Page 1

LEXSEE 90 NORTHWESTERN U.L REV. 1084
Copyright (c) 1996 Northwestern University Law Review
Northwestern University Law Review
90 Nw. U.L. Rev. 1084
LENGTH: 19089 words
REPLY: ALL BENEFITS, NO COSTS: THE GRAND ILLUSION OF MIRANDA'S DEFENDERS
NAME: Paul G. Cassell *
BIO:
* Associate Professor of Law, University of Utah College of Law. I wish to especially thank the editors of
the Review for arranging this debate and Professor Stephen Schulhofer for cheerfully and vigorously
participating. Joe Grano, Richard Leo, Michael McConnell, Wayne McCormack, and Lee Teitelbaum provided
helpful suggestions on an earlier draft. This Article was supported by the University of Utah College of Law
Research Fund. Because of production deadlines, Professor Schulhofer did not have an opportunity to review all
of this rejoinder before his article was sent to the printer. He has graciously agreed to give me, at least on this
occasion, the last word.
SUMMARY:
... After I wrote my Article on the social costs of Miranda and possible alternatives to it, I sent a copy to Professor
Stephen Schulhofer for "a few comments," if he had time. ... Turning next to the specific studies, Professor Schulhofer
contends that my estimate of the post-Miranda confession rate is too high. While Professor Schulhofer makes a number
of "site specific" adjustments to my data on Miranda-induced changes in confession rates, only three are quantitatively
important: eliminating data from New York showing significant declines in the confession rate, adding an estimate for
Los Angeles showinga significant increase in the rate, and shrinking the Philadelphia numbers by using a different
baseline for comparison. ... Professor Schulhofer ignores contrary data from Seaside City for 1964 to 1966, where
confession rates were not falling but rising, and from Detroit, where confession rates were falling only about 0.7% per
year from 1961 to 1965, about one-third of his estimate. ... It also falls within the 1% to 5% estimate of Miranda's cost
from the ABA "Special Committee" survey, upon which Schulhofer himself relies. ... As Professor Grano has written,
"by giving the suspect power to prevent questioning even before it begins, ... Miranda gave the law of confessions a
"single focus - protection of the suspect.' ...
TEXT:
[*1084]
After I wrote my Article on the social costs of Miranda and possible alternatives to it, n1 I sent a copy to
Professor Stephen Schulhofer for "a few comments," if he had time. I had little right to hope for something on the order
of the 64-page response he produced. n2 I am very appreciative that, as one of the leading defenders of Miranda, he
has undertaken not simply to comment on my position but to respond fully and thoughtfully to it.

Page 2
90 Nw. U.L. Rev. 1084, *1084

For anyone uninitiated to the Miranda literature, it must come as something of a surprise to read his conclusion that
the Supreme Court's most famous and controversial criminal law decision has "substantial benefits and vanishingly
small costs." How can something that makes so little difference be so important? In the real world, alas, things are more
complicated. While Professor Schulhofer and the conventional academic wisdom have it that Miranda is virtually
costless, the facts suggest otherwise. And once we see through this grand illusion, the hard tradeoffs of Miranda require
reasoned justification, which, due to the "no costs" illusion, has yet to be provided. Indeed, considering a reasonable
alternative, the costs of Miranda cannot be justified.
Typically a rejoinder moves immediately to points of difference. But because our points of agreement are of some
importance, it is valuable to highlight them first. Professor Schulhofer agrees that "the size of the [Miranda cost] does
matter" and that, as a means of ap- [*1085] praising that cost, the enterprise of reviewing the studies done
before-and-after Miranda "deserves to be taken seriously." n3 With respect to the specific studies, he admits that most
of the reliable studies performed in the wake of Miranda showed sizable decreases in the number of confessions. n4
This must be regarded as a significant acknowledgment, since we have heretofore been told that these studies showed
no such thing. n5 Recognizing that Miranda creates some costs, Professor Schulhofer further agrees that Miranda can
be replaced by an alternative, as long as it serves "equally well to dispel compelling interrogation pressures, prevent
brutality, and give clear guidance to the police." n6 He also concedes that the centerpiece of my proposed alternative
(the videotaping of interrogations) would be "an extremely valuable tool - for both the police and the suspect." n7
Where, then, are the salient points of disagreement? While I conclude that Miranda causes the loss of roughly 3.8%
of all criminal cases each year and about an equal number of more lenient plea bargains, Professor Schulhofer conducts
a "reanalysis" and concludes that Miranda's costs total only 1.1% when compared to an interrogation regime without
warnings or 0.78% when compared to a regime with warnings. n8 He calls his recalculated cost "for all practical
purposes ... zero." n9 Yet this small percentage can, as the Supreme Court has explained, "mask a large number of
felons" who are released. n10 For the record, in 1993 alone, using Schulhofer's own 1.1% estimate, Miranda's "zero"
effect cost society criminal cases against 8100 violent criminals and 22,000 property offenders for crimes covered by
the FBI's crime index and 160,000 cases for crimes outside the crime index. n11 To put the so-called "zero" cost into
perspective, it is within the estimated range of the lost cases from the search and seizure exclusionary rule, n12 long
considered a "social cost" of some importance. n13 [*1086]
In interpreting his figures, Professor Schulhofer goes even further and aggressively argues that they represent the
maximum extent of Miranda's aftermath. n14 He concludes firmly that his calculation "substantially overstates
Miranda's current effect" and that the real cost today "must be placed far lower." n15 Of course, one who undertakes to
derive an upper bound estimate of the decision's effects must follow a methodology suited to that task. In particular,
when faced with competing, reasonable alternative interpretations of data, one must consistently adopt the interpretation
suggesting the greatest costs. n16 Yet at a number of critical junctures in his article, Schulhofer indulges every
presumption against harm from Miranda. An illustration of his one-sided methodology comes from his discussion of
social science statistics. Schulhofer notes the obvious point that research like the Miranda studies "generates "soft'
estimates with a wide range of error." n17 From this uncontroverted premise, Schulhofer draws the conclusion that
cost calculations resting on such estimates "can create a deceptive illusion" of large costs from Miranda. n18 But
Schulhofer does not acknowledge that, in view of the range of error, they are equally likely to create a "deceptive
illusion" of low or nonexistent costs. Throughout his article, Schulhofer adopts this kind of partisan stance. Far from
calculating Miranda's greatest possible harm, he instead simply reads the data resourcefully to put Miranda in the best
possible light and to shrink Miranda's costs to a level he considers acceptable, as I demonstrate in Part I of this
rejoinder.
In Part II, I take up Professor Schulhofer's challenge to the desirability of my proposal to modify Miranda by
replacing its waiver and questioning cut-off rules with a mandatory videotaping regime. On this issue, his analysis
inflates the benefits of Miranda while overlooking the real virtues of videotaping. [*1087]

Page 3
90 Nw. U.L. Rev. 1084, *1087

I. Miranda's Costs

A. Events Since Miranda

In an effort to deflect the import of the before-and-after studies even before he begins analyzing them, Professor
Schulhofer contends that the cost of Miranda today "is almost certainly far lower than the costs incurred in the
immediate post-Miranda period." n19 Schulhofer argues that when the studies were conducted (typically just a few
months after the decision) police were hampered by lack of opportunity to adjust their interviewing procedures to
Miranda's requirements. I discussed the same conjecture in my initial article, but noted the competing possibility that
police might not have implemented all of the harmful features of the Miranda regime in just a few months. n20
Professor Schulhofer dismisses my argument, contending that I offer "no evidence - empirical, impressionistic, or
otherwise - to support [my] conjecture that confession rates in the selected high-compliance cities might have declined
after the initial post-Miranda period." n21 That police accommodation might exacerbate Miranda's harms is,
Schulhofer argues, "an imaginative but hollow speculation" because the confession rate data comes from jurisdictions
"in which compliance with Miranda was reasonably rapid and complete ...." n22 On close examination, however, only
"imaginative but hollow speculation" supports Schulhofer's claim. The most important evidence on police compliance
comes from the observational study in New Haven, which found that police failed to follow many of the important parts
of the Miranda regime at the time of the study. n23 The failure to follow the waiver requirement is particularly
significant because the empirical evidence suggests this is the single most damaging feature of Miranda. n24 Thus, the
New Haven study quite clearly does not involve "rapid and complete" implementation of Miranda. n25 The other stud[*1088] ies I relied upon were not observational studies and accordingly contain little (if any) specific discussion about
whether police were in fact following all the Miranda rules. n26 Nor do they provide any data about whether courts
suppressed confessions at trial because of Miranda violations. One must therefore speculate to conclude that police
procedures rapidly changed to fall into line with Miranda. Strongly arguing against such conjecture is the fact that the
available studies with detailed information about police interrogation immediately after Miranda all report significant
deviations, not "complete" implementation. n27 If anything, then, the enterprise of reviewing the [*1089]
before-and-after studies will produce a cost estimate that is too low, not too high.
How else might we assess quantitatively Professor Schulhofer's claim that because of police accommodation over
time Miranda's damage is "far lower" today than in the immediate post-Miranda period? In an article published in 1987,
Professor Schulhofer suggested that we might look to data on the crime "clearance" rate (the rate at which police solve
crimes) as a measure of Miranda's effects. In fact, Schulhofer said in that 1987 article that, while some of the
before-and-after studies suggested declining confession rates after Miranda, within "a year or two" clearance "rates were
thought to be returning to pre-Miranda levels." n28
We need not speculate about people's thoughts on whether clearance rates returned to pre-Miranda levels. FBI data
from 1950 to date are available in the Uniform Crime Reports. n29 Figure 1 depicts the data for the clearance rate for
violent crimes, the area where Miranda would have its greatest impact. n30 As can be seen, the violent crime clearance
rate fell sharply following Miranda and, far from stabilizing in 1966 or even 1967, continued to plummet in 1968 and to
fall slightly in 1969. n31 More striking, however, is the long-term perspective on Miranda's costs visible from this
graph. Coinciding exactly with the decision, violent crime clearance rates fell from about 60% before Miranda to about
45% just a few years later. They have hovered [*1090] around 45% ever since. In other words, looking just to the
clearance rate data - as Professor Schulhofer himself suggested in 1987 - one would conclude that about one out of
every four violent crimes that was "cleared" before Miranda was not cleared after! n32 Of course, more research on
other possibly competing causal factors is warranted. n33 But, at the very least, we have strong evidence suggesting
two conclu- [*1091] sions: first, the before-and-after studies likely understate Miranda's costs because they do not
capture adverse effects from 1967 through 1969; n34 second, over the long haul, law enforcement never recovered

Page 4
90 Nw. U.L. Rev. 1084, *1091

from the blow inflicted by Miranda.
[SEE FIGURE 1 IN ORIGINAL]
An alternate way of assessing Professor Schulhofer's claim that confession rates have risen since the
before-and-after studies is to compare recent data on confession rates with pre-Miranda rates. The relevant data suggest,
consistent with the clearance rate figures, that confession rates today are lower than before Miranda. The most recent
data come from a 1994 study of Salt Lake County by Bret Hayman and me, which found that suspects confessed or
gave incriminating statements in 37.4% of all cases presented by police for prosecution. n35 Similarly, a 1993 study
by Professor Richard Leo found an incriminating statement rate of about 64% in custodial interrogation by detectives in
California. n36 To be compared to other Miranda studies, Leo's sample needs to be adjusted to reflect suspects who are
[*1092] never questioned, suspects questioned in less productive noncustodial settings, and suspects questioned not
only by detectives but also by less effective patrol officers. n37 Such adjustments produce a confession rate of
somewhere below 38.7%, n38 quite close to what we found. Other data from the 1970s suggest confession rates in the
range of 20.3% to 51.3%. n39 Generally speaking, these rates are well below the average rate reported before Miranda,
which is roughly around 55% to 60%. n40 Thus, as with the clearance rate data, the confession rate data suggest that
police effectiveness in obtaining statements has not improved since the before-and-after studies were conducted.
B. "Site Specific" Adjustments

