American Psychologist 48:553-71. Yant, M. 1991. Presumed Guilty: When Innocent People Are Wrongly Convicted. Buffalo, NY: Prometheus.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014 708 WRONGFUL CONVICTIONS

Yuille, J.C. 1993. "We Must Study Forensic Eyewitnesses to Know About Them." American Psychologist 48:572-73.

Downloaded by [University of Chicago Library] at 04:21 11 December 2014