Skip navigation

Uclalr Cassell Empirical Study of Miranda Effects on Police Interrogations Feb 1996

Download original document:
Brief thumbnail
This text is machine-read, and may contain errors. Check the original document to verify accuracy.
Page 1

LEXSEE 43 UCLA L. REV. 839
Copyright (c) The Regents of the University of California 1996.
UCLA Law Review
February, 1996
43 UCLA L. Rev. 839
LENGTH: 36405 words
DIALOGUE ON MIRANDA: POLICE INTERROGATION IN THE 1990S: AN EMPIRICAL STUDY OF THE
EFFECTS OF MIRANDA
NAME: Paul G. Cassell * and Bret S. Hayman **
BIO:
* Associate Professor of Law, University of Utah College of Law. Formerly Assistant U.S. Attorney, U.S.
Dept. of Justice 1988-91; Associate Deputy Attorney General, U.S. Dept. of Justice 1986-88. B.A., 1981,
Stanford University; J.D., 1984, Stanford University.
** Third year law student, University of Utah College of Law. B.S., 1993, University of Utah.
Helpful comments on an earlier draft were provided by Lee Teitelbaum, Yale Kamisar, Bill Stuntz, and
George Thomas. We gratefully acknowledge the assistance of Brendan McCullagh (class of '95), who helped to
collect the data described in this Article; Karen L. Korevaar, who provided valuable insight on statistical
analysis and ran our statistical significance tests; and Kathleen Morgan, who salvaged this project by sorting all
of our data. We also gratefully acknowledge the assistance of District Attorney Neil Gunnarson and County
Attorney David E. Yocom and their able felony prosecutors, who graciously allowed us to collect the data.
Finally, we appreciate the editors of the UCLA Law Review for arranging to publish this Article as a dialogue
with Professor Thomas's preceding and following essays. This Article was supported by the University of Utah
College of Law Research Fund and the University of Utah Research Committee.
SUMMARY:
... Miranda v. Arizona is the Supreme Court's most famous criminal law decision, spelling out the requirements for
police interrogation of criminal suspects. ... D quantifies the effect on police questioning of such factors as the strength
of the evidence against a suspect, prior criminal record, tape recording, race, gender, and type of police officer. ... The
result of questioning was "successful" when the police: (1) obtained a written confession; (2) obtained a verbal
confession; (3) obtained an incriminating statement; or (4) locked a suspect into a false alibi. ... Drawn from this case,
the Beheler warning consists of telling a suspect that he is not under arrest and is free to leave during the "interview." ...
However, in all likelihood, an interrogation involving a confession or incriminating statement takes longer than an
interrogation where the suspect has little to say. ... Thus, under Professor Thomas's view of the world, the critical
subject for empirical inquiry should be not simply whether the confession rate fell after Miranda but why it fell -- in
other words, to what degree that reduction stems from eliminating unconstitutional coercion or from prophylactic
overkill preventing otherwise constitutional questioning techniques. ... The adjusted 38.7% confession rate derived from

Page 2
43 UCLA L. Rev. 839, *

Professor Leo's data is almost indistinguishable from our 33.3% rate. ...
TEXT:
[*839] [*840] INTRODUCTION
Miranda v. Arizona n1 is the Supreme Court's most famous criminal law decision, n2 spelling out the
requirements for police interrogation of criminal suspects. Among other things, the decision required police officers to
deliver warnings to suspects about their right to remain silent and to obtain a waiver of these rights before beginning
any custodial interrogation. From the day the Court issued its decree in 1966, questions loomed about the new rules'
effects on police questioning. In the immediate wake of the decision, a handful of researchers attempted to provide
answers with empirical studies. n3 But interest in the subject quickly waned, and research in the last couple of decades
has been virtually nonexistent.
In this "empirical desert," n4 we have little knowledge about what police interrogation looked like shortly after
Miranda, much less what it looks like today. How many suspects waive their Miranda rights? How many confess? How
important are confessions to the outcome of prosecutions? Even the most informed observers can offer little beyond
speculation on these fundamental subjects. A few careful scholars have recognized the pressing need for more research
in this area. Perhaps the best example [*841] comes from the preceding article by Professor George Thomas, subtitled
"A Plea for More (and Better) Empirical Evidence" on Miranda. n5 There, Professor Thomas explains that "by the
mid-1970s, the issue of Miranda's real-world effect seemed to disappear. . . . [This issue] should be at the core of the
debate about Miranda." n6 More often, however, legal academics have not been as careful as Professor Thomas and
have simply made broad claims about our knowledge of police interrogation. For example, one of the nation's leading
hornbooks on criminal procedure reports that "defendants given the warnings seldom request counsel" and that "about
as many confessions are obtained by giving the Miranda rights as were gotten before the Miranda decision." n7 Other
prominent academics have proffered similar opinions. n8 Often these assertions have been in service of the position
that Miranda has not harmed the investigation and prosecution of criminal cases. Equally often, these claims are
couched in vague language that, upon closer inspection, says little. For instance, take the statement that suspects seldom
request counsel. Do they request counsel in 1% of all cases? In 30%? No one knows the answer.
This ignorance about police questioning has important public policy ramifications. The Supreme Court now
characterizes the Miranda requirements not as a constitutional command but rather as a "prophylactic" rule. n9 Under
this view, Miranda rests on cost-benefit analysis in which its costs in reducing the effectiveness of police interrogation
are assayed against its benefits in protecting suspects from abuses. n10 Given that the entire [*842] doctrine now rests
on a pragmatic foundation, it is curious to discover that virtually nothing is known about the decision's real world
effects. Without practical information, how can the Court say whether the balance struck with the Miranda rules was (or
remains) the right one?
To shed light on these important questions, we conducted a study in Salt Lake County, Utah, a large urban area. We
gathered information on police questioning and suspects' confessions n11 in a sample of more than 200 cases in the
summer of 1994. This Article presents our data on such subjects as the frequency of waivers of rights and confessions,
police compliance with Miranda, and the role of confessions in the outcome of prosecutions. Part I briefly reviews the
existing literature, which supports Professor Thomas's conclusion that we know little about confessions. Part II sets
forth the methodology used in this survey. Part III reports our specific findings. Part III.A assesses the extent to which
police question suspects, obtain waivers of Miranda rights, and gain incriminating statements. We found that a
surprisingly large number of suspects (about 21%) are never questioned, that a substantial fraction of suspects (about
16%) invoke their Miranda rights, and that only about 33% of all suspects give confessions. The 33% confession rate
we found is lower than the rates reported in studies done before Miranda and conducted in countries that do not follow
the Miranda requirements, suggesting that Miranda has reduced the confession rate. n12 Part III.B examines the
frequency with which police question suspects outside the Miranda regime in noncustodial and other settings, finding
that police have shifted to noncustodial questioning to some extent and that other exceptions to Miranda are
quantitatively unimportant. Part III.C reviews police compliance with the Miranda rules, finding compliance to be high.

Page 3
43 UCLA L. Rev. 839, *842

Part III.D quantifies the effect on police questioning of such factors as the strength of the evidence against a suspect,
prior criminal record, tape recording, race, gender, and type of police officer. Part III.E analyzes the effect of
confessions on the ultimate outcome of criminal cases, concluding that confessions make it significantly more likely
that a prosecutor will file charges and that the defendant will be convicted of those charges as filed. Part IV assesses the
importance of these findings for the on-going debate about the wisdom of the Miranda requirements. [*843] It
concludes by expanding on Professor Thomas's plea for additional research on Miranda and explaining how the
Miranda decision, despite the promises of the Court and its defenders, has yet to be empirically justified as the proper
balance between the competing interests of criminal suspects and society at large.
I. PREVIOUS EMPIRICAL RESEARCH ON POLICE QUESTIONING
A dearth of empirical research on police interrogations is no recent phenomenon. When the Supreme Court began
to consider new restrictions on questioning in the 1960s, it had amazingly little empirical information to draw upon. In
the late 1950s, the American Bar Foundation conducted field research in several cities, but presented little quantitative
analysis. n13 The first hard data on the subject was published in 1962 by Professor Edward L. Barrett, Jr., who studied
police practices in two California cities. n14 Unfortunately his study contained summary information only on the length
of interrogations and the frequency of confessions. n15 Thus, it is not surprising that the briefs in Escobedo v. Illinois,
n16 a 1964 case involving a general challenge to the Court's jurisprudence on police questioning, contained no
references to empirical studies. n17 Apart from a passing reference to Barrett's study, n18 the Escobedo opinions are
likewise devoid of empirical citations.
The Court's opaque ruling in Escobedo (suggesting that once police have begun to "focus" on a suspect they cannot
deny him access to his attorney n19 ) created a great deal of interest in -- and uncertainty about -- the law governing
police questioning. It did not, however, generate much new [*844] information on police questioning. n20 Two years
later, one commentator noted, "With all the hundreds of thousands of foundation and government dollars that are being
spent these days on what is known as 'the crime problem,' I find it incomprehensible that some dribbles have not been
spent upon a dispassionate survey concerning the significance of police interrogation." n21
In 1966, as the participants in the Miranda case prepared to file their briefs, they had little empirical information to
draw upon. To solve this problem, the National District Attorney's Association asked prosecutors in large cities to
collect information on the prevalence of confessions and their importance in obtaining convictions. The Association
reported its results in a forty-five-page appendix to its amicus brief in the case. n22 The ACLU countered with a brief
that relied heavily on police interrogation "manuals." n23
The police interrogation manuals became the centerpiece of what realworld support there was for the majority
opinion in Miranda. The Court's opinion did cite the famous 1931 Wickersham report on police abuse n24 and
anecdotal examples of police misconduct. n25 But aside from a passing nod acknowledging, "a gap in our knowledge
as to what in fact goes on in the interrogation rooms," n26 the decision rested on the supposed representativeness of
police manuals -- not scholarly investigation. n27 This led Justice [*845] Harlan to complain in dissent that the
Court's decision lacked "the vast advantage of empirical data." n28
Miranda sparked a brief flurry of interest in police interrogation, as both proponents and opponents of the decision
sought to bolster their positions. Perhaps the most famous study was published by the editors of the Yale Law Journal,
who camped out in the New Haven police department during the summer after the decision and reported on the
interrogations they observed. n29 A short postscript to the study assessed the use of Miranda warnings in the
questioning of draft protestors at Yale. n30 Other academic researchers conducted post-Miranda analysis of police case
files to determine the "before-and-after" prevalence of confessions in Pittsburgh, n31 the District of Columbia, n32 and
pseudonymous "Seaside City" (a suburb of Los Angeles). n33 Prosecutors undertook similar surveys in New York
County, n34 Philadelphia, n35 Los Angeles, n36 Brooklyn, n37 New Orleans, n38 Kansas City, n39 and Chicago.
n40

Page 4
43 UCLA L. Rev. 839, *845

Apart from these before-and-after assessments, only a few other studies on confessions were done in the wake of
Miranda, although they did not involve before-and-after assessments. In 1967, the Vera Institute surveyed [*846]
police interrogation practices in Manhattan. n41 In the same year, Neal Milner analyzed the impact of Miranda on
police attitudes in four Wisconsin cities. n42 In 1968, David W. Neubauer gathered statistics on the role of confessions
in prosecutions in Midwestern "Prairie City." n43 In 1969, law student Lawrence S. Leiken gathered information from
prisoners in Denver on whether they had given statements to police. n44
The best reading of these older studies is that confession rates fell substantially after Miranda, as one of the present
authors has argued at length elsewhere. n45 Professor Thomas, however, concludes that we must accept the "null
hypothesis that Miranda has made no statistically significant difference in the confession rate." n46 His conclusion
stems from just four studies -- New Haven, District of Columbia, Denver, Seaside City -- which in his view outweigh
the contrary finding of a fifth study in Pittsburgh, which reported a significant decline in confessions after Miranda.
n47
In reaching his conclusion, Professor Thomas may have been led astray on several points. The New Haven study,
contrary to the conclusions of its authors, reveals a substantial 16.0% drop in admissible confessions after Miranda.
n48 The D.C. study cannot be regarded as a true test of the [*847] Miranda procedures because the police there, at the
time of the study, did not follow the most important Miranda requirements. n49 Moreover, contrary to Professor
Thomas's reading, when the study's data is reanalyzed to compare suspects who received some Miranda warnings
compared with those who did not, one finds that confessions fell after Miranda. n50 The Denver study also sheds no
light on the harmful features of Miranda. n51 And the Seaside City study is anomalous in several respects. n52
Professor Thomas also possibly overlooks several studies showing harmful effects from Miranda. In particular,
Professor Thomas fails to consider the assessments from New York County, Philadelphia, New Orleans, Kansas City,
Brooklyn, and Chicago, all of which suggest that [*848] Miranda had an inhibiting effect on suspects. n53 The
findings of these studies are consistent with other bodies of data: a comparison of low confession rates in this country
after Miranda with higher rates in Britain and Canada; n54 the drop in violent crime clearance rates that accompanied
the imposition of the Miranda rules; n55 and reports of police officers about the harmful effects of the new rules. n56
A broader view of the data also raises questions about Professor Thomas's tentative "conjecture" that Miranda warnings
had a "double-effect" of simultaneously encouraging suspects to talk to police while discouraging admissions of guilt.
n57
Regardless of how the earlier studies are interpreted, however, we share Professor Thomas's concern with the lack
of research on critical Miranda issues since then. As Professor Thomas points out, in about 1970, interest [*849] in the
subject of confessions died. n58 Virtually no empirical studies on police interrogations have been published since. Only
a few haphazard statistics on confessions are available, typically gathered for other purposes (such as research on plea
bargaining). n59 The only recent study on interrogation is one by Professor Richard A. Leo in the San Francisco Bay
Area. n60 His innovative research involved observing more than 100 interrogations by detectives and provides an
unparalleled source of modern information about police tactics and methods. Because of the study's narrow focus on
station house interrogation by detectives, however, it provides no information on the broader course of police
questioning (or nonquestioning) in cases outside the station-house by officers other than detectives. n61
The lack of data in this country should be contrasted with Britain, where an extensive quantitative literature on
police questioning developed in recent years. Several Royal Commissions on Criminal Justice have supported research
on the subject, n62 and researchers have compiled statistics on many aspects of British police questioning. n63
As the thirtieth anniversary of Miranda draws near, then, research on its effects appears long overdue. As Professor
Thomas concisely explains in the concluding paragraph of his article, "We need more empirical evidence. What we
have so far raises more questions than it answers." n64
[*850] II. SOURCES OF DATA AND METHODOLOGY

Page 5
43 UCLA L. Rev. 839, *850

To answer some of the questions, we undertook an empirical survey of confessions and their importance. We
selected the Salt Lake County District Attorney's Office for the research because of its proximity and because of the
willingness of the prosecutors in the office to allow the study. n65 Salt Lake County has been the site of previous
criminal justice research, which found that it was quite typical in its processing of criminal cases. n66 Because
single-area studies raise questions of generalizability, it may be useful to set forth information about the county.
Salt Lake County is a primarily urban and suburban area comprised of Salt Lake City and several other cities and
incorporated areas. The residents of the county total 725,956, virtually all of whom live in what the Census Bureau
classifies as "urban" (that is, other-than-rural) areas. n67 About 10% of the population is of Hispanic origin or minority
race. n68 The County is policed by various law enforcement agencies, including the Salt Lake City Police Department,
the Salt Lake County Sheriff's Office, and various other smaller city police departments.
While there might be a tendency to think of Salt Lake County as a pastoral area free from the ravages of crime, the
county in fact has crime rates similar to other areas. n69 In 1993, 17,789 FBI "index" crimes (murder, rape, robbery,
aggravated assault, arson, burglary, vehicle theft, and larceny) were reported in Salt Lake County, resulting in a crime
rate of 55.9 per 1000 population, n70 just slightly above the national average of 54.8. n71 Within the county, Salt Lake
City proper has a higher crime rate than [*851] other cities of similar size n72 and, in contrast to other areas, crime
increased in 1994. n73 Like other urban areas, Salt Lake County has a developing gang problem. n74 Salt Lake
County's crimes, however, are less often violent than the national average. n75
We collected the basic data for the study by attending "screening" sessions held at the Salt Lake County District
Attorney's office during a six-week period in the summer of 1994. n76 Prosecutors in the District Attorney's Office
screen all felony cases for prosecutive merit. Misdemeanors are reviewed by a separate section, located in a separate
office building. Our sample was drawn from the felony screening process. n77
Through the gracious consent of the District Attorney's Office, researchers n78 were allowed to attend all
screenings unless the case raised some special sensitivity. The screening session is a forty-five-minute interview by the
prosecutor of the police officer concerning the evidence supporting the filing of charges. Screenings take place soon
after the officer completes investigation of the case. For suspects in custody, the screening takes place on the next
working day.
During the screening session, the researchers filled out survey forms regarding relevant case characteristics. n79
Over the summer, the researchers were allowed to observe all but four screenings out of more than two hundred they
went to attend. n80 Overall, the researchers estimate that they attended approximately 70%-80% of the screenings held
during the six-week period. n81 The main reason they missed screenings was simultaneous scheduling. When faced
with a conflict, the researchers went first [*852] to screenings for crimes of violence and last to screenings for drug
offenses. n82 The researchers also may have missed a few unscheduled "special" screenings for high profile cases and
a few other screenings for child abuse n83 and small drug cases. n84 As a result, the sample of cases drawn here
should generally reflect the caseload of the District Attorney's Office, with the exception that high profile, child abuse,
and narcotics cases may be underrepresented.
There were 219 suspects in our sample. Most (86.8%) were male. n85 Most (70.2%) had prior records. n86 The
age of the suspects ranged from 18 n87 to 63, with the median around 24. The racial composition of the sample was:
white (68.2%); Hispanic (23.9%); African-American (6.0%); and other (2.0%). n88
The most serious charge under consideration during the screening is indicated on Table 1. Most of the charges were
property crimes while the balance were violent crimes. n89
Table 1 -- Charges Considered Against Suspect
Charge
No.
VIOLENT CRIMES

91

%
41.6%

Page 6
43 UCLA L. Rev. 839, *852

Murder

3

Attempted Murder

4

Robbery

9

Kidnapping

1

Felony Assault

19

Sex Crime-Adult Victim

12

Sex Crime-Child Victim

6

Child Abuse

2

Drugs

15

Fleeing/Escape

9

Weapons

5

Arson

3

Other

3

PROPERTY CRIMES

128

58.4%

Theft

49

Burglary

35

Auto Theft

18

Larceny/Forgery

24

Other
TOTAL

2
219

100.0%

[*853] [*854] III. THE INTERROGATION PROCESS AND ITS OUTCOMES
The subparts below report our findings on various aspects of police interrogation. Before doing that, it may be
useful to set out our plan of attack. In each subpart, we set out the existing literature on the particular subject, describe
our finding on the question at hand, and finally draw our (admittedly more subjective) conclusions from the data.
A. The Frequency of Questioning, Waivers, and Confessions
1. Questioning and Nonquestioning
One important issue that has not been the subject of much empirical study is the frequency with which suspects are
questioned. Not every person who is arrested will be questioned. n90 In our sample, police questioned 79.0% of the
suspects. n91 This means that a surprisingly large percentage (21.0%) were not questioned.
Our 1994 data that a significant proportion of suspects are never questioned, read in conjunction with older studies,
support the hypothesis that questioning rates have declined somewhat since Miranda. n92 Our results parallel a 1979
study of police practices in two cities, which found that police never questioned 18.5% of arrested burglary suspects in
Jacksonville, Florida, and 20.1% of such suspects in San Diego, California. n93 However, older studies from around
the time of Miranda suggest that almost every suspect was questioned then. The Pittsburgh study found that, at least for
cases handled by the Detective Branch, virtually all suspects were questioned. n94 Likewise, the Seaside City study
reported data for cases in which suspects were "actually arrested and incarcerated," n95 [*855] finding that in such
cases officers failed to question suspects in only 2% of pre-Miranda cases. n96 Two California cities in 1961 had
confession rates of 58% and 88%, n97 implying interrogation rates in excess of these figures. In Philadelphia in 1964,

Page 7
43 UCLA L. Rev. 839, *855

an estimated 90% of arrested suspects gave statements, n98 implying an interrogation rate in excess of 90%. Similarly,
in Brooklyn, only 10% of suspects for serious crimes refused to give statements, n99 implying that police were
questioning more than 90%. The New Haven study found that detectives questioned all but two of 90 suspects (97.8%)
over the summer of 1966, at a time when New Haven police appeared to be following pre-Miranda operating
procedures. n100 The only clear exception to the older studies reporting a very high questioning rate appears to be the
1967 Washington, D.C. study, which reported that only 55% of suspects were interrogated before Miranda. n101
Of course, even if interrogation rates have declined over time, a remaining question would be whether this is due to
Miranda. n102 Here the data directly on point are quite limited. However, the only two studies to address this question
specifically report lower interrogation rates after Miranda. In Seaside City, the interrogation rate dropped from 98%
before Miranda to 92% after; in the District of Columbia, the interrogation rate apparently went from 55% before
Miranda to 48% after. n103
[*856] Given that the only data on point suggest lower interrogation rates after Miranda, it is useful to try and
identify mechanisms that might cause this to happen. More generally, one would think that police would interrogate
every suspect, especially in view of the fact that a suspect is "the most knowledgeable and unimpeachable source of
information about his past conduct." n104 Why is it that police fail to question so many suspects?
To shed light on these subjects, we collected information on why police failed to interrogate suspects (as reported
by the officer at screening). Our data are contained in Table 2. As can be seen, in two cases (4.9% of the nonquestioning
cases), the reason for not questioning was a belief that the suspect would invoke Miranda. n105 In two other cases
(4.9% of the nonquestioning cases), the police cited the fact that a suspect had an attorney as the obstacle to
questioning, a reason that has a possible connection with Miranda. n106 Our data thus confirm what is suggested in the
earlier studies: that some suspects will not be questioned due to Miranda. n107
[*857] The most often-given reason for nonquestioning was that a suspect's whereabouts was unknown at the time
police screened the case. In many of these cases, it appeared that officers sought the filing of charges because they
wanted an arrest warrant for a hard-to-locate suspect entered into the law enforcement computer system. When the
suspect was later picked up (as he inevitably would be, for a minor traffic violation or some other scrape with the law),
the warrant would pop up and the suspect would be detained. Because we drew our sample at the initial screening, it
was possible that police later interviewed some of these suspects when they were apprehended. To collect data on such
interviews, we attempted to follow up with the screening officer to determine if police later interrogated. If so, we noted
the results and moved the suspect into our "questioned" category. While it is possible we missed a few of these
post-screening interrogations, the officers also reported that such interrogation was frequently difficult. In many cases,
the investigating officer (the only one with sufficient information to effectively question a suspect) could not be
informed of the arrest in a timely fashion. n108 In others, the suspect would receive routine assignment of counsel
before questioning was possible, a reason for not questioning that might be attributable to Miranda. n109
The second most common reason for failure to question was the officer's belief that the case against the suspect was
overwhelming. One possible conclusion to be drawn is that police agencies should be more diligent in attempting to
obtain incriminating statements. It is likely that prosecutors, faced with the task of ultimately obtaining convictions,
would be less sanguine about the assumption that so many cases were "over-whelming." Moreover, in a few cases the
officer pled the press of "other business" or assumed unproductiveness as reasons for not questioning. n110 These
reasons can also be read to support Professor Thomas's suggestion [*858] that how badly police want a confession is
an important variable in the confession equation. n111
In sum, our data fit the trend of lower interrogation rates after Miranda and suggest that Miranda has played at least
some role in that drop.
Table 2 -- Reasons for Not Questioning Suspects (N=46; 5 unavailable)
Reason
No.
%

Page 8
43 UCLA L. Rev. 839, *858

Suspect's Whereabouts Unknown

14

34.1%

Overwhelming Case

11

26.8%

Suspect Intoxicated

7

17.1%

No Opportunity Given Press of Business

2

4.9%

Belief Suspect Would Invoke Miranda

2

4.9%

Suspect Already Had an Attorney

2

4.9%

Questioning Thought Unproductive

1

2.4%

Suspect in Rehabilitation Program

1

2.4%

Suspect Injured

1

2.4%

TOTAL

41

100.0%

2. Invocations of Miranda Rights
One of the most important questions about the Miranda regime is how often suspects invoke their Miranda rights,
preventing any police questioning. A suspect can claim Miranda rights in two ways. First, he can refuse at the start of
an interview to waive his rights (including the right to [*859] remain silent and the right to counsel), thus precluding
any interview. Second, even if he initially waives his rights, he can assert them at any point in the interview. n112 If a
suspect asserts Miranda rights, police questioning must stop. The effect of these questioning cutoff rules is important,
because it appears that most of Miranda's harms stem from the cutoff rules, not the more famous Miranda warnings.
n113
Surprisingly very little information is available on such a fundamental subject. The previously published evidence
on invocations has been collected elsewhere. n114 The evidence, although generally quite dated, suggests that about
20% of all suspects invoke their Miranda rights.
We gathered data on how often and in what ways suspects asserted their Miranda rights. Table 3 shows that of
suspects given their Miranda rights, 83.7% waived them. n115 Reflecting the practices of the local law enforcement
agencies, virtually all of these waivers were verbal rather than written. At the same time, 16.3% invoked their rights. Of
the twenty-one suspects who invoked their rights, nine invoked their right to an attorney (two even before Miranda
warnings could be read n116 ), six invoked their right not to make a statement, and six either refused to execute a
waiver or otherwise invoked their rights. For these purposes, a suspect who partially waived his rights (that is, agreed to
talk about some charges but not others), was classified as waiving his rights. n117
[*860] Table 3 -- Invocation of Rights
Suspect's Response to Miranda Rights
WAIVED
Throughout

No.