Turning next to the specific studies, Professor Schulhofer contends that my estimate of the post-Miranda
confession rate is too high. While Professor Schulhofer makes a number of "site specific" adjustments to my data on
Miranda-induced changes in confession rates, only three are quantitatively important: eliminating data from New York
showing significant declines in the confession rate, adding an estimate for Los Angeles showing a significant increase in
the rate, and shrinking the Philadelphia numbers by using a different baseline for comparison. n41 While one could
question other aspects of Schulhofer's reanalysis, n42 I will now turn to New York and Los Angeles, as the [*1093]
Philadelphia revision does not require extended analysis. n43
1. New York. - In my article, I include data from a study by District Attorney Frank Hogan of New York County,
New York, who reported that confessions used in presentations to the grand jury fell after Miranda from 49% to 14.5% a 34.5% fall. Criticizing my "literal reading of [Hogan's] testimony," n44 Professor Schulhofer argues that Hogan must
have meant something different than what he said and that his post-Miranda confession rate is too small because of the
retroactive effect of Miranda. We have each set out our positions at some length for the reader to evaluate. n45 What is
worth adding in this rejoinder is exploration of the following question: Even assuming that Professor Schulhofer is right
that Hogan's 14.5% post-Miranda confession rate is "artificially deflated," n46 is it deflated to any signifi- [*1094]
cant degree?
Professor Schulhofer appears disinterested in this subject, concluding only that "the Hogan study must be
excluded." n47 But we know more about what interrogation looked like in the immediate post-Miranda period in New
York City than we know about any other city. We have not only Hogan's study, but two separate Vera Institute studies
showing low confession rates in sample sizes of hundreds of suspects, and District Attorney Aaron Koota's study in
adjoining Brooklyn, also revealing few confessions. n48 Considered together the data show that Hogan's rate could not
have been deflated by much.
Turning first to the 1967 Vera Study showing low post-Miranda confession rates, Professor Schulhofer discounts
the data because a significant fraction of the sample reflected misdemeanants who were less likely to confess. n49 The
obvious solution to this concern is to take the misdemeanants out. The Vera Study involved two samples, one drawn
from Manhattan and a second drawn from the Twentieth Precinct. An offense-specific table of confession rates is
reported for both samples. It is a simple matter to construct a set of the more serious offenses that would be directly
comparable to Hogan's offenses. For example, selecting the offenses of felony assault (including assault with a

Page 5
90 Nw. U.L. Rev. 1084, *1094

dangerous weapon), burglary, forcible rape, assault and robbery, robbery, grand larceny (including auto theft), felony
drugs, and arson produces the result for the Manhattan sample that 3.1% of suspects confessed and 13% gave
admissions. n50 This is a success rate of 16.1%, just a stone's throw away from Hogan's "artificially deflated" 14.5%.
The same offenses for the smaller Twentieth Precinct sample produce an admission rate of 24.8%, n51 somewhat
above Hogan's but still well below Hogan's pre-Miranda rate of 49%. Substi- [*1095] tuting the Manhattan data for
Hogan's leaves an "undeflated" post-Miranda confession-rate drop of 32.9%; substituting the Twentieth Precinct data
leaves a drop of 24.2%. n52
Further suggestion of a substantial decrease in the post-Miranda New York confession rate comes from Brooklyn.
As Professor Schulhofer notes, New York City police patrol both sides of the East River in Brooklyn and Manhattan.
n53 District Attorney Aaron Koota reported that the post-Miranda confession rate in Brooklyn was 29.5%. n54
Assuming that officers in neighboring Manhattan were no more successful, that would be a substantial 19.5% below the
pre-Miranda confession rate reported for Manhattan. n55
Although he does not discuss how the Brooklyn data help validate Hogan's conclusion, Professor Schulhofer
wishes to exclude the Brooklyn study as well. The Brooklyn statement rate fell from an estimated 90% before Miranda
to 59% after. Applying a 50% ratio of incriminating statements to total statements, n56 the confession rate dropped
from 45% to 29.5%, a substantial drop of 15.5%. n57 In an elaborate set of speculations, Professor Schulhofer notes
that my assumed 50% ratio of incriminating statements to total statements might have increased over time, rendering
my total number of incriminating statements too low. n58 But, of course, the converse is possible as well: the total
number of incriminating statements to total statements might have dropped over time, rendering my total number of
incriminating statements too high. n59 Schulhofer does not appear to even consider this second possibility, asserting
only that "Cassell's own sources indicate that the proportion of statements that were incriminating rose after Miranda
because suspects who made purely exculpatory statements before Miranda chose to remain silent ...." n60 On
examination of his cross-references, Cassell's own sources shrink to but one source: [*1096] the Hogan testimony on
the experience of the Homicide Bureau after Miranda. n61 Hogan has no concrete figures on this particular point; n62
the only hard data available (from Seaside City across a broad range of offenses) found no statistically significant
difference in the ratio of incriminating statements to total statements before and after Miranda. n63
Professor Schulhofer also questions Koota's reliance on an estimate for the pre-Miranda statement rate, noting
correctly that this means the before-and-after data sets were collected in different ways. n64 While this can be a
problem if some specific reason suggests that the two sets measured different things, n65 it is not a reason for
automatically rejecting such evidence, particularly given Professor Schulhofer's point that we "must include all the
usable observations" when we have so few to analyze. n66 In any event, recalling the comparability of the Brooklyn
and Manhattan police forces validates the Brooklyn pre-Miranda estimate of a 45% confession rate. Professor
Schulhofer does not dispute Hogan's 49% pre-Miranda rate, which is quite close.
What is striking about all these before-and-after calculations - Hogan before to Hogan after, Hogan before to Vera
after, Hogan [*1097] before to Koota after, Koota before to Hogan after - is that they all produce the same conclusion:
confession rates fell dramatically in New York after Miranda. n67 This is exactly what one would expect in the
nation's largest city given Professor Schulhofer's view that Miranda harmed police in big cities the most. n68 Professor
Schulhofer's bottom-line conclusion (that on average the confession rate fell only 6% after Miranda) would be changed
significantly if double-digit drops in the confession rate from New York City were included. But his arguments fail to
establish that the New York data should be exluded, much less that there are "decisive reasons" n69 for so doing.
2. Los Angeles. - The other quantitatively important "reanalysis" of my confession-rate figures comes from
Professor Schulhofer's decision to include data from Los Angeles, at least in his calculation of Miranda's costs
compared to the costs of a modified warnings regime. (Schulhofer concedes that for a calculation of Miranda's costs
compared to the costs of the voluntariness regime, the L.A. data "should be discarded" because of a California precursor
to Miranda, that required police to give warnings even before Miranda. n70 ) He aggregates two separate surveys and
then concludes that, within thirty-four days of Miranda, n71 confession rates there rose 10%. Earlier Schulhofer

Page 6
90 Nw. U.L. Rev. 1084, *1098

[*1098] called New York's reported confession-rate decline an "exaggerated outlier" that should be excluded. n72
What, then, are we to make of his decision to include the only reported rise in confession rates (a substantial rise at
that), particularly when he apparently agrees that Miranda's negative effects would be more pronounced in the nation's
major urban areas and suggests that a sharp rise in confessions in such a short time is unlikely? n73 We must regard
the Los Angeles surveys with great suspicion, even before examining their underlying methodology.
The underlying methodology renders the study unusable. As I pointed out in my first article, the pre-Miranda
survey measured only "confessions and admissions" while the separate, post-Miranda survey measured the broader
category of "confession, admission or other statement." n74 Professor Schulhofer apparently agrees that this is what
the original documentation of the study says; but - once again - we are not to read it too literally. Schulhofer speculates
that not only did the post-Miranda data cover "confessions, admissions, or other statements," n75 but so did the
pre-Miranda data. The reason he gives is remarkably weak. Schulhofer writes that the description of the pre-Miranda
category "was not taken from the [actual] questionnaire but only from the "summary sheets' used by the law clerk."
n76 Speculation that something different might have been on the questionnaires remains just that - speculation. The
description from the "summary sheets" is the only direct description available.
Moreover, compelling reasons indicate that the words mean what they say. Schulhofer glides over the fact that the
Los Angeles "study" is in fact two distinct surveys from discrete periods that must be cobbled together to produce a
before-and-after figure. The first Los Angeles survey was designed to measure "the effect of the Dorado decision upon
the District Attorney's operation." n77 The report of the study in January 1966 gives no indication of any thought of a
later follow-up survey. n78 This distinguishes the Los Angeles data from, for [*1099] example, the data from
Pittsburgh and Philadelphia, which involved continuous sampling of cases under the same conditions, which were then
divided into pre- and post-Miranda groupings. n79
That two distinct surveys were done at two different times suggests that the questionnaires may have asked
different questions. I argued in my initial article that the Office "apparently redesigned the study questionnaire." n80
Schulhofer responds that this is a purely speculative claim and that it is more plausible to assume the questionnaires
were not materially different. n81 How then does one explain that the pre-Miranda survey produced data on issues not
covered in the post-Miranda survey and vice versa? n82 The form must have been redesigned.
And, in fact, it was.
Pursuing Professor Schulhofer's claim, I discovered definitive proof of the redesign in a document released in 1966
by the Los Angeles District Attorney's Office and available at the University of Utah College of Law Library. n83 This
document contains a copy of one of the two actual questionnaires used: the post-Miranda questionnaire. n84 This
questionnaire does not ask the questions that would have been necessary to collect data on issues covered in the
pre-Miranda survey. n85 It simply could not have been used in the pre-Miranda survey. Moreover, the questionnaire
asks questions on specific, Miranda-related issues, such as: "Was [a] "Miranda type' admonition given to defendant?"
n86 A questionnaire with such a question could not have been written before the decision was handed down. Different
questionnaires asking different questions about different subjects destroy any effort to piece together meaningful
before-and-after data from [*1100] Los Angeles. n87
Professor Schulhofer's only remaining argument is that, even if the two questionnaires asked questions "worded
somewhat differently," n88 the category of "other statements" would not have collected information on, well, other
statements. He reasons that, because the data were drawn from police requests for complaints, only incriminating
statements such as confessions and admissions would have been included. But in all likelihood, the materials attached to
the requests for complaints would have been raw police reports about the investigation of the case, which a prosecutor
could review to determine whether to file charges. The notion that certain types of statements "would not be attached to
requests for issuance of a complaint" n89 suggests a punctiliousness that is at odds with the realities of police work in
a large city.

Page 7
90 Nw. U.L. Rev. 1084, *1100

There are still further reasons for excluding the Los Angeles data. n90 In particular, if we simply take the surveys
as producing data on the subjects they described, the pre-Miranda "confession[-]admission" rate was 40% and the
post-Miranda total statement rate ("confessions, admissions, and other statements") was 50%. Applying the previously
discussed one-half ratio of incriminating statements to all statements, n91 the post-Miranda confession-admission rate
would be 25% (one-half of 50%) - a 15% drop from pre-Miranda levels. This literal, sensible reading of the Los
Angeles study transforms it from a bizarre outlier suggesting a striking increase in confession rates imme- [*1101]
diately after Miranda to a study showing a confession rate decline of almost exactly what was found in other major
cities.
In sum, while Professor Schulhofer gamely maintains that "the Los Angeles study is one of the least vulnerable of
the eleven surveyed here," n92 it is plainly among the most vulnerable. His decision to include the purported increase
in the Los Angeles confession rate skews his results and is inappropriate, particularly for one claiming to calculate an
upper-bound estimate of Miranda's costs.
C. Time Trends