%

108
103

79.8%

5

3.9%

Changed Mind Later
INVOKED

83.7%

21

16.3%

Asked for Attorney

9

7.0%

Invoked Right to Silence

6

4.7%

Refused to Waive/Other

6

4.7%

TOTAL

129

100.0%

Page 9
43 UCLA L. Rev. 839, *860

As also shown in Table 3, five suspects who initially waived their rights changed their minds later and invoked
their rights during the interview. Of these five suspects, three asked for an attorney during the interview (two after
giving incriminating statements, one after giving a denial with an explanation), and two asserted their right to remain
silent during the interview (one after giving a flat denial, one after giving a denial with an explanation). Thus, three of
these five suspects invoked their rights before successful police interrogation. If these three are added to suspects who
invoked their rights initially, then a total of 18.6% of the suspects in our sample who were given Miranda rights
invoked them before police succeeded in obtaining incriminating information. The percentage viewed as invoking their
rights shrinks slightly if we add to the suspects given Miranda warnings those who were never given the warnings. If
these suspects are added in, then 12.1% of all the suspects in our sample who were questioned invoked their rights
before police were successful in interrogation. n118
[*861] One characterization issue was how to treat suspects who were interviewed several times, invoking their
rights in one interview, but not in another. We had five such cases, four involving invocations in the second interview,
one in the first. We included three of these cases in our "invoked" list because that seemed to describe the course of
questioning more accurately. n119 We thought the other two cases were best treated as not having been asked for a
waiver of rights. n120
Finally, in none of our cases did the police continue questioning a suspect after an invocation of Miranda rights.
This finding has considerable importance in the on-going debate about Miranda's scope. The Court crafted a limitation
to the Miranda rule in 1971, holding that prosecutors could use statements obtained in violation of Miranda for
impeachment. n121 Because of this exception, prestigious academic commentators have advanced the suggestion that
police would deliberately flout Miranda to obtain impeaching statements. n122 We found no evidence that this was
occurring. n123
Determining whether the frequency of invocation of rights we found is "high" or "low" is a matter of judgment that
will not be pursued at [*862] length here. It bears noting, however, that a 12.1% invocation of rights figure has
substantial public policy implications. If our figure is typical of the nation as a whole, then each year approximately
300,000 criminal suspects for FBI index crimes invoke their rights before successful police questioning. n124
3. Confessions, Incriminating Statements, and Denials
Perhaps the critical issue in the debate over Miranda is how often today suspects confess or otherwise make
incriminating statements. Surprisingly little information is available on this subject, despite frequent definitive
pronouncements that many suspects still confess. n125 In this subsection, we first explain our methodology for
classifying "confessions," "incriminating statements," and "denials." We then report our findings and conclude that our
confession rate is lower than the rate that prevailed before Miranda. We end with a brief discussion of other purposes
for interrogation besides obtaining incriminating statements.
a. Classifications
We gathered detailed information on the frequency of confessions, incriminating statements, denials, and other
results of police questioning. While categorizing statements is always difficult, n126 we divided the outcomes into
"successful" and "unsuccessful" categories -- looking at the results from law enforcement's point of view. n127 The
result of questioning was "successful" when the police: (1) obtained a written confession; (2) obtained a verbal
confession; (3) obtained an incriminating statement; or (4) locked a suspect into a false alibi.
[*863] A written or verbal "confession" was "any substantial acknowledgment by the suspect that he or she
committed the crime." n128 Simply put, the suspect had to say in essence "I did it." For example, in one case police
questioned a suspect about the theft of various checks. He admitted cashing the checks for two female friends and
acknowledged that he knew they were stolen. We classified this statement as a confession. A "confession" could include
statements in mitigation. In the example above, we classified the suspect's statements as a confession even though he

Page 10
43 UCLA L. Rev. 839, *863

said that "the women were using me" and that he had derived no money from the crime.
Our next category was for "incriminating statements." Because this category is somewhat subjective and because its
scope makes a critical difference in determining the scope of police success statistics, it is worth explaining in detail. An
"incriminating statement" was "any statement that tends to establish guilt of the accused or from which with other facts
guilt may be inferred or which tends to disprove some anticipated defense." n129
We found that such statements generally occurred in one of three situations. First, a suspect might give a statement
that linked him to the crime or crime scene without admitting his guilt. For example, in an attempted homicide case the
police asked the suspect if a certain car, known to have been involved in the crime, was his. He responded that it was,
but denied any involvement in the crime. As another example, a suspect was questioned about the theft of a gun. He
admitted having the gun and acknowledged that he thought the gun was stolen. However, he denied stealing the gun.
We classified these statements as "incriminating."
A second kind of incriminating statement was a partial admission of guilt. For example, in one case involving a
burglary, a suspect admitted being in the house unlawfully, but said he was there because of something the victim had
done to his girlfriend, not to steal anything. We classified these statements as "incriminating."
Third, a suspect might make statements to the police that tended to call into question his truthfulness or make him
appear suspicious. For example, in one case an officer suspected a car was stolen and pulled over the suspect. The
officer asked the suspect certain questions concerning his license, registration, and his relationship to the car. The
suspect was visibly shaken by the questions and gave stammering and sometimes contradictory answers. We classified
these statements as "incriminating."
[*864] After confessions and incriminating statements, the final category of "successful" outcomes was for a
suspect who was "locked into a false alibi." This category was for a suspect who gave statements that committed him to
a scenario and the prosecution had substantial evidence that the scenario was untrue. n130 In evaluating our
classifications of statements, it should be noted that we classified some "on the fence" statements as "incriminating"
when they might have legitimately been classified as denials of some type. n131
On the other side of the coin, an interrogation was "unsuccessful" when the police obtained from the suspect: (1) a
"flat denial"; (2) a "denial with explanation"; or (3) "other," unincriminating statements.
A "flat denial" was the statement "I didn't do it" or "I don't know what you're talking about" and essentially nothing
else. n132
A "denial with explanation" involved a suspect who denied the crime and provided "additional statements regarding
circumstances surrounding the offense, an alibi, or mitigating version." n133 These were possibly the most difficult
statements to classify, because many denials with [*865] explanation, expansively viewed, might have somehow
connected the suspect to the crime scene or called into question his truthfulness, making it an "incriminating" statement.
We used this category for statements by defendants denying involvement in the crime, but involving more than the
simple statement "I didn't do it." We also used the category for suspects' statements whose primary import was not to
assist police in proving their case but rather in describing the suspects' actions in a way that involved no criminal
wrongdoing. In close cases where the researchers were unsure whether to catalog a statement as "incriminating" or as a
"denial with explanation," they relied on the view of the police officer and prosecutor screening the case.
The most common denial with explanation was a statement by a suspect claiming authorization to do what would
have been a crime without the authorization. In particular, in a number of cases suspects caught in possession of
"stolen" goods said that the goods were legitimately obtained. For example, a woman was caught leaving the victim's
home with a television in her hand. She told police that the victim allowed her to take the television to pay off a debt.
The main effect of this statement was to deny an essential element of the offense of theft. n134 Along the same lines, in
several stolen vehicle cases, suspects stopped in stolen cars denied knowing that the car was stolen or claimed to have

Page 11
43 UCLA L. Rev. 839, *865

recently purchased the car. n135 A related kind of statement came from suspects claiming authority to be where they
were. For example, police caught a suspect in a home not his [*866] own. In response to the police question "What are
you doing here?," the suspect said he had been told he could go in. The main thrust of the statement was to deny an
element of the offense of burglary. n136 We classified all of these statements as "denials with explanations." These
statements themselves did not add anything to what the police already knew and could readily prove. To the contrary,
the statements were all likely to be the centerpiece of a defense to the charges. Of the forty denials with explanation,
roughly fourteen were of this type.
The next most common denial with explanation occurred in sex crimes, where suspects admitted sex acts but said
there was consent. For example, a suspect accused of rape said that the victim had "come on to him" and that they had
consensual sex. This statement denied an essential element of the crime n137 and, again, would likely be featured by
the defense, not the prosecution. Roughly nine cases fit this description. n138
The remaining denials with explanations could not be conveniently categorized. We give a few examples so that the
reader can get the flavor of this category. In three cases, suspects in assault cases said they were fighting only to defend
themselves. (Two of the suspects were arrested in the same case for fighting with each other.) In an attempted homicide
case, a suspect admitted being with the shooter (a fact that was never in doubt -- both the car and another person in the
car were identified as being at the scene), but claimed he did not shoot anyone and that he did not know the shooter was
going to fire at the victim. In a child abuse case, a father claimed that burn marks on his son's arms were not, as police
alleged, cigarette burns, but rather accidental burns from a camping trip. After an arrest for a drug deal, two suspects
each claimed that they had never seen the other suspect before. All of these statements were not helpful to police and
therefore classified as denials with explanation, [*867] consistent with the earlier studies. n139 In the real world of
criminal prosecutions, the difference between the usefulness of such denials with explanation, as compared to
incriminating statements, is probably quite substantial. n140
The catch-all category for "other" statements involved declarations that were not incriminatory and could not be
otherwise categorized. The few statements in this category typically involved incoherent or rambling [*868] remarks
from suspects. An example would be a suspect who, when responding to questions by police, gave answers that were
nonsensical or unresponsive to the questions asked (usually because of intoxication).
About one out of every ten suspects was questioned more than once. n141 For purposes of statistical analysis in
this study, we categorized all suspects by the most fruitful result of questioning. For example, if the suspect gave a
denial in the first interview but a confession in the second, he was categorized as "confessing" rather than "denying." By
including only the greatest result in multiple interviews, we may have given a slightly exaggerated impression of police
effectiveness.
b. Findings
The number of incriminating statements and denials we found are shown in Table 4. Overall, 9.5% of the suspects
invoked their rights, 33.3% were successfully questioned, 36.1% were questioned unsuccessfully, and 21.0% were not
questioned. n142 Because some might argue that only suspects who were in fact questioned should be included in
determining a confession rate, n143 we also report percentages for only those suspects who were questioned. As can
also been seen in Table 4, of all police interviews, 12.1% produced an immediate invocation of Miranda rights, 42.2%
were successful, and 45.7% were unsuccessful. It should also be noted that, however measured, our "success" rate is
artificially inflated because of the point at which our sample was drawn. We sampled cases from screening sessions
where police officers believed they had gathered sufficient evidence for prosecution. As a consequence, our sample
excludes a significant number of cases that police thought were too weak to warrant prosecution -- including cases that
were too weak because of an unsuccessful out-come of police questioning. n144 We heard from police officers that
many [*869] cases were "prescreened" and never made it to the prosecutor's office for review. Police officers knew
that these cases were simply insufficient to trouble taking to the prosecutor's office.

Page 12
43 UCLA L. Rev. 839, *869

Table 4 -- Results (N=219 overall; 173 for those questioned)
Questioned
Overall
Only
No.

%

%

INVOKED RIGHTS

21

12.1%

9.5%

SUCCESSFUL

73

42.2%

33.3%

Written Confession

5

2.9%

2.3%

Verbal Confession

42

24.3%

19.2%

Incriminating

26

15.0%

11.9%

0

0.0%

0.0%

Locked into False Alibi
UNSUCCESSFUL

79

Flat Denial

34

19.7%

15.5%

Denial w/Explanation

40

23.1%

18.3%

5

2.9%

2.3%

Other
NOT QUESTIONED

46

TOTAL

219

45.7%

36.1%

21.0%
100.0%

100.0%

We also gathered more specific data on what suspects actually said during questioning. As also shown in Table 4,
of all suspects, 21.5% gave written or verbal confessions, while 11.9% provided incriminating statements. None of the
interrogations locked the defendant into a false alibi. While it has been suggested that a noticeable number of suspects
tie [*870] themselves to an unlikely story, n145 to our knowledge no one has tried to quantify the prevalence of this
outcome. Our quantitative data suggests that this outcome is not as important as these previous qualitative assessments
have suggested.
We also examined whether suspects for particular types of crimes were more likely to confess. The limited previous
research on this issue, from studies conducted in the 1960s, suggests that suspects are less likely to confess to violent
crimes than to property crimes. n146 Our data trend in this direction; as shown in Table 5, police were less successful
in interrogating suspects for crimes of violence than for property crimes, although the result is not statistically
significant at the conventional 95% confidence level. n147
Table 5 -- Questioning Result by Type of Crime
Type of
Invoked
Questioning
Questioning
Crime
Violent

Rights

Successful

Total

Unsuccessful

No.

%

No.

%

No.

%

No.

%

9

12.7%

25

35.2%

37

52.1%

71

100.0%

12

11.8%

48

47.1%

42

41.2%

102

100.0%

21

12.1%

73

42.2%

79

45.7%

173

100.0%

Crime
Property
Crime
Total

[*871] c. Comparison with Pre-Miranda Confession Rates
Although our success rate is inflated, it is lower than success rates found in this country before Miranda, suggesting

Page 13
43 UCLA L. Rev. 839, *871

that Miranda has hampered law enforcement efforts to obtain incriminating statements. In an earlier article, one of the
present authors collected the available pre- and post-Miranda information on confession rates in this country. n148
Although broad generalizations are hazardous, that evidence suggests that interrogations were successful, very roughly
speaking, in about 55% to 60% of interrogations conducted before the Miranda decision. For example, the earliest
academic study in this country reported confession rates of 88.1% and 58.1% in two cities in California in 1960. n149
Similarly, a 1961 survey in Detroit reported a 60.8% confession rate, which fell slightly to 58.0% in 1965. n150 In New
Haven, the confession rate was about 58%-63% in 1960. n151 These figures deserve special attention in calculating a
pre-Miranda confession rate, n152 because they avoid the problem of anticipatory implementation of Miranda in
various jurisdictions. In particular, confession rates after June 1964 might be dampened by the Supreme Court's
Escobedo decision, which led some police to adopt Miranda-style warnings even before the decision. n153 Other
commentators have agreed that the pre-Miranda confession rate was at least 55% to 60%. n154 Our 33.3% overall
success rate (and even our 42.2% questioning success rate) is [*872] well below the 55%-60% estimated pre-Miranda
rate and, therefore, is consistent with the hypothesis that Miranda has harmed the confession rate.
In his carefully-argued rejoinder to our Article, Professor Thomas spends most of his time discussing this
conclusion, contending that confession rates have not changed since Miranda -- a conclusion he labels the "steady-state"
hypothesis. In support of this hypothesis, he makes an extensive series of elaborate adjustments to our "confession rate,"
arguing that our confession rate is actually 54%, within the pre-Miranda range. n155 However, these adjustments are,
at least for purposes of comparing to the pre-Miranda data, inappropriate.
Professor Thomas's adjustments begin from the major premise that the earlier studies apply only to custodial
interrogation. n156 Thomas then asserts that we have compared our apples (a study with both custodial and
noncustodial interrogations) to these oranges (studies of just custodial interrogation). Agreeing with our
intuitively-appealing finding that noncustodial interrogations are less successful in obtaining confessions (as we discuss
below n157 ), Thomas concludes that, for comparison purposes, our figures need to be adjusted upwards from 42% to
48% by excluding noncustodial interrogations. n158
We compared apples to apples. The pre-Miranda studies we relied upon generally appear to include both custodial
and noncustodial interrogations. For example, the District of Columbia study involved defendants "who had been
subjected to arrest procedures." n159 Of course, suspects who were arrested could include suspects arrested after a
noncustodial interview. Likewise, the Seaside City study involved suspects who "were actually arrested and
incarcerated," n160 a pool that therefore would include those arrested and incarcerated after noncustodial interviews.
The closest Professor Thomas comes to supporting his assertion is with the Pittsburgh study, where the authors reported
that "the questioning occurs at the time the suspect is taken into custody." n161 But whether they were simply [*873]
describing the standard practices of the Pittsburgh Detective Branch n162 or only a limitation of the study to custodial
interrogations is unclear. To be compared directly to these studies, our figures should be adjusted -- at most -- by
looking at "arrested" suspects, an adjustment that does not make much of a difference. n163
In addition, our rough estimate of a 55%-60% pre-Miranda confession rate rests on a much broader limited
universe of studies than Thomas discusses. In particular, our estimate rests on a comparison to twenty pre-Miranda
confession rates that have been collected elsewhere. n164 It is readily apparent that many of these studies simply
examined cases in which charges had been filed, regardless of whether the suspect was interrogated in custody, outside
of custody, or even not at all. For example, a New York County study involved confessions "used in presenting cases to
[the] grand jury." n165 That methodology would pick up all confessions, regardless of whether they were derived from
custodial or noncustodial interrogations. Similarly, the Brooklyn study involved all suspects for particular crime
categories, regardless of whether they were interrogated in custody or out. n166 It may be the case that, due to
Miranda, fewer of our interrogations were custodial than in the earlier studies. n167 But for Thomas to establish
incomparability, he needs to show that noncustodial interrogations were positively excluded from these studies,
something he has not done. Using both custodial and noncustodial interrogation, as we have done, is therefore the best
way to assure consistency with the pre-Miranda studies.

Page 14
43 UCLA L. Rev. 839, *873

While Professor Thomas has failed to establish inconsistency with the earlier studies, the implication of his minor
premise -- that noncustodial interrogations should be excluded because they are less effective n168 -- is important. If
the minor premise is true, then to prove that the confession rate [*874] is lower today, one need only establish that
(other things remaining equal) police have shifted to more noncustodial interrogations since Miranda. That police have
so shifted is supported by a considerable body of evidence, including the recommendations of the leading police
interrogation manual, n169 reports from those who have observed police in action, n170 court cases suggesting
intentional noncustodial interviews, n171 and academic commentators. n172 If this body of evidence is believed,
n173 then [*875] Miranda has reduced the confession rate and the only subject left for discussion is how much it has
done so.
On the subject of comparing apples to apples, we believe that Professor Thomas has introduced some oranges into
the equation. In revising our confession rate, Professor Thomas considers only the number of confessions obtained from
those suspects actually questioned, excluding those who were never questioned. Because 21% of the suspects in our
sample were not questioned, n174 this exclusion inflates our confession rate substantially from 33% to 42%, the
starting point for Professor Thomas's calculations. n175 But here Thomas is quite clearly comparing apples to oranges.
So far as we can tell, all of the pre-Miranda studies included suspects who were not questioned in the sample from
which their confession rates were determined. n176 Thus, for purposes of comparing our confession rate to
pre-Miranda rates, all of his calculations start from a base line that is far too high. n177
In a last attempt to show that our 33.3% incriminating statement rate is too low, Professor Thomas points to other
recent studies. However, our incriminating statement rate fits comfortably with this data. In particular, a 1979 National
Institute of Justice study of Jacksonville, Florida and San Diego, California, reported confession rates of 32.9% and
20.3% and, throwing in statements admitting being at the scene, overall statement [*876] rates of 51.3% and 36.6%.
n178 A 1977 study of six cities reported a "confession" rate of 40.3% in all cases. n179 A study by Professor Leo in the
California Bay Area reported an overall interrogation success rate of 63.8%. Because it is so much higher than all the
other studies, Leo's figure seems anomalous. It turns out, as we explain in more detail in Appendix B, n180 that the
figure is inflated (for purposes of comparing to other studies) because of three important factors: It includes only
suspects who were in fact interrogated; it includes only interrogations performed in stationhouse custody; and it
includes only interrogations by experienced detectives. n181 Adjusting for these factors produces an incriminating
statement rate of 37.0%, not far from ours. Clearly, caution is appropriate in drawing conclusions from a single study.
But because our confession rate and, properly read, all the other recently-reported confession rates fall well below the
estimated pre-Miranda confession rate of 55%-60%, the inference that Miranda has depressed the confession rate seems
quite reasonable.
d. Comparisons with Confession Rates in Other Countries
A different route to the same conclusion is to compare our success rate with rates in other countries that do not
follow the Miranda requirements. The available evidence suggests that British and Canadian confession rates without
Miranda are in excess of 60%, sometimes substantially [*877] so. n182 Comparing such rates to our 33.3% rate also
suggests that the American Miranda rules have an inhibiting effect on suspects.
In his reply, Professor Thomas suggests that this methodology cannot be accepted because the United States,
Britain, and Canada do not "share a core of relevant characteristics," at least in so far as police interrogation practices
are concerned. n183 We thought agreement existed that these countries could be compared generally. With respect to
criminal practices in general, Warren Court defenders have long trotted out comparative analysis when it served their
purposes. n184 Similarly, Craig Bradley devotes much his recent book, The Failure of the Criminal Procedure
Revolution, to comparing American practices to, among others, British and Canadian practices. n185 With respect to
police questioning in particular, the instructiveness of comparative research often has been recognized. Professor Van
Kessel, in perhaps the leading article on American and British interrogation practices concludes that "English and
American systems have the same common core" which should allow for profitable comparisons. n186 This is not
simply academic musing without grounding in the real world. The Miranda rules were created, in part, based on

Page 15
43 UCLA L. Rev. 839, *877

overseas analogies. Chief Justice Warren's opinion argued that foreign jurisdictions [*878] followed comparable rules
with no ill effects. n187 More recently, Justice O'Connor observed that "the learning of [England and certain other
countries] was important to development of the initial Miranda rule. It therefore should be of equal importance in
establishing the scope of the Miranda exclusionary rule today." n188
Professor Thomas amplifies his comparability concern with a list of purported differences between this country and
Canada and Britain which must be "quantified and explained" before any foreign comparisons can be accepted. n189
This challenge leaves us feeling a little like Dorothy in the Wizard of Oz, who is told she must bring back the
broomstick of the Wicked Witch of the West before her request will be considered. Just as the Wizard never thought
Dorothy would return, we doubt whether Professor Thomas really expects us to be able to "quantify" such variables as
America's allegedly "more skeptical attitude toward authority in general." n190 But let us take a stab at answering the
one objection he raises that is subject to quantification: America's higher violent crime rate (and hence, given lower
confession rates for violent crimes, n191 possibly lower confession rates overall). This should pose no obstacle for
comparisons of confession rates studies. One need not hold constant the proportion of violent crimes overall, n192 but
only the proportion of violent crimes in the confession study samples. Such an adjustment could not possibly begin to
explain the substantially lower confession rate in Salt Lake County. Our questioning success rate for property crimes is
47%, n193 well below that reported in British and Canadian studies including both property and violent crimes. This
suggests that, far from remaining in a "steady state" after Miranda, the American confession rate has fallen.
[*879] It is important, however, to make one final point about Thomas's "steady-state" hypothesis. Despite the
billing, it is not a "steady-state" theory at all. Instead, Professor Thomas concedes that data from our study and other
sources support the conclusion that after Miranda "more suspects invoke their right to remain silent," resulting in some
lost confessions among this group. n194 In Thomas's view, however, the overall confession rate has remained the same
because, among the remaining suspects, more make incriminating statements than before Miranda. n195 This is no
simple "steady state" theory. Instead, truth in labeling would suggest a different moniker: the
"more-invocations-offset-by-more-incriminating-statements" theory.
Unpacked in this fashion, a problem emerges: If lost confessions result because more suspects invoke after
Miranda, then a true "steady-state" theory would suggest fewer confessions -- unless someone can demonstrate the
offsetting more-incriminating-statements part of the equation. Here the burden of proof quite clearly rests on Professor
Thomas, and he candidly admits that he has not proven his theory. n196 Moreover, the causal mechanism for this offset
would be a curious one. Thomas suggests that Miranda created an "incentive to tell more of the truth. The Miranda
warnings might make suspects believe they can avoid being prosecuted only if they tell a story close enough to the truth
that the police will be fooled into letting them go. But those stories will often turn into incriminating statements." n197
Proving this unique theory will, we think, require empirical proof that seems unlikely to develop. The specific data to
date on this point seem to undercut Thomas's theory, not support it. n198 And, while academics and researchers have
been exploring why suspects confess for some time, no one has heretofore suggested that the "incentive to tell more of
the truth" is a significant factor. n199
A true "steady-state" theory, therefore, supports our conclusion: Because Miranda prevents police from obtaining
some confessions (through [*880] its waiver and questioning cut-off rules), police obtain fewer confessions overall.
4. Other Results from Interrogations
Our data also suggest that interrogation is useful for law enforcement for collateral reasons other than obtaining
incriminating statements. n200 Of suspects questioned, 13.9% gave "other information" that was useful to law
enforcement, n201 a figure that likely understates the true extent of the ancillary results police derive from questioning.
n202
Our data on collateral purposes for interrogation shed light on a subject that has been generating great academic
commentary of late -- the extent to which police officers question suspects to obtain "fruits" of crimes. Professor Yale

Page 16
43 UCLA L. Rev. 839, *880

Kamisar, in a recently published debate with Professor Akhil Reed Amar and Renee B. Lettow, appears to argue that "a
principal purpose -- if not the primary purpose -- of interrogation is to obtain information such as the location of
physical evidence." n203 The notion that a secondary purpose of interrogation (finding fruits) would be more important
to police than the primary purpose (obtaining incriminating statements) seems questionable, and we found no evidence
to support it. n204 Examining the cases involving "other information," police rarely [*881] obtained incriminating
fruits. The most common type of other information (found in five cases) was a statement implicating an accomplice.
n205 In three cases, police recovered stolen property as the result of interrogation, but their primary concern seemed to
have been restitution for victims rather than building cases for prosecution. In two cases, the police recovered a weapon
through interrogation, although even here it is not clear if the primary police motivation was to strengthen a case against
the interrogated suspect with this fruit. n206 In view of this data, it seems hard to argue that finding fruits is a
"principal" purpose of interrogation.
B. Questioning Inside and Outside the Miranda Regime
One question that has not been the subject of any substantial empirical research is the extent to which police
questioning falls inside or outside the Miranda regime and whether this makes any difference to ultimate outcomes.
This section reviews various types of police questioning not covered by Miranda, specifically noncustodial questioning
and public safety questioning. It also discusses volunteered statements that police obtain without questioning.
1. Noncustodial Questioning
The Miranda rules cover only "custodial" interrogation. n207 Evidence suggests that police have adjusted to
Miranda by shifting to noncustodial "interviews" to skirt Miranda's requirements. n208 In talking to police officers, the
researchers found some anecdotal evidence supporting this view. In a few screenings, officers (mostly from one large
department) referred to giving suspects a "Beheler" warning, as in "I gave him Beheler." This is a reference to the
Supreme Court's decision in California v. Beheler, n209 which held that a suspect was not in custody when he
"voluntarily agreed to accompany police to the station house [and] the [*882] police specifically told [him] that he was
not under arrest." n210 Drawn from this case, the Beheler warning consists of telling a suspect that he is not under
arrest and is free to leave during the "interview." This sophisticated approach suggests that at least some police forces
know the difference between custodial and noncustodial interrogations and structure their investigations accordingly.
On the other hand, in a few cases in our sample it appeared that officers Mirandized when not required to do so.
One department, the Special Victims Unit (which handled primarily sex offenses), apparently had a policy of
Mirandizing every noncustodial interviewee. n211 One possible reason was provided by a prosecutor: "It doesn't hurt
with our Supreme Court, because they may well find it to be custodial setting where no other reasonable person would
find it to be custody." The prosecutor's observation has a basis in the Utah case law. In a muddled series of opinions, the
Utah Supreme Court has possibly held that the standards for determining "custody" in Utah are more
defendant-protective than under federal law. n212
To date, no one has quantified how often police interview in noncustodial settings. n213 In our sample, 69.9% of
the interviews were [*883] custodial while 30.1% were noncustodial. n214 Of the noncustodial interviews, 40.3%
were at the scene, 26.9% were field investigations, and 32.7% were arranged interviews (that is, interviews where police
officers had previously contacted the suspects to set up an interview time). n215
Even if police are able to avoid Miranda's requirements by conducting various noncustodial interviews, the
question would remain whether such interviews are less effective in obtaining incriminating information. We found that
police were less successful in noncustodial interviews, as noted in Table 6 n216 -- a result that was statistically
significant. n217 One hypothesis to account for this difference might be that police officers more frequently question
suspects in custody for presumptively more dangerous crimes of violence and therefore might try harder to get a
confession. n218 In our sample, however, property offenders were slightly more likely to be interviewed in custody:
72.5% of the interviews with property suspects were in custody as compared to 66.2% of the interviews with violent

Page 17
43 UCLA L. Rev. 839, *883

suspects. n219
Table 6 -- Questioning Result by Custodial Status
Kind of

Questioning

Questioning

Questioning

Successful

Unsuccessful

No.

%

No.

%

Total
No.

%

Custodial

58

56.9%

44

43.1%

102

100.0%

Noncustodial

15

30.0%

35

70.0%

50

100.0%

Total

73

47.7%

79

51.6%

152

100.0%

[*884] An interesting side note is a technique used by some detectives for speedy noncustodial interviews. In six
cases in our sample, the investigating detective simply called the suspect on the telephone -- a clearly noncustodial
interview! Setting aside one anomalous interview involving an innocent suspect, n220 in two of the remaining five
cases the detective succeeded in obtaining a verbal confession. Indeed, in one case, the detective not only obtained a
confession but even convinced the suspect in Oregon to come back to Utah and surrender himself. All of the telephone
interviews involved property crimes, especially small thefts and larcenies. Why do detectives resort to mere telephone
interviews? One detective said, "I make telephone interviews whenever I can. I'm way too busy to talk to them in
person."
A conclusion that might be drawn from our data is that police have adapted to Miranda to some extent by shifting
to noncustodial interviews. However, the value of that tactic appears to be mitigated by the lower success rates of
noncustodial interviews. Also, the very fact that police have tried to shift suggests, contrary to the view of some
defenders of Miranda, n221 that interrogating police officers believe the Miranda rules are harmful to their efforts.
2. Public Safety Questioning
In various decisions since 1966, the Supreme Court has created exceptions to the Miranda doctrine. Critics of these
decisions have suggested that the exceptions dramatically scale back the Miranda regime. n222 For example, after the
Court's 1984 decision recognizing a "public safety" exception to Miranda's requirements, n223 articles proclaimed the
"dissolution" of Miranda n224 because the exception "carv[ed] a gaping hole." n225
[*885] We made an effort to determine the extent to which such exceptions play a part in everyday law
enforcement. With respect to public safety questioning, only one of 173 suspects was even arguably questioned without
Miranda warnings in such circumstances, and even that case probably did not fall under the Miranda requirements.
n226 The Court also recently held that "routine booking" questions were not subject to the Miranda requirements. n227
Our sample contained no instances of statements being obtained through such questioning. This finding that the
Miranda exceptions are rarely used n228 supports the suggestion made elsewhere that Miranda has not been modified
in ways likely to result in significantly more successful police questioning. n229
3. Volunteered Statements
Another way in which police officers obtain statements outside the Miranda requirements is from suspects
"volunteering" incriminating information. In Miranda itself, the Court held that "volunteered statements of any kind are
not barred by the Fifth Amendment . . . ." n230 Suspects who volunteer statements, even without the benefit of
Miranda warnings or waivers, can find those statements used against them in court. n231
In our study, 19 of 217 of suspects (8.8%) volunteered statements without any prompting or questioning by the
officers. Thirteen of these suspects volunteered incriminating information (four verbal confessions and five
incriminating statements), while an additional six volunteered denials of some type. Perhaps the most interesting aspect
of these unbidden statements is that, for nine of the thirteen suspects, the volunteered statement was the only

Page 18
43 UCLA L. Rev. 839, *885

incriminating statement available to police. In one case, for [*886] example, police saw a person throwing rocks at a
window. They chased him and saw him throw something into some bushes. After police caught the suspect, they found
a gun in the bushes. On his way to the station, the suspect received his Miranda rights and invoked them. Later the
suspect volunteered that he had bought the gun on the streets and was carrying it for his protection. He was charged
with carrying a concealed weapon.
What made the thirteen suspects volunteer incriminating statements to police? While our small numbers make
meaningful statistical significance tests impossible, it appeared these volunteered statements might be associated with
three factors: strong evidence, violent crimes, and prior criminal records. Turning first to strength of the evidence, of the
thirteen suspects volunteering incriminating statements, we had data on the strength of the evidence for nine. n232 For
eight of these nine, the strength of the evidence was strong or overwhelming -- a good indication that the suspects
thought they had nothing to lose from volunteering information. Turning next to violent versus property crimes, ten of
the thirteen suspects (76.9%) volunteering incriminating statements were involved with violent crimes (as compared to
41.6% involved with violent crimes in the overall sample). n233 Finally, ten of the thirteen suspects (76.9%)
volunteering incriminating statements had prior criminal records (as compared to 69.1% in the overall sample). n234
One interesting aspect of volunteered statements is that some suspects who invoked their rights had a hard time
remaining quiet. We found six suspects who invoked their rights at some point during the questioning also volunteered
statements to police. Three of these suspects invoked their rights immediately after being read explanations of Miranda
and later voluntarily gave one sentence statements to police explaining their behavior. n235 In one case, the suspect
volunteered a denial with explanation even before Miranda was read; he then received his Miranda rights, gave a brief
denial with explanation, and invoked his rights. In the final two cases, the suspects invoked their Miranda rights during
the interviews and subsequently gave volunteered statements to police that matched the information they gave before
invoking.
[*887] Suspects who were unsuccessfully questioned but later volunteered incriminating statements presented a
categorization issue for us: Should they be treated as "successful" or "unsuccessful" outcomes to police questioning?
Because our study was designed to answer questions about police questioning, we followed the latter course and
categorized these suspects by what happened during questioning, not by what they volunteered. We were fortified in
this view by the fact that the volunteered statements were generally viewed as unimportant in the processing of the
cases -- in seven of these nine cases the statements were regarded as either unnecessary or relatively unimportant. n236
Because others might disagree with this categorization, we have also included as Appendix C to this Article a list of the
cases involving volunteered statements and the important factors associated with them. This will allow anyone
unsatisfied with our categorization to assess what difference it makes. We have also highlighted the difference that this
categorization makes in a few places where it seemed particularly salient. n237
C. Police Compliance with Miranda
A critical question about the Miranda rules is whether police obey them. The empirical evidence from the late
1960s suggests that, at least immediately following Miranda, police compliance varied widely. n238 Although little
research has been conducted since then, the prevailing view now seems to be that police adhere to the Miranda
requirements. n239 This section sets forth the evidence we gathered on this issue, which supports the same conclusion.
[*888] 1. Researcher Impressions of Police Compliance
The qualitative impression of the researchers was that police almost always followed the Miranda requirements.
n240 It appeared that officers generally read Miranda warnings and waivers from a card. n241 For example, Salt Lake
City police carry a laminated card which reads:
(Front side)

Page 19
43 UCLA L. Rev. 839, *888

Miranda Warning
1. You have the right to remain silent.
2. Anything you say can and will be used against you in a court of law.
3. You have the right to talk to a lawyer and have him present with you while you are being questioned.
4. If you cannot afford to hire a lawyer, one will be appointed to represent you before you answer any
questions, if you wish.
5. If you wish to answer questions now without contacting a lawyer or without a lawyer present, you
have the right to stop answering questions at any time. n242
[*889] (Back side)
Waiver
After the warning and in order to secure a waiver, the following questions should be asked and an
affirmative reply secured to each question.
1. Do you understand each of these rights I have explained to you?
2. Having these rights in mind, do you wish to talk to us now?
Although the researchers were instructed to pay close attention to identifying possible Miranda violations, at most
three arguable cases of noncompliance (out of a sample of 173 questionings) were discovered, with only one case of
clear noncompliance.
The only clear case of Miranda noncompliance involved a suspect who stole a $ 300 trailer. He was arrested by an
inexperienced police officer, who questioned the suspect at the scene without reading him Miranda warnings. The
suspect denied having stolen the trailer, claiming he had bought it. This outcome made the probable Miranda violation a
moot point because there was no incriminating information for the prosecution to use. The suspect later pled guilty to a
misdemeanor theft charge.
Turning next to the two arguable cases of noncompliance, one suspect equivocated a bit before waiving his rights.
A suppression motion was filed, and the court found the waiver valid. In another case, a suspect said he did not want to
invoke his Miranda rights but also did not want to talk "in formal terms." Police continued to talk to him (presumably
"informally"), and he admitted his involvement with a stolen social security check. No charges were filed because the
government had been reimbursed. Our data thus support the emerging consensus that police play by the Miranda rules.
2. Motions to Suppress
Because our study rests on information gathered at police-prosecutor screening sessions, one might argue that our
methodology would automatically miss information about police misconduct. After all, cops and [*890] prosecutors
are not likely to swiftly admit wrongdoing, particularly in the presence of an outside researcher. But our initial
impression of police compliance was confirmed by the later paucity of motions to suppress the statements police
obtained. Previous studies have found that motions to suppress under Miranda are rare and successful motions even
rarer. n243
In our sample, of cases in which police obtained incriminating statements and charges were filed, defendants filed

Page 20
43 UCLA L. Rev. 839, *890

suppression motions in 4.8%. n244 All three of the suppression motions alleged only that police committed a technical
Miranda violation, not that they coerced an "involuntary" confession under more general Fifth Amendment doctrine.
n245 The allegations involved disputes over when Miranda "custody" began, n246 with one dispute also involving the
validity of the waiver. n247 All three motions were denied. Our data thus support earlier findings that successful
Miranda suppression motions are rare. n248
[*891] 3. Length and Timing of Questioning
Another way of examining police conduct is to review the length of police interrogation. It might be that police
obtain confessions only by protracted questioning. n249 There is very limited American data on how long
interrogations last. The 1960 study of two California cities found that most interrogations by detectives lasted less than
an hour. n250 The 1967 Vera Institute Study in Manhattan found that 95% of the interrogations (all of which were
recorded) took less than twenty minutes. n251 The 1993 Bay Area study found that most interrogations lasted less than
an hour. n252
To gather data on length of questioning, the researchers asked the officer screening the case to estimate how long
the questioning of a suspect lasted. Because the officers were estimating interrogation length, our data on this point
should be considered particularly "soft." n253 Nonetheless, as shown in Table 7, most suspects appear to have been
questioned only thirty minutes or less. Table 7 also reveals a very modest (statistically insignificant) correlation
between length of questioning and questioning success. n254 However, in all likelihood, an interrogation involving a
confession or incriminating statement takes longer than an interrogation where the suspect has little to say. Even with a
statistically significant association, it would be difficult to say that lengthy interrogations produce confessions. In fact,
causality might run the other way -- confessions might produce lengthy interrogations. [*892]
Table 7 -- Questioning Result by Length of Questioning
(N = 173; 87 unavailable)
Length of

Questioning

Questioning

Questioning

Successful

Unsuccessful

No.

%

No.

%

Total
No.

%

</= 5 Minutes

6

33.3%

12

66.7%

18

100.0%

6-10 Minutes

11

40.7%

16

59.3%

27

100.0%

11-15 Minutes

5

62.5%

3

37.5%

8

100.0%

16-30 Minutes

15

68.2%

7

31.8%

22

100.0%

31-60 Minutes

4

40.0%

6

60.0%

10

100.0%

> 60 Minutes

1

100.0%

0

0.0%

1

100.0%

42

48.8%

44

51.2%

86

100.0%

Total

One point from the data seems unassailable, however. The overwhelming majority of interrogations are quite
limited in duration. Of eighty-six interrogations for which data were available, only eleven extended beyond thirty
minutes and only one of these extended beyond an hour. This suggests that extended incommunicado questioning is
exceedingly rare, if it exists at all.
To get another handle on interrogation coerciveness, we also collected data on the times of the day when police
questioned suspects. For cases in which data were available, n255 most interrogations occurred during the daytime.
There was no obvious pattern between late night interrogation and interrogation success.
D. Factors Bearing on the Success of Questioning

Page 21
43 UCLA L. Rev. 839, *892

While the confession literature contains much speculation about variables that might relate to police success in
questioning, surprisingly [*893] little hard data exist. This subpart reviews several factors that might be thought to
make a difference to police success.
1. Strength of the Evidence
One hypothesis is that police are more likely to obtain a confession when they have considerable evidence against a
suspect. A correlation between the evidence the police have and their success in interrogations has been reported in the
one American study on this point. n256 Suspects confronted with significant evidence against them presumably see
little reason for denying the allegations. n257
We found a relation between strength of the evidence and successful questioning in our survey. We collected data
on the strength of the evidence against the suspect at the time of the questioning. The researchers classified the strength
of the evidence as "overwhelming," "strong," "moderate," "weak," or "nonexistent" based on their overall assessment of
the case. The police were more successful as the evidence strengthened, as shown in Table 8. n258 This may explain
why police frequently confront suspects with the evidence against them. The 1993 Bay Area study found that police did
this in 85% of all cases, making it the most common interrogation tactic (other than appealing to a suspect's self
interest). n259 The 1967 Vera study in Manhattan found that police [*894] confront suspects with the available
evidence with "great frequency." n260 Our data suggest that this tactic is a successful one. n261
Table 8 -- Questioning Result by Strength of Evidence
(N = 173; 30 unavailable)
Strength

Invoked

Questioning

Questioning

of Evidence

Rights

Successful

Unsuccessful

No.
Overwhelm.

%

No.

%

No.

%

Total
No.

%

3

33.3%

5

55.6%

1

11.1%

9

100%

12

16.7%

34

47.2%

26

36.1%

72

100%

Moderate

4

9.3%

18

41.9%

21

48.8%

43

100%

Weak

0

0.0%

5

26.3%

14

73.7%

19

100%

Total

19

13.3%

62

43.4%

62

43.4%

143

100%

Strong

Our data also suggest a complicating issue for Professor Thomas's hypothesis that "how much the police think they
need a confession" is an important factor in explaining whether a confession is obtained. n262 If such an effect exists,
its empirical confirmation may be obscured by the fact that suspects confess more often when the evidence is
overwhelmingly against them. Disentangling these two effects promises to be quite difficult.
[*895] 2. Prior Criminal Record
Frequently it is argued that suspects who are wise to the ways of the system are less likely to confess. n263 The
available empirical evidence, mostly from the 1960s, supports this hypothesis. n264
To test this hypothesis, we gathered data on suspects' criminal records. For our purposes, we defined a "criminal
record" as a previous adult arrest, including a misdemeanor arrest. A prior arrest probably, though not necessarily,
meant that a suspect had previously received Miranda warnings. The alternative measure of criminal record -- previous
conviction -- was impractical because of the difficulties in collecting such data. n265
Our survey found that suspects with a prior criminal record were slightly more likely to invoke their Miranda
rights, as shown in Table 9, although the small sample size means the result is not statistically significant. n266 As to
those who did not invoke, there was virtually no difference in whether police were successful. One possible way of

Page 22
43 UCLA L. Rev. 839, *895

reconciling our failure to find any prior record effect with contrary studies is our definition of "criminal record," which
included misdemeanor arrests that might not have involved delivery of the Miranda warnings. As a result, it is possible
[*896] our "prior record" category was too expansive to capture the effect of a serious criminal record. n267
Table 9 -- Questioning Result by Prior Record
(N = 173; 8 unavailable)
Prior

Invoked

Questioning

Questioning

Record

Rights

Successful

Unsuccessful

No.
Record

%

No.

%

No.

%

Total
No.

%

16

14.0%

49

43.0%

49

43.0%

114

100.0%

None

5

9.8%

22

43.1%

24

47.1%

51

100.0%

Total

21

12.7%

71

43.0%

73

44.2%

165

100.0%

3. Recording
A pressing question in the debate over Miranda is whether the recording of interrogations affects the confession
rate. Several commentators have argued that Miranda should be largely replaced by a requirement that police audiotape
or videotape interrogations. n268 Others have argued that Miranda should not be replaced, but rather supplemented by
such a rule. n269 Such proposals have received relatively little attention, however, because of concerns from law
enforcement agencies that a recording would inhibit suspects. n270
The available empirical evidence on whether taping affects suspects is quite limited. A massive 1993 National
Institute of Justice study of [*897] police videotaping reported only qualitative assessments of this issue. n271 Of the
police agencies surveyed (which selectively videotaped interrogations), 8.6% thought suspects were more willing to talk
to police, 63.1% thought there was no difference, while 28.3% reported suspects less willing to talk. n272 While these
figures suggest that videotaping was not deleterious to questioning efforts, the study did not report any hard numbers on
suspects' willingness to talk.
The only published American data on suspects' volubility while being taped come from the 1967 Vera Institute
study in Manhattan. The study compared audiotaped police interrogations in one New York City precinct with standard
interrogations in other comparable precincts. Police obtained more admissions in the taped interrogations. n273 Studies
from Britain and Canada also suggest that videotaping does not harm confession rates. n274 However, a study from
Tasmania, Australia concluded that taping was an "inhibiting factor in the interrogation of suspects." n275 Based on the
Tasmanian study, the leading American police interrogation manual recommends against taping interrogations. n276
To shed light on the controversy over taping, we collected data on whether interviews were recorded. Some police
agencies in Salt Lake County record interrogations, although the decision of when to record is left to the discretion of
the individual officer. n277
[*898] Our data suggest that recording does not inhibit confessions. Table 10 shows that officers obtained
statements as frequently when the interviews were recorded than when they were not. n278 It should be noted that our
data on recording do not come from a "controlled" sample -- that is, the cases in the recording category may not have
been randomly distributed. It is possible that police agencies more often tape interviews in cases involving particular
types of crimes n279 or victims, n280 and that this biased the percentage of "successful" outcomes we observed in the
recording category. In particular, it may be possible that the police particularly wanted confessions in the cases that they
were taping. If police are more often successful when interrogating seriously, as Professor Thomas suggests, n281 and
if police more often tape interrogations when they are serious, then our methodology could produce a spurious
correlation between recording and interrogation success. At the very least, however, our data provide no support for the

Page 23
43 UCLA L. Rev. 839, *898

suggestion that recording inhibits suspects.
Table 10 -- Effect of Recording on Questioning
(N = 173; 31 unavailable)
Questioning

Invoked

Questioning

Questioning

Recorded?

Rights

Successful

Unsuccessful

No.

%

No.

%

No.

Total

%

No.

%

Yes

1

5.0%

12

60.0%

8

40.0%

20

100.0%

No

19

15.6%

55

45.1%

67

54.9%

122

100.0%

Total

20

14.1%

67

47.2%

75

52.8%

142

100.0%

[*899] To our knowledge, no one has published data on whether the recording of interrogations increases the
likelihood of suspects' invoking their rights. It is plausible that the recording equipment itself could serve as a reminder
to the suspect that statements could be used again later, and would therefore inhibit the suspect from talking at all. The
effect of recording on invocations is also shown in Table 10. Again, we found nothing suggesting that recording has an
inhibiting effect. n282
The fact that outcomes of recorded interrogations look about the same as unrecorded interrogations has an
important implication about police perjury. If misrepresentations about obtaining waivers and confessions are
widespread, one would expect to find different results when police actions are recorded. n283 Our data suggest,
therefore, that police do not lie with any frequency about obtaining waivers or confessions, contrary to intimations to
that effect advanced by some academics. n284
4. Gender and Race
One relatively unexplored subject is whether gender or race produce any differences in confession rates. Turning
first to gender, to our knowledge, there is no existing American data on whether gender plays any role in interrogation
outcomes. n285 While some police investigators have suggested that women are harder to question successfully, n286
we found a trend suggesting that women were more likely to confess, as shown in Table [*900] 11, although the small
number of women in our sample renders the result statistically insignificant at the conventional level. n287
Table 11 -- Questioning Result by Gender
Gender

Invoked

Questioning

Questioning

Rights

Successful

Unsuccessful

No.
Male
Female
Total

%

No.

%

No.

%

Total
No.

%

18

12.0%

60

40.0%

72

48.0%

150

100.0%

3

13.0%

13

56.5%

7

30.4%

23

100.0%

21

12.1%

73

42.2%

79

45.7%

173

100.0%

Turning next to race, only limited research has been done. In 1967 the New Haven study found whites and
African-Americans were equally likely to confess. n288 In 1969, the Denver study found that Spanish-Americans were
more likely to make statements and give confessions than whites or African-Americans. n289 One British study found
no racial difference in willingness to make statements. n290
Our data on race are shown in Table 12. Because most of the suspects in our sample were white, we have
insufficient data to have a [*901] realistic chance of detecting any variations. Our limited data suggest no significant
racial differences. n291
Table 12 -- Questioning Result by Race

Page 24
43 UCLA L. Rev. 839, *901

(N = 173; 14 unavailable)
Race of

Invoked

Questioning

Questioning

Suspect

Rights

Successful

Unsuccessful

No.
White

%

No.

%

No.

%

Total
No.

%

14

13.0%

51

47.2%

43

39.8%

108

100.0%

Afr.-Am.

1

11.1%

3

33.3%

5

55.6%

9

100.0%

Hispanic

6

15.8%

15

39.5%

17

44.7%

38

100.0%

Other

0

0.0%

2

50.0%

2

50.0%

4

100.0%

Total

21

13.2%

71

44.7%

67

42.1%

159

100.0%

5. Patrol Officers and Detectives
One other factor that might bear on the success of interrogation is the type of officer who interrogates. It might be
that those most familiar with interrogation tactics are more successful. Some data suggesting this conclusion comes
from a National Institute of Justice study, which found that in both Jacksonville, Florida and San Diego, California,
detectives obtained a higher percentage of confessions than did patrol officers. n292
In Salt Lake County, both patrol officers and detectives interview suspects. Patrol officers generally handle the
questioning of arrested suspects for less serious felonies. Detectives generally handle the questioning of suspects for
more serious crimes and for crimes involving investigation. In a "jail case" (a case in which the suspect is being held in
jail), detectives will typically move quickly to interview a suspect before arraignment and the assignment of counsel. In
other cases, detectives may move more slowly.
[*902] We found that detectives were more successful at obtaining statements than were patrol officers, as shown
in Table 13. n293 One possible interpretation of this data is that detectives are more skilled in interrogation techniques
and are, accordingly, more successful in questioning. An alternative explanation for our finding is that detectives
typically interview suspects at a different point in the process from patrol officers. Detectives might be more likely to
interview a suspect in custody than a street officer would. Since custodial interviews are more often successful, n294
this might explain our finding. To test this hypothesis, we compared success rate by type of officer, limited to custodial
interrogations only. n295 Controlling for this difference by examining custodial interrogations only, the trend toward
greater detective success remained, but the result was not statistically significant (perhaps because of the smaller sample
size). n296
Table 13 -- Questioning Result by Officer Type
Officer

Invoked

Questioning

Questioning

Rights

Successful

Unsuccessful

No.

%

No.

%

No.

%

Total
No.

%

Patrol

11

13.8%

25

31.3%

44

55.0%

80

100.0%

Detective

10

11.0%

47

51.6%

34

37.4%

91

100.0%

Other

0

0.0%

1

50.0%

1

50.0%

2

100.0%

Total

21

12.1%

73

42.2%

79

45.7%

173

100.0%

[*903] 6. Multiple Interviews

Page 25
43 UCLA L. Rev. 839, *903

One other factor that might bear on police success was whether a suspect was questioned more than once. While
this might seem like an obvious tactic, no research has been done in this country on the frequency of multiple
interviews. n297
We found that police interviewed 9.2% of the questioned suspects in our sample (sixteen in all) more than once.
n298 All sixteen suspects were given Miranda rights at some point. Fifteen of the sixteen suspects were charged with
the crimes for which they were being questioned. n299
The most interesting aspect of the sixteen multiple interviews was their effect on police success in questioning. In
four, the police obtained a better result in the second interview, n300 a finding that tends to corroborate Professor
Thomas's suggestion that police interest in obtaining a confession is an important variable. Part of the success in later
interviews might be attributed to different questioners: Often the original interrogator was a patrol officer, while the
second, more successful interrogator was a detective. n301 In seven multiple interviews, the second interview did not
[*904] change the result of the first. n302 In one instance, the second interview actually yielded a poorer result. In this
particular case, the suspect provided incriminating statements to the police in the first, non-custodial (and thus
non-Mirandized) interview. During the second, custodial interview, the suspect received Miranda warnings and gave a
denial with explanation. Finally, four of the suspects invoked their Miranda rights to preclude questioning during a
second interview. n303
To determine what kinds of cases might prompt police to conduct a second interview, we examined whether
multiple interviews were more often found for crimes of violence or for suspects with criminal records. The percentages
of multiple interviews in these categories corresponded almost exactly to the overall percentages in our sample.
Finally, we found no suggestion that police have used multiple interviews to circumvent Miranda's requirements.
This theoretical possibility arises because of the Supreme Court's ruling in Oregon v. Elstad, n304 which held that
where police first obtained a confession from a suspect in violation of Miranda and then followed it up with a properly
Mirandized confession, the second confession was admissible. Academic critics of the decision argued that it gave
police "the power to reduce Miranda to a virtual nullity by routinely withholding the warnings until after the suspect
gives the desired information." n305 In none of our cases did police employ this tactic.
E. Confessions and the Outcomes of Criminal Prosecutions
In the ongoing debate over Miranda, a critical issue is whether confessions are important in determining the
outcome of criminal cases. In the Miranda opinion itself, the Court stated in vague terms that there had [*905] been
"overstatement of the 'need' for confessions." n306 Since then, critics of Miranda have responded that confessions are
often necessary to convict criminals; defenders of the decision respond that confessions are only rarely needed. n307
To shed light on this debate, we collected data on confession importance in three ways. First, we asked prosecutors
for their assessment of the importance of confessions. Second, we evaluated whether confessions affected prosecutors'
decisions to file charges. Finally, we quantified the effect of confessions on the ultimate outcomes of cases that were
filed. Each of these measures suggested that confessions play a significant role in obtaining convictions.
1. Prosecutors' Assessments of Confession Importance
Some limited empirical work on the question of importance of confessions is available, with studies suggesting that
confessions are needed to convict in about 23 to 26% of criminal cases, n308 although some have argued for a lower
figure. n309 A possible flaw in these earlier studies is that they have generally rested on academic assessments of the
utility of confessions to criminal prosecutions. It might be argued that professors underpredict how often confessions
are needed to see a case through to conviction in the hurly-burly world of criminal prosecutions. It is interesting that in
the two studies reporting police assessment of confession importance along with academic assessment, the police
thought confessions were [*906] more often needed. n310 Also, much of the data on confession importance are now

Page 26
43 UCLA L. Rev. 839, *906

quite dated. In view of the suggestions that the criminal justice system today is more hostile to police and prosecutors,
n311 research into the role of confessions is even more imperative.
To obtain an assessment of confession importance, we decided to ask prosecutors directly. After all, prosecutors are
perhaps in the best position to know how confessions affect the prosecution of criminal cases. To collect our data, at the
conclusion of a screening involving statements from a suspect, the researcher asked the screening prosecutor: "What
role (if any) do you think the confession/statement will play in obtaining a conviction?" n312 Using terminology from
the Yale Law Journal study of interrogations, n313 the researcher asked the prosecutor to categorize the statements as
"essential," "important," "not important," or "not necessary." Following the Yale Law Journal, responses in the first two
categories were combined into an interrogation "necessary" category while the last two were combined into an
interrogation "unnecessary" category.
Our data on the importance of confessions appear in Table 14. n314 As can be seen, in 61.0% of cases with
incriminating statements the prosecutors identified the statements as necessary for conviction, that is, as either essential
or important in obtaining a favorable outcome. This figure is considerably higher than figures provided in other studies,
which supports either (or both) of the hypotheses advanced earlier: that those familiar with the vagaries of the criminal
justice system think confessions are more important than academics, and that the difficulty of obtaining convictions in
the criminal justice system has increased in recent decades. [*907]
Table 14 -- Importance of Confessions
(N=73; 14 unavailable)
NECESSARY

No.

%

36

61.0%

Essential

13

22.0%

Important

23

39.0%

NOT NECESSARY

23

Relatively Unimportant

39.0%
21

35.6%

2

3.4%

Unnecessary
TOTAL

59

100.0%

2. Relation of Confessions to Filing Charges
To see whether the prosecutors' overall assessments of the importance of confessions were accurate, we also
gathered quantitative information on the effect of confessions on the processing of criminal cases. The first opportunity
for a confession to influence the disposition of a criminal case is the prosecutor's decision whether to even file charges.
While prosecutorial charging decisions are widely recognized as a critical juncture in the criminal justice process, n315
American research on the relationship between interrogation outcome and the filing of criminal charges is quite limited.
n316
[*908] We found prosecutors were more likely to file charges in cases in which they had incriminating statements,
as shown in Table 15. Prosecutors filed charges in 87.5% of the cases in which they had such statements, as compared
to 81.0% when suspects invoked their rights, 78.3% when suspects were not questioned, and 74.0% when questioning
was unsuccessful. n317 Our study may miss some of the effects of incriminating statements on charging. We assessed
only the effect of confessions on the prosecutor's decision of whether to file charges, not on the decision of what
charges to file. It is possible that the absence of a confession in a case led prosecutors to file reduced charges.
Table 15 -- Relation of Statements to Filing Charges
(N=219; 3 unavailable)

Page 27
43 UCLA L. Rev. 839, *908

Questioning

Whether Charges Filed

Result
Filed
No.
Invoked

Not Filed
%

No.