Professor Schulhofer next contends that determining Miranda's costs from the before-and-after studies overlooks
"trends and other causes" that, coincidentally, were operating to depress the confession rate in exactly the years when
the studies were conducted. n93 The "other causes" include such things as the Supreme Court's 1961 decision in Mapp
v. Ohio n94 governing search and seizure rules and the 1963 decision in Gideon v. Wainwright n95 regarding the
right to counsel. Surprisingly, these decisions indirectly continue to create new adverse effects on police questioning as
late as 1967 and 1968 (thereby infecting the studies), although Schulhofer argues simultaneously that Miranda's direct
impact on police interrogation dissipated "quickly." n96 Even more surprisingly, the indirect effects of these decisions
from the early 1960s apparently take precedence over the direct effects of Escobedo v. Illinois, n97 the Supreme
Court's 1964 decision squarely covering police interrogation. Of course, it would be unfair to call Escobedo's adverse
effect on confessions an alternate cause, since (as a precursor to Miranda) it is properly regarded as a harbinger of
Miranda's costs. If the trend resulted from Escobedo depressing the confession rate, studies begun after that decision
would use a pre-Miranda baseline too low to fully capture Miranda's costs.
Professor Schulhofer acknowledges this problem, explaining that "Miranda's requirements were anticipated in
some jurisdictions before the decision was announced." n98 As a corrective, he promises, at least "where the evidence
permits," to make a comparison of interrogation under the Miranda rules with interrogation under "traditional
voluntariness standards." n99 Attempting to fulfill this promise, Professor Schulhofer reports a total cost figure for
Miranda measured [*1102] against what he labels "regimes without warnings." n100 Contrary to the billing,
however, not even a single one of the five studies on which he relies for this calculation clearly involved police
questioning without warnings n101 - as would be required for a true comparison to the voluntariness regime.
Schulhofer's approach becomes even more confusing when we examine how the five studies were selected. He agrees
that, for purposes of comparing Miranda to the voluntariness test, the Los Angeles "data should be discarded" because
of California's anticipatory implementation of Miranda. n102 Inconsistently, however, Schulhofer then includes data
from Seaside City (an immediately adjoining Los Angeles suburb with overlapping police jurisdiction n103 ) in
determining these costs. n104 Why data from one jurisdiction should be included while data from an adjacent
jurisdiction is excluded remains unclear. n105 Also remaining unclear is how this cost figure can be touted as
representing "the high-water mark of Miranda's negative impact." n106 Because part of the Miranda regime was in
place when the studies started, the resulting before-and-after cost figure plainly fails to capture all of Miranda's adverse
effects, as one of Schulhofer's footnotes seems to admit. n107 [*1103]
As a fallback argument, Professor Schulhofer suggests that we should not really care about a true, pre-Escobedo
voluntariness baseline "for policy purposes" because warnings have often been a prominent feature of alternatives to
Miranda, n108 including my proposed alternative. n109 But a number of vigorous supporters urge a return to
voluntariness standards. n110 They are unlikely to be satisfied by Schulhofer's position that a modified warning and

Page 8
90 Nw. U.L. Rev. 1084, *1103

videotape regime would eliminate only some of Miranda's problems. "Don't stop halfway - throw out the whole thing"
is likely to be their ready response. Moreover, it is noteworthy for "policy" purposes that the policy of the people of the
United States - embodied in a federal statute adopted by Congress - is the pre-Miranda voluntariness standard. n111
While the statute has largely been ignored by federal prosecutors, interest in it will likely grow in light of Justice
Scalia's recent opinion highlighting its provisions. n112
In any event, even accepting Professor Schulhofer's claims that we can ignore Escobedo and the need for a true
voluntariness baseline and that we should look for "trends" depressing the confession rate [*1104] for reasons other
than restrictions on police, the only data supporting Professor Schulhofer's negative time trend come from the New
Haven study. The study reported "very cautiously" that there appeared to have been a 10 to 15% decline in the number
of incriminating statements from 1960 to the summer of 1966. n113 From this, Schulhofer extrapolates beyond the
data and simply assumes that a steady 2% decline per year continued thereafter. However, the study's authors noted that
their sample was biased in three respects, n114 and that "each of these factors might explain the apparent decrease."
n115 Moreover, the New Haven data are not the only time trend data available on confession rates. n116 Professor
Schulhofer ignores contrary data from Seaside City for 1964 to 1966, where confession rates were not falling but rising,
n117 and from Detroit, where confession rates were falling only about 0.7% per year from 1961 to 1965, about
one-third of his estimate. n118 If we use all three studies, n119 the average time trend is up, not down, n120 and the
proper adjustment (especially for someone calculating a maximum cost to Miranda) is to increase my cost figure.
D. Weighted Average Adjustments

Professor Schulhofer argues next that my cost estimate rests on an "overrepresentation of large urban areas" among
the before-and-after studies. n121 While acknowledging that "the data are not strong enough to permit confident
conclusions," n122 he nonetheless claims that Miranda's overall cost figure should be reduced by about a quarter
through construction of an average weighted to give more impact [*1105] to the studies from less populated areas.
n123 In assessing the validity of this discount, it is important to "look behind the curtain" and examine how that
weighting takes place. Schulhofer groups most of the studies showing substantial declines in the confession rate
(Pittsburgh, Philadelphia, along with Kansas City and Los Angeles) into the "large city" group and assigns them a
weight of about one-third of the total. This leaves just two cities (New Haven and "Seaside City") serving as
representatives for two-thirds of the United States. Such a weighting is hard to justify, particularly in view of the fact
that the Seaside City data must be regarded with suspicion. n124
A more fundamental problem is, however, that Professor Schulhofer offers no principled basis for determining
when an adjustment should be made to the overall Miranda cost figure. As pointed out in my initial article, my
methodology (adopted by Schulhofer) makes six generalizing assumptions. n125 Schulhofer has chosen to refine one
of those six generalizations in light of additional available empirical evidence. But what about the other five
generalizations? Information is available that could be used to refine at least three of them in ways that would increase
Miranda's costs. Substantial data suggest that Miranda has more serious effects in violent crime categories, that
confessions are more often necessary for violent crimes, and that the need for confessions has increase in recent years.
n126 Professor Schulhofer does not question any of this empirical evidence. A consistent effort to adjust the raw
Miranda cost figure must take the bitter with the sweet. Adjustments in these areas would likely more than offset the
decrease associated with Schulhofer's weighting for big city effects. n127 In any event, Schulhofer's "pick-and-choose"
approach to [*1106] produce a downward adjustment of the raw figures is arbitrary and inappropriate, particularly
since he promises to calculate the maximum extent of Miranda's harms.
E. Police Concern About Miranda

In an attempt to bolster his reading of the Miranda studies, Professor Schulhofer asserts that law enforcement
professionals in general and police officers in particular "pervasively share[]" the view that Miranda poses no

Page 9
90 Nw. U.L. Rev. 1084, *1106

significant problems. n128 This bold claim is said to be supported by, among other things, a "widespread and
consistent" body of police "testimony." n129 He made identical claims in his 1987 article. n130 Given the sweeping
nature of this assertion, it is surprising to discover that with one exception (discussed below n131 ) Professor
Schulhofer cites nothing but anecdotal evidence in support of his assessment of law enforcement views on Miranda,
failing to acknowledge that such evidence is unsurprisingly conflicting. n132 On examination some of his citations are
downright curious as proof of "consistent" satisfaction. For instance, Professor Schulhofer's 1987 article cites Los
Angeles District Attorney Evelle Younger's approval of Miranda in 1967; n133 but Younger's view was plainly
atypical, as he was criticized by many other prosecutors for "breaking ranks" on the Miranda issue. n134 In 1976,
then-California Attorney General Younger joined twenty-one other Attorneys General to urge the Supreme Court to
"abandon" Miranda. n135 In light of what Professor Schulhofer [*1107] calls "pervasive" satisfaction with Miranda,
it is also curious to read his report that "attempts to evade the Miranda requirements or "bend' them to the limit remain
commonplace in some police precincts." n136 Why police should work so often to evade something that causes so few
problems is not explained. n137
Any serious attempt to establish a "widespread and consistent" belief in Miranda's benign effects must rely not on
such data as hit-or-miss newspaper interviews, but rather systematic surveys of police opinions. Such surveys as exist
generally support, if anything, the opposite conclusion - that Miranda reduced police success in obtaining confessions.
Cyril Robinson conducted a nationwide survey in 1966 of police and prosecutors on the effects of Escobedo and
Miranda. Most police and prosecutors who responded thought that the percentage of suspects who refused to make a
statement had increased and that the percentage of suspects confessing had decreased after police were required by
Escobedo to warn suspects of their rights. n138
In New Haven, the Yale students interviewed the detectives involved in most of the interrogations they observed
during the summer of 1966 and an additional twenty-five detectives six months later. n139 The students reported that
"the detectives unanimously believe [Miranda] will unjustifiably [help the suspect]." n140 They also reported that "the
detectives continually told us that the decision would hurt their clearance rate and they would therefore look
inefficient." n141
Otis Stephens and his colleagues surveyed law enforcement officers in Knoxville, Tennessee and Macon, Georgia
in 1969 and 1970. n142 Virtually all of the officers surveyed believed that Supreme Court decisions had adversely
affected their work and most attributed [*1108] this negative influence first and foremost to Miranda. n143
Law student Gary L. Wolfstone sent letters in 1970 to police chiefs and prosecutors in each state and the District of
Columbia. Most agreed that Miranda raised obstacles to law enforcement. n144
In "Seaside City," James Witt interviewed forty-three police detectives sometime before 1973. n145 Witt reported
that the detectives "were in almost complete agreement over the effect that the Miranda warnings were having on the
outputs of formal interrogation. Most believed that they were getting many fewer confessions, admissions and
statements." n146
Finally, an unpublished 1987 telephone survey of the membership of the Police Executive Research Forum found
that their members favored "reconsideration of Miranda and some modification." n147
All of these surveys support the conclusion that police officers believed that Miranda reduced the confession rate.
n148 To support the notion of "widespread and consistent" accommodation to Miranda, Professor Schulhofer cites but
one survey - by a "Special Committee" [*1109] of the American Bar Association. n149 While it concluded that
Miranda is "not considered troublesome" by the police, the 1988 telephone survey on which this conclusion is based is
equivocal because of the phrasing of its questions. n150 Properly read, however, the data clearly refute the notion that
there is "widespread" agreement that police have become accommodated to Miranda. n151
Even taking the conclusions at face value, it is difficult to know what weight to give to assessments made by police

Page 10
90 Nw. U.L. Rev. 1084, *1109

in 1988 that Miranda had no adverse effects. How would they know what difference Miranda makes? "Virtually all of
today's police interrogators have known no law other than Miranda." n152 Moreover, it must be remembered that this
survey was conducted against the backdrop of the conventional wisdom, which was that Miranda had no adverse
effects. It seems likely that police officers in 1988, with no direct pre-Miranda experience to draw upon, might well
have concluded that things then were about the same as things had always been. Finally, police might like Miranda for
other reasons, even if it reduced [*1110] confessions. n153
One last point should be made about the ABA survey. It is a peculiar citation in support of the argument that my
3.8% estimate of Miranda's costs is too high. In fact, the prosecutors in the survey estimated that Miranda's costs fell in
the 1% to 5% range, corresponding with my 3.8% estimate, but apparently not with Schulhofer's estimate of 0.78%.
n154 Commenting on the survey's finding, Professor Craig Bradley observed that, even though 1% to 5% is a small
percentage, it "could be described as demonstrating, contrary to the ABA's own characterization of it, disturbingly high
numbers of cases lost due to evidentiary exclusion." n155
In sum, Professor Schulhofer's claim that police "consistently" conclude that Miranda produced no adverse effects
is untenable.
F. Miranda's Adverse Effect on Plea Bargaining
In my article, I reviewed the considerable empirical evidence supporting the position that lost confessions would not
only produce the outright "loss" of some criminal cases, but also, in addition, would weaken bargaining positions for
prosecutors and, hence, lead to more favorable plea bargains for defendants. n156 I calculated that these costs would
roughly equal the number of convictions that were lost outright. This point about the need for confessions to convict did
not seem particularly controversial to me. Indeed, writing in response to my claim in another journal that prosecutors
regarded confessions as [*1111] necessary for conviction in perhaps 61% of all cases, n157 Professor George
Thomas generally conceded the point and suggested that "I doubt that Miranda supporters would contest the proposition
that confessions are important to prosecutions." n158
Professor Schulhofer, however, does contest my position, arguing that the dynamics of plea bargaining "tend to
offset the numerically estimated loss of convictions." n159 Professor Schulhofer ultimately concludes that, of the cases
in which confessions disappeared because of Miranda, the "best measure of necessity rate" is 19%, n160 somewhat
lower than my already conservative estimate off 24%. n161 He maintains, however, that even his lower necessity
figure "significantly overstates" the true cost. n162 Schulhofer's argument is that ameliorative mechanisms in the
criminal justice system cause some of the cases that might otherwise be "lost" due to the missing confession to in fact be
"won" (or at least salvaged to some extent) through plea bargaining to lesser charges. n163 I raised the same caveat in
my original article. n164 As I explained there, however, it is important to even-handedly recognize not only that lost
cases might be won but that also that won cases might be lost. Schulhofer's lengthy effort to reduce Miranda's impact
fails to even acknowledge this competing possibility. n165 This point as- [*1112] sumes some importance when we
recognize that, on the assumption that only 19% of confessions are "necessary" for conviction, there are many more
"won" cases to lose than "lost" cases to win. Schulhofer's claim requires us to believe that in none of the 81% of the
cases in which a confession was regarded as "unnecessary" for conviction would the defendant obtain a dismissal from
the weaker bargaining position of the prosecution. This is in some tension with the common understanding of plea
bargaining dynamics, in which - as Schulhofer himself has noted - negotiations "depend on the likelihood of conviction
that is governed primarily by the admissible evidence available to the prosecution and defense." n166 The available
statistics also reveal that dismissals are more likely in cases without confession n167 and that generally defendants
obtained better deals when prosecutors were operating without confessions. n168
If we look at the specific studies that Professor Schulhofer relies upon, it becomes quite clear that dismissals
among some of the "won" cases would occur. Two of the three studies Schulhofer uses for a necessity estimate relied on
an "evidence-investigation" scale for evaluating the need for confessions. n169 On that scale, the study's authors