%

Total
No.

%

17

81.0%

4

19.0%

21

100.0%

63

87.5%

9

12.5%

72

100.0%

57

74.0%

20

26.0%

77

100.0%

36

78.3%

10

21.7%

46

100.0%

173

80.1%

43

19.9%

216

100.0%

Rights
Questioning
Successful
Questioning
Unsuccessful
Never
Questioned
Total

[*909] Our finding that confessions affect the decision to charge has an important implication for the debate about
Miranda's effects. Some scholars have argued that Miranda had no adverse effect on law enforcement because, it is
said, conviction rates did not decline after Miranda. n318 In response, it has been argued that prosecutors would screen
out cases weakened by Miranda before filing charges, which would preclude any finding of a conviction rate decline.
n319 Our findings support the response, suggesting that prosecutorial screening may mask Miranda's adverse effects on
conviction rates.
3. Relation of Confessions to Ultimate Outcomes
Intuitively one would think that confessions would have a profound effect on processing of the criminal cases.
After all, a confession is powerful evidence of guilt that would allow prosecutors to force favorable dispositions to such
cases. The limited previous research on this subject suggests this conclusion. n320
Our data support the intuition. Defendants who confessed were more likely to be convicted -- and more likely to be
convicted of more serious charges -- than those who did not. n321
Before turning to our specific data, it is important to explain the various possible outcomes to criminal cases in Salt
Lake County. A case could go to trial, where a defendant would be convicted or acquitted. But, as is probably true in
most jurisdictions in this country, most cases in our sample were resolved before trial in various ways. Prosecutors
could dismiss charges before trial, the most favorable outcome for a defendant. The next most favorable pre-trial
outcome for a defendant is to enter a plea in abeyance. Under such a plea, a defendant enters a plea of guilty, but [*910]
charges are dropped if the defendant does not commit any new criminal offenses within a set period of time. n322
Other pre-trial outcomes involve plea bargains, which hinge on the offense penalty structure. Utah follows the
typical approach of dividing felonies into various classes: first degree, second degree, and third degree. n323 Each class
of offense is associated with a possible penalty range. n324 Utah follows an essentially indeterminate sentencing
scheme, under which actual sentences within the penalty range are determined by a parole board. n325 Defendants
convicted on multiple counts generally serve concurrent sentences.
As a result of this penalty structure, the level of the offense of conviction -- not the number of convictions -- is the
primary concern in plea bargaining. Although the District Attorney's Office has no hard-and-fast plea bargaining
guidelines, n326 a common plea offer in an otherwise unexceptional case was to reduce the charges by one level. For
example, a prosecutor might offer a defendant charged with burglarizing a dwelling (a second degree felony) a plea

Page 28
43 UCLA L. Rev. 839, *910

bargain to burglarizing a building (a third degree felony). n327
[*911] For plea bargains, we categorized the outcome based on the relation between the level of the most serious
charge filed against a defendant and the level of the most serious charge to which he pled guilty. For example, a plea to
a charge reduced from a first degree to a second degree felony was classified as a "Plea-1," to a third degree felony as
"Plea-2," and so on. Other researchers studying plea bargaining have adopted similar approaches. n328
With this terminology in mind, the effect of questioning success on ultimate outcomes is shown in Table 16.
Defendants whom police successfully questioned were less likely to receive concessions in plea bargaining. Of suspects
whom police successfully questioned, 30.6% pled to charges at the same level as initially filed, compared to only 15.4%
for suspects invoking Miranda rights, 9.4% for suspects questioned unsuccessfully, and 10.8% for suspects not
questioned. n329 Along the same lines, defendants whom police successfully questioned were generally less likely to
have charges dropped. Of suspects whom police successfully questioned, 9.6% either had charges dismissed or entered
pleas in abeyance, as compared to 7.7% for suspects who invoked, 30.2% for suspects unsuccessfully questioned, and
21.6% for suspects not questioned. n330 It is also interesting that three of the five trials in our sample involved suspects
for whom questioning was unsuccessful. n331
Table 16 -- Relation of Statements to Outcomes
(N=172; 12 unavailable)
Result of

Outcomes

Questioning
Dismissed
No.

%

Abeyance
No.

%

Plea-3
No.

Plea-2

%

No.

%

Invoked Rights

1

7.7%

0 0.0%

0 0.0%

1

7.7%

Questioning

3

4.8%

3 4.8%

0 0.0%

5

8.1%

15

28.3%

1 1.9%

1 1.9%

8

15.1%

6

16.2%

2 5.4%

1 2.7%

4

10.8%

25

15.2%

6 3.6%

2 1.2%

18

10.9%

Successful
Questioning
Unsuccessful
Not Questioned
Total

Table 16 -- Relation of Statements to Outcomes
(N=172; 12 unavailable)
Result of

Outcomes

Questioning
Plea-1
No.
Invoked Rights

%

Plea-0
No.

%

Convicted

Acquitted

No.

No.

%

%

Total
No.

%

9

69.2%

2

15.4%

0 0.0%

0 0.0%

13 100%

30

48.4%

19

30.6%

2 3.2%

0 0.0%

62 100%

20

37.7%

5

9.4%

2 3.8%

1 1.9%

53 100%

Not Questioned

20

54.1%

4

10.8%

0 0.0%

0 0.0%

37 100%

Total

79

47.9%

30

18.2%

4 2.4%

1 0.6% 165 100%

Questioning
Successful
Questioning
Unsuccessful

Page 29
43 UCLA L. Rev. 839, *911

[*912] [*913] A subject of recent interest is the outcome of cases of suspects who invoked their rights. n332 The
small number of suspects in our sample who invoked their Miranda rights makes it difficult to say much or report
meaningful statistical significance tests about their dispositions compared to other categories of suspects. n333
Suspects who invoked their rights appeared to be less likely to be convicted as charged than suspects who gave
incriminating statements (15.4% vs. 30.0%), although this result was not statistically significant. n334 Those who
invoked also appeared less likely to have charges dismissed or to enter pleas in abeyance than those who were
questioned unsuccessfully, although again this result lacked statistical significance. While this might indicate that those
who invoke do not benefit from their invocations, an alternate explanation is possible. Although we did not have
complete data on the criminal records of those who invoked their rights, it appeared that the records were generally
substantial. n335 If prosecutors more aggressively prosecute those with serious criminal records (as seems likely n336
), and if the suspects who invoke generally have more serious records, n337 then simply looking at the outcomes of
those cases without controlling for criminal record may fail to discover any effect that the [*914] invocation of rights
has in producing less successful prosecutions. n338 Because of our incomplete data on the seriousness of prior record,
we could not analyze this issue.
A final way of assessing the importance of confessions to outcomes is to look at the difference in the overall
conviction rate, comparing those who confessed with those who did not. This incorporates the effects of confessions on
both the charging process and the adjudication process. As shown in Table 17, in our survey, when police successfully
questioned a suspect, that suspect was likely to be convicted of some charge in 78.9% of the cases; when they
unsuccessfully questioned a suspect, that suspect was likely to be convicted in only 49.3% of the cases. n339 The
difference in the rates for successful versus unsuccessful questioning is therefore 29.6%. n340 Comparing suspects
whom police successfully questioned with all other suspects (that is, adding those who invoked and those who were not
questioned to those who were unsuccessfully questioned) the conviction difference is 22.2% (78.9% vs. 56.7%), a
statistically significant difference. n341
[SEE TABLE IN ORIGINAL]
[*915] [*916] One question that might be asked about all these differentials is whether they are explained by the
factual innocence of defendants, rather than the role of confessions. Presumably innocent defendants are both less likely
to confess and less likely to be convicted. Studies on confessions almost invariably ignore this issue, simply assuming
that the suspects were in fact guilty. n342
While exact quantification of guilt is impossible, it seems unlikely that innocent suspects make much difference in
our figures. We had only one clear case of a suspect who was innocent. n343 This is unsurprising, as in cases of clear
innocence, police would not be seeking to have prosecutors file charges. Indeed, our sample probably has fewer
innocent suspects than other studies because we drew from cases that police had "pre-screened" for prosecutorial merit.
n344 As to the other cases, it is perhaps enough to note the generally accepted wisdom that almost all defendants who
are charged are in fact guilty. n345
In sum, our data suggest that police questioning plays an important role in the ultimate disposition of criminal
cases.
IV. IMPLICATIONS FOR THE CONTINUING CONTROVERSY OVER MIRANDA
Although our survey presents data on many aspects of police interrogation, it still leaves many questions
unanswered. Throughout we have highlighted questions that we think future researchers can profitably explore. On all
of the subjects we investigated, it is possible that Salt Lake [*917] County might not be comparable to other
jurisdictions. Moreover, our small sample size meant that we could detect only the most substantial variations in the
way various factors affect questioning. Most of our findings are consistent with the limited earlier research, but we also
report a few disparate results. n346 Further research in the area of police interrogation is the only way to illuminate
these issues.

Page 30
43 UCLA L. Rev. 839, *917

While our study has these limitations, it may be useful to generalize from our findings (coupled with other available
research) to more universal conclusions about the on-going debate over Miranda's real world effects. n347 In recent
cases, the Court has attempted to calibrate the Miranda rules by weighing costs and benefits. n348 Taken on these
pragmatic terms, n349 it is fair to ask whether the Court's balance is the right one. To honestly answer that question
requires empirical evidence of Miranda's effects -- evidence that our study begins to supply.
Turning first to society's interests, significant concern remains that Miranda has seriously hampered the prosecution
and conviction of criminals. The frequency of confessions in our 1994 sample appears to be much lower than the
frequencies in samples drawn before Miranda. Only 33.3% of all of the suspects in our sample gave incriminating
statements; only 42.2% of the suspects who were questioned gave incriminating statements. n350 Of suspects asked for
waivers of Miranda rights, 16.3% declined to give them, thereby completely preventing any police questioning, no
matter how restrained or reasonable. n351 An indication that the possibility of invocations was a real concern to police
was their shift in some cases to [*918] noncustodial interviews, which obviated the need to secure Miranda waivers
but also appeared to reduce police success. n352
Our data also suggest that Miranda's lost confessions would not have gilded already golden prosecution cases.
Rather, the absence of confessions appears to make an important difference in the real world processing of criminal
cases. Our survey supports the conclusion that if police can obtain an incriminating statement from a suspect, that
suspect is more likely to be convicted on more serious criminal charges. On the other side of the coin, if police are
unsuccessful, it appears the suspect is more likely to "walk." n353
Further research on Miranda's costs is undoubtedly warranted. Perhaps the most pressing area is the frequency of
confessions. On this point, our study suffers from a lack of pre-Miranda data in Salt Lake County. We can only infer by
looking to other jurisdictions, both domestic and foreign, that our 33.3% confession rate is lower than prevailed before
Miranda. To shed light on this question, it might be interesting to return to a jurisdiction where pre-Miranda data are
available and see what the confession rate looks like today. But the general thrust of our study, reviewed in light of
other data, n354 seems clear: Miranda imposes costs on society by reducing the number of confessions and,
consequently, the success of criminal prosecutions.
Given these tangible costs, what compensating benefits has Miranda provided? Professor Thomas's approach to this
issue is worth noting. Making a virtue out of vice, he concludes that any demonstrated adverse effect on confessions is
evidence either that Miranda is needed to prevent unconstitutional coercion or to treat suspects equally. n355 In his
view, a Miranda-induced drop in the confession rate simply means that coercion disappeared or that suspects who were
previously unaware of their rights now had that information. But, respectfully, this view is too simple.
[*919] Turning directly to the coercion argument -- the only one subject to a real empirical test n356 -- a drop in
the confession rate would appear to be evidence that Miranda reduced pressure in the interrogation room. But the
critical question for the Miranda debate is whether the lower confession rate stems from eliminating coercion that the
Constitution forbids or from pressure that it permits. The Court now admits that the "prophylactic" Miranda doctrine
"sweeps more broadly than the Fifth Amendment" n357 and is "overbroad in that its application excludes some
statements made during custodial interrogations that are not in fact coercive." n358 A suspect has no legitimate interest
under the Constitution in avoiding confessing; rather, a suspect is entitled to be free from unconstitutional coercion.
Thus, under Professor Thomas's view of the world, the critical subject for empirical inquiry should be not simply
whether the confession rate fell after Miranda but why it fell -- in other words, to what degree that reduction stems from
eliminating unconstitutional coercion or from prophylactic overkill preventing otherwise constitutional questioning
techniques.
Our study suggests that Miranda goes well beyond what is necessary to prevent unconstitutionally coercive
questioning methods. For starters, we found that a significant number of suspects could not be questioned at all because
they invoked their Miranda rights -- that is, they simply declined to give police permission to ask questions. It seems
hard to argue that any questioning at all without a waiver is inherently coercive, unless one views custodial

Page 31
43 UCLA L. Rev. 839, *919

interrogation itself as creating coercion. The Miranda court made such a finding. But this conclusion, as Professor
Thomas has written elsewhere,
has been Miranda's Achilles' heel. Justice White quickly found the ultimate hard case: the police ask the
single question, "Do you have anything to say?" To find coercion there, White argued, is contrary to
common sense and is "patently unsound. . . ." Justice White's common-sense repudiation of inherent
coercion seems . . . as sound as any attempt to demonstrate inherent coercion . . . . n359
[*920] One possible answer to Justice White's hypothetical is that it is "not a 'real' case; the 'real' cases involved
sustained questioning." n360 This is an empirical justification subject to empirical confirmation. Our evidence suggests
otherwise. About half of incriminating statements in our sample were obtained through fifteen minutes of questioning or
less and a number of incriminating statements were obtained in five minutes or less. n361 Again, it seems hard to argue
that all of these statements result from inherent compulsion.
An additional suggestion that Miranda extends too far in its quest to prevent police misconduct comes from our
finding that police now appear to follow the Court's commands scrupulously. n362 With only one possible exception,
in none of the cases in our sample did police depart from the Miranda requirements, n363 much less engage in coercive
questioning that would violate the narrower "voluntariness" rules of the Fifth Amendment. Instead, police officers seem
carefully to "toe the line" set by Miranda. An interesting example comes from officers who referred to giving "Beheler
warnings," citing a specific Supreme Court case as authority for their actions. n364 Other officers went beyond the
Miranda requirements. n365 Moreover, the new doctrinal exceptions and limitations to Miranda, alleged by academics
to leave "gaping holes" in the rules, in fact make little difference in the real world of police questioning. Police do not
undercut Miranda by interviewing first without the required warnings. n366 Nor do they push the bounds of Miranda
by frequently invoking the "public safety" and "booking questions" exceptions. n367 In light of this police compliance,
the kind of prophylactic "buffer zone" Miranda envisioned in the 1960s might be unnecessary in the 1990s and beyond.
[*921] In our view, then, the benefits of Miranda seem slim while the costs seem substantial. We readily concede
-- again -- that our conclusions must be tentative. At each point in our analysis, additional research would be quite
helpful. We therefore happily join Professor Thomas's plea for more empirical research. We would add to his list of
research projects exploration of reasonable alternatives to the Miranda rules, which might prevent unconstitutional
police coercion while at the same time increasing the confession rate. On that score, it is noteworthy that we found no
evidence that recording police questioning, the most prominently-offered alternative to Miranda, n368 had harmful
effects on police success.
But this continuing need for research underscores a more general point. Professor Thomas wonders "why a
constitutional interpretation that will soon be thirty years old must justify itself with empirical evidence." n369 The
answer is that this is how the Court has justified (or at least has purported to have justified) the decision. The Court now
tells us that Miranda is "a carefully crafted balance designed to fully protect both the defendant's and society's
interests." n370 This is not a rhetorical device from a conservative Court designed to undercut the decision: As
Professor Yale Kamisar reminds us, striking a balance "is the way Miranda's defenders -- not its critics -- have talked
about the case for the past twenty years." n371 Yet without an empirical foundation, the Court's balancing of interests
is illusory. n372 If there is no empirical answer to the question of how many criminals the doctrine sets free, how can
the Court blithely assert -- as it has in various cases -- that the benefits of the doctrine outweigh a cost of unknown
magnitude? n373 From the start, the Court developed [*922] the Miranda rules without empirical support, as this
Article's review of the scanty available empirical research makes clear. n374 Justice Harlan exposed this deficiency in
his Miranda dissent, explaining that the Court's precipitous imposition of the Miranda rules precluded other legislative
action that "would have the vast advantage of empirical data and comprehensive study." n375 In applying the Miranda
rules since then, the Court has had precious little information on what Miranda did for suspects, much less to law
enforcement. Despite the Court's promise that it has "carefully" considered society's interests, in truth neither the Court
nor Miranda's academic defenders has ever undertaken a substantial, empirically based cost-benefit assessment of the

Page 32
43 UCLA L. Rev. 839, *922

Miranda rules and possible alternatives to them. n376 Indeed, in seems fair to number among Miranda's costs its
preemptive effect, which has prevented research in different states on various ways of regulating police questioning.
n377
As Miranda's thirtieth anniversary approaches, these deficiencies in the empirical underpinnings of Miranda must
be regarded as extraordinary. The Miranda rules stem not from constitutional command, but rather from cost-benefit
prophylaxis. Yet despite nearly three decades to make their case, Miranda's defenders have yet to provide research
supporting the assertion that the decision's social costs are outweighed by its benefits. This failure leaves the entire
Miranda doctrine apparently resting on nothing other than the personal intuitions or unarticulated assumptions of
Supreme Court Justices about how the rules have operated in the real world. Perhaps Miranda's defenders are simply
exercising their right to remain silent. But in the face of that silence, the rest of us will draw the reasonable inference -an empirical case for Miranda does not exist.
APPENDIX A -- SURVEY FORMS
DATE: ___/___/___
CASE SURVEY FORM
Date Sheet No.
[Separate form for each suspect]
SUSPECT'S NAME[Exact Spelling]: ___ (Last) ___ (First) ___ (Initial)
AGE: ___ SEX: [] M [] F RACE: [] White [] Black [] Hispanic [] Pacific [] Other
TYPE OF CRIME (most serious):
[010] [] Murder [070] [] Sex Offense [110] [] Theft
[020] [] Attempted Murder
[071] [] Adult Victim [120] [] Burglary
[030] [] Robbery
[072] [] Child Victim [130] [] Auto Theft
[040] [] Kidnapping [080] [] Drug Trafficking [140] [] Fraud
[050] [] Felony Assault [090] [] Fugitive [150] [] Larceny/Forgery
[060] [] Weapons [100] [] Arson [160] [] Other: ___
CHARGE UNDER CONSIDERATION: ___
Describe Briefly (if property offense, provide dollar amount involved) ___
SUSPECT PREVIOUSLY KNOWN TO VICTIM? [] Yes [] No [] N/A
PROSECUTOR'S NAME: ___
INVESTIGATOR'S NAME: ___ PHONE NO. ___
INVESTIGATOR'S AGENCY: ___ SQUAD: ___
HOW WAS THE SUSPECT CAUGHT? ___
PRIOR CRIMINAL RECORD? [] Yes [] No
Number of prior felony arrests: [] 1 [] 2-4 [] 5-9 [] 10+
Number of prior felony convictions: [] 1 [] 2-4 [] 5 or more

Page 33
43 UCLA L. Rev. 839, *922

Did the county attorney FILE CHARGES in this case? [] Yes [] No
If No, WHY NOT? ___ [] Further Investigation
Did the POLICE TRY TO QUESTION the suspect? [] Yes [] No
If POLICE DIDN'T TRY TO QUESTION the suspect, reason was:
[010] [] No need - case against suspect overwhelming
[020] [] Belief suspect would invoke Miranda rights
[030] [] Reason to believe questioning would be unproductive
[040] [] Suspect intoxicated when arrested
[050] [] Suspect mentally unstable/physically injured
[060] [] No opportunity given press of other business
[070] [] Suspect already had attorney
[080] [] Other: ___
Before any questioning, EVIDENCE AGAINST suspect was:
[] Overwhelming [] Strong [] Moderate [] Weak [] Virtually nonexistent
[] Check this box if MULTIPLE INTERVIEWS attempted; complete form for each: [] 1st [] 2nd [] 3rd
If INVESTIGATOR TRIED TO QUESTION suspect:
Was the INVESTIGATOR a: [] street cop [] detective [] other: ___
Suspect questioned: [] IN CUSTODY [] not in custody -- at station [] not in custody, other
Not in custody = [] scene of crime [] field investigation [] arranged interview
In custody = [] when arrested [] on way [] at station
WHERE was the suspect questioned: ___
Were Miranda RIGHTS READ to the suspect: [] Yes [] No
Was the questioning done: [] in person [] over the telephone
Was the questioning: [] recorded [] not recorded
Was the suspect asked to WAIVE Miranda rights? [] Yes [] No
If yes, did suspect AGREE TO WAIVE rights? [] Yes [] No
If yes, was waiver [] verbal [] in writing

Page 34
43 UCLA L. Rev. 839, *922

If suspect did not agree to waive his rights, did questioning STOP? [] Yes [] No [] Parnal
If no, reason: ___
Was a statement VOLUNTEERED without questioning? [] Yes
If yes, describe: ___
Was a statement made during "BOOKING"? Describe: ___ [] Yes
PUBLIC SAFETY questioning? Describe: ___ [] Yes
Did the suspect ask for an ATTORNEY at some point? [] Yes [] No
If yes, was the attorney requested [] at the start [] during interview
If yes, did questioning STOP at that time? [] Yes [] No
If no, reason: ___
[] Suspect brought attorney to interview
Did the suspect assert a RIGHT TO REMAIN SILENT at some point? [] Yes [] No
If yes, did he assert [] at the start [] during interview
If yes, did questioning stop at that time? [] Yes [] No
If no, reason: ___
Did the suspect ANSWER QUESTIONS about the crime? [] Yes [] No
If yes, describe: ___
[] Written confession [] Locked into false alibi
[] Verbal confession [] Flat denial
[] Incriminating statements: ___ [] Denial/with explanation
___ [] Other: ___
[] Other useful information: (codefendants, location of stolen property, etc.) ___
HOW LONG did questioning take [in minutes] ___
WHAT TIME did questioning begin [military time] ___
If the suspect confessed/made statements, what ROLE (if any) do you think the confession/statements will play in
obtaining a conviction:
[] Essential [] Important [] Relatively Unimportant [] Unnecessary
FOLLOWUP CASE SURVEY FORM

Page 35
43 UCLA L. Rev. 839, *922

Date Sheet No.
DATE: ___/___/___
[Separate form for each suspect]
SUSPECT'S NAME[Exact Spelling]: ___ (Last) ___ (First) ___ (Initial)
District Court Docket Number ___
Circuit Court Docket Number (if known) ___
County Attorney Number (if known) ___
1. What charges were filled: Number Filed ___
Level: [] 1st degree [] 2nd degree [] 3rd degree [] A mis [] B mis [] Lower
2. Did the defendant file a motion to suppress statements/confession? [] Yes [] No
Describe briefly: ___
3. If so, did the motion allege: [] Miranda violation only
[] Coerced confession only
[] Both
[] Other: ___
4. Was suppression motion granted? [] Yes, in whole [] Yes, in part [] No [] Resolved before
If granted in whole or in part describe: ___
5. Was the case ultimately dismissed? [] Yes [] No [] Diversion [] Other
6. Did the defendant pled guilty or no contest? [] Yes [] No
If yes, did he pled to: [] Charge at most serious level
[] Charge at -1 level
[] Charge at -2 level
[] Charge at -3 level
[] Other: ___
7. If the defendant went to trial,
Check this box if multiple trials: [] 1st [] 2nd [] 3rd
Was he: [] Acquitted [] Convicted [] Mistrial [] Hung Jury

Page 36
43 UCLA L. Rev. 839, *922

If convicted, convicted of: [] Charge at most serious level
[] Charge at -1 level
[] Charge at -2 level
[] Charge at -3 level
[] Other: ___
8. If plea/convicted, convicted of (indicate number of counts): ___
[*923] [*924] [*925] [*926] APPENDIX B -- COMPARISON OF LEO STUDY TO CASSELL/HAYMAN
STUDY
In his reply to our Article, Professor Thomas prominently cites Professor Richard Leo's recent study of confessions.
Leo observed, either in person or on videotape, 182 interrogations in the Bay Area of California in 1993. n378 He
found that 64.3% of the suspects gave incriminating information of some type. n379 Leo himself is cautious in
claiming that this statement rate can be directly compared to rates from older studies. n380 However, Thomas and
others n381 have made such an unabashed inference. Thomas also argues that Leo's figure should be reconciled with
ours. Accordingly, it may be useful to compare Leo's figure specifically with the pre-Miranda studies and our study.
For comparison purposes, both the pre-Miranda studies and our study included suspects who were not questioned,
n382 while Leo's sample is artificially limited to cases in which police actually interrogated a suspect. n383
Accordingly, Leo's figure needs to be adjusted to reflect the presumably significant fraction of cases where police never
obtain a confession because they never interrogate. We found that 21% of suspects were never interrogated, n384 a
figure that is quite close to the only other recent data. n385 Applying our figure for suspects not interrogated to Leo's
reduces his incriminating statement rate to 50.8%. n386
The next adjustment that needs to be made is to recognize that Leo's study is limited to custodial interrogations. Yet
both the earlier studies and our study generally included custodial and noncustodial interrogations. n387 Looking at
noncustodial interrogations today is particularly [*927] important because (as Leo himself has noted) police may be
shifting to noncustodial interrogations to avoid Miranda's onerous requirements. n388
Noncustodial interrogation appears to be less effective in obtaining confessions. Our data, the only specific data on
this subject, were that 56.9% of custodial interrogations succeeded, a statistically significant difference from the 30.0%
success rate of noncustodial interrogations. n389 We also found that 30.1% of all interrogations were noncustodial.
Applying our figure of 30.1% interrogations not in custody to Leo's figures n390 and assuming that noncustodial
interrogation was less effective to the same extent that we found n391 reduces Leo's incriminating statement rate to
44.8%. n392
A final quantitative adjustment stems from the fact that Leo's study covered only interrogations by experienced
detectives, and not interrogations by patrol or other police officers. Most of the pre-Miranda data involved
interrogations by police officers of all types. n393 Our study also [*928] involved the broader sweep of all police
work. n394 We found a statistically significant difference between the success of detectives and patrol officers in
obtaining statements, n395 a result that is consistent with the only other recent data on this subject. n396 The clear
trend in our data remained in this direction even when controlling for the fact that detectives were more likely to
question suspects in custody than were patrol officers, although the result was not quite statistically significant. n397
We also found that slightly more than half of all custodial questioning was done by detectives, n398 a [*929] result
slightly higher than reported in the only other recent data. n399 Applying our ratio of detective-to-patrol-officer
custodial interviews to Leo's study and applying our lower effectiveness for the patrol officer interviews n400 lowers
Leo's incriminating statement rate to 38.7%. n401

Page 37
43 UCLA L. Rev. 839, *929

The adjusted 38.7% confession rate derived from Professor Leo's data is almost indistinguishable from our 33.3%
rate. A couple of other adjustments to Leo's data, however, might lower his figure even more to close the gap. First, Leo
found that police in the Bay Area continued interrogations in a few cases after suspects invoked their Miranda rights.
n402 In three interrogations, they obtained incriminating statements in violation of Miranda. n403 The pre-Miranda
studies, of course, do not exclude confessions for violation of Miranda. Our study contained no Miranda-suppressible
incriminating statements. n404 Thus, for comparison purposes, his figure must be reduced slightly to account for
Miranda violations. Second, Leo used a very broad definition of what was incriminating. The category of "some
incriminating information" was apparently expansively defined as embracing "implausible or contradictory denials that
the detectives believed corroborated other evidence pointing to the suspect's guilt or that could be used successfully to
impeach a suspect's credibility . . . ." n405 Many things said by a suspect could be viewed by the detectives as potential
"impeaching" information about credibility even when they were not really incriminatory. n406 Supporting the
interpretation that some of these statements [*930] were not really incriminating are the facts that many of the suspects
in Leo's sample (31%) were not even charged n407 and that prosecutors are particularly likely to charge suspects where
they have strong incriminating statements. n408 Supporting this interpretation is the lower (and relatively more
objective) figure of 24.2% "confessions" in Leo's data, which corresponds roughly to the "confession" figures in other
studies that suggest lower overall incriminating statement rates. n409 Finally, the study's sampling methodology also
may have overstated the number of incriminating statements. n410
Taking all of these factors into account, Professor Leo's study suggests an overall incriminating statement rate
somewhere below 38.7%, quite close to our 33.3% rate and much lower than the rates generally reported in
pre-Miranda studies.
[*931] APPENDIX C -- VOLUNTEERED STATEMENTS
This table reports the features of the nine cases in which suspects volunteered incriminating information and the
volunteered statement was the only incriminating statement obtained.
Suspect
Questioning
Volunteered
Role of
Type of
Number

Result

Statement

Crime

H29a

not quest.

incrim.

rel. unimp.

violent

H33

not quest.

incrim.

rel. unimp.

violent

H36

not quest.

confess

rel. unimp.

violent

H115

not quest.

confess

unnec.

prop.

H160

not quest.

confess

rel. unimp.

prop.

M2

not quest.

incrim.

ess.

violent

M11

not quest.

incrim.

unnec.

violent

H137

invoke

confess

n/a

prop.

H152

invoke

incrim.

rel. unimp.

violent

Suspect

Prior

Strength of

Number

Record

Evidence

Charged

Outcome

H29a

no

n/a

yes

plea-1

H33

yes

overwhelm.

yes

plea-1

H36

yes

overwhelm.

yes

plea-1

H115

yes

overwhelm.

yes

n/a

H160

no

strong

yes

plea-1

Page 38
43 UCLA L. Rev. 839, *931

M2

yes

n/a

yes

dismis'd

M11

yes

strong

yes

n/a

H137

yes

strong

yes

plea-1

H152

yes

n/a

yes

n/a

Legal Topics:
For related research and practice materials, see the following legal topics:
Criminal Law & ProcedureInterrogationMiranda RightsCustodial InterrogationCriminal Law &
ProcedureInterrogationMiranda RightsSelf-Incrimination PrivilegeCriminal Law &
ProcedureInterrogationNoncustodial Confessions & Statements
FOOTNOTES:

n1 384 U.S. 436 (1966).
n2 A 1974 survey of lawyers by the American Bar Association found that Miranda was the third most
notable decision of all time, trailing only Marbury v. Madison and United States v. Nixon, and leading Brown v.
Board of Education. JETHRO K. LIEBERMAN, MILESTONES! 200 YEARS OF AMERICAN LAW:
MILESTONES IN OUR LEGAL HISTORY vii (1976).
n3 The studies are collected in Paul G. Cassell, Miranda's Social Costs: An Empirical Reassessment, 90
NW. U. L. REV. 387 (forthcoming 1996).
n4 H. RICHARD UVILLER, TEMPERED ZEAL: A COLUMBIA LAW PROFESSOR'S YEAR ON THE
STREETS WITH THE NEW YORK CITY POLICE 198 (1988).
n5 George C. Thomas III, Is Miranda a Real-World Failure? A Plea for More (and Better) Empirical
Evidence, 43 UCLA L. REV. 821 (1996).
n6 Id. at 822; see also LAWRENCE S. WRIGHTSMAN & SAUL M. KASSIN, CONFESSIONS IN THE
COURTROOM ix (1993) ("For reasons that are unclear to us . . . the topic of confession evidence has been
almost completely ignored by psychologists and other social scientists."); Richard A. Leo, Police Interrogation
and Social Control, 3 SOC. & LEGAL STUD. 93, 98 (1994) ("We know very little about what actually happens
during custodial interrogation.").
n7 1 WAYNE R. LAFAVE & JEROLD H. ISRAEL, CRIMINAL PROCEDURE § 6.5(c), at 484 (1984 &
Supp. 1991).
n8 See, e.g., JEROME H. SKOLNICK & JAMES J. FYFE, ABOVE THE LAW: POLICE AND THE
EXCESSIVE USE OF FORCE 58 (1993) (noting that suspects "usually waive their right to remain silent");
CHRISTOPHER SLOBOGIN, CRIMINAL PROCEDURE: REGULATION OF POLICE INVESTIGATION
347 (1993) ("In many jurisdictions, Miranda has had little effect on the confession rate . . . . The suspect often

Page 39
43 UCLA L. Rev. 839, *931

pays the warnings no heed . . . .").
n9 See, e.g., Davis v. United States, 114 S. Ct. 2350, 2355 (1994) (Miranda rule "is not itself required by
the Fifth Amendment's prohibition on coerced confessions, but is instead justified only by reference to its
prophylactic purpose."). See generally JOSEPH GRANO, CONFESSIONS, TRUTH, AND THE LAW 173-98
(1993) (explaining subconstitutional status of Miranda rules).
n10 See, e.g., New York v. Quarles, 467 U.S. 649, 657 (1984) (not applying Miranda rules to "public safety
questioning" because such situations "pos[e] a threat to the public safety [that] outweighs the need for the
prophylactic rule protecting the Fifth Amendment's privilege against self-incrimination").
n11 For convenience, when not discussing specific statistics, this Article will generally use the term
"confession" as broadly encompassing all forms of incriminating statements.
n12 In reaching this conclusion, we also respond to Professor Thomas's "steady state" theory of confessions,
which he explicates in a thoughtful rejoinder to our Article. See George C. Thomas III, Plain Talk About the
Miranda Empirical Debate: A "Steady-State" Theory of Confessions, 43 UCLA L. REV. 933 (1996).
n13 See 1-7 AMERICAN BAR FOUNDATION, THE ADMINISTRATION OF CRIMINAL JUSTICE IN
THE UNITED STATES: PILOT PROJECT REPORT (1957). I am indebted to Professor Richard A. Leo for
calling this research to my attention.
n14 Edward L. Barrett, Jr., Police Practices and the Law -- from Arrest to Release or Charge, 50 CAL. L.
REV. 11 (1962).
n15 Id. at 41-44.
n16 378 U.S. 478 (1964).
n17 See Briefs for Petitioner, Respondent, and the ACLU, Escobedo v. Illinois, 378 U.S. 478 (1964) (No.
615). The brief for respondent Illinois did contain a reference to an empirical study on indigent defendants'
receipt of counsel after arraignment. See Brief for Respondent, supra, at 24 (citing Yale Kamisar & Jesse H.
Choper, The Right to Counsel in Minnesota: Some Field Findings and Legal-Policy Observations, 48 MINN. L.
REV. 1 (1963)). The ACLU's brief contained references to police interrogation manuals. See Brief for the
ACLU, supra, at 4-8.
n18 Escobedo, 378 U.S. at 488 n.10.
n19 See CHARLES H. WHITEBREAD & CHRISTOPHER SLOBOGIN, CRIMINAL PROCEDURE: AN
ANALYSIS OF CASES AND CONCEPTS 379 (3d ed. 1993) ("Escobedo did not provide much guidance to the
court or the police.").
n20 The only specific published studies seem to be: Criminal Investigation Div., Detroit Police Dept.,

Page 40
43 UCLA L. Rev. 839, *931

Confessions in Felony Prosecutions for the Year of 1961 as Compared to January 20, 1965 Through October 31,
1965 (Dec. 13, 1965) (reporting percentage of confessions for various kinds of crimes), reprinted in Theodore
Souris, Stop and Frisk or Arrest and Search -- The Use and Misuse of Euphemisms, 57 J. CRIM. L.,
CRIMINOLOGY & POLICE SCI. 251, 263-64 (1966); Nathan R. Sobel, The Exclusionary Rules in the Law of
Confessions, A Legal Perspective -- A Practical Perspective, N.Y. L.J., Nov. 22, 1965, at 1; see also WAYNE
R. LAFAVE, ARREST: THE DECISION TO TAKE A SUSPECT INTO CUSTODY (1965) (reporting
observational information about arrests).
n21 See Richard H. Kuh, The "Rest of Us" in the "Policing the Police" Controversy, 57 J. CRIM. L.,
CRIMINOLOGY & POLICE SCI. 244, 247 (1966).
n22 Brief of National District Attorneys Association, Amicus Curiae, Miranda v. Arizona, 384 U.S. 436
(1966) (No. 759). Unfortunately, the methodology for collecting this data is unclear, and compilers of the data
acknowledged that "many difficulties were encountered due to insufficient information in the records." Id. at 4a.
n23 Brief of the ACLU, Amicus Curiae, Miranda v. Arizona, 384 U.S. 436 (1966) (No. 759).
n24 Miranda, 384 U.S. at 445 (citing IV NATIONAL COMM'N ON LAW OBSERVANCE AND
ENFORCEMENT, REPORT ON LAWLESSNESS IN LAW ENFORCEMENT (1931)).
n25 Id. at 446 & n.7 (citing Commission on Civil Rights, Justice pts. 5, 17 (1961) and various court
opinions).
n26 Id. at 448.
n27 The only other arguably empirically-based discussion was of FBI questioning practices, id. at 483-86,
and questioning practices of other countries, id. at 486-90.
n28 Id. at 524 (Harlan, J., dissenting).
n29 Project, Interrogations in New Haven: The Impact of Miranda, 76 YALE L.J. 1519 (1967) [hereinafter
Yale Project].
n30 John Griffiths & Richard E. Ayres, A Postscript to the Miranda Project: Interrogation of Draft
Protestors, 77 YALE L.J. 300 (1967).
n31 Richard H. Seeburger & R. Stanton Wettick, Jr., Miranda in Pittsburgh -- A Statistical Study, 29 U.
PITT. L. REV. 1 (1967).
n32 Richard J. Medalie et al., Custodial Police Interrogation in our Nation's Capital: The Attempt to
Implement Miranda, 66 MICH. L. REV. 1347 (1968).

Page 41
43 UCLA L. Rev. 839, *931

n33 James W. Witt, Non-Coercive Interrogation and the Administration of Criminal Justice: The Impact of
Miranda on Police Effectuality, 64 J. CRIM. L. & CRIMINOLOGY 320, 322 (1973).
n34 Controlling Crime Through More Effective Law Enforcement: Hearings Before the Subcomm. on
Criminal Laws and Procedures of the Senate Comm. on the Judiciary, 90th Cong., 1st Sess. 1, 1120 (1967)
(statement of New York County District Attorney Frank S. Hogan) [hereinafter Controlling Crime Hearings].
n35 Id. at 200 (statement of Philadelphia District Attorney Arlen Specter).
n36 Id. at 341 (statement of Los Angeles District Attorney Evelle J. Younger).
n37 Id. at 219 (statement of Kings County District Attorney Aaron E. Koota).
n38 AMERICAN LAW INST., A MODEL CODE OF PRE-ARRAIGNMENT PROCEDURE: STUDY
DRAFT NO. 1, at 140 (1968).
n39 Wayne E. Green, Police vs. "Miranda": Has the Supreme Court Really Hampered Law Enforcement?,
WALL ST. J., Dec. 15, 1966, at 16.
n40 See Donald Janson, Homicides Increase in Chicago, but Confessions Drop by 50%, N.Y. TIMES
(Chicago), July 24, 1967, at 24.
n41 VERA INST. OF JUSTICE, MONITORED INTERROGATIONS PROJECT FINAL REPORT:
STATISTICAL ANALYSIS (1967) [hereinafter VERA INST., MONITORED INTERROGATIONS]; VERA
INST. OF JUSTICE, TAPING POLICE INTERROGATIONS IN THE 20TH PRECINCT, N.Y.P.D. (1967)
[hereinafter VERA INST., TAPING POLICE]. Curiously, the research was never formally published, but is
available only in type-written manuscripts.
n42 NEAL A. MILNER, THE COURT AND LOCAL LAW ENFORCEMENT: THE IMPACT OF
MIRANDA (1971).
n43 DAVID W. NEUBAUER, CRIMINAL JUSTICE IN MIDDLE AMERICA viii (1974).
n44 Lawrence S. Leiken, Police Interrogation in Colorado: The Implementation of Miranda, 47 DEN. L.J. 1
(1970).
n45 See Cassell, supra note 3, at 395-417 (concluding that the reliable studies suggest confessions fell
16.1% after Miranda); see also WHITEBREAD & SLOBOGIN, supra note 19, at 381 (concluding studies
conducted in jurisdictions complying with Miranda "found a significant (10 to 20%) drop in confessions after
Miranda went into effect"). But compare Stephen J. Schulhofer, Miranda's Practical Effect: Substantial Benefits
and Vanishingly Small Social Costs, 90 NW. U. L. REV. 500, 544-47 (forthcoming 1996) (criticizing Cassell's
figures and suggesting that the studies support the conclusion that confessions fell after Miranda somewhere
between 4.1% and 6.4%) with Paul G. Cassell, All Benefits, No Costs: The Grand Illusion of Miranda's

Page 42
43 UCLA L. Rev. 839, *931

Defenders, 90 NW. U. L. REV. (forthcoming 1996) (responding to Schulhofer).
n46 Thomas, supra note 5, at 833.
n47 See id. at 827-31.
n48 The authors of the New Haven study reported that 3 of 81 suspects were apparently prevented from
talking and perhaps confessing by the Miranda warnings and that an additional 10 suspects gave incriminating
statements only after police violated Miranda's questioning cutoff rules. See Cassell, supra note 3, at 408-49
(discussing Yale Project, supra note 29, at 1571, 1578). This means that the study is properly read to support a
16.0% (13/81) drop in admissible confessions after Miranda. Id.; see also Schulhofer, supra note 45, at 530
(agreeing that New Haven study should be read to support at least a 12.3% (10/81) drop in admissible
confessions).
In a reply to our Article, Professor Thomas suggests that our reasoning ignores the "counterfactual" because
it holds constant the behavior of the suspects who received warnings while counting the costs of confessions
obtained in violation of the Miranda questioning cutoff rules. Thomas, supra note 12, at 940. Thus, Thomas
argues we have overlooked the possibility that the warnings might "encourage some suspects to incriminate
themselves." Id. But we are simply reporting what the Yale editors who observed the interrogations reported.
They saw no indication that warnings were facilitating confessions. See Yale Project, supra note 29, at 1571
(discussing behavior of suspects affected by warnings; none apparently induced to talk more by Miranda
warnings).
n49 Professor Thomas acknowledges that the D.C. police were not generally following Miranda's
requirements, but believes that the study can be salvaged by examining confession rates of those suspects who
actually received Miranda warnings with those who did not. See Thomas, supra note 5, at 829-30. But at best
this methodology would reveal the effect of the warnings; it could not reveal the effect of Miranda's other
requirements, such as the questioning cutoff rules. See Cassell, supra note 3, at 410 (noting D.C. police did not
follow Miranda waiver requirements and questioning cutoff rules); see also Schulhofer, supra note 45, at 531
(agreeing with Cassell that "the study's post-Miranda data comes from a period before the D.C. police were fully
implementing Miranda safeguards"). This point assumes some importance when coupled with the fact that most
of Miranda's harms appear to stem not from the warnings, but rather from its other requirements. See infra note
113 (explaining why questioning cutoff rules, not warnings, have the most substantial effect on questioning).
The D.C. study is also suspect because its findings appear to have been presented in a misleading way. See
Richard A. Leo, Police Interrogation in America: A Study of Violence, Civility and Social Change 321 & n.17,
337 & n.25 (1994) (unpublished Ph.D. dissertation, University of California at Berkeley).
n50 See Cassell, supra note 3, at 411 (finding that 41% of suspects who received some of the Miranda
warnings confessed as compared to 55% who received no such warning).
n51 Professor Thomas recognizes that the Denver study suggests that the warning inhibited the making of
admissions, but correctly points out that the key findings are statistically insignificant in large part because of
the tiny sample size (50 suspects total). Thomas, supra note 5, at 828. But, in any event, the study appears to
focus on warnings, not the more important waiver and questioning cutoff rules.
n52 See Cassell, supra note 3, at 405 (discussing problems with Seaside City data); see also Cassell, supra

Page 43
43 UCLA L. Rev. 839, *931

note 45 (discussing why Seaside City must be regarded as anomalous by defenders of Miranda). But see
Schulhofer, supra note 45, at 528-30 (defending the study).
n53 See Cassell, supra note 3, at 399-405, 412-14 (discussing data from these studies); Cassell, supra note
45 (same). But cf. Schulhofer, supra note 45, at 517-38 (criticizing reliance on data from New York County,
New Orleans, Brooklyn, agreeing that Philadelphia and Kansas City data showed drops in the confession rate,
and arguing that Los Angeles data showed a rise in confession rates). To be sure, some of these studies were
conducted by prosecutors, not law professors; others rest on estimates, not hard data. But rejecting these studies
out of hand, particularly in view of the paucity of data, seems the wrong approach. Cf. Thomas, supra note 12, at
948 (noting risk of idiosyncracy in individual studies and concluding that "we need several studies, from which
an average can be obtained").
n54 See infra note 182-192 and accompanying text.
n55 See Cassell, supra note 45 (graphing clearance rates for violent crimes after Miranda); Paul G. Cassell,
The Costs of the Miranda Mandate: A Lesson in the Dangers from Inflexible, "Prophylactic" Supreme Court
Creations, 28 ARIZ. ST. L.J. (forthcoming 1996) (discussing clearance rate data).
n56 See Cassell, supra note 45 (collecting opinion research on police attitudes about Miranda). But see
Schulhofer, supra note 45, at 507-11 (arguing that police reported accommodation to Miranda).
n57 Thomas's interesting speculation is based on data from "Seaside City," which showed that suspects
were less likely to make an outright admission after Miranda but more likely to make incriminating statements.
Thomas, supra note 5, at 831 (citing Witt, supra note 33, at 325 tbl. 3). Thomas suggests that the Miranda
warnings would encourage statements because, among other reasons, suspects "might think that their willingness
to talk in the face of the warnings demonstrates their innocence"; on the other hand, the warnings would
discourage admissions because "even the most superficial understanding of the warnings communicates that the
suspect's words will be used in court." Thomas, supra note 5, at 831. However, Thomas's interpretation does not
fit the Seaside City data. Police there apparently began giving warnings to suspects in January 1965, when
People v. Dorado, 398 P.2d 361 (Cal.), cert. denied, 381 U.S. 937 (1965), required them to do so. See Witt,
supra note 33, at 325 n.41. In 1965, oral incriminating statements fell sharply. See id. at 325 tbl. 3. Thomas's
conjecture also conflicts with data from other jurisdictions. See Controlling Crime Hearings, supra note 34, at
161 (reporting decline in statements in New York County in homicide cases after Miranda); id. at 200-01
(reporting that statements in Philadelphia fell both after Miranda and also after other earlier decisions requiring
police warnings); id. at 161 (reporting that statements in Brooklyn fell after Miranda); Cassell, supra note 3, at
411-12 (reanalyzing data from the District of Columbia suggesting that statements dropped after Miranda);
Green, supra note 39, at 16 (reporting that statements in Kansas City fell about 12% after Miranda).
n58 See Thomas, supra note 5, at 822.
n59 See, e.g., FLOYD FEENEY ET AL., ARRESTS WITHOUT CONVICTION: HOW OFTEN THEY
OCCUR AND WHY (1983); FLOYD FEENEY & ADRIANNE WEIR, THE PREVENTION AND CONTROL
OF ROBBERY (1974); Gary D. LaFree, Adversarial and Nonadversarial Justice: A Comparison of Guilty Pleas
and Trials, 23 CRIMINOLOGY 289, 302 (1985).

Page 44
43 UCLA L. Rev. 839, *931

n60 Leo, supra note 49. Leo's informative dissertation will be published as Richard A. Leo, Inside the
Interrogation Room: A Qualitative and Quantitative Analysis of Contemporary American Police Interrogation
Practices, 86 J. CRIM. L. & CRIMINOLOGY (forthcoming 1996), and Richard A. Leo, The Impact of Miranda
Revisited, 86 J. CRIM. L. & CRIMINOLOGY (forthcoming 1996).
n61 See infra notes 378-410 (Appendix B) (discussing the Leo study in more detail).
n62 See generally ROYAL COMM'N ON CRIMINAL JUSTICE, REPORT (1993) (describing criminal
justice questions in Britain).
n63 See, e.g., DAVID BROWN, DETENTION AT THE POLICE STATION UNDER THE POLICE AND
CRIMINAL EVIDENCE ACT 1984 (1989); BARRIE L. IRVING, POLICE INTERROGATION: A CASE
STUDY OF CURRENT PRACTICE (1980); BARRIE L. IRVING & IAN K. McKENZIE, POLICE
INTERROGATION: THE EFFECTS OF THE POLICE AND CRIMINAL EVIDENCE ACT 1984 (1989);
MICHAEL McCONVILLE, CORROBORATION AND CONFESSIONS: THE IMPACT OF A RULE
REQUIRING THAT NO CONVICTION CAN BE SUSTAINED ON THE BASIS OF CONFESSION
EVIDENCE ALONE (1993); PAUL SOFTLEY, POLICE INTERROGATION: AN OBSERVATIONAL
STUDY IN FOUR POLICE STATIONS (Royal Comm'n on Criminal Procedure Research Study No. 4, 1980);
Barry Mitchell, Confessions and Police Interrogation of Suspects, 1983 CRIM. L. REV. 596; Stephen Moston et
al., The Incidence, Antecedents and Consequences of the Use of the Right to Silence During Police Questioning,
3 CRIM. BEHAV. & MENTAL HEALTH 30 (1993).
n64 Thomas, supra note 5, at 837.
n65 Unattributed descriptions of the practices of law enforcement and prosecuting agencies in this Article
are based on personal interviews with members of the affected agencies during and following the collection of
the data.
n66 See, e.g., KATHLEEN B. BROSI, INSTITUTE FOR LAW AND SOCIAL RESEARCH, A
CROSS-CITY COMPARISON OF FELONY CASE PROCESSING 7, 9, 10, 12, 15, 24 (1979) (finding that Salt
Lake City was in the middle of five jurisdictions in its rejections at screening and that dispositions after arrest
looked similar to other jurisdictions); see also John D. O'Connell & C. Dean Larsen, Note, Detention, Arrest,
and Salt Lake City Police Practices, 1965 UTAH L. REV. 593 (reporting results of field research on Salt Lake
City police practices).
n67 BUREAU OF ECONOMICS AND BUSINESS RESEARCH, UNIVERSITY OF UTAH,
STATISTICAL ABSTRACT OF UTAH 1993, at 25 (1994) (citing Bureau of the Census information).
n68 Id. at 24 tbl. 18.
n69 See generally Paul G. Cassell, Balancing the Scales of Justice: The Case for and the Effects of Utah's
Victims' Rights Amendment, 1994 UTAH L. REV. 1373, 1376-79 (collecting statewide statistics on this point).
n70 DEPARTMENT OF PUBLIC SAFETY, BUREAU OF CRIMINAL IDENTIFICATION, CRIME IN

Page 45
43 UCLA L. Rev. 839, *931

UTAH 1994, at 17 (1994) [hereinafter cited as CRIME IN UTAH].
n71 See U.S. DEPT. OF JUSTICE, UNIFORM CRIME REPORTS, CRIME IN THE UNITED STATES
1993, at 5 (1994) [hereinafter cited as 1993 UNIFORM CRIME REPORTS].
n72 Survey Shows S.L. with High Crime Rate for City Its Size, SALT LAKE TRIB., Apr. 27, 1993, at B10.
n73 S.L. Crime Up, Bucks U.S. Trend, SALT LAKE TRIB., May 22, 1995, at A1.
n74 See Anti-Gang Bills Pass in Special Session, SALT LAKE TRIB., Oct. 17, 1993, at B4.
n75 Compare CRIME IN UTAH, supra note 70, at 17 (Salt Lake County violent crime rate of 3.7 per 1000)
with 1993 UNIFORM CRIME REPORTS, supra note 71, at 10 (national violent crime rate of 7.5 per 1000).
n76 Prosecutors told us that their summer caseload was not significantly different than during other parts of
the year.
n77 In a few cases, the screening resulted in the decision to file misdemeanor rather than felony charges.
We continued to follow the disposition of these cases, which were a tiny fraction of our sample.
n78 Most of the data were collected by one of the authors (Hayman) and a second-year law student. The
other of the authors (Cassell) collected data for the last day of the study.
n79 The survey forms used to collect the data are set forth in Appendix A, infra.
n80 The four cases included a sensitive rape case and a homicide case involving a suspect with complicated
psychiatric problems. These cases did not appear, on the surface, to pose major Miranda issues.
n81 This estimate does not include cases involving drug sweeps in a downtown park. See infra note 84.
n82 Confessions may be less important in drug cases. See SKOLNICK & FYFE, supra note 8, at 194;
UVILLER, supra note 4, at 199.
n83 Some screenings for child abuse cases were held in the Salt Lake Children's Justice Center. See UTAH
CODE ANN. §§ 67-5b-101 to-107 (Supp. 1995) (describing Children's Justice Centers).
n84 Small drug cases are underrepresented in part because we excluded "Pioneer Park" cases from our
sample. Pioneer Park is an area in Salt Lake City of concentrated drug trafficking. See Brian Maffly, Crime and
Commerce: They Coexist Painfully Near Pioneer Park, SALT LAKE TRIB., Oct. 23, 1994, at B1. Salt Lake
City police conduct concerted "buy-bust" operations in the park, in which an undercover officer purchases drugs
while under surveillance from the other officers, who then join the undercover officer in an arrest. In these cases,

Page 46
43 UCLA L. Rev. 839, *931

the officers do not typically question suspects for a combination of reasons: the cases are "slam-dunks," the
street dealers rarely provide useful cooperation against higher-ups, and language problems (many of the suspects
are immigrants from Mexico) make interrogation difficult. The researchers attended six Pioneer Park screenings.
In none of the six was the suspect interrogated. At that time, concerned that the sheer numbers of such cases
would hamper reasonable statistical analysis, we decided to exclude these cases from the sample and not to
attend any further Pioneer Park screenings. The effect of our decision was to increase the percentage of
confessions we found in our sample.
n85 190 males of 219 suspects.
n86 146 with prior records of 208 suspects, with information unavailable on an additional 11 suspects.
n87 Offenders under the age of 18 are handled by Salt Lake County's juvenile court system. While juvenile
offenders can be "certified" and tried as adults, see UTAH CODE ANN. § 78-3a-25 (1953 & Supp. 1995), such
cases are rare and were not present in our sample.
n88 137 white, 49 Hispanic, 12 African-American, 4 other suspects, with information unavailable on an
additional 17 suspects.
n89 Naturally enough, categorizing offenses into only two categories requires some generalizations. For
example, we have included drug offenses under the heading of "violent" offenses. For a similar categorization,
see NEUBAUER, supra note 43, at 104.
n90 See UVILLER, supra note 4, at 199.
n91 173 questioned of 219 suspects.
n92 See Gordon Van Kessel, The Suspect as a Source of Testimonial Evidence: A Comparison of the
English and American Approaches, 38 HASTINGS L.J. 1, 116-17 (1986) (advancing this hypothesis and
collecting studies). But see Schulhofer, supra note 45, at 34 ("There is no apparent reason why Miranda would
prevent police from making an initial attempt to question a suspect . . . .").
n93 FEENEY ET AL., supra note 59, at 13 tbl. 15-2 (Jacksonville 37/200; San Diego 44/219).
n94 See Seeburger & Wettick, supra note 31, at 7, 23; see also id. at 13 tbl. 3 (all suspects divided into one
of two categories: refused to talk or willing to talk). It is possible that the interrogation rate statistics were
skewed since the study involved "the more difficult cases for which investigation was deemed necessary by the
police." Id. at 7. Unfortunately there is no way to determine the interrogation rate for cases not referred to the
Detective Branch.
n95 Witt, supra note 33, at 323.
n96 Id. at 325 tbl. 3.