Page 11
90 Nw. U.L. Rev. 1084, *1112

regarded a confession as "unnecessary" when, for example, police had available against a defendant only "some" other
evidence - or even "no" other evidence - but the study's authors believed that other investigation by the police was
"available and probably adequate." n170 [*1113] Of course, in the real world of harried and underfunded criminal
investigations, n171 investigative avenues that are only "probably adequate" will not always deliver the goods against
a defendant. It therefore seems virtually indisputable that at least some of what Schulhofer is counting as
bird-in-the-hand "won" cases would be lost when the police entered the bushes with only "probably adequate" leads to
follow. To establish a net "offset" to the number of lost confessions, Schulhofer must demonstrate why these won cases
that end up lost would not roughly equal, or even exceed, the number of lost cases that he claims can be salvaged
through plea bargaining, a demonstration he does not even attempt.
As an additional argument against tallying plea bargaining costs, Professor Schulhofer maintains that, even if a
defendant obtains a better plea bargain because of the lack of a confession, "the justice system has other ameliorative
mechanisms to deploy ...." n172 For example, Schulhofer suggests that judges might impose an equally severe
sentence despite the more favorable plea bargain. n173 I outlined the same theory previously, citing Professor
Schulhofer's distinguished article in this area. n174 At the same time, however, I explained that the evidence on this
point was conflicting and cited two extensive quantitative studies that concluded that charge reductions affected
sentencing disposition. n175 Professor Schulhofer does not respond to this conflicting empirical data. Professor
Schulhofer also speculates that suspects who fail to confess might actually receive higher sentences because of their
failure to act contritely. n176 Here again, I noted the same possibility initially, n177 but reported that four specific
empirical studies had reviewed this issue and had found no evidence in support of it. n178 Schulhofer does not move
beyond academic theorizing and discuss this hard data. Professor Schulhofer's theories are skillfully argued. But in view
of his failure to grapple with the contrary empirical studies, his claim that the plea bargaining and other practices offset
Miranda's costs must be viewed skeptically. [*1114]
G. The Big Picture

Both Professor Schulhofer and I have argued spiritedly, yet respectfully, for our particular interpretations of
individual studies and bits of data. But in concluding this section, it might be useful to step back from assessing the
particular blows landed in this hand-to-hand (or at least word processor-to-word processor) combat and ask the
overarching question: Apart from who has the better of the argument on individual points, which overall Miranda cost
estimate makes the most sense when viewed in perspective?
In making that assessment, it is first useful to consider which theory corresponds with what common sense
suggests about Miranda's harms. My reading of the data is simply that after Miranda placed significant new restrictions
on police questioning, the efficacy of that questioning significantly declined - the expected conclusion. My reading also
corresponds with systematic surveys of police officers with direct experience in questioning suspects both with and
without the Miranda restrictions, who reported that the Miranda rules posed significant problems from them. n179
Professor Schulhofer's reading that Miranda had very little effect is at odds with both what we would expect logically
and what police in fact reported after the decision.
With respect to the specific before-and-after studies, my reading puts them all into a neat pattern, with specific
differences in results mostly explained by the size of the city. n180 Professor Schulhofer's reading scatters them into
no discernable pattern with confession rates decreasing substantially in some cities, e.g., Pittsburgh; remaining the same
elsewhere, e.g., Seaside City; and even rising substantially in one, Los Angeles. My conclusion that American
confession rates fell 16% after Miranda compares quite favorably with data from Britain and Canada, where confession
rates without the Miranda rules appear to have been at least 16% higher than in this country. n181 Professor
Schulhofer's reading requires some (as-of-yet unoffered) explanation for the difference. My reading also corresponds
with the 10% to 15% fall in British confession rates when a regime similar to Miranda's was imposed. n182 Again,
Schulhofer's requires us to come up with an explanation.

Page 12
90 Nw. U.L. Rev. 1084, *1114

My reading of the data is also consistent with such cross-checks as other data sources allow. In particular, my
estimate that police lose about 3.8% of all cases from Miranda fits comfortably with the clearance rate data, which
suggest that violent crime clearance rates de- [*1115] clined about 4% in the first year after the new rule. n183 It also
falls within the 1% to 5% estimate of Miranda's cost from the ABA "Special Committee" survey, n184 upon which
Schulhofer himself relies.
Entirely apart from the trees, the forest's picture is quite clear: the best reading of the available empirical evidence
is that Miranda has depressed the confession rate substantially and has thus imposed significant social costs.
II. Alternatives to Miranda

A. Miranda's "Prophylactic" Status

To mitigate some of the costs of Miranda as we approach the twenty-first century, I urged that we use videotaping
to regulate police interrogations, retain most warnings of rights, and modify the features of Miranda that the empirical
evidence identifies as most harmful. n185 Professor Schulhofer criticizes this proposal as a "straightforward political
horsetrade" and, seizing the constitutional high ground, replies that, because Miranda proceeds from constitutional
principles, its "defenders (and the courts) are not free to "deal' in this fashion." n186 But "horsetrading" is just a
colorful and pejorative term for balancing competing concerns, which is the way virtually everyone defends Miranda.
As Professor Yale Kamisar has written, striking a balance "is the way Miranda's defenders, not its critics, have talked
about the case for the past twenty years." n187 The Supreme Court now describes the doctrine in precisely those terms
- as "a carefully crafted balance designed to fully protect both the defendant's and society's interests." n188
Professor Schulhofer attempts to fortify his constitutional position with the contention that police interrogation will
almost invariably "violate the Fifth Amendment bar on the use of compelling pressure, at least in the absence of
safeguards sufficient to dispel that pressure." n189 He has advanced a quite similar argument previously, n190 and it
is unsatisfying to find that his position is so static. n191 Particu- [*1116] larly lacking is discussion of recent work by
Professor Joseph Grano, a forceful doctrinal critic of Miranda, who has published a brilliant book devoted in large
measure to refuting Schulhofer's earlier position. n192 Although Grano's book was referenced as the basis for my
abbreviated doctrinal discussion, n193 Schulhofer repeats without refinement arguments that Grano has, at least in my
view, n194 refuted. For example, Professor Schulhofer criticizes my reading of Miranda as a subconstitutional,
prophylactic decision by arguing: "A reading of judicial dicta that has the Court simultaneously criticizing Miranda for
its lack of judicial restraint, and then repeatedly exercising an unprecedented power to overturn state proceedings that
fully complied with the Constitution, is surely paradoxical," n195 to say the least. To say the least, Professor Grano
makes exactly the same point a central part of his critique of the doctrine, calling it Miranda's "legitimacy" problem
n196 and noting that "the Court has never sought to justify or explain its assumption of such judicial power." n197
Likewise misplaced is Schulhofer's more general suggestion that, under my reading of current Fifth Amendment
doctrine, "large bodies [*1117] of noninterrogation case law become uncomprehensible." n198 But under the
standard reading - indeed, under any reading at all - it is clear that much of the doctrine is inconsistent. As Professor
Akhil Amar and Renee Lettow outline in their powerful new article entitled "Fifth Amendment First Principles," almost
every word in the self-incriminating clause "sits atop considerable confusion or perversion because courts do not yet
understand how the words fit together, or what big idea(s) might underlie the clause." n199 Both Grano specifically
and Amar and Lettow at a general level have offered more effective replies to Schulhofer's doctrinal position than I
could hope to craft here, and I will simply rely on them.
In any event, after his brief attempt to elevate Miranda to constitutional status, it turns out that Professor
Schulhofer is willing to "horsetrade" on the Miranda rules. He agrees that Miranda can be constitutionally replaced with

Page 13
90 Nw. U.L. Rev. 1084, *1117

an alternative, so long as the replacement serves equally well to dispel compelling interrogation pressures, prevent
brutality, and give clear guidance to the police. n200 Even after admitting the possibility of alternatives, however,
Schulhofer is unwilling to follow his logic to its conclusion and to consider all reasonable options. For instance, he
attacks my proposed elimination of the automatic questioning cut-off rules as "an explicit recipe for involuntary
custodial interrogation, in direct violation of the Fifth Amendment." n201 But this reading cannot be reconciled with
the fact that the Supreme Court has already allowed "direct violation" of Miranda rules on the basis of cost-benefit
analysis in creating a "public safety" exception to Miranda, among other holdings. n202 In establishing the exception,
the Court clearly explained that "absent actual coercion by the officer, there is no constitutional imperative requiring the
exclu- [*1118] sion of the evidence ...." n203 Exceptions like the one for public safety are not in "direct violation" of
the Fifth Amendment because they violate only Miranda's "prophylactic," that is, subconstitutional rules. n204 As a
result, my proposed modification, like the public safety exception, should be assessed on the basis of costs and benefits.
It is to this issue that we now turn.
B. Videotaping Versus Miranda

In assessing my proposed alternative, it is important to remind ourselves that the relevant comparison is not to a
perfect world, but to Miranda with all its flaws and warts. As any good horsetrader would do, Professor Schulhofer
gives us a rather romanticized account of his merchandise. But as I noted earlier, there appears to be "general agreement
among writers on the subject that Miranda is an inept means of protecting the rights of suspects." n205 Indeed, the
defects of Miranda have been powerfully outlined in another article by Professor Schulhofer himself. n206
In a 1981 article, Professor Schulhofer concluded that Miranda failed to "deliver[] even a fraction of what it seems
to promise." n207 Most seriously, Miranda's entire warnings-and-waivers apparatus rests on police accounts of what
happened during secret questioning inside the stationhouse. As Professor Schulhofer wrote then:
The subject of the "swearing contest" is saddest of all because this problem was the central obstacle to effective
judicial control of the interrogation process. In Miranda itself the Court referred in several places to the problem of
secrecy in police interrogation... Yet Miranda does nothing whatsoever to mitigate the pitfalls of the swearing contest.
The heralded warnings need not be given by a disinterested person, and the Court required no objective proof to
corroborate claims that they were given in proper form by the police... The most that can be said is that by requiring the
prosecution to bear a "heavy burden" of proving waiver, and by dropping a strong hint about the State's ability to furnish
objective corroboration, the Court was perhaps laying the groundwork for tackling the swearing contest in the future ...
One has to wonder, however, whether in seeking a truly effective package of prophylactic rules, the Court should have
started where Miranda starts and postponed what Miranda postponed. n208
[*1119] A solid regime of videotaping does exactly what Professor Schulhofer suggested: it tackles, not postpones,
dealing with "the central obstacle" to effective judicial control of police interrogations. Concerns that Miranda does
nothing to end the secrecy of interrogations are widely shared. n209 Consequently, my proposed alternative should
also prove quite appealing to those, like Amar and Lettow, who are concerned that, "Despite Miranda's promise to open
up the black box of the police station, it did not require that lawyers, magistrates, or even tape or video recorders be
present in the interrogation rooms." n210 My proposed alternative might even prove appealing to critics of Amar and
Lettow like Yale Kamisar, who has written that "not even the otherwise-bold Miranda Court was bold enough to require
law enforcement officers, whenever feasible, to make audio or video recordings of how the now-familiar warnings are
delivered, how the suspect responds, or how the questioning proceeds." n211
In this journal, Professor Schulhofer attempts to dispose of videotaping on the gallop, devoting less than a page to
reaching the conclusion that "though an excellent idea, [it] does not meet the constitutional concerns about compulsion
to which the Miranda safe- [*1120] guards are addressed." n212 But surely the venerable idea of taping interrogations
n213 deserves more attention. Professor Schulhofer has but two comments on the taping-is-better strand of my
argument. He suggests, first, that while taping would prevent physical abuse of suspects, it cannot." n214 But, of