Page 47
43 UCLA L. Rev. 839, *931

n97 Barrett, supra note 14, at 43-44.
n98 Controlling Crime Hearings, supra note 34, at 200.
n99 Id. at 223.
n100 Yale Project, supra note 29, at 1589 tbl. F-7 (reporting that 2 of 90 suspects were "not questioned");
see also Cassell, supra note 3, at 407-08 (arguing that the New Haven study is really a pre-Miranda study
because the police were following pre-Miranda procedures).
The Yale editors cautioned that they observed only felony cases brought to the station house for
questioning, Yale Project, supra note 29, at 1587, and reported (without specifying a basis for the finding) that
"no statement was requested in over 25 per cent of the cases," id. at 1574. The editors observed that police rarely
interrogated in misdemeanor cases. For example, the Special Services Division (vice squad) questioned only 8
suspects out of 200 arrests. Id. at 1587. The Division handled predominantly misdemeanor cases. Id. at 1587
n.179.
n101 Medalie et al., supra note 32, at 1364. This data was based on interviews with defendants. Id. It is hard
to tell whether the study's "no interrogation" category included suspects who were in fact questioned but
requested counsel.
n102 In his reply to our Article, Professor Thomas appropriately highlights this point. See Thomas, supra
note 12, at 944-46.
n103 Witt, supra note 33, at 325; Medalie et al., supra note 32, at 1365 tbl. 5. Professor Thomas contends
that the change in the interrogation rate in the District of Columbia should be disregarded because it is not
statistically significant. Thomas, supra note 12, at 944 n.50. However, to reach this conclusion he shrinks the
sample size of the study by throwing out post-Miranda defendants with counsel on the grounds that they were
not included in the pre-Miranda sample. Id. Nothing in the report of the D.C. study supports this conjecture.
Instead, the study strongly suggests consistent pre-and post-Miranda sampling. See Medalie et al., supra note
32, at 1351, 1354 (noting that a series of interviews with District of Columbia defendants was divided into
pre-Miranda and post-Miranda categories based on date of arrest).
n104 Arizona v. Fulminante, 111 S. Ct. 1246, 1257 (1991).
n105 For example, one case involving police belief that a suspect would invoke Miranda arose when the
mother of a suspect told police her son would not talk without an attorney. The detectives left their card with the
mother in case the son wanted to talk. He never called.
n106 The two cases involved suspects who had not yet been charged or detained, suggesting that the
prophylactic Miranda right to counsel had not yet technically attached. Stansbury v. California, 114 S. Ct. 1526,
1529 (1994) (per curiam) (discussing custody as triggering event for Miranda right to counsel). However, courts
have suggested that questioning in such circumstances is troublesome, particularly under rules of ethics
applicable to prosecutors and their agents. See, e.g., United States v. Foley, 735 F.2d 45, 47-48 (2d Cir. 1984). It
seems fair to conclude that some of the questions about questioning in such circumstances has arisen from the

Page 48
43 UCLA L. Rev. 839, *931

backdrop of the Miranda rules. See, e.g., id. at 48 (citing Miranda-related cases as grounds for concern).
n107 In his reply to our study, Professor Thomas essentially agrees that at least some suspects will not be
questioned due to Miranda, but argues that the extent of that effect is small. See Thomas, supra note 12, at 945.
Agreeing that some suspects are not questioned, however, is an important concession: It means that Miranda has
reduced the interrogation rate (and, inevitably, the confession rate) and the only subject remaining for discussion
is the size of that effect. Moreover, Thomas's quantification of the size of the Miranda effect we found is too
small. He includes only cases in which Miranda was directly identified as the reason for not questioning, failing
to consider cases in which Miranda's indirect effects may have prevented questioning. See supra note 105, 106;
infra note 109. It is also important to remember that our study is not the only one to estimate the interrogation
rate drop after Miranda. Other data suggest a larger drop in the interrogation rate. See supra note 103 and
accompanying text.
n108 The effect of logistical difficulties of arranging for interviews has been noted in other studies. See,
e.g., FEENEY ET AL., supra note 59, at 143 (finding that hours worked by detectives and physical layout of jail
make interrogations more difficult in San Diego than in Jacksonville).
n109 See OFFICE OF LEGAL POLICY, U.S. DEP'T OF JUSTICE, REPORT TO THE ATTORNEY
GENERAL: THE LAW OF PRE-TRIAL INTERROGATION 126 (1986) [hereinafter OLP PRE-TRIAL
INTERROGATION REPORT] (recounting case of murderer who went free due to Miranda because confession
was obtained after routine assignment of counsel) reprinted in 22 U. MICH. J.L. REF. 437 (1989).
n110 Cf. Witt, supra note 33, at 325 n.44 (noting that Miranda allows "unenterprising detectives . . . [to]
rationalize their indolence"); see also supra note 84 (giving example of cases where questioning was rare).
n111 Thomas, supra note 5, at 823-24, 826, 832-33.
n112 See generally LAFAVE & ISRAEL, supra note 7, at 479-83. Cf. Davis v. United States, 114 S. Ct.
2350 (1994) (assertion of right to counsel after waiver must be unambiguous).
n113 See Cassell, supra note 3, at 492-96 (collecting evidence showing that warnings had small impact but
waiver requirements had large impact).
This point may be underemphasized in Professor Thomas's article. For instance, the article calls for research
on the behavior of suspects after they have received judicial warnings at the initial appearance. Thomas, supra
note 5, at 834 (calling for research "to determine what part, if any, judicial warnings played in the interrogation
process"). While this might be a useful project to undertake in its own right, it will shed little light on the costs
of Miranda, which stem in significant part from suspects invoking their rights before initial appearance.
n114 See Cassell, supra note 3, at 495 n.623.
n115 108 of 129 suspects asked to waive rights.
n116 Both had prior criminal records.

Page 49
43 UCLA L. Rev. 839, *931

n117 Cf. ROGER LEND, THE RIGHT TO SILENCE IN POLICE INTERROGATION 16-17 (Royal
Comm'n on Criminal Justice Research Study No. 10, 1993) (employing similar methodology). There was one
such case in our sample involving a suspect who was suspected of a series of check forgeries. He agreed to talk
to police about some of the forgeries but not others. Police obtained incriminating statements from him.
n118 See infra Table 4.
n119 In two of the three cases, the suspects were questioned at the scene and gave denials. At the station
house, they invoked their rights. In the third case, a suspect was questioned and invoked her rights. She later
called a probation officer and a short discussion of a non-incriminating nature followed.
n120 In the first (a domestic violence case), police questioned a suspect at the scene non-custodially, where
he admitted that he stabbed his wife. They then took him to the police station and gave him Miranda. He refused
to waive his rights for follow-up questioning. In the second (an attempted murder case), a patrol officer at the
scene asked a suspect a few desultory questions without Miranda. By the time the case reached a detective for
interrogation, the suspect had been assigned an attorney.
n121 Harris v. New York, 401 U.S. 222 (1971).
n122 See, e.g., Albert W. Alschuler, Failed Pragmatism: Reflections on the Burger Court, 100 HARV. L.
REV. 1436, 1442-43 (1987) (suggesting police will ignore invocations of Miranda rights); Geoffrey R. Stone,
The Miranda Doctrine in the Burger Court, 1977 SUP. CT. REV. 99, 113 (commenting that Harris "seems to
carry at least the potential seriously to undercut the incentive of the police to comply with the dictates of
Miranda") (footnote omitted); Alan M. Dershowitz & John H. Ely, Comment, Harris v. New York: Some
Anxious Observations on the Candor and Logic of the Emerging Nixon Majority, 80 YALE L.J. 1198, 1220
(1971) ("Under the Harris rule, what possible incentive would the police have to comply with Miranda . . . .");
see also Harris v. New York, 401 U.S. 222, 232 (1971) (Brennan, J., dissenting) ("I fear that today's holding will
seriously undermine the achievement of [the] objective [of deterring improper police questioning].").
n123 It might be argued that because our study relies on police recounting of events, it would fail to capture
questioning in violation of Miranda. After all, police might be unwilling to admit deviations from Miranda.
However, the relative infrequency of Miranda suppression motions, see infra note 243 and accompanying text,
corroborates the law enforcement version.
n124 This figure is derived by multiplying 12.1% by the 2,800,000 arrests for FBI index crimes for 1993.
See 1993 UNIFORM CRIME REPORTS, supra note 71, at 217 tbl. 29. FBI index crimes are: murder, rape,
robbery, assault, burglary, larceny, and vehicle theft. Our sample includes all felonies, not just index crimes.
n125 See, e.g., Yale Kamisar, Edward L. Barrett, Jr.: The Critic with "That Quality of Judiciousness
Demanded of the Court Itself," 20 U.C. DAVIS L. REV. 191, 210 (1987) ("It cannot be denied that for the past
twenty years suspects have continued to make incriminating statements with great frequency.").
n126 See Thomas, supra note 5 (noting importance of distinguishing between confessions and incriminating
statements); see also THOMAS GRISSO, JUVENILES' WAIVER OF RIGHTS: LEGAL AND

Page 50
43 UCLA L. Rev. 839, *931

PSYCHOLOGICAL COMPETENCE 35 n.23 (1981) (noting difficulty of classifying "admissions,"
"confessions," etc.); McCONVILLE, supra note 63, at 30 (discussing problems in defining "confession").
n127 Similar terminology is employed in Yale Project, supra note 29, at 1643-47.
n128 Survey protocol at 4 (copy on file with Cassell).
n129 Id.
n130 Id. The intent of this category was to capture cases where the prosecutor had direct evidence that a
particular alibi scenario was untrue. Of course, prosecutors who convict defendants with alibis always have
substantial indirect evidence that the alibi is untrue -- otherwise the prosecution would fail.
n131 For example, in the case just discussed, we classified a suspect's stammering and hesitating answers as
"incriminating," even though they did not substantively implicate him in the auto theft, because their tone made
the officer believe something was amiss. Acting on his suspicions, the officer checked and the suspect was
arrested and charged with auto theft. A second example comes from a first degree murder case. During police
questioning the suspect admitted being at the scene -- a fact never really in doubt -- but stated that he could not
remember anything else. Eventually the suspect pled to a reduced manslaughter charge. Thus, his
"incriminating" statements did not end up being notably useful in prosecuting the crime charged. A final
example of an "on the fence" incriminating statement occurred in a domestic violence case where the victim
claimed her boyfriend hit her. The suspect responded to police questioning by stating that the two were fighting
but he never struck her. We classified the statement as incriminating because the defendant admitted that he and
the victim were fighting. The district attorney's office did not prosecute the alleged assault due to evidence
problems.
n132 Survey protocol at 4 (copy on file with Cassell).
n133 Id. Professor Thomas calls this classification scheme "more problematic than most." Thomas, supra
note 12, at 949. Yet we used this classification because it appeared in ANDREW SANDERS ET AL., ADVICE
AND ASSISTANCE AT POLICE STATIONS AND THE 24 HOUR DUTY SOLICITOR SCHEME 135
(1989), a widely-cited British study on confessions that appeared to be more sophisticated than other American
studies on point.
n134 See UTAH CODE ANN. § 76-6-404 (1995) ("A person commits theft if he obtains or exercises
unauthorized control over the property of another with a purpose to deprive him thereof.") (emphasis added).
n135 Professor Thomas questions whether more than a very few suspects would testify falsely, asking,
"How many defendants would falsely testify at trial that the car was recently purchased?" Thomas, supra note
12, at 950. But, in fact, testimony by defendants claiming to have purchased stolen cars is routine. See, e.g.,
State v. Larocco, 742 P.2d 89, 90 (Utah Ct. App. 1990) (noting that the defendant "consistently claimed he had
purchased the [allegedly stolen] Mustang"). According to the supervisor of the section of the Salt Lake County
District Attorney's Office responsible for prosecuting stolen car cases, a claim of legitimate purchase is "the
most common defense we get." Telephone Interview with Deputy District Attorney Ernest Jones (Jan. 10, 1996).

Page 51
43 UCLA L. Rev. 839, *931

Defeating such explanations is a "50/50 proposition" because sometimes the jury will credit the story, or at least
be unable to reject it beyond a reasonable doubt. Id. Viewed in this real world context, statements of the form "I
bought the car" cannot realistically be viewed as "incriminating."
n136 See UTAH CODE ANN. § 76-6-202 (1995) ("A person is guilty of burglary if he enters or remains
unlawfully in a building or any portion of a building with intent to commit a felony or theft or commit an assault
on any person.").
n137 See id. § 76-5-402 (1995) ("A person commits rape when the actor has sexual intercourse with another
person without the victim's consent.").
n138 Two of the nine involved unlawful sexual fondling, where the suspects claimed that the touching was
accidental.
n139 Professor Thomas claims that as many as two-thirds of our denials with explanations might have been
categorized as incriminating statements by earlier researchers. Thomas, supra note 12, at 951. This assertion is
untenable. The earlier studies defined incriminating statements as those that were helpful to police in
prosecuting cases. See, e.g., Seeburger & Wettick, supra note 31, at 10 ("We did not include as confessions any
statement which contained certain self-incriminating statements but were primarily self-serving in nature and of
little value to the police . . . .") (emphasis added); Witt, supra note 33, at 325 n.43 (including "oral incriminating
evidence or other useful material for conviction through interrogation") (emphasis added). Likewise, our study
made helpfulness to the police the touchstone of incriminating statements. With this in mind, Thomas's
suggestion that statements like "I fought back in self-defense" or "I bought the [allegedly stolen] car" would be
regarded as helpful to the prosecution, Thomas, supra note 12, at 949-50, seems dubious.
It also should be noted that Professor Thomas parses our study for incriminating statements we might have
miscategorized as denials, but not the reverse, offsetting possibility: denials that we miscategorized as
incriminating. Cf. supra note 131 (explaining that statements were classified as "incriminating" even though
they were not substantive admissions of a crime). Of course, Professor Thomas is free to conduct such
flyspecking of our denial category only because we have provided more information about our classifications
than any other previous confessions study. We hope that our extensive explication will provide a new standard
of disclosure to which future researchers on confessions will adhere.
n140 Confirming that our "denials with explanation" were not really "incriminating" statements is the
ultimate outcome of such cases: Those who gave denials with explanations were significantly less likely to be
convicted than those who gave incriminating statements. Of suspects who gave denials with explanations, 18
were convicted of some charge (47%) while 20 were not. Of suspects who gave incriminating statements, 17
were convicted of some charge (72%) while 8 were not. (The chi-square statistic for the difference was
significant at the .05 level.) Even among those convicted of some charge, those who gave only denials (rather
than incriminating statements) fared better. Of convicted suspects who gave denials with explanations, three
(17%) were convicted of an offense as serious as the one with which they were charged and nine (50%) were
convicted of an offense only one level lower. See infra notes 322-328 and accompanying text (describing
methodology for determining case outcomes). Of convicted suspects who gave incriminating statements, four
(24%) were convicted of an offense as serous as the one with which they were charged and 12 (71%) were
convicted of an offense only one level lower. Professor Thomas despairs of the "near impossibility" of making a
distinction between incriminating and exculpatory statements. But these findings -- that suspects who gave a
denial with explanation were less likely to be convicted of some charge and, even when convicted, more likely

Page 52
43 UCLA L. Rev. 839, *931

to be able to escape with a conviction to a lower charge than those who gave incriminating statements -- are
precisely the results one would expect if we have properly distinguished between incriminating and exculpatory
statements. Thus, we must strongly disagree with Professor Thomas's suggestion that "for future research, a
better system is one that includes all explanations as potentially incriminating." Thomas, supra note 12, at 951.
Rather, we believe -- and our statistics confirm -- that lumping such statements together will obscure the
important real world difference between incriminating and exculpatory statements.
n141 See infra notes 297-305 and accompanying text (discussing multiple interviews of suspects).
n142 Police also were fortunate enough to receive thirteen "volunteered" incriminating statements,
including nine statements from suspects who had not otherwise provided incriminating information. See infra
notes 230-237 and accompanying text (discussing volunteered statements). Adding these nine suspects into the
total would produce the result that 37.4% of our cases contained incriminating statements.
n143 We address this issue infra notes 174-177 and accompanying text. In view of this concern, for
subsequent tables dealing with variables affecting interrogation success, we exclude suspects who were never
questioned.
n144 Our success rate would also have been lower if we included more routine drug cases, where police
officers often did not interrogate. See supra note 84 (discussing exclusion of park drug sweep cases from
sample).
n145 See VERA INST., TAPING POLICE, supra note 41, at 62; Controlling Crime Hearings, supra note
34, at 199, 205 (statement of Arlen Specter, District Attorney for the City and County of Philadelphia,
Pennsylvania).
n146 See generally Cassell, supra note 3, at 463-64.
n147 The chi-square statistic is not significant at .05 level, but is significant at .10 level. For the chi-square
statistics here and elsewhere in this Article, we have focused on factors that might produce statistically
significant differences in successful and unsuccessful questioning and have therefore excluded the invocation
category (unless specifically noted otherwise). We report chisquare results with some trepidation, because our
small sample size may make such tests misleading. In particular, a small sample size means that all but the most
substantial differences will be found to be not statistically significant. Cf. Medalie et al., supra note 32, at 1414
(declining to report statistical significant tests at all on these grounds); Thomas, supra note 5, at 823 n.7 (using
.10 significance level in view of paucity of data). For the crimes included in the categories of "violent" and
"property" crimes, see supra Table 1.
n148 See Cassell, supra note 3, at 458-59 & tbl. 3.
n149 Barrett, supra note 14, at 43-44.
n150 Souris, supra note 20, at 264.

Page 53
43 UCLA L. Rev. 839, *931

n151 See Cassell, supra note 3, at 406 (discussing Yale Project, supra note 29, at 1573, 1644).
n152 The only other early figures come from the NDAA Brief in Miranda, the methodology of which is
unclear. See supra note 22.
n153 See, e.g., Cassell, supra note 3, at 403 (noting that estimated confession rate fell in Philadelphia after
Escobedo and again after a pre-Miranda decision by the Third Circuit). See generally Cassell, supra note 45
(discussing influence of Escobedo in confession rate figures).
n154 See SLOBOGIN, supra note 8, at 6 (Supp. 1995) (concluding that a 64% confession rate is
"comparable to pre-Miranda confession rates").
Curiously, Professor Thomas ignores these pre-Escobedo studies in calculating a pre-Miranda confession
rate. Rejecting our estimate, he proposes instead an estimated range of 45%-53% based solely on Cassell's table
of data from seven jurisdictions in which data from both before and after Miranda is available. See Thomas,
supra note 12, at 935-36 & n.12 (citing Cassell, supra note 3, at 418 & tbl. 1). But in determining the confession
rate before Miranda, there is no sound reason for arbitrarily confining the database to jurisdictions in which we
can determine what the confession rate was after. Indeed, the inevitable consequence of such a limitation is to
use only the most problematic data, that is, only the data obtained closest to Miranda and thus most likely to be
infected by anticipatory implementation problems. In any event, simply including the five confession rates
mentioned in the text above with the seven studies Professor Thomas uses produces an estimated pre-Miranda
confession rate of 56%, which lies within our estimate but outside of his.
n155 Thomas, supra note 12, at 936, 952.
n156 Id. at 947.
n157 See infra notes 216-221 and accompanying text.
n158 Thomas, supra note 12, at 952.
n159 Medalie et al., supra note 32, at 1351.
n160 Witt, supra note 33, at 323 ("This eliminated all cases in which suspects were detained for questioning
but never incarcerated.").
n161 Seeburger & Wettick, supra note 31, at 7.
n162 See id. ("It is standard practice for the [Detective] Branch to question every suspect in all cases in
which it makes an arrest. The questioning occurs at the time the suspect is taken into custody.").
n163 Our Table 15, infra, divides suspects into those against whom charges were filed and those against
whom they were not. Looking only at the category for charges filed, the "questioning successful" rate is 63/173

Page 54
43 UCLA L. Rev. 839, *931

or 36.4%, still well below pre-Miranda levels.
n164 See Cassell, supra note 3, at 459 tbl. 3.
n165 Controlling Crime Hearings, supra note 34, at 1120.
n166 Id. at 223 (referring to "suspects" in these categories of crimes); see also id. at 224 (giving example of
an interview of a suspect in his home where "custody" was disputed).
n167 See infra notes 169-172 and accompanying text (collecting evidence that Miranda has led to more
noncustodial interrogations).
n168 Thomas, supra note 12 (noting that our data "confirm that non-custodial settings produce fewer
successful interrogations") (footnote omitted).
n169 FRED E. INBAU ET AL., CRIMINAL INTERROGATION AND CONFESSIONS 234 (3d ed. 1986)
(The Supreme Court's decision in Mathiason "is an excellent illustration of the advisability of arranging,
whenever feasible, for an interrogation opportunity based upon a consensual situation. A suspect who has
willingly consented to come to or be taken to the place of interrogation . . . is not in custody."); cf. FRED E.
INBAU & JOHN E. REID, CRIMINAL INTERROGATION AND CONFESSION (1962) (pre-Miranda edition
of same manual that apparently does not discuss need to avoid interrogating in custody).
n170 See, e.g., H. RICHARD UVILLER, TEMPERED ZEAL: A COLUMBIA LAW PROFESSOR'S
YEAR ON THE STREETS WITH THE NEW YORK CITY POLICE 52 (1988) (reporting a case where the
police "didn't really want to arrest him at that point; they figured if he wasn't under arrest, they could 'interview'
him without giving him Miranda warnings"); Jerome H. Skolnick & Richard A. Leo, The Ethics of Deceptive
Interrogation, CRIM. JUSTICE ETHICS, Winter/Spring 1992, at 5 ("Police will question suspects in a
'non-custodial' setting . . . so as to circumvent the necessity of rendering warnings. This is the most fundamental,
and perhaps the most overlooked, deceptive stratagem police employ.").
n171 See, e.g., United States v. J.H.H., 22 F.3d 821, 831 (8th Cir. 1994) (suspect told that questioning was
voluntary, that he was not under arrest, and that he was free to leave at any time, apparently taking advantage of
Supreme Court decision finding similar questioning noncustodial in California v. Beheler, 463 U.S. 1121, 1125
(1983) (per curiam)); see also infra notes 209-210 and accompanying text (reporting use of Beheler warnings in
Salt Lake County); cf. Oregon v. Mathiason, 429 U.S. 492, 499 n.5 (1977) (Marshall, J., dissenting) (suggesting
that in the context of traffic stops "police officers can circumvent Miranda by deliberately postponing the
official 'arrest' and the giving of Miranda warnings until the necessary incriminating statements have been
obtained").
n172 See, e.g., JEROME H. SKOLNICK & JAMES J. FYFE, ABOVE THE LAW: POLICE AND THE
EXCESSIVE USE OF FORCE 58-59 (1993) ("Police who are seeking admissions have learned how to
interview rather than interrogate. . . . 'Interviews' do not require Miranda warnings . . . ."); Mark Berger,
Compromise and Continuity: Miranda Waivers, Confession, Admissibility, and the Retention of Interrogation
Protections, 49 U. ILL. L. REV. 1007, 1020 (1988) (police will be able to undertake interrogation "in a

Page 55
43 UCLA L. Rev. 839, *931

non-custodial environment in order to avoid having to administer warnings and secure waivers"); Irene M.
Rosenberg & Yale L. Rosenberg, A Modest Proposal for the Abolition of Custodial Confessions, 68 N.C. L.
REV. 69, 112 (1989) (suggesting that police will circumvent Miranda by conducting coercive noncustodial
interviews in suspect's home); see also Yale Kamisar, "Custodial Interrogation" Within the Meaning of Miranda,
in CRIMINAL LAW AND THE CONSTITUTION -- SOURCES AND COMMENTARIES 335, 341 (Jerold H.
Israel & Yale Kamisar eds., 1968) ("I think it is quite legitimate to read Miranda as encouraging the police to
engage more extensively in pre-arrest, pre-custody, pre-restraint questioning.") (quoting Kamisar's remarks at a
1966 CLE conference on Miranda).
n173 Professor Thomas derides the evidence as consisting of "nothing but anecdotes and police
interrogation manuals." Thomas, supra note 12, at 947 n.60. But the fact remains that this is the only available
evidence on this issue because of the dearth of pre-Miranda empirical studies distinguishing between custodial
and noncustodial interrogations. Cf. id. at 943 (promising to "accept the Cassell-Hayman challenge to render a
judgment on the available evidence").
n174 See supra notes 91-104 and accompanying text.
n175 See Thomas, supra note 12, at 952 tbl.
n176 See Medalie et al., supra note 32, at 1365 tbl. 5, 1373 tbl. 9 (reporting statement rates in which a
substantial proportion of suspects were never questioned); Seeburger & Wettick, supra note 31, at 7 ("For
almost every file which we examined, the Detective Branch attempted to obtain a confession from all the
suspects involved.") (emphasis added); Witt, supra note 33, at 325 tbl. 3 (reporting confession rate data with
category for "suspect not questioned"); supra notes 164-167 and accompanying text (collecting many studies
which determined number of confessions present in criminal cases that were filed, without regard to whether
suspects were questioned or not); see also Yale Project, supra note 30, at 1589 tbl. F-7 (reporting that two
percent of the suspects sampled were "not questioned" after Miranda).
n177 Professor Thomas responds that this means that the Salt Lake City confession rate is lower than
pre-Miranda rates, in part, because the interrogation rate is lower. See Thomas, supra note 12, at 946. But the
critical point is that, contrary to Thomas's "steady-state" hypothesis, post-Miranda confession rates are in fact
lower. We can then, of course, debate causes for that fact. Cf. supra notes 92-103 and accompanying text
(arguing that lower interrogation rates are attributable, at least in part, to Miranda). It is important to remember
that the total number of confessions -- our subject of inquiry -- is determined not just by suspect's responses to
Miranda but by police responses as well. To borrow an economic analogy, one must assess not only the
"supply" of confessions but also the "demand" for them. Cf. Stephen J. Markman & Paul G. Cassell, Protecting
the Innocent: A Response to the Bedau-Radelet Study, 41 STAN. L. REV. 121, 155 (1988) (noting importance of
measuring supply and demand for crime in deterrence studies).
n178 FEENEY ET AL., supra note 59, at 142. Professor Thomas adjusts these figures upwards by
excluding suspects who were never questioned. Thomas, supra note 12, at 954. This adjustment is inappropriate.
See supra notes 174-177 and accompanying text.
n179 LaFree, supra note 59, at 298. Professor Thomas notes that the study used a narrow definition of
"confession" that, in his view, lowered the confession rate. Thomas, supra note 12, at 955. A compensating
effect, however, was that the study's sample was drawn from court cases, not police interrogations. Id. at 298 tbl.