Page 14
90 Nw. U.L. Rev. 1084, *1120

course, neither can Miranda. n215 The relevant question is whether taping is as good (or better) than Miranda on this
measure.
He next argues that "the tape lays down no rules of behavior and gives the police no guidance about the tactics
they may use." n216 But this indictment is plainly overbroad as well, since it lies against the Miranda doctrine too.
Under Miranda, police know that they need to obtain a waiver; but after that, they operate under the same "notoriously
vague due process concepts" that Schulhofer argues would be deficient under my proposal. n217 A good example of
Miranda's open-endedness comes from interrogation manual authors Fred Inbau and John Reid, who were able to write
in their post-Miranda edition that, after a waiver of rights, "all but a few of the interrogation tactics and techniques
presentedin our earlier publication are still valid." n218 Another example comes from Schulhofer's argu- [*1121]
ment that videotaping would "not prevent psychological pressure." n219 But Miranda fails to give guidance on this
subject as well. n220 Given the way police questioning has developed under Miranda, it is hard to understand why we
should regard videotaping - with its permanent record and possibility of after-the-fact judicial (and administrative)
review of police tactics - as worse than the current approach.
Even more telling than the weaknesses in his argument against taping as a replacement for Miranda is Professor
Schulhofer's failure to join the issue of its other substantial benefits. Schulhofer appears to agree with the extended
argument that videotaping would be more effective than Miranda in preventing police brutality and identifying fine
points of coercion, n221 but he offers no specific explanation for his conclusion that this benefit is outweighed by what
he sees as the costs of modifying Miranda. Nor does he offer any response to my argument that videotaping would
eliminate the problem of false confessions that convict the innocent. n222 Indeed, the nation's leading expert on false
confessions, Richard Ofshe, recently concluded that the problem is "fixable," but only if police interrogations are
electronically recorded. n223
Instead of engaging on these issues, Professor Schulhofer moves swiftly and predictably to the one part of my
proposal that is designed to reduce Miranda's social costs: the abolition of the prophylactic rules that require an
affirmative waiver of rights and give suspects an unqualified right to avoid questioning. Here Professor Schulhofer's
approach to the Miranda compromise is revealed to be intransigent. In essence, his stance is: "What's mine is mine, and
what's yours is negotiable." Having conceded that "videotaping is an extremely valuable tool" for the suspect, n224 he
should be willing to admit, since Miranda is a "carefully crafted balance," that at least some part of current doctrine can
be modified to reduce societal costs. Yet he puts not even a single reform on the bargaining table and, indeed, suggests
[*1122] only that videotaping might be piled on top of Miranda's requirements. n225
Professor Schulhofer then paints an exaggerated picture of my proposal's costs. Today about 80% of suspects
waive their rights and submit to questioning. n226 It is hard to see how they are seriously disadvantaged by
eliminating a rule that they do not employ. Indeed, recalling that my proposal includes videotaping as a safeguard, these
suspects will almost surely be better off than being interrogated in complete secret. n227 For most interrogations, then,
the cost calculation in fact generates net benefits, without even factoring in the possible societal benefit of additional
convictions.
This reduces Professor Schulhofer to making the most out of the disadvantages to the 20% of suspects who, under
current rules, either refuse to waive their rights or ask at some point for questioning to stop. n228 The available
empirical evidence suggests that most of these suspects fall into the first category, refusing an initial waiver, rather than
the second, invoking questioning cut-off rules. n229 Like the old children's game, police seeking Miranda waivers
must essentially say "mother, may I?" before asking questions, and a number of suspects simply decline to give
permission. It is hard to regard such initial invocations as anything other than a social cost. As one defender of Miranda
put it, the prohibition on asking even reasonable (and, under my proposal, videotaped) questions is "Miranda's Achilles'
heel." n230 Justice White exposed the problem in his Miranda dissent, explaining that it is "patently unsound" to
exclude the answer to a question such as "Did you kill your wife?" on the theory that, without a formal waiver, the
answer is somehow compelled. n231 Indeed, the whole notion of giving suspects a right to avoid any questioning can
rest on nothing other than a one-sided balance. As Professor Grano has written, "by giving the suspect power to prevent

Page 15
90 Nw. U.L. Rev. 1084, *1122

questioning even before it begins, ... Miranda gave the law of confessions a "single focus - protection of the suspect.' "
n232 [*1123]
Professor Schulhofer spends little time justifying the requirement that police obtain initial waivers, concentrating
instead, not on the majority of suspects who waive their rights or even the majority of "invokers" who simply decline to
consent to questioning, but rather on suspects who specifically ask for questioning to stop at some point. n233 Under
my proposal, police would not have to fold up the tent just because a suspect said the magic words "I want a lawyer."
n234 Of course, these suspects would not be compelled to answer police questions; they would only be compelled to
listen. As others have suggested, a detective would not be "required to cease questioning merely because the suspect has
expressed an initial reluctance to answer" questions. n235 Additional limitations could be placed on how much of a
"pitch" an officer could make for a confession, n236 all of which could be carefully reviewed later on videotape. The
legitimate social costs from such an approach, not including, of course, the fact that some additional suspects might
confess, can hardly be considered substantial.
Even if preventing police from asking suspects reasonable questions is considered a net benefit (at least on the
peculiar Miranda ledger), and even if the virtues of videotaping are not more substantial, we come to perhaps the
greatest difficulty for Miranda's defenders in opposing reasonable alternatives like mine: How are we to assess any such
benefits against the undeniable social costs of criminals escaping conviction? Having gone beyond the boundaries of the
Fifth Amendment on the basis of cost-benefit prophylaxis, n237 where is the metric that explains how a family's pain
when a murderer escapes justice is outweighed by the benefit of giving a suspect the power to stop questioning? What is
the calculus that concludes that a [*1124] victim's suffering when a rapist remains on the streets is less important than
the need to obtain an explicit waiver of a suspect's rights? Miranda's paladins have no acceptable answers to these
questions, n238 which may explain why they have worked so hard for so long to maintain the conventional wisdom
about Miranda that only a "negligible" and invariably unspecified number of criminals go free as a result. Without this
grand illusion, it would be necessary for Miranda's academic defenders to confront explicitly the costly sacrifices that
inhere in the decision no less than in any other compromise social policy.
As we now observe Miranda's thirtieth anniversary, it is time to have a full and fair debate about the decision about its costs, its benefits, and its alternatives. It will no longer do to maintain that the decision does not release many
criminals; compelling evidence indicates otherwise. In view of that fact, we must also no longer indulge in the fantasy
that Miranda strikes the best balance between competing concerns; surely we must have learned something in the last
generation that could serve as a useful improvement by reducing these costs. Indeed, videotaping offers a genuine
possibility that we can craft an alternative that not only significantly helps society but also better protects suspects
against unconstitutional abuse. When Miranda's defenders recognize these points, we can then move on to the urgently
pressing tasks of reforming Miranda and striking a better balance between legitimate and competing concerns.
Legal Topics:
For related research and practice materials, see the following legal topics:
Criminal Law & ProcedureInterrogationMiranda RightsCustodial InterrogationCriminal Law &
ProcedureInterrogationNoncustodial Confessions & StatementsGovernmentsLocal GovernmentsPolice Power
FOOTNOTES:

n1. Paul G. Cassell, Miranda's Social Costs: An Empirical Reassessment, 90 Nw. U. L. Rev. 387 (1996).
n2. Stephen J. Schulhofer, Miranda's Practical Effect: Substantial Benefits and Vanishingly Small Social
Costs, 90 Nw. U. L. Rev. 500 (1996).

Page 16
90 Nw. U.L. Rev. 1084, *1124

n3. See id. at 505.
n4. Id. at 539 (Table 1) (noting that confessions fell after Miranda in Pittsburgh, Philadelphia, New Haven,
and Kansas City).
n5. See, e.g., Wayne R. LaFave & Jerold H. Israel, Criminal Procedure 6.5(c), at 316 (2d ed. 1992)
("Various studies have indicated that ... about as many confessions are obtained by giving the Miranda rights as
were gotten before the Miranda decision."); American Bar Assoc. Comm. on Criminal Justice in a Free Soc'y,
Criminal Justice in Crisis 64 (1988) (concluding that the Pittsburgh study "is the only study that found that
Miranda had an impact").
n6. Schulhofer, supra note 2, at 556.
n7. Id.
n8. Id. at 545 (Table 2).
n9. Id. at 547.
n10. See United States v. Leon, 468 U.S. 897, 908 n.6 (1984).
n11. See Cassell, supra note 1, at 438-40 (explaining extrapolation methodology).
n12. See id. at 484 (citing statistics supporting the conclusion that the exclusionary rule results in the release
of between 0.6% and 2.35% of felony arrestees).
n13. See generally Akhil R. Amar, Fourth Amendment First Principles, 107 Harv. L. Rev. 757, 785-800
(1994) (discussing problems created by the exclusionary rule).
n14. See, e.g., Schulhofer, supra note 2, at 502, 506, 544 (arguing that Miranda's costs are "at most" 0.78%
of all cases); id. at 546 (claiming that there is "overestimation" in his cost estimate); id. at 510 (arguing that
methodology shows the "high-water mark of Miranda's negative impact").
n15. Id. at 546-47.
n16. One must also undertake to assess all the costs of the decision, not simply its costs in terms of lost
cases. As I noted in my article, Miranda likely has other costs such as consuming police and judicial resources
and undermining public confidence in the criminal justice system. See Cassell, supra note 1, at 437 n.292 (citing
Gordon Van Kessel, The Suspect as a Source of Testimonial Evidence: A Comparison of the English and
American Approaches, 38 Hastings L.J. 1, 129 (1986)). Professor Schulhofer makes no substantial effort to
assess these other costs.

Page 17
90 Nw. U.L. Rev. 1084, *1124

n17. Schulhofer, supra note 2, at 546.
n18. Id.
n19. Schulhofer, supra note 2, at 508.
n20. See Cassell, supra note 1, at 453-54. In my initial article, I also argued that over time suspects might
have adjusted their practices to Miranda, depressing the confession rate even more than detected in the
before-and-after studies. See id. at 450 & n.362 (citing LaFave & Israel hornbook, supra note 5, 6.5(c), at 484
n.30, as suggesting that"now that Miranda has become part of our culture and presumably the rights declared
therein are more widely perceived by the public at large" it is "unclear" where confession rates have remained
close to pre-Miranda levels). Schulhofer does not respond to this argument.
n21. Schulhofer, supra note 2, at 508.
n22. Id. (emphasis added).
n23. See Cassell, supra note 1, at 407-09, 454 n.384 (discussing Project, Interrogations in New Haven: The
Impact of Miranda, 76 Yale L.J. 1519, 1552, 1555 (1967) [hereinafter Yale Project]).
n24. See id. at 494-96 (collecting evidence on this point).
n25. As a partial offset to this problem, Schulhofer and I estimate that admissible confessions fell between
12% and 16% in New Haven because of suspects affirmatively invoking their rights during questioning. See
Cassell, supra note 1, at 408-09; Schulhofer, supra note 2, at 530. That reconstruction, however, captures only
suspects bold enough to affirmatively assert their rights, not the presumably larger number who simply decline
to execute waivers requested by the police. See infra note 222 (collecting evidence that many more suspects
simply decline to waive their rights than affirmatively assert them during questioning). Thus, even the
reconstructed New Haven confession rate does not begin to capture all the damage from "complete" compliance
with Miranda.
n26. See Controlling Crime Through More Effective Law Enforcement: Hearings Before the Subcomm. on
Criminal Laws and Procedure of the Senate Comm. on the Judiciary, 90th Cong., 1st Sess. 200-01 (1967)
[hereinafter Controlling Crime Hearings] (statement of Philadelphia District Attorney Arlen Specter) (reporting
only that in Philadelphia "on June 17, 1966, the District Attorney's Office provided the Police Department with
guidelines on warnings to be given and questions to be asked in light of the Miranda decisions [and that w]hen
the requisite warnings were given, these statistics followed"); id. at 1120 (statement of New York County
District Attorney Frank Hogan) (reporting statistics for what happened "after Miranda"); 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, 325 (1973) (noting only that "July 1, 1966 was the date when
the Seaside City detective began to observe the Miranda precepts"); Wayne E. Green, Police vs. "Miranda": Has
the Supreme Court Really Hampered Law Enforcement, Wall St. J. Dec. 15, 1966, at 16 (not describing police
implementation of Miranda in Kansas City); Controlling Crime Hearings, supra, at 223 (statement of Kings
County, New York, District Attorney Aaron Koota) (reporting statistics for June to September 1966 without

Page 18
90 Nw. U.L. Rev. 1084, *1124

description of police implementation practices). The only study to provide even limited information about police
compliance is the Pittsburgh study, in which the study's authors report in conclusory fashion (apparently based
on information in police files) that the detectives followed Miranda's requirements. See Richard H. Seeburger &
R. Stanton Wettick, Jr., Miranda in Pittsburgh - A Statistical Study, 29 U. Pitt. L. Rev. 1, 8 (1967).
n27. In New Haven, students observed detectives in the summer of 1966 and reported substantial police
deviations from Miranda, as noted earlier. See supra note 23. In the District of Columbia in 1966 and 1967,
interviews with defendants and defense attorneys revealed substantial noncompliance with Miranda. See Cassell,
supra note 1, at 410 (discussing Richard J. Medalie et al., Custodial Police Interrogation in Our Nation's Capital:
The Attempt to Implement Miranda, 66 Mich. L. Rev. 1347 (1968)). (The problem with noncompliance in the
Medalie study is so substantial that both Schulhofer and I discard the study's data entirely in determining a
Miranda cost figure.) Finally, in New York City in 1967, the Vera Institute collected tape recordings of a
substantial number of interrogations, which revealed many police practices that Miranda condemns. See Vera
Inst. of Justice, Taping Police Interrogations in the 20th Precinct, N.Y.P.D. 16-47 (1967); see also Vera Inst. of
Justice, Monitored Interrogations Project: Final Report-Statistical Analysis 75-76 (1967) [hereinafter Vera
Study] (in different sample, stating that the assumption that Miranda was complied with "is open to serious
question"). Because these three studies are the only ones that used an observational or quasi-observational
methodology, they are the only ones that provide concrete information about whether police followed Miranda.
The other studies simply collected data about the outcomes of police interrogation, not the methods. There is no
basis for reading them as demonstrating "complete" implementation of Miranda. Cf. Schulhofer, supra note 2, at
513 (noting that Miranda was "implemented slowly and imperfectly in many jurisdictions").
n28. Stephen J. Schulhofer, Reconsidering Miranda, 54 U. Chi. L. Rev. 435, 456 (1987) [hereinafter
Schulhofer, Reconsidering Miranda]; see also Stephen J. Schulhofer, The Fifth Amendment at Justice: A Reply,
54 U. Chi. L. Rev. 950, 954 n.17 (1987) (arguing that clearance rates coupled with other evidence refute notion
that Miranda has harmed law enforcement). Schulhofer does not address clearance rates in his current article.
Schulhofer's 1987 clearance-rate argument was tied to other evidence, specifically "conviction rates" and "the
impressions of law enforcement officers in the field" that, Schulhofer argued, "considered together" produce a
"reasonably reliable picture." Id. Interestingly, Schulhofer does not repeat the 1987 conviction-rate argument
here. See also Cassell, supra note 1, at 396-98 (explaining why conviction rates are a poor measure of Miranda's
effects). Nor does the police "impressions" strand of his 1987 argument seem tenable. See infra notes 128-55
and accompanying text (noting that almost all of the systematic surveys of police impressions found that police
thought Miranda was harmful).
n29. See Federal Bureau of Investigation, Crime in the United States: Uniform Crime Reports 1950-1993
(1951-1994) (tables for clearance rates). The data for 1950 to 1974 are helpfully collected in James A. Fox,
Forecasting Crime Data: An Econometric Analysis 86 (1978) (Table A-1).
n30. See Cassell, supra note 1, at 463-64 (collecting evidence that Miranda had the most substantial effect
for suspects charged with violent crimes). Violent crimes are murder, robbery, forcible rape, and assault.
n31. To be clear, the point on the graph marked "Miranda" is the last pre-Miranda data point - 1965. The
exact data from the relevant years are: 1964 - 59.5%, 1965 - 59.5%, 1966 - 55.4%, 1967 - 51.0%, 1968 - 46.9%,
1969 - 45.6%.
n32. It seems likely that clearance rates understate Miranda's effects. See Cassell, supra note 1, at 398-99

Page 19
90 Nw. U.L. Rev. 1084, *1124

(arguing that clearance rates are "an understated measure of Miranda's effects").
n33. The standard statistical technique for sorting out competing effects is multiple regression analysis. I am
in the process of conducting such analysis of violent crime clearance rates, controlling for relevant variables
such as the number of crimes, the number of police officers, and the funds spent on law enforcement.
Preliminary results indicate that Miranda had a substantial, statistically significant effect on violent crime
clearance rates. Even without resort to such techniques, however, it is clear that the obvious competing factor
from the 1960s - rising crime rates that might overwhelm police efforts to clear crimes - does not seem to be the
explanation. Crime rates began rising before Miranda and continued rising after the 1966-68 period. See Paul G.
Cassell, The Costs of the Miranda Mandate: A Lesson in the Dangers of Inflexible, "Prophylactic" Supreme
Court Inventions, 28 Ariz. St. L.J. (forthcoming 1996) (graphing number of crimes against crime clearance rates,
which reveals no sudden change in crime rates confined to the 1966 to 1968 period); see also Fox, supra note 29,
at 10, 11, 14 (finding that violent crime clearance rates, unlike crime rates, changed sharply in the mid-1960s
and then stabilized). See generally Proceedings of the Fifteenth Annual National Student Federalist Society
Symposium on Law and Public Policy - 1996, 20 Harv. J.L. & Pub. Pol'y (forthcoming 1997) (debate between
Cassell and Schulhofer on the effect of crime rates on the post-Miranda drop in clearance rates).
Since rising crimes rates cannot explain away the post-Miranda drop in clearance rates, the challenge for
those who cling to the notion that the decision did not harm law enforcement is to provide an alternative "X
factor" that explains the change. See Cassell, supra (discussing and dismissing other X factors besides Miranda).
Because my regression equations are still preliminary, I will happily receive any suggestions as to other
variables that should be included.
n34. Most of the before-and-after studies were concluded between December 1966 and July 1967. But the
clearance rate data suggest that Miranda's harmful effects continued to unfold through 1967 and 1968.
n35. See Paul G. Cassell & Bret S. Hayman, Police Interrogation in the 1990s: An Empirical Study of the
Effects of Miranda, 43 UCLA L. Rev. 839, 868 & n.142 (1996). This figure includes incriminating statements
volunteered by suspects, which may slightly inflate the rate for purposes of comparing to other studies.
Excluding volunteered statements drops the rate to 33.3%. Id. Even this rate may be inflated. See id. at 868-69
(noting that police may have "prescreened" cases without confessions before presenting them to the District
Attorney's Office).
Professor Schulhofer argues that our confession rate is deflated because our denominator includes suspects
who were never questioned. Schulhofer, supra note 2, at 509 n.28. But all of the Miranda studies from the 1960s
included such suspects in their samples. See Cassell & Hayman, supra, at 875 n.176 (collecting citations from
relevant studies). Schulhofer also complains that our study included both custodial and noncustodial
interrogations. Schulhofer, supra note 2, at 509 n.28. But here again, the studies done around the time of
Miranda generally included both types of interrogation. See Cassell & Hayman, supra, at 872-73 (collecting
citations from relevant studies). Finally, Schulhofer, supra note 2, at 509 n.28, argues that our definition of
"incriminating" statements was different than earlier studies. But, in fact, we tracked accepted categorizations.
See Cassell & Hayman, supra, at 867 n.139 (explaining consistency of classification methodology with earlier
studies). But cf. George C. Thomas III, Plain Talk about the Miranda Empirical Debate: A "Steady-State"
Theory of Confessions, 43 UCLA L. Rev. 933, 953 (1996) (arguing that our confession rate is comparable to
that found by pre-Miranda studies).
n36. Richard A. Leo, Police Interrogation in America: A Study of Violence, Civility, and Social Change
(1994) (unpublished Ph.D. dissertation, University of California (Berkeley)). Leo's informative dissertation will

Page 20
90 Nw. U.L. Rev. 1084, *1124

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).
n37. Cassell & Hayman, supra note 35, at 926-29.
n38. Id. at 929-30.
n39. Id. at 875-76 (collecting studies from Jacksonville, Florida; San Diego, California; and six other cites
from around the country).
n40. See id. at 871-72 and accompanying text (collecting relevant studies). But cf. Thomas, supra note 35,
at 936 n.12 (arguing for a lower pre-Miranda average). These pre-Miranda averages may understate Miranda's
effects because they rest on data collected in jurisdictions already following some of the Miranda requirements
as the results of anticipatory court decisions. See infra notes 93-97 and accompanying text.
n41. Accepting all of Schulhofer's "site specific" adjustments save for these three leaves a post-Miranda
confession-rate drop of 15.9%, virtually indistinguishable from my 16.1%.
n42. For example, Schulhofer recognizes (as I did) that a 2% drop in confessions in "Seaside City" is not
statistically significant, and therefore adjusts the figure to 0%. Schulhofer, supra note 2, at 529. But Schulhofer
at the same time hypothesizes a 2% yearly drop in the confession rate during this period from "other causes."
See infra notes 113-20 and accompanying text. Accordingly, the null hypothesis is not a stable confession rate
but a 2% annual drop, for a total drop of roughly 4% in Seaside City during the two post-Miranda years covered
by the study. Therefore, under Schulhofer's accounting, the Seaside City figure must be readjusted to -4% to
avoid doublecounting when he makes the "other cause" adjustment. As another example, in analyzing data from
New Haven, Schulhofer argues that for several of the suspects I count as failing to confess due to Miranda,
"[we] have no reason to assume" that they would have "confessed without warnings." Schulhofer, supra note 2,
at 530 (emphasis omitted). But that appears to be the judgment that the study's authors made. See Yale Project,
supra note 23, at 1573 (reporting that warnings affected "the interrogation result in several cases") (emphasis
added). Schulhofer also criticizes my inclusion of data from New Orleans on the ground that they rest on an
estimated pre-Miranda confession rate. See Schulhofer, supra note 2, at 531. But the net effect of including an
11.2% drop from New Orleans was to decrease my cost estimate, which rested on an average 16.1% drop. While
including New Orleans still seems proper to me, see infra note 66 and accompanying text (explaining why
estimates can properly be used), the neteffect of excluding New Orleans is to increase my estimate of Miranda's
costs.
n43. Schulhofer adjusts the Philadelphia numbers in part because they rest on estimates, a methodology I
defend at infra notes 64-66 and accompanying text, and because they assume a 50% ratio of incriminating to
total statements, an assumption I defend at infra notes 56-63 and accompanying text. Schulhofer also adjusts the
Philadelphia numbers by selecting for comparison a different baseline than the one I used. Schulhofer, supra
note 2, at 526-27. Given the choice, however, between the pre-October 1965 regime (when Philadelphia police
gave limited advice of rights) and the post-October 1965 regime (when they gave an advice-of-counsel
warning), he picks the latter, concluding that "it seems unlikely that the slightly stronger wording of the ...
warning would by itself produce a significant difference in the confession rate . ..." Id. at 526 n.100. But
Schulhofer offers no other explanation for the fact that confession rates fell after police introduced the

Page 21
90 Nw. U.L. Rev. 1084, *1124

right-to-counsel warning, see Cassell, supra note 1, at 402, and other empirical evidence associates right to
counsel warnings with fewer confessions, id. at 494-96 & n.623. Because my proposed alternative dispenses
with the right to counsel warning entirely, the proper baseline is questioning without such a warning.
In his discussion of the Philadelphia study, Schulhofer also tries to describe as pre-Escobedo "standard
practice" a case in which a suspect was questioned after being "stripped naked." Schulhofer, supra note 2, at 526
(discussing Commonwealth ex rel. Kern v. Banmiller, 187 A.2d 185, 190 (Pa. Super. Ct. 1962), cert. denied, 374
U.S. 852 (1963)). It seems a bit of a stretch to cite police questioning from 1959 as "immediately prior to
Escobedo." See Schulhofer, supra note 2, at 526; cf. id. at 531 n.127 (arguing that 1961 data "would be a poor
proxy for the relevant pre-Miranda baseline"). In any event, on close inspection, the dramatic "standard practice"
(in Pittsburgh, not Philadelphia) of "questioning suspects after they had been stripped naked" turns out to be the
more prosaic questioning of a suspect after he entered a secure area and had undergone a routine (hence,
"standard practice") strip search for security reasons. Banmiller, 187 A.2d at 190.
n44. Schulhofer, supra note 2, at 520.
n45. Compare Cassell, supra note 1, at 399-402 with Schulhofer, supra note 2, at 517-24; compare also
Stephen J. Markman, The Fifth Amendment and Custodial Questioning: A Response to "Reconsidering
Miranda," 54 U. Chi. L. Rev. 938, 946 n.19 (1987) (arguing that Hogan's conclusions are valid) with Schulhofer,
Reconsidering Miranda, supra note 28, at 457 (arguing that Hogan's conclusions are invalid).
n46. Schulhofer, supra note 2, at 523.
n47. Id. Instead of excluding the data entirely, however, we could assume generously that a third of Hogan's
post-Miranda sample was contaminated by inadmissible pre-Miranda interrogations that were not presented to
the grand jury. That would transform Hogan's 14.5% post-Miranda confession rate into a 21.8% confession rate
(1/3 x 0% + 2/3 x true confession rate = 14.5% reported confession rate), still a very significant 27.2% drop
from the pre-Miranda rate.
n48. Vera Study, supra note 27; Controlling Crime Hearings, supra note 26, at 205, 223 (statement of
District Attorney Aaron Koota).
n49. Schulhofer, supra note 2, at 523. Schulhofer also notes that the Vera data involve police interrogations,
not grand jury presentations, and thus are not identical to Hogan's data. I note the same qualification in my
original article. See Cassell, supra note 1, at 401 n.76. However, Schulhofer overstates the problem in suggesting
that cases in the Vera sample dropped before grand jury presentation would have a confession rate that is "very
low (or zero)." Schulhofer, supra note 2, at 523. The only available empirical evidence is to the contrary. See
Cassell & Hayman, supra note 35, at 908 (Table 15) (finding 20.5% of suspects not charged had given
incriminating statements as compared to 36.6% of suspects charged); see also Leo, supra note 36, at 294 (Table
16) (finding that 24% of suspects successfully interrogated were not charged).
n50. See Vera Study, supra note 27, at 26 (Table X).
n51. See id. at 47 (Table XXIII). The table does not differentiate between confessions and admissions.