Page 56
43 UCLA L. Rev. 839, *931

2. Because cases are more likely to result in charges and verdicts if a defendant confesses, see infra notes
315-341 (finding that incriminating statements affect the processing of criminal cases), the overall confession
rate in police interrogations was certainly below the 40.3% reported. Sorting out these two offsetting effects
precisely is impossible; but their countervailing nature suggests that simply relying on the reported 40.3% figure
is reasonable.
n180 See infra notes 378-410 and accompanying text.
n181 Of course, this is not a criticism of the study itself. Leo's study is probably the single best source of
information about what goes on during custodial interrogation by experienced detectives, which was the subject
the study was designed to explore. The study is applied beyond its intended purposes, however, if used as a
measure of overall police success in obtaining confessions.
n182 See Cassell, supra note 3, at 418-22 (collecting available data); see also Van Kessel, supra note 92, at
127-28 (exhaustively reviewing available studies and concluding that most American studies find suspect
admissions to be below 50% and that "a substantially greater percentage of English than American suspects
subjected to questioning make damaging statements").
n183 Thomas, supra note 12, at 942.
n184 See, e.g., Yale Kamisar, How to Use, Abuse -- and Fight Back with -- Crime Statistics, 25 OKLA. L.
REV. 239, 242, 248-49 (1972) (contending that rising crime rates in England and Wales during late 1960s
disprove connection to the American "revolution in criminal procedure"); Yale Kamisar, On the Tactics of
Police-Prosecution Oriented Critics of the Courts, 49 CORNELL L.Q. 436, 460-62 (1964) (chiding Chicago
police chief's criticism of Warren Court's restriction as contributing to increasing crime rates for failing to
recognize that crime rates were increasing in England and Wales).
n185 See CRAIG M. BRADLEY, THE FAILURE OF THE CRIMINAL PROCEDURE REVOLUTION
96-108 (England and Wales), 112-17 (Canada) (1993). Bradley makes the valuable point that "while the
Supreme Court's cooption of [the criminal procedure] field has essentially eliminated the states as 'laboratories,'
where new ideas can be tested, foreign countries can play the same role." Id. at 95. Indeed, foreign countries
may be the only possibility for real comparisons in the Miranda area. It appears that all jurisdictions in the
United States follow the critical questioning cutoff rules prescribed by Miranda. Accordingly, despite Thomas's
plea for "cross-jurisdictional" comparison of Miranda's effects, Thomas, supra note 5, at 834-35, there is no
possibility of a domestic assessment of the effect of this critical feature of the doctrine. See supra note 109 and
accompanying text (noting centrality of questioning cutoff rules to Miranda's costs). See generally Cassell,
supra note 3, at 498-99 (noting that Miranda's "greatest cost" is that it has blocked experimentation on
alternative ways of regulating interrogation).
n186 Van Kessel, supra note 92, at 8.
n187 See Miranda v. Arizona, 384 U.S. 436, 486-89 (1966) (discussing regulations of interrogation
practices in England, Scotland, India, and Ceylon).

Page 57
43 UCLA L. Rev. 839, *931

n188 New York v. Quarles, 467 U.S. 649, 673 (1984) (O'Connor, J., concurring and dissenting) (footnote
omitted).
n189 Thomas, supra note 12, at 942.
n190 Id. One response is to note that British confession rates fell from over 60% before 1986 to between
40% and 50% when the Miranda-style Police and Criminal Evidence Act was imposed on British police in that
year. See Cassell, supra note 3, at 420-21. British respect for authority cannot explain the drop in confession
rates; only the view that restrictions on police questioning lead to lower confession rates seems plausible. See id.
n191 See supra notes 146-147 and accompanying text.
n192 On that score, in any event, Salt Lake County might be a comparable jurisdiction for comparison. See
supra note 75 and accompanying text (noting that Salt Lake County has lower violent crime rates than the
American average).
n193 See supra Table 5. The overall success rate (including suspects not questioned) would be even lower.
n194 Thomas, supra note 12, at 956.
n195 Id.
n196 Id.
n197 Id.
n198 See supra note 57 (collecting data suggesting that incriminating statements fell after delivery of
warnings).
n199 Cf., e.g., Leiken, supra note 44, at 16-26 (reporting factors bearing on the "decision to talk" based on
interviews with interrogated prisoners; incentive to tell the truth not mentioned); Leo, supra note 49, at 282-84
(discussing why interrogations succeed; incentive to tell the truth not mentioned); Yale Project, supra note 29, at
1570-73 (reporting observational assessments of interrogations; incentive to tell the truth not mentioned).
n200 See, e.g., infra note 343 (questioning of one person produces information about real perpetrator of
crime).
n201 Of 173 suspects overall, 24 suspects gave other information.
n202 Because we gathered our data from screening sessions where the agenda was whether to initiate
criminal charges against a particular suspect, officers likely reported only information that was relevant for this

Page 58
43 UCLA L. Rev. 839, *931

purpose. Thus, results gleaned from interrogation that were useful for other purposes (e.g., prosecutions of other
persons or other crimes) would not have been regularly reported. Moreover, because our data came from such
sessions, they would not capture cases where interrogation provided information leading to other enhanced
charges. For example, if the police were prepared to charge a suspect with a burglary before interrogation and a
murder afterwards, from the perspective of a screening session the case would appear to involve only a murder
charge.
n203 Yale Kamisar, On the "Fruits" of Miranda Violations, Coerced Confessions, and Compelled
Testimony, 93 MICH. L. REV. 929, 1000 (1995); see also id. at 933 n.17, 1000 n.329 (collecting commentary
supporting this view); cf. Akhil R. Amar & Renee B. Lettow, Fifth Amendment First Principles: The
Self-Incrimination Clause, 93 MICH. L. REV. 857 (1995) (discussing fruits doctrine in general).
n204 Two other studies have also measured collateral purposes of interrogation. Neither reported that
uncovering "fruits" was a significant purpose. See Witt, supra note 33, at 326-28; Yale Project, supra note 29, at
1593-97. An attitudinal study found that police, and prosecutors as well, viewed obtaining a confession (or its
corollary -- determining that a suspect was innocent) as far and away the primary purpose for interrogation,
while obtaining other leads was a more remote concern. See Cyril D. Robinson, Police and Prosecutor Practices
and Attitudes Relating to Interrogation as Revealed by Pre- and Post-Miranda Questionnaires: A Construct of
Police Capacity to Comply, 1968 DUKE L.J. 425, 437-438 tbl. 1C, 440 tbl. 1D, 445-46 tbls. 1G & 1H.
n205 Accord Witt, supra note 33, at 327 (finding that identification of accomplices is an important collateral
purpose); Yale Project, supra note 29, at 1593-94 (same).
n206 In one case, a murder weapon was recovered. However, the interrogation that produced the gun was of
a person who was viewed more as a witness than a suspect. Charges against this person were later dismissed
pursuant to a diversion agreement. In the other case, the police wanted to recover a knife in a domestic violence
case, apparently to avoid further incidents. Charges were never filed in the case.
n207 See Stansbury v. California, 114 S. Ct. 1526 (1994) (per curiam).
n208 See supra notes 169-172 and accompanying text.
n209 463 U.S. 1121, 1125 (1983) (per curiam).
n210 Id. at 1122.
n211 Other agencies follow a different approach. For example, one large department trained its field
officers not to give Miranda prematurely because of the possibility of the suspect invoking his rights. As a
result, typically in this department the only officers to give Miranda were detectives, who conducted most
post-arrest custodial interrogations.
n212 In Salt Lake City v. Carner, 664 P.2d 1168, 1171 (Utah 1983), the Utah Supreme Court set out a
four-part test for determining "custody" under Miranda. Then, the United States Supreme Court handed down
California v. Beheler, 463 U.S. 1121, 1125 (1983), and Berkemer v. McCarty, 468 U.S. 420, 441-42 (1984),

Page 59
43 UCLA L. Rev. 839, *931

which established a more-relaxed custody standard, focusing on whether the suspect was in essence under arrest.
Faced with conflicting opinions, in 1993, the Utah Supreme Court held that its earlier decision in Carner
prevailed because it rested in part on the Utah Constitution, art. I, § 12 (state protection against
self-incrimination), and therefore was not controlled by federal interpretations. State v. Wood, 868 P.2d 70, 82
(Utah 1993); see also State v. Mirquet, No. 930098, slip op. at 3-4 (Utah June 30, 1995) (applying Carner). But
the Utah Supreme Court apparently did not recognize that after Carner, it had specifically held that the Utah
self-incrimination provision was no broader than the federal provision. American Fork City v. Crosgrove, 701
P.2d 1069 (Utah 1985). Indeed, the Utah Supreme Court has never even squarely held that Miranda is part of the
state self-incrimination provision and has historically construed the provision as turning on considerations of
voluntariness, rather than compliance with Miranda nuances. See, e.g., State v. Mares, 192 P.2d 861, 870-71
(Utah 1948) (explaining state voluntariness doctrine). See generally Paul G. Cassell, The Mysterious Creation of
Search and Seizure Exclusionary Rules Under State Constitutions: The Utah Example, 1993 UTAH L. REV.
751 passim (explaining difficulties with interpreting Utah constitutional provisions beyond federal standards);
Paul G. Cassell, Search and Seizure and the Utah Constitution: The Irrelevance of the Antipolygamy Raids,
1995 B.Y.U. L. REV. 1 passim (1995) (same).
n213 The available qualitative evidence suggests that noncustodial interrogations have increased since
Miranda. See supra notes 169-172 and accompanying text.
n214 121 of 173 custodial; 52 of 173 noncustodial. The researchers coded a case as "custodial" where they
thought, under prevailing Miranda doctrine, a court would conclude that the suspect was in custody.
n215 21 at the scene; 14 field investigations; 17 arranged interviews.
n216 For clarity of presentation, we have omitted invocations from this table because virtually all of them
would fall into the custodial category. It is interesting, however, that we found two suspects who invoked their
rights in non-custodial interviews.
n217 The chi-square is significant at the .01 level.
n218 See Thomas, supra note 5, at 832-33 (advancing similar hypothesis).
n219 74 of 102 property offenders questioned in custody; 47 of 71 violent offenders questioned in custody.
n220 This interview is described infra note 343.
n221 See, e.g., Tamar Jacoby, Fighting Crime by the Rules: Why Cops Like Miranda, NEWSWEEK, July
18, 1988, at 53; Schulhofer, supra note 45, at 12 n.32 (suggesting that "Miranda, now that police have
assimilated it and learned to use it, may actually generate some net increase in the rate of successful
interrogations").
n222 See, e.g., Mathew Lippman, A Commentary on Inbau and Manak's "Miranda v. Arizona -- Is It Worth
the Costs?," THE PROSECUTOR, Spring 1989, at 35, 37 ("The Court has eviscerated the decision to the point
of evaporation.").

Page 60
43 UCLA L. Rev. 839, *931

n223 New York v. Quarles, 467 U.S. 649 (1984).
n224 See, e.g., Mary M. Keating, Note, New York v. Quarles: The Dissolution of Miranda, 30 VILL. L.
REV. 441 (1985).
n225 Irene M. Rosenberg & Yale L. Rosenberg, A Modest Proposal for the Abolition of Custodial
Confessions, 68 N.C. L. REV. 69, 82 (1989).
n226 The case involved an officer who approached a woman suspected of being an accomplice to an armed
robber. The officer asked for consent to search her purse for a weapon. After obtaining consent, the officer
searched the purse and found stolen checks. This prompted volunteered incriminating statements from the
woman. The case probably would not be viewed as a public safety questioning by the courts, as there was no
"interrogation" to trigger the Miranda requirements. See Rhode Island v. Innis, 446 U.S. 291 (1980) (Miranda
warnings required only for "interrogation").
n227 Pennsylvania v. Muniz, 496 U.S. 582, 600-02 (1990).
n228 Of course, even though the exceptions have a small quantitative impact, they might still be justified on
their own merits. See, e.g., William T. Pizzi, The Privilege Against Self-Incrimination in a Rescue Situation, 76
J. CRIM. L. & CRIMINOLOGY 567 (1985) (vigorously defending public safety exception).
n229 See Cassell, supra note 3, at 460-62.
n230 Miranda v. Arizona, 384 U.S. 436, 478 (1966).
n231 See, e.g., U.S. v. Gonzalez, 875 F.2d 875, 881 (D.C. Cir. 1989) (holding spontaneous statements made
at the moment of arrest admissible even without Miranda warnings).
n232 For discussion of how we assessed the strength of the evidence, see infra note 256 and accompanying
text.
n233 91 of 219. See generally supra Table 1.
n234 114 of 165. See generally infra Table 9.
n235 Of these statements, one would have been classified as a verbal confession, one as an incriminating
statement, and one as an "other statement" unrelated to the crime.
n236 See infra notes 312-313 and accompanying text (describing methodology for assessing statement
importance).

Page 61
43 UCLA L. Rev. 839, *931

n237 See, e.g., supra note 142 and accompanying text (noting the difference this categorization makes for
the total number of "successful" outcomes in the sample).
n238 Compare, e.g., Seeburger & Wettick, supra note 31, at 8 (suggesting that Pittsburgh police complied
with Miranda) with Medalie et al., supra note 32, at 1362-70 (suggesting that D.C. police failed to implement
Miranda).
n239 See, e.g., LAFAVE & ISRAEL, supra note 7, § 6.5, at 484 (suggesting that officers "regularly" advise
suspects of their rights); Stephen J. Schulhofer, Reconsidering Miranda, 54 U. CHI. L. REV. 435, 456 & n.56
(1987) (suggesting that the degree of compliance with Miranda "has been high").
n240 Cf. Dallin H. Oaks, Studying the Exclusionary Rule in Search and Seizure, 37 U. CHI. L. REV. 665,
722 (1970) ("The predominant incentive for interrogation is to obtain evidence for use in court. Consequently,
police conduct in this area is likely to be responsive to judicial rules governing the admissibility of that
evidence.").
n241 Based on the descriptions of the Miranda warnings, the researchers found no cases of "participating"
Miranda. See Peter W. Lewis & Harry E. Allen, "Participating Miranda": An Attempt to Subvert Certain
Constitutional Safeguards, 23 CRIME & DELINQUENCY 75 (1977) (describing delivery of Miranda warnings
interwoven into discussion with defendants). Nor were any such cases alleged in motions to suppress by
defendants.
n242 This is the so-called "fifth warning." It is not strictly required by Miranda, see INBAU, supra note
169, at 221-22, and its use is discouraged by the leading police interrogation manual, see id. at 222.
n243 See COMPTROLLER GEN. OF THE U.S., IMPACT OF THE EXCLUSIONARY RULE ON
FEDERAL CRIMINAL PROSECUTIONS 8 (1979) (finding that 4.4% of federal defendants filed motion to
suppress confessions); FEENEY ET AL., supra note 59, at 144 (finding in two-city study that motions to
suppress confessions were filed in 1.0% of all arrests and that 50.0% of these were granted, for a total of 0.5%
successful motions); Peter F. Nardulli, The Societal Cost of the Exclusionary Rule: An Empirical Assessment,
1983 AM. B. FOUND. RES. J. 585, 595, 597, 598 (finding in nine-county survey that motions to suppress
confessions were filed in 6.6% of all cases, and that 2.3% of these were granted, for a total of 0.16% successful
motions); Peter F. Nardulli, The Societal Costs of the Exclusionary Rule Revisited, 1987 U. ILL. L. REV. 223,
228, 231, 232 (finding in Chicago survey that motions to suppress confessions were filed in 1.1% of all cases
and that 6.1% of these were granted, for a total of 0.04% successful motions); see also MICHAEL ZANDER &
PAUL HENDERSON, ROYAL COMM'N ON CRIMINAL JUSTICE, CROWN COURT STUDY (1993)
(finding that confessions were challenged in about 5% of cases in British sample).
n244 Three motions in 63 filed cases with successful outcomes to questioning. Expressed as a percentage of
total cases filed, suppression motions were filed in 1.7% (3 motions in 173 filed cases). One theoretical
possibility was that defendants pled guilty before they had an opportunity to file a motion to suppress. Although
we were unable to interview defense attorneys about whether this took place in any of our cases, discussions
with prosecutors indicated that such a sequence of events would be extraordinary.
n245 See supra note 9 and accompanying text (discussing difference between Miranda requirements and

Page 62
43 UCLA L. Rev. 839, *931

Fifth Amendment voluntariness requirements).
n246 All three of the motions sought to suppress statements that were obtained at the scene of the crime,
rather than through station house interrogation.
n247 Interestingly, none of the interrogations was recorded. Cf. Cassell, supra note 3, at 488 (suggesting
that recording of interrogations might eliminate "swearing contests" about what happened).
n248 This finding should not be confused with support for the proposition that Miranda has had no ill
effects on law enforcement. See id. at 391-94 (explaining why analysis of the suppression motions is the wrong
way to measure Miranda's overall effects).
n249 See, e.g., George C. Thomas III, An Assault on the Temple of Miranda, 85 J. CRIM. L. &
CRIMINOLOGY 807, 821 (1995) (noting position of Yale Kamisar that real world interrogations involve
"sustained questioning").
n250 Barrett, supra note 14, at 42-43.
n251 VERA INST., TAPING POLICE, supra note 41, at 49.
n252 Leo, supra note 49, at 267.
n253 We also had a very large number of cases in which data was unavailable.
n254 The chi-square statistic is not significant at the .05 level but is significant at the .10 level. Our findings
on this point may shed some light on Professor Thomas's suggestion that police interest in obtaining a
confession is an important factor in whether a confession is obtained. See Thomas, supra note 5, at text
accompanying notes 43-44. The length of interrogation might serve as a surrogate for police interest.
n255 We had great difficulty collecting data on this subject, as many officers were uncertain as to exactly
what time interrogations took place.
n256 See Yale Project, supra note 29, at 1646-47. More extensive data from Britain suggests the same
conclusion. See IRVING & MCKENZIE, supra note 63, at 95; McCONVILLE, supra note 63, at 29; cf.
MICHAEL McCONVILLE & JACQUELINE HODGSON, CUSTODIAL LEGAL ADVICE AND THE
RIGHT TO SILENCE 182 (Royal Comm'n Criminal Justice Research Study No. 16, 1993) (suggesting
problems with collecting data on this subject); Stephen Moston et al., The Effects of Case Characteristics on
Suspect Behaviour During Police Questioning, 32 BRIT. J. CRIMINOLOGY 23, 34-36 (1992).
n257 See UVILLER, supra note 4, at 200-01 (advancing a similar hypothesis); Gisli H. Gudjonsson &
Hannes Petursson, Custodial Interrogation: Why Do Suspects Confess and How Does It Relate to Their Crime,
Attitude and Personality?, 12 PERSONALITY & INDIVIDUAL DIFFERENCES 295, 303 (1991) (studying
Icelandic prisoners and finding that "the most common reasons for the confession related to perceptions of proof

Page 63
43 UCLA L. Rev. 839, *931

and the offender consequently did not see much point in denying the offence").
n258 The chi-square statistic is significant at the .01 level.
n259 Leo, supra note 49, at 265 tbl. 5.
n260 VERA INST., TAPING POLICE, supra note 41, at 53.
n261 An alternative explanation for our finding is that the pool of suspects in the category of "weak"
evidence might include more innocent suspects than other categories. Since innocent suspects are obviously less
likely to confess, this effect would produce a spurious correlation between the strength of the evidence and
confessions. Because of the methodology used in this study, however, the number of innocents in the sample is
probably quite small. See infra note 343 and accompanying text (sample drawn from cases pre-screened and
presented for prosecution).
n262 Thomas, supra note 5, at 823-24.
n263 See Philip B. Heymann, The Defendant's Rights vs. Police Efficiency: An Important Symbol of
Fairness, N.Y. TIMES, Jan. 25, 1987, at E5 ("Every hardened criminal . . . knows his rights and isn't about to
confess.").
n264 See NEUBAUER, supra note 43, at 105 tbl. 2 (finding that only 36% of those with a prior felony
conviction confessed, compared to 59% without); Leiken, supra note 44, at 21 tbl. 4 (finding suspects with nine
or more previous arrests slightly less likely to give a confession); Leo, supra note 49, at 277 (finding suspects
with a felony record four times as likely to invoke Miranda rights); Yale Project, supra note 29, at 1644 tbl. A
(finding interrogation successful for 29% of the suspects with a previous arrest, compared to 60% without).
British data also suggests that suspects with criminal records are less likely to talk. See ROYAL COMM'N
ON CRIMINAL JUSTICE, REPORT 51 (1993) (reporting police found experienced criminals less likely to
answer questions); SOFTLEY, supra note 63, at 69, 75 (observing that suspects with a criminal record were
significantly more likely to exercise right to silence and to request counsel); Moston et al., supra note 256, at 38
tbl. 4 (finding that 21% of suspects with criminal history stayed silent as compared with only 11% without). But
cf. id. at 39 tbl. 7 (finding that for those with legal advice criminal history had a marked effect on the rate of
admission).
n265 At the initial screening, police officers typically had available a National Criminal Information Center
("NCIC") "rap sheet" for each suspect. Not infrequently the rap sheet would list arrests with no indication of
disposition.
n266 The chi-square statistic (pr = .387) is not significant.
n267 See Leo, supra note 49, at 276-77 (finding a small difference between suspects with no record and
those with a misdemeanor record in willingness to invoke Miranda rights but finding a big difference between
those with no record and those with a felony record).

Page 64
43 UCLA L. Rev. 839, *931

n268 See OLP PRE-TRIAL INTERROGATION REPORT, supra note 109, at 105; Cassell, supra note 3, at
486-92; Phillip E. Johnson, A Statutory Replacement for the Miranda Doctrine, 24 AM. CRIM. L. REV. 303,
306, 313 (1987).
n269 See, e.g., BRADLEY, supra note 138, at 85; Yale Kamisar, Foreword: Brewer v. Williams -- A Hard
Look at a Discomfiting Record, 66 GEO. L.J. 209, 236-43 (1977); Schulhofer, supra note 45, at 66-67; George
Dix, Putting Suspects' Confessions on Videotape, MANHATTAN LAW., Apr. 25, 1989, at 12.
n270 See Paul G. Cassell & Stephen J. Markman, Miranda's Hidden Costs, NAT'L REV., Dec. 25, 1995, at
30, 33 (noting law enforcement reluctance to adopt videotaping).
n271 WILLIAM A. GELLER, POLICE VIDEOTAPING OF SUSPECT INTERROGATIONS AND
CONFESSIONS: A PRELIMINARY EXAMINATION OF ISSUES AND PRACTICES -- A REPORT TO
THE NATIONAL INSTITUTE OF JUSTICE (1993).
n272 Id. at 107.
n273 VERA INST., MONITORED INTERROGATIONS, supra note 41, at 53.
n274 See CAROLE F. WILLIS ET AL., THE TAPE-RECORDING OF POLICE INTERVIEWS WITH
SUSPECTS: A SECOND INTERIM REPORT 34-35, 73 (1988) (British study finding that police obtained more
confessions when interviews were taped); Alan Grant, The Audio-Visual Taping of Police Interviews with
Suspects and Accused Persons by Halton Regional Police Force, Ontario, Canada -- An Evaluation Final Report
80 (1987) (unpublished study done for the Law Reform Commission of Canada) (on file with author)
(concluding from empirical study that the "videotaping process does not appear to inhibit suspects from making
confessions"); cf. Cassell, supra note 3, at 491 n.602 (noting problems with statistics in Grant's study).
n275 LUPO PRINS, SPECIAL STUDY GRANT'S SCHEME FOR MEMBERS OF AUSTRALASIAN
POLICE FORCES: SCIENTIFIC AND TECHNICAL AIDS TO POLICE INTERROGATION 145 (1983).
n276 INBAU, supra note 169, at 176-78.
n277 The Utah Supreme Court had previously encouraged recording of custodial interrogations. State v.
Carter, 776 P.2d 886, 891 (Utah 1989). During the period of our study, a case was pending before the Utah
Supreme Court in which the defendant argued that the court should formalize that encouragement and mandate
recording of custodial interrogations. After our data collection was finished, the Utah Supreme Court rejected
that contention. State v. Villarreal, 889 P.2d 419, 427 (Utah 1995).
n278 The hypothesis that recording actually helps interrogation was not proven, as the chisquare statistic (pr
= .211) was not statistically significant.
n279 In our sample, the recorded interrogations involved a higher percentage of violent crimes (particularly
sex crimes) than the overall sample.

Page 65
43 UCLA L. Rev. 839, *931

n280 Police agencies apparently tape statements more frequently in cases involving child victims. This
might be because videotaped statements of children are admissible in certain cases, see UTAH R. CRIM. P. 15.5
(1995) (authorizing admission of visual recording of statement of child victim or witness of sexual or physical
abuse if certain conditions are met), and, as a result, investigators in these kinds of cases have greater access to
videorecording equipment.
n281 See Thomas, supra note 5, at 832-33.
n282 One possible problem with our data should be noted: We did not have a specific box on our survey
form for suspects who objected to recording. See infra Appendix A (survey form). However, the researchers do
not recall any such case in police descriptions of questioning.
n283 There were no allegations in our sample of tape tampering or other recording problems.
n284 See, e.g., Charles J. Ogletree, Are Confessions Really Good for the Soul?: A Proposal to Mirandize
Miranda, 100 HARV. L. REV. 1826, 1843 (1987); Rosenberg & Rosenberg, supra note 225, at 98; cf. JEROME
H. SKOLNICK & JAMES J. FYFE, ABOVE THE LAW: POLICE AND THE EXCESSIVE USE OF FORCE
49, 62-64 (1993) (discussing police perjury more generally); Myron W. Orfield, Jr., Deterrence, Perjury, and
the Heater Factor: An Exclusionary Rule in the Chicago Criminal Courts, 63 U. COLO. L. REV. 75, 95-114
(1992) (discussing police perjury in search and seizure context).
n285 There is one British study, which reported that women were less likely to "exercise silence" when
questioned by police (28.5% for males vs. 9.1% for females), although the small sample size meant the
difference was not statistically significant. McCONVILLE & HODGSON, supra note 256, at 177-78.
n286 See William Hart, The Subtle Art of Persuasion, POLICE MAG., Jan. 1981, at 17.
n287 The chi-square statistic is not significant at the .05 level, but is significant at the .10 level. A possible
explanation for this difference is that women are less likely to commit violent crimes than are men, see 1993
UNIFORM CRIME REPORTS, supra note 71, at 222 (reporting violent crime arrests of males outnumber
arrests of females by a 6 to 1 ratio), and confessions are less likely for crimes of violence, see supra note 143
and accompanying text. To see whether this explained the difference, we compared success rates by gender,
limited to property crimes only. Even controlling for this possibility, the trend remained that women were more
likely to confess. For males suspected of property crimes, 38 were questioned successfully, and 37 were
questioned unsuccessfully. For females suspected of property crimes, 10 were questioned successfully, and five
were questioned unsuccessfully. The chi-square statistic is not significant at .05 level, but is significant at the .10
level.
n288 Yale Project, supra note 29, at 1644-45 (finding no statistically significant relation between race and
interrogation success).
n289 Leiken, supra note 44, at 13; see also GRISSO, supra note 100, at 34, 37 (finding black juvenile
offenders interrogated less often than whites).