Page 22
90 Nw. U.L. Rev. 1084, *1124

n52. Forty-nine percent pre-Miranda rate minus 16.6% and 25.6% post-Miranda rates, respectively.
n53. Schulhofer, supra note 2, at 518.
n54. Cassell, supra note 1, at 413-14 (discussing Controlling Crime Hearings, supra note 26).
n55. See also infra note 59 (offering another way of reconciling the Brooklyn and Manhattan data).
n56. See Cassell, supra note 1, at 403 (defending this assumption).
n57. Id. at 413-14.
n58. Schulhofer, supra note 2, at 526.
n59. For illustrative purposes, Schulhofer notes that a 10% rise in the proportion of incriminating statements
from 40% before Miranda to 50% after Miranda, would mean that the confession rate drop in Brooklyn was only
6.5%, not the 15.5% a steady 50% ratio generates. Id. at 533. On the same methodology, a 10% fall in the
proportion of incriminating statements would mean a confession rate drop of about 21% (90% pre-Miranda
statement rate x 50% incriminating statement rate - 59% post-Miranda statement rate x 40% incriminating
statement rate = 21.4%). Interestingly, this brings the Brooklyn data more in line with the data from neighboring
Manhattan and thus may provide a partial explanation for the slightly differing figures.
n60. Id. at 525.
n61. See id. at 522 & n.83-84 (citing Controlling Crime Hearings, supra note 26, at 1122).
n62. Hogan reports no specific figures on the ratio of incriminating statements to total statements,
suggesting only that before Miranda suspects were more likely to "protest their innocence." Controlling Crime
Hearings, supra note 26, at 1122. This conclusion was based on a pre-Miranda estimate. Cf. Schulhofer, supra
note 2, at 525 (arguing that reliance on "crude, ballpark estimate based on recollections and conversations" is an
inappropriate methodology). It is also limited to homicide offenses. Cf. id. at 541 n.172 (arguing that Hogan's
data on homicide confession necessity "cannot be generalized to other crimes").
N63. The Seaside data show that, while the percentage of incriminating statements rose very slightly after
Miranda (from 76.1% to 80.6%), this was not a statistically significant difference, for Pr. = .55. Cf. Schulhofer,
supra note 2, at 529 (rejecting evidence of a post-Miranda drop in the confession rate in Seaside City because it
"was not statistically significant"). Discussing this data, Professor Thomas concludes that it supports the
conclusion that Miranda "simultaneously encourages suspects to answer police questions while discouraging
admissions." George C. Thomas III, Is Miranda a Real-World Failure? A Plea for More (and Better) Empirical
Evidence, 43 UCLA L. Rev. 821, 831 (1996).
n64. Schulhofer, supra note 2, at 532-33.

Page 23
90 Nw. U.L. Rev. 1084, *1124

n65. See infra notes 74-87 and accompanying text (discussing problems in two different data sets in Los
Angeles).
n66. Schulhofer, supra note 2, at 534. It bears noting that Schulhofer himself uses an estimate for an
important part of his cost calculation. To reduce Miranda's costs, Schulhofer estimates thatfrom June 1966 until
1968, confession rates were declining 2% each year. Id. at 511-12. While there are substantial problems with
this figure, see infra notes 113-20 and accompanying text (criticizing the adjustment), the important point for
present purposes is that it is simply an estimate. See Schulhofer, supra note 2, at 512 (arguing that a 2% yearly
drop can serve as a "rough approximation" of the trend). Schulhofer also implicitly accepts the validity of
estimates of confession rates in relying heavily on police impressions of their success in obtaining confessions
after Miranda. See, e.g., id. at 501 (relying on "perception" of law enforcement professionals); id. at 503-04
(discussing "informal impressions of law enforcement personnel").
n67. Also supporting this conclusion is Hogan's report that fewer homicide suspects made statements after
Miranda. See Cassell, supra note 1, at 401. In response, Professor Schulhofer finds it noteworthy that Hogan's
separately calculated post-Miranda confession rate for homicide is 35%, higher than his 14.5% rate for other,
less serious crimes in the grand jury study. Schulhofer, supra note 2, at 522 ("The 35% post-Miranda confession
rate is more than double the 14.5% post-Miranda rate recorded in the problematic grand jury figures."). Yet
Schulhofer also notes that the 1967 Vera Study in New York found that "in the more serious specific offenses,
the percentage [of suspects] making statements was considerably higher than in the comparatively less serious
offenses." Id. at 523 (citing Vera Study, supra note 27, at 22-23). Indeed, for homicide, in the Vera Study's
Manhattan sample, the police success rate was 43%; in the Twentieth Precinct sample, it was 71%. See Vera
Study, supra note 27, at 26 (Table X), 47 (Table XXIII). Thus, comparing high confession rates for homicide to
lower confession rates for all crimes is truly comparing apples to oranges. Schulhofer also responds by reading
Hogan's testimony as suggesting only that homicide suspects made fewer statements after Miranda, not fewer
incriminating statements. See Schulhofer, supra note 2, at 522. But, read in context, it is quite clear that Hogan
said that such incriminating statements were harder to obtain. In the next two sentences after the sentence cited
by Schulhofer, see id. (citing Controlling Crime Hearings, supra note 27, at 1122), Hogan discusses other
statistics for "confessions and admissions" as establishing a higher pre-Miranda baseline than the 35% he found.
Controlling Crime Hearings, supra note 27, at 1122 (reporting that 85% of pre-Miranda capital cases in New
York State involved confessions or admissions and that 76% of pre-Miranda capital cases in New York County
involved confessions or admissions).
n68. See Schulhofer, supra note 2, at 539-40 (discussed at infra notes 121-27 and accompanying text).
n69. Id. at 538.
n70. Id. at 534; see also infra notes 102-05 and accompanying text. The California case is People v. Dorado,
398 P.2d 361 (Cal.) (en banc), cert. denied, 381 U.S. 937 (1965).
n71. See Controlling Crime Hearings, supra note 26, at 344 (describing survey dates).
n72. Schulhofer, supra note 2, at 518.
n73. See id. at 534, 540.

Page 24
90 Nw. U.L. Rev. 1084, *1124

n74. See Cassell, supra note 1, at 415-16 (discussing Controlling Crime Hearings, supra note 26, at 347, 350
(emphasis added)). Professor Schulhofer agrees that this is a common problem with the Miranda studies.
Schulhofer, supra note 2, at 511 ("Many of the before-and-after studies concerning Miranda are critically flawed
because ... the kinds of statements counted as confessions were not identical in the before and after periods.").
n75. This phrase is repeatedly placed in quotations in the study description, so there seems little doubt that it
comes from the questionnaires themselves. See, e.g., Controlling Crime Hearings, supra note 26, at 347, 348.
n76. See Schulhofer, supra note 2, at 535.
n77. See Controlling Crime Hearings, supra note 26, at 349.
n78. See id. at 349-52 (reprinting memorandum describing pre-Miranda survey; no mention of any planned
follow-up).
n79. Hogan's New York County study also appears to have been performed on a continuous sample basis
from December 1965 to December 1966, although data are excluded for June, the month when Miranda was
handed down. Controlling Crime Hearings, supra note 26, at 1120.
n80. See Cassell, supra note 1, at 415.
n81. See Schulhofer, supra note 2, at 535.
n82. For example, the pre-Miranda survey collected data on "sufficient evidence without confession or
admission to sustain conviction"; no such data are reported for the post-Miranda survey. Compare Controlling
Crime Hearings, supra note 26, at 350 (reporting results of pre-Miranda survey) with id. at 347 (reporting results
of post-Miranda survey). Conversely, the post-Miranda survey collected data on the number of cases in which a
suspect provided a "confession, admission or other statement" and a "Miranda admonition [was] given"; no
comparable data are reported for the pre-Miranda survey. Compare id. at 350 (reporting results of pre-Miranda
survey) with id. at 347 (reporting results of post-Miranda survey). See also Cassell, supra note 1, at 425-26
(noting differences in pre- and post-Miranda data on necessity for confessions in Los Angeles data).
n83. Office of the Dist. Attorney, County of Los Angeles, Dorado-Miranda Survey (1966).
n84. Id. (complaint stage questionnaire).
n85. For instance, the questionnaire contains no question regarding whether at the complaint stage there
would be sufficient evidence without a confession to sustain a conviction. See id.
n86. Id.
n87. A purely derivative description is found in Evelle J. Younger, Interrogation of Criminal Defendants -

Page 25
90 Nw. U.L. Rev. 1084, *1124

Some Views of Miranda v. Arizona, 35 Fordham L. Rev. 255 (1966) (citing internal memoranda as authority on
various points). Schulhofer finds it "jarring" that I pay more attention to the survey instruments and other
first-hand reports than to Younger's second-hand recounting. Schulhofer, supra note 2, at 536. But, of course, a
primary source should take precedence over a secondary one when there is a conflict. Schulhofer also claims this
approach is inconsistent with my use of Hogan's summary of his data as an interpretative device. Id. But we
have no original survey instruments for Hogan's study. Thus, Hogan's summary becomes the "original" we must
work with.
n88. Schulhofer, supra note 2, at 536.
n89. Id.
n90. I also argued that some of the post-Miranda data may have effectively been pre-Miranda data. Cassell,
supra note 1, at 416. Schulhofer replies that, because requests for complaints follow quickly on the heels of
arrest, all of the post-Miranda complaint data would have involved arrests made after June 13, 1966, the date
Miranda was decided. Schulhofer, supra note 2, at 537-38. My point was intended, however, to be a broader one:
since it could have taken a week or two to "get the word out" to law enforcement officers about how to comply
with Miranda, some of the officers may have been following pre-Miranda routines for arrests made after June
13th, particularly with respect to the important waiver and questioning cut-off rules. While Schulhofer
complains that I should lodge a similar complaint against other studies, id., the L.A. study is much more
sensitive to this problem because of its extremely short data-collection period. See Controlling Crime Hearings,
supra note 26, at 344 (noting that data collection started several days after June 21, 1966 and stopped on July 15,
1966).
n91. See supra note 56 and accompanying text.
n92. Schulhofer, supra note 2, at 538.
n93. Id. at 511-13, 540-41, 545 (Table 2).
n94. 367 U.S. 643 (1961).
n95. 372 U.S. 335 (1963).
n96. Schulhofer, supra note 2, at 507.
n97. 378 U.S. 478, 488 (1964).
n98. See Schulhofer, supra note 2, at 511.
n99. Id. at 515.