Page 66
43 UCLA L. Rev. 839, *931

n290 See McCONVILLE & HODGSON, supra note 256, at 177-78 (no substantial difference in "exercise
of silence" between whites and Afro-Caribbeans); cf. SANDERS ET AL., supra note 132, at 37 tbl. 3.22
(reporting that "take-up" rates on solicitor assistance during questioning was 32.8% for Afro-Caribbeans, 26.7%
for whites, and 18.1% for Asians).
n291 The chi-square statistic (pr. = .921) is not significant.
n292 FEENEY ET AL., supra note 59, at 144 tbl. 15-3.
n293 The chi-square statistic is significant at the .05 level (not including "other" category).
n294 See supra notes 214-217 and accompanying text.
n295 Even with this limitation, there might still be differences in questioning circumstances between patrol
officers and detectives. A patrol officer might more frequently interview a suspect "in custody" on the street than
a detective, for whom custodial interviews are almost always at the station house.
n296 For patrol officers, 11 suspects invoked, 19 were questioned successfully, and 23 were questioned
unsuccessfully. For detectives, 8 suspects invoked, 37 were questioned successfully, and 20 were questioned
unsuccessfully. The chi-square statistic (pr. = .125) is not statistically significant.
n297 One British study found that more than 40% of arrested suspects were questioned more than once.
IRVING, supra note 63, at 103. Another British study reported that about 5% of suspects were interviewed in
custody twice. McCONVILLE & HODGSON, supra note 256, at 174.
n298 16 of 173 suspects. Ten suspects were originally questioned by patrol officers and later questioned by
detectives; five suspects were questioned by detectives in all interviews. Six suspects were in custody for both
interviews; ten suspects were not in custody during the first interview but were in custody during the second;
one suspect was not in custody for both interviews (but even he was given his Miranda warnings during this
second interview).
Only one suspect was interrogated more than twice. The suspect was questioned three times about his
involvement in an auto theft, all by a detective. The first two were separate custodial interviews at the scene of
the crime, while the third was a custodial interrogation at the station. The suspect originally gave incriminating
statements to the police during the first interview, which called his truthfulness into question. But in the
subsequent interviews, the suspect simply gave denials with explanations concerning his involvement. It appears
that the longest interrogation lasted only fifteen minutes.
n299 The one exception was in a domestic violence case in which the victim declined to cooperate with
police.
n300 Two suspects offered flat denials in their first interviews and then gave verbal confessions in their
second interviews. One suspect refused to answer questions in his first, noncustodial interview, but then gave
incriminating statements in his second interview. One suspect who gave incriminating statements in his first

Page 67
43 UCLA L. Rev. 839, *931

interview gave a verbal confession in his second interrogation.
n301 Three of four suspects. See supra notes 293-296 and accompanying text (finding detectives more often
successful in questioning than patrol officers).
n302 In three cases the suspect gave a verbal confession at each interview: One suspect gave incriminating
statements at both interviews; two suspects gave denials with explanations at both interviews; and a final suspect
gave a flat denial at both interviews.
n303 In one case, the suspect obtained an attorney between the first and second interviews. In the other
cases, the suspects simply invoked their Miranda rights. In two of these cases, the suspect had not been in
custody during the first interrogation, and hence had not received Miranda warnings.
n304 470 U.S. 298 (1985).
n305 Ogletree, supra note 284, at 1840; see Rosenberg & Rosenberg, supra note 225, at 95 (arguing that
Elstad gives "officers a blueprint for obtaining inculpatory statements that will withstand judicial scrutiny
despite circumvention of the spirit if not the letter of Miranda").
n306 Miranda v. Arizona, 384 U.S. 436, 481 (1966).
n307 Compare INBAU, supra note 169, at xiv ("Many criminal cases . . . are capable of solution only by
means of an admission or confession from the guilty individual or . . . [from] the questioning of other criminal
suspects.") with James Fyfe, The Rule Is Rarely an Issue in Criminal Cases, NEWS PRESS (Fort Myers, Fla.),
Feb. 8, 1987, at A23 ("Confessions rarely have anything to do with convictions because all the evidence needed
to convict is usually in long before anybody confesses.").
n308 See Cassell, supra note 3, at 422-37 (collecting available research).
n309 See Schulhofer, supra note 45, at 541-44 (arguing that the necessity rate is at most around 19%). But
see Cassell, supra note 45 (responding to Schulhofer).
n310 See 2 FEENEY & WEIR, supra note 59, at 42 (finding that police sergeants thought confessions were
essential more often than did academic researchers); Yale Project, supra note 29, at 1591-92 (finding that
detectives thought confessions necessary more often than did Yale editors).
n311 See Cassell, supra note 3, at 468-70.
n312 See infra Appendix A (survey forms).
n313 See Yale Project, supra note 29, at 1582-83. We used the Yale terminology rather than our own to
avoid suggestions that we biased our outcome by asking slanted questions. The Yale terminology was also

Page 68
43 UCLA L. Rev. 839, *931

adopted in Witt, supra note 33, at 324.
n314 We have less data on this subject than on others because the researchers had to ask prosecutors a
question in the course of screening sessions rather than passively record information that was already being
provided. Sometimes a convenient opportunity for this question was unavailable.
n315 See, e.g., BROSI, supra note 66, at 12 (reporting findings on prosecutorial screenings).
n316 The Bay Area study found that successfully interrogated suspects were more likely to be charged than
those who were unsuccessfully interrogated. Leo, supra note 49, at 294 tbl. 16. The study also found that
suspects who waived or invoked their Miranda rights were equally likely to be charged. Id. at 278 tbl. 10. An
older study by L.A. District Attorney Evelle Younger found that complaints were issued in 65.3% of cases
where the defendant had given a statement but in only 49.9% of cases in which no such statement was available.
See Controlling Crime Hearings, supra note 34, at 347 (figures derived from worksheets in the complaint-stage
table); cf. Cassell, supra note 3, at 426 (noting the definitional problems in Younger's study). A British study
found that "suspects who confess are far more likely to be charged than suspects who deny the offence or remain
silent." SANDERS ET AL., supra note 133, at 148.
n317 Comparing those successfully questioned to all others produced a definite trend that just missed
significance at the .05 level, but was significant at the .10 level (pr = .054). Comparing those successfully
questioned to those unsuccessfully questioned was chi-square significant at the .05 level.
n318 See, e.g., Schulhofer, supra note 239, at 456.
n319 See, e.g., Stephen J. Markman, The Fifth Amendment and Custodial Questioning: A Response to
"Reconsidering Miranda," 54 U. CHI. L. REV. 938, 946-47 n.20 (1987).
n320 See PETER F. NARDULLI ET AL., THE TENOR OF JUSTICE: CRIMINAL COURTS AND THE
GUILTY PLEA PROCESS 236, 254 tbl. 8.13 (1988) (finding that some effect of confessions on plea
bargaining); NEUBAUER, supra note 43, at 109-10 (finding that defendants who confessed received fewer
concessions in plea bargaining); Yale Project, supra note 29, at 1608 (finding defendants who confessed were
less successful in bargaining for lower charges); see also Leo, supra note 49, at 293 (finding confessing
defendants less likely to have their cases dismissed and more likely to have their cases resolved by plea
bargaining).
n321 This is not the same, of course, as definitively proving that confessions caused this effect. It is possible
that other factors that correlated with confessions were responsible.
n322 See generally UTAH CODE ANN. §§ 77-2a-1 to -4 (1995) (setting forth the standards for pleas in
abeyance). It is also possible for a defendant to be "diverted" after charging through a formal diversion
agreement. See generally id. §§ 77-2-2(2), 77-2-5 (1995). There were no such cases in our sample.
n323 Id. § 76-3-103(1) (1995). Utah also has a fourth category for capital felonies. Id. There were no capital
cases in our sample. Misdemeanors are likewise divided into three classes: Class A, Class B, and Class C. Id. §

Page 69
43 UCLA L. Rev. 839, *931

76-3-104 (1994).
n324 Id. § 76-3-203 (1995) (generally speaking, first degree felonies are punishable by five years to life;
second degree by one to fifteen years; and third degree felonies by probation to five years). Utah also has a few
mandatory minimum sentences, primarily associated with sex and firearms offenses.
n325 For this reason, we did not collect data on whether confessions affected sentences ultimately served by
defendants. To do so would have unduly prolonged our study, as complete data would not be available until
parole dates were set for all defendants. Nor did we collect data on sentencing recommendations, an area that is
sometimes the subject of plea bargaining.
n326 The office had a policy prohibiting plea bargaining unless there was some reason justifying it. In
practice, the exception appeared to be at least as important as the rule.
n327 See UTAH CODE ANN. § 76-6-202 (1995) (defining burglary offenses).
n328 See, e.g., Hans Zeisel, The Disposition of Felony Arrests, 1981 AM. B. FOUND. RES. J. 407, 437-40.
n329 Comparing those who were successfully questioned to all others combined produced a chi-square
result significant at the .01 level; comparing to those unsuccessfully questioned produced a chi-square result that
just missed significance at the .05 level (pr = .055). Because a case can be made for doing the statistical
significance test either way, we have reported both numbers here and report both such tests for all other analyses
of Tables 16 and 17, infra.
n330 Comparing those who were successfully questioned to all others combined produced a chi-square
result significant at the .05 level; comparing to those unsuccessfully questioned produced a chi-square result
significant at the .01 level.
n331 One more suspect who was questioned unsuccessfully went to trial. During the trial, he pled to
reduced charges (at the -1 level).
n332 Compare Leo, supra note 49, at 394-95 & n.4 (arguing that the proper measure of Miranda's effects is
to compare ultimate outcomes for those who invoked their Miranda rights with those who waived their rights)
with Cassell, supra note 3, at 432 n.267 (arguing that such an approach is unlikely to capture all of Miranda's
effects).
n333 Our "invoked" category was also unusual in that it appeared to take much longer to resolve these
cases. The only theory we could think of was that those who invoked might be "hard cases" who would fight
criminal charges more aggressively all the way through the system.
n334 The chi-square result is not significant (pr = .189).
n335 Difficulties in collecting prior record data are discussed supra note 265 (noting difficulty of

Page 70
43 UCLA L. Rev. 839, *931

determining convictions). Of twenty-one suspects who invoked, sixteen had criminal records. We had no further
information on four of these suspects. Of the remaining twelve, one had five or more convictions, two had at
least two convictions, three had ten or more arrests, three had at least two arrests, and three had one arrest (at
least two of whom were ultimately convicted). These kinds of records seemed to be more serious than the
average criminal record in our sample.
n336 See, e.g., NEUBAUER, supra note 43, at 218.
n337 See supra note 264 and accompanying text (noting that studies find suspects with prior records more
likely to invoke).
n338 For example, the Bay Area study found no difference in punishment between those who invoked their
rights and those who waived them. Leo, supra note 49, at 281. But the same study found that those who invoked
were four times more likely to have felony records. Id. at 277. Without controlling for this effect, it is difficult to
reach conclusions about the unique contribution of invocations to sentencing outcomes.
n339 The chi-square is significant at the .01 level.
n340 The conviction difference is important because it might plausibly serve as a measure of the need for
confessions to convict. See Cassell, supra note 3, at 430.
n341 The chi-square is significant at .01 level.
n342 The only exception is the New Haven study, which suggested that two of ninety suspects were
believed to be innocent by the study's authors. Yale Project, supra note 29, at 1586.
n343 The fact that even this one case ended up in our sample was a quirk. Originally a boyfriend was a
suspect in an arson. When police questioned him over the phone, he gave them information about his old
girlfriend (and even secretly recorded a conversation in which she admitted the crime -- confirming that "heaven
has no rage like love to hatred turned"). When the case was screened, the researcher recorded the results of
questioning for both the boyfriend and the girlfriend.
n344 See supra note 144 and accompanying text (discussing police pre-screening).
n345 See generally ALAN M. DERSHOWITZ, THE BEST DEFENSE xxi (1982) (proposing "Rule I" of
the "justice game" as "almost all criminal defendants are, in fact, guilty"); Cassell, supra note 3, at 480-81
(discussing small percentage of convictions of innocent persons). Lest this position be misconstrued, we should
make clear our belief that suspects are entitled to a presumption of innocence in a court of law and that cases of
innocents ensnared in the criminal justice system are great tragedies. We also acknowledge that a few innocent
persons are in fact convicted, a few more are charged, and even more suspected by police. Our claim here is only
that factually innocent defendants constitute a tiny percentage of cases that pass through the courts in Salt Lake
County.

Page 71
43 UCLA L. Rev. 839, *931

n346 See, e.g., supra notes 263-267 and accompanying text (finding prior criminal record had no bearing on
questioning success).
n347 Such generalizations have a venerable tradition in the confessions literature. See, e.g., Yale Project,
supra note 29, at 1533 (generalizing from 100 cases in New Haven to the entire country).
n348 See, e.g., Moran v. Burbine, 475 U.S. 412, 427 (1986) (rejecting rule requiring police to inform
suspect when lawyer attempts contact because the "minimal benefit . . . would come at a substantial cost to
society's legitimate and substantial interest in securing admissions of guilt"); Oregon v. Elstad, 470 U.S. 298,
312 (1985) (rejecting rule against continued questioning after confession obtained in violation of Miranda
because "this immunity comes at a high cost to legitimate law enforcement activity, while adding little desirable
protection to the individual's interest in not being compelled to testify against himself.").
n349 Cf. GRANO, supra note 9, at 202 (arguing that Miranda is vulnerable apart from pragmatic
considerations).
n350 See supra notes 142-181 and accompanying text (reporting frequency of successful outcomes to
interrogations and offering comparison to pre-Miranda statistics). Of course, the proposition that confession
rates are lower than before Miranda is different than the proposition that confession rates are low. One could
believe that a 33.3% success rate is "high" and accept the proposition that it is lower than pre-Miranda rates.
n351 See supra notes 115-119 and accompanying text (reporting findings on invocations).
n352 See supra notes 214-217 and accompanying text (reporting findings on noncustodial questioning).
n353 See supra notes 306-345 and accompanying text (reporting findings on confessions and ultimate
outcomes).
n354 See supra notes 45-56 and accompanying text (noting that before-and-after studies, declining crime
clearance rates, and police perceptions all suggest that Miranda reduced confessions).
n355 See Thomas, supra note 5, at 821-23.
n356 The "equality" rationale for Miranda has been, in our view, effectively refuted by Professor Joseph
Grano. See GRANO, supra note 9, at 32-38. Reviewing Grano's book, Professor Thomas concedes that if
empirical evidence demonstrates a substantial loss of confessions from Miranda, the costs to society "could
outweigh the equality objective." Thomas, supra note 249, at 826.
n357 Oregon v. Elstad, 470 U.S. 298, 306 (1985).
n358 New York v. Quarles, 467 U.S. 649, 684 n.7 (1984) (Marshall, J., dissenting).

Page 72
43 UCLA L. Rev. 839, *931

n359 Thomas, supra note 249, at 819, 821; see also id. at 826-27.
n360 Id. at 821 & n.21 (noting view of Professor Yale Kamisar).
n361 See supra Table 7. This data involves both custodial and noncustodial questioning. Other limitations
to the data are discussed at supra notes 253-255 and accompanying text.
n362 Attributing the absence of unconstitutional questioning to Miranda requires proof that such
questioning often took place before Miranda and that Miranda helped eliminate it. The empirical evidence on
both points, however, is quite weak. See Cassell, supra note 3, at 473-78 (discussing whether Miranda resulted
in the disappearance of coercion).
n363 See supra note 243 and accompanying text (finding only one case of 173 in which Miranda was not
followed).
n364 See supra notes 209-210 and accompanying text (describing use of Beheler warnings).
n365 See supra notes 211-212 and accompanying text (describing the giving of Miranda warnings by
officers not required to do so).
n366 See supra notes 304-305 and accompanying text (discussing compliance with Miranda in multiple
interview situation).
n367 See supra notes 222-237 and accompanying text (discussing infrequent use of these exceptions to
Miranda).
n368 See supra note 268 and accompanying text.
n369 Thomas, supra note 12, at 935.
n370 Moran v. Burbine, 475 U.S. 412, 433 n.4 (1986).
n371 Yale Kamisar, The "Police Practice" Phases of the Criminal Process and the Three Phases of the
Burger Court, in THE BURGER YEARS 143, 150 (Herman Schwartz ed., 1987).
n372 Cf. Thomas Y. Davies, A Hard Look at What We Know (and Still Need to Learn) About the "Costs" of
the Exclusionary Rule: The NIJ Study and Other Studies of "Lost" Arrests, 1983 AM. B. FOUND. RES. J. 611,
626 (noting that "there is no empirical content to this 'balancing' approach to considering the costs and benefits"
of the search and seizure exclusionary rule); James B. White, Forgotten Points in the "Exclusionary Rule"
Debate, 81 MICH. L. REV. 1273, 1282 (1983) (suggesting that in the absence of real empirical evidence, an
assessment of the costs and benefits of the exclusionary rule "must rest not upon those grounds, but upon prior
dispositions or unarticulated intuitions that are never justified") (footnote omitted).

Page 73
43 UCLA L. Rev. 839, *931

n373 See, e.g., Minnick v. Mississippi, 498 U.S. 146, 151 (1990) (asserting that the benefit of "specificity"
of the Edwards prophylactic rule "has been thought to outweigh the burdens that the decision in Miranda
imposes on law enforcement agencies and the courts by requiring the suppression of trustworthy and highly
probative evidence even though the confession might be voluntary under traditional Fifth Amendment analysis")
(quoting Fare v. Michael C., 442 U.S. 707, 718 (1979)).
n374 See supra notes 13-28 (discussing dearth of empirical information on police interrogation at the time
of Miranda).
n375 Miranda v. Arizona, 384 U.S. 436, 524 (1966) (Harlan, J., dissenting). Justice Harlan noted that
"studies are also being conducted by . . . the Georgetown Law Center, and by others equipped to do practical
research." Id. at 523 (footnote omitted).
n376 Professor Schulhofer's article, written after this Article was drafted, is the most notable effort in this
regard. Schulhofer, supra note 45. Yet his analysis is quite problematic. See generally Cassell, supra note 45.
n377 See supra note 185. Professor Thomas's limited research agenda illustrates the point. See Thomas,
supra note 5, at 833-37 (outlining research agenda). He is unable to recommend real comparative projects
(comparing, for instance, what happens when police do and do not request waivers of rights), but instead must
suggest projects that would at best provide inferential information on Miranda (such as, whether judicial
warnings at initial appearances make a difference in confessions).
n378 Leo, supra note 49.
n379 Id. at 268 tbl. 7.
n380 Id. at 269 n.5 (comparing to post-Miranda confession rates but noting this must be done "with some
caution").
n381 See, e.g., Schulhofer, supra note 45 (arguing that Leo's 64% confession rate is "much higher than most
of the estimates available in the pre-Miranda period").
n382 See supra note 176.
n383 Leo, supra note 49, at 255 (study consists of 182 interrogations); see also id. at 455-69 (describing
process of waiting at the station house for detectives to begin interrogations).
n384 See supra notes 91-104 and accompanying text.
n385 FEENEY ET AL., supra note 59, at 143 tbl. 15-2 (finding 18.5% of arrested burglary suspects in
Jacksonville, Florida, and 20.1% in San Diego, California not interrogated).

Page 74
43 UCLA L. Rev. 839, *931

n386 Leo reports 117 incriminating statements from 182 suspects interrogated. Leo, supra note 49, at 268
tbl. 7. Assuming an additional 48 suspects never questioned would approximate a 21% noninterrogation rate and
produce a 50.9% (117/(182 + 48)) incriminating statement rate.
n387 See supra notes 159-167.
n388 See supra note 170 (citing article co-authored by Leo on this point); see also supra notes 169-172 and
accompanying text (collecting evidence that police are shifting to noncustodial interrogations).
n389 See supra notes 216-217 and accompanying text.
n390 Leo reported 182 interrogations. See supra note 386. Applying a 30.1% noncustodial interrogation rate
would mean that, for comparison purposes, 55 interrogations should be regarded as not in custody.
n391 One measure of the extent of effectiveness is to divide our success rate for noncustodial interrogations
by the success rate for custodial interrogation. For this purpose, an invocation is an "unsuccessful" outcome. To
determine an effectiveness rate, one can add to our data supra Table 6 (custodial vs. noncustodial questioning
success rate), data on invocations by custodial status, see supra Table 3; supra note 216. This produces the result
that custodial interrogations were successful 58 times and unsuccessful 63 times (44 denials + 19 invocations), a
success rate of 47.9%. Noncustodial interrogations were successful 15 times and unsuccessful 37 times (35
denials + 2 invocations), a success rate of 28.8%. This means that noncustodial interrogation is about 60.1%
(28.8%/47.9%) as effective as noncustodial interrogation.
n392 Reducing Leo's 64.2% success rate to adjust for the lower effectiveness for custodial interrogations,
see supra note 391, results in a 38.6% success rate (64.2% x 60.1%) for noncustodial interrogations. In the 55
assumed noncustodial interrogations, see supra note 390, 21 incriminating statements would be obtained (55 x
38.6%), while in the remaining 127 custodial interrogations (182 total interrogations -- 55 noncustodial
interrogations), a total of 82 incriminating statements would be obtained (127 x 64.2%). Dividing the 103 total
incriminating statements (82 + 21) by the 230 suspects assumed to have been questioned and not questioned, see
supra note 9, results in a 44.8% (103/230) incriminating statement rate.
n393 See Controlling Crime Hearings, supra note 34, at 1120 (New York County study assessed "almost all
felony cases"); id. at 223, 224 (Kings County study involves all suspects in various categories of crimes;
providing illustration of questioning by "police officer" after vehicle stop); Brief of National District Attorneys
Association, supra note 22, at 6a-7a (collecting data from eight cities based on confessions present in
"prosecutions" without regard for detective/police officer distinctions); Barrett, supra note 14, at 41 (reporting
information for interrogation in City B, California by "all interrogation, including . . . arresting officer");
Medalie et al., supra note 32, at 1362 (study assessed response of the D.C. "police department"); cf. Controlling
Crime Hearings, supra, at 200 (discussing statistics for Philadelphia "police" and "Police Department" but
noting that statistics compiled by "Detective Division of the Philadelphia Police Department"); Witt, supra note
33, at 324, 325 (study based on files from "Seaside City Police Department" but referring to procedures followed
by Seaside City "detectives"). But see Barrett, supra note 14, at 41 (reporting interrogation in City A, California
by "members of the detective division"); Seeburger & Wettick, supra note 31, at 7 (study includes only case
"investigated by the Detective Branch"); Souris, supra note 20, at 264 (Detroit data based on "prosecutions
handled by the Specialized Bureaus of the Criminal Investigation Division"); Yale Project, supra note 29, at
1527 (observations made by students "in the detective division"). Our count of these citations is that data from

Page 75
43 UCLA L. Rev. 839, *931

eleven of the cities involved interrogations by all officers, four interrogations by detectives only, and data from
two could not be classified.
n394 See supra note 293 and accompanying text.
n395 See id.
n396 See FEENEY ET AL., supra note 59, at 144 tbl. 15-3 (detectives obtained a higher percentage of
confessions than patrol officers in Jacksonville and San Diego).
n397 Patrol officer success in custodial interviews was 35.8%. See supra note 296 (19/53 successful).
Detective success in custodial interviews was 56.9%. See id. (37/65 successful). This difference just missed
statistical significance at the 90% confidence level. See id. Because we have already adjusted Leo's figures to
reflect lower effectiveness of noncustodial interviews, some may argue that any further adjustment for lower
effectiveness of patrol officer interviews is "double counting." We nonetheless believe that the adjustment is
appropriate for several reasons: (1) The lack of statistical significance may well stem from the small sample size
available when restricted to custodial interviews by patrol officers. See supra note 147 (noting difficulty with
small sample size). Indeed, when a 20% difference in success rates (35.8% vs. 56.9%) comes back as
statistically insignificant, it is clear that the sample size permits only the most substantial variations to be found
to be significant. (2) Common sense suggests that experienced detectives would be more successful in obtaining
confessions. Most of the detectives in Leo's sample had been police officers for ten to twenty years. Leo, supra
note 49, at 259. It seems reasonable to assume that these officers would be better interrogators, either from
experience or training, than the general run of patrol officers. Indeed, Leo discusses interrogation training
seminars that "certify police detectives as professionals who command exclusive control over a specialized body
of knowledge." Id. at 108. (3) In our study, patrol officers frequently conducted "custodial" interviews "on the
street" after arrest, probably a less productive setting for questioning than the station house, where custodial
interviews by detectives were frequently conducted. (4) Professor Thomas has challenged us to reconcile our
study with Leo's. The adjustment for detective effectiveness is one way of doing so. However, as a partial
response to double counting concerns, we have performed this adjustment only for custodial interrogations,
leaving in place the assumption that detectives and patrol officers are equally effective in noncustodial settings.
n398 See supra note 296 (custodial questioning done by detectives 65 times; patrol officers 53 times; a ratio
of 55.1% of questioning done by detectives).
n399 See FEENEY ET AL., supra note 59, at 144 tbl. 15-3 (more than half of interrogations conducted by
patrol officers in Jacksonville and San Diego).
n400 One measure of reduced effectiveness is to divide the success rate of patrol officer questioning by the
success rate of detective questioning. Using figures only for custodial questioning, see supra note 397, the ratio
of patrol officer success to detective success is 62.9% (35.8%/56.9%).
n401 Taking the 82 incriminating statements in custodial interrogations out of the 230 suspects overall, as
calculated earlier, see supra note 391, and multiplying by the 55.1% ratio of detective interviews to patrol
interviews, see supra note 398, produces the result that 45 would have been obtained by detectives and 37 by
patrol officers. Adjusting for the fact that the patrol officers would have obtained only 62.9% as many

Page 76
43 UCLA L. Rev. 839, *931

incriminating statements in the 37 successful interrogations, see supra note 400, means that the 37 incriminating
statements obtained by patrol officers should be reduced to 23 incriminating statements (37 x 62.9%). The total
success rate is, therefore, 38.7% ((45 + 23 custodial successes + 21 noncustodial successes)/230 suspects
overall).
n402 Leo, supra note 49, at 263.
n403 Internet message from Richard A. Leo to Paul G. Cassell, Nov. 8, 1995 (on file with Cassell).
n404 See supra notes 244-247 and accompanying text.
n405 Leo, supra note 49, at 268 n.4 (emphasis added).
n406 Compounding this problem is Leo's failure to have a "denial with explanation" category, like the
category we used in our study. See supra notes 139-140 (explaining importance of this category). As a result, he
probably classified some such denials as incriminating statements.
n407 Leo, supra note 49, at 273.
n408 See supra note 317 and accompanying text (only 12.5% of suspects successfully interrogated not
charged).
n409 Compare supra notes 144-145 and accompanying text (reporting a 27.2% "confession" rate) and
FEENEY ET AL., supra note 59, at 142 tbl. 15-1 (reporting a 32.9% confession rate in Jacksonville and 20.3%
in San Diego) with Leo, supra note 49, at 268 tbl. 7 (24.2% of suspects gave confessions, 17.6% gave partial
admissions, and 22.5% gave "incriminating statements").
n410 Almost a third of Leo's sample (60/182) consisted of videotapes of "interrogations performed" by two
Bay Area police departments in cases that were no longer pending. Leo, supra note 49, at 452, 474 n.10. Neither
department had a policy of necessarily storing videotapes. Id. at 474 n.10. It seems quite likely, therefore, that
the videos on hand for academic analysis included only interrogations that "got off the ground," not
interrogations where suspects promptly invoked their rights or were otherwise generally uncooperative. This
would artificially increase the percentage of incriminating statements Leo found in his sample.