Page 26
90 Nw. U.L. Rev. 1084, *1124

n100. Id. at 545 (Table 2) (reciting data from Pittsburgh, Philadelphia, Seaside City, New Haven, and
Kansas City).
n101. See Controlling Crime Hearings, supra note 26, at 200 (noting warnings given in Philadelphia after
Escobedo and Russo); Seeburger & Wettick, supra note 26, at 8 (noting that Pittsburgh detectives advised
suspects of their right to counsel after Escobedo with the advice woven into interrogation); Witt, supra note 26,
at 325 & n.41 (noting warnings given, People v. Dorado, 398 P.2d 361 (Cal. 1965), which expanded on
Escobedo); Yale Project, supra note 23, at 1573 (noting that New Haven police claimed "they had advised on the
right to silence since Escobedo"); Wayne E. Green, Police vs. "Miranda": Has the Supreme Court Really
Hampered Law Enforcement, Wall St. J., Dec. 15, 1966, at 16 (not describing whether warnings were given in
Kansas City before Miranda). See generally 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, 482 (reporting that after Escobedo but before Miranda 90% of police and
prosecutors said they advised suspects of their right to silence).
n102. Schulhofer, supra note 2, at 534.
n103. See Witt, supra note 26, at 323.
n104. Schulhofer, supra note 2, at 539 (Table 1).
n105. Compare id. at 534 (conceding that an objection to using Los Angeles data in comparison of Miranda
to 1960s voluntariness test "would be well-founded") with id. at 529 n.113 (arguing that including Seaside City
data seems "a reasonable approach"). The net effect of Schulhofer's decision to include Seaside City data is to
decrease his Miranda cost calculation by about one fifth (12.1% drop in confessions without Seaside City data
vs. 9.7% drop as calculated by Schulhofer with Seaside City data).
n106. Id. at 510; see also supra note 14 (collecting citations asserting that Schulhofer's cost estimates reflect
the maximum possible cost of Miranda).
n107. See Schulhofer, supra note 2, at 539 (Table 1 n.*) (concluding that the effect of including data from
Seaside City and Philadelphia may be "to somewhat understate the difference between Miranda and a regime
without warnings"). Schulhofer also argues that the extent of this problem is limited. Id. But the important point
remains that Schulhofer's cost figure is too low, not too high.
n108. Schulhofer, supra note 2, at 515.
n109. Professor Schulhofer attempts to capitalize on the fact that I have not proposed the complete abolition
of Miranda and that my alternative fails to fully eradicate all of the decision's harms. Somewhat curiously for a
defender of Miranda, Schulhofer suggests that "the victims of crime ... should be uppermost in our minds" and
wonders what I would "say to the families" of the victims "whose attackers would be released by [my]
proposal?" Id. at 546, 563. I would tell these families that I have done my utmost to reduce Miranda's harms by
proposing modification of its harmful requirements, in the face of a tenacious defense of the status quo by
Miranda's defenders. (I have actually made such presentations on reforming Miranda to several crime victims'

Page 27
90 Nw. U.L. Rev. 1084, *1124

organizations.) Since Schulhofer has asked such a question, perhaps fairness permits me one in return: What
would he say to the undoubtedly much larger number of victims' families who will continue to see criminals
escape justice under the outdated Miranda regime?
n110. See, e.g., Joseph D. Grano, Confessions, Truth, and the Law 199-222 (1993); Gerald M. Caplan,
Questioning Miranda, 38 Vand. L. Rev. 1417, 1467-76 (1985); Fred E. Inbau, Over-Reaction - The Mischief of
Miranda v. Arizona, 73 J. Crim. L. & Criminology 797, 809-10 (1982).
n111. See 18 U.S.C. 3501 (1994); see also S. Rep. No. 1097, 90th Cong., 2d Sess. 40-52 (1968), reprinted
in 1968 U.S.C.C.A.N. 2123, 2127-38.
n112. See Davis v. United States, 114 S. Ct. 2350, 2358 (1994) (Scalia, J., concurring) (noting that failure to
enforce 3501 may have produced "the acquittal and the nonprosecution of many dangerous felons, enabling
them to continue their depredations upon our citizens"); see also Brief of Amicus Curiae Washington Legal
Foundation at 7 & passim, Davis v. United States, 114 S. Ct. 2350 (1994) (No. 92-1949) (arguing that 3501
replaced Miranda with voluntariness standards in federal cases); Paul G. Cassell & Joseph Grano, A Federal
Statute that Overrules Miranda: A New Argument for Federal Prosecutors in Confession Cases, 2 Crim. Prac. L.
Rep. 145 (1994); Paul G. Cassell, Tossing Out the Law on Confessions, Legal Times, Apr. 24, 1995, at 26
(same). During the Oversight of the Solicitor General's Office: Hearing Before the Senate Committee on the
Judiciary, 104th Cong., 1st Sess. (Nov. 14, 1995), Paul G. Cassell has also argued that the Department of
Justice's failure to defend 3501 raises serious legal and ethical questions.
n113. Yale Project, supra note 23, at 1573.
n114. The sample contained more serious crimes, more evidence against defendants, and more juvenile
suspects. Id. at 1573-74.
n115. Id. at 1574.
n116. Time trend data on clearance rates also cut against his conclusion; clearance rates were stable from
1964 to 1965. See supra note 31.
n117. See Witt, supra note 26, at 325 (Table 3), discussed in Cassell, supra note 1, at 454. The reported
confession rates were as follows: 1964-67%; 1965-70%; 1966-77%. Some of the 1966 data are post-Miranda.
By my calculations, adjusting the 1966 rate to reflect only pre-Miranda cases leaves a rate of 74%, up 4% from
the previous year.
n118. See Theodore Souris, Stop and Frisk or Arrest and Search - The Use and Misuse of Euphemisms, 57
J. Crim. L., Criminology & Police Sci. 251, 255 (1966) (2.8% decline in confessions from 1961 to 1965),
discussed in Cassell, supra note 1, at 454 n.385. A conflicting account of the study reports that confessions
actually rose in Detroit between 1961 and 1965. See Cassell, supra note 1, at 494 n.618 (citing Yale Project,
supra note 23, at 1641).
n119. Cf. Schulhofer, supra note 2, at 534 (noting that to make causal inferences from data "we must

Page 28
90 Nw. U.L. Rev. 1084, *1124

include all the usable observations").
n120. The average trend is New Haven (-2.0%) plus Seaside City (+3.0%, and maybe more if the 1966 data
are used, see supra note 117) plus Detroit (-0.7%) equaling an average +0.1% increase.
n121. See Schulhofer, supra note 2, at 540.
n122. Id. at 539.
n123. Id. at 545 (Table 2) (adjusting my confession-rate decline from 16.1% to 12.0%). While this
adjustment means that the absolute number of lost cases from Miranda is smaller than estimated, it also means
that they disproportionately harm inner city residents, who may be unfairly burdened by the problems of crime.
See generally Charles Murray, Losing Ground: American Social Policy, 1950-1980, at 119-20 (1984); Douglas
S. Massey, Getting Away with Murder: Segregation and Violent Crime in Urban America, 143 U. Pa. L. Rev.
1203 (1995); Adam Walinsky, The Crisis of Public Order, Atlantic Monthly, July 1995, at 39, 53.
n124. See Cassell, supra note 1, at 405-06. Under Schulhofer's view of the world, Seaside City must be
regarded as even more anomalous because (1) its pre-Miranda confession rate was rising when it should have
been falling, see supra note 117 and accompanying text, and (2) police officers reported contrary to the research
findings that they were getting "many fewer confessions after Miranda," see infra note 146 and accompanying
text.
n125. See Cassell, supra note 1, at 446-47 (noting generalizations for (1) confession rate reductions and (2)
confession rate necessity across the dimensions of (a) geography, (b) time, and (c) criminal offense categories).
n126. See id. at 463-66, 466-70, 470-71.
n127. For example, to adjust for the increasing need for confessions over time, one could use the most
recent data on confessions necessity - the 61% figure from my 1994 Salt Lake County study. See Cassell &
Hayman, supra note 35, at 38. Using it to refine the cost figure would easily offset Schulhofer's large city
adjustment: (16% drop in confessions adjusted to 12% for large city representation) x (24% confession necessity
rate adjusted to 61% because of increased necessity over time) = 7.3% total cost figure, which is more than the
3.8% total cost figure originally estimated.
n128. Schulhofer, supra note 2, at 501.
n129. Id. at 507.
n130. Schulhofer, Reconsidering Miranda, supra note 28, at 456 (arguing that the "view that Miranda posed
no barrier to effective law enforcement had become widely accepted" by "prominent law enforcement
officials").
n131. See infra notes 148-55 and accompanying text (discussing ABA "Special Committee" study).

Page 29
90 Nw. U.L. Rev. 1084, *1124

n132. Compare Schulhofer, supra note 2, at 501 n.4 with Pete Yost, Police Executive Backs Miranda Ruling
Review: Meese Urged to Seek Reversal by Court, Wash. Post, Jan. 23, 1987, at A17 (reporting executive
director of the International Association of Chiefs of Police concludes that Miranda warnings "have significantly
reduced the number of voluntary confessions over the years") and Stephen Wermiel, Miranda Ruling Continues
to Fall Under Attack: Some Critics See It as Law Enforcement Barrier, Wall St. J., Sept. 8, 1987, at 72.
n133. See Schulhofer, supra note 28, at 456.
n134. See Yale Kamisar, How to Use, Abuse - and Fight Back With - Crime Statistics, 25 Okla. L. Rev.
239, 255 (1972) (noting a news story suggesting that at the 1967 Annual Convention of the National District
Attorneys Association "many" of Younger's colleagues showed "resentment" toward him "for having broken
ranks with them on Miranda").
n135. See Amicus Curiae Brief of Americans for Effective Law Enforcement et al., Brewer v. Williams,
430 U.S. 387 (1977) (No. 74-1263). In his current article, Professor Schulhofer cites this brief as a reason for
believing "Younger had no reason to understate Miranda's effects." Schulhofer, supra note 2, at 537.
n136. Schulhofer, supra note 2, at 562.
n137. Cf. Cassell & Hayman, supra note 35, at 884 (discussing police attempts to lawfully avoid the
Miranda regime by shifting to noncustodial interrogations and concluding that "the very fact that police have
tried to shift suggests, contrary to the view of some defenders of Miranda, that interrogating police officers
believe the Miranda rules are harmful to their efforts"); Jerome H. Skolnick & Richard A. Leo, The Ethics of
Deceptive Interrogation, Crim. Just. Ethics, Winter/Spring 1992, at 5 (noting police use of noncustodial
interviews to "circumvent" the Miranda rules).
n138. 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, 465 (Table 12).
n139. Yale Project, supra note 23, at 1528.
n140. Id. at 1611.
n141. Id. at 1612 n.265.
n142. See Otis H. Stephens et al., Law Enforcement and the Supreme Court: Police Perceptions of the
Miranda Requirements, 39 Tenn. L. Rev. 407 (1972); see also Otis H. Stephens, Jr., The Supreme Court and
Confessions of Guilt (1973).
n143. Stephens et al., supra note 142, at 420 (Table IV) (finding that 98% believed Court decisions created
adverse affects and that 58% attributed this first and foremost to Miranda). Seventy-four percent said that advice
of rights had an adverse effect on investigations. Id. at 424 (Table VIII). In individual interviews, the officers

Page 30
90 Nw. U.L. Rev. 1084, *1124

surveyed generally gave negative assessments of Miranda. Id. at 426-29. In light of these findings, Stephens's
conclusion that his survey showed little impact from Miranda, see id. at 430-31, is hard to understand.
n144. Gary L. Wolfstone, Miranda - A Survey of Its Impact, 7 Prosecutor 26, 27 (1971).
n145. Witt's article was published in 1973. Witt, supra note 26. He appears to have begun collecting his data
sometime after 1968.
n146. Id. at 325. It might be argued that all of the studies discussed so far were conducted too close to
Miranda to reveal police accommodation to the decision. However, in his 1987 article, Professor Schulhofer
claimed that the accommodation effect would manifest itself at least by the early 1970s. See Schulhofer, supra
note 28, at 456 ("By the early 1970s ... the view that Miranda posed no barrier to effective law enforcement had
become widely accepted ...."). In Schulhofer's current article, he places the date at which accommodation
became apparent at "since the mid-1970s." Schulhofer, supra note 2, at 509. If accomodation appeared by such
dates, it is hard to understand why the Stephens survey (1969-70), the Wolfstone survey (1970), and the Witt
Survey (1964-68) found such widespread concern about Miranda and why 22 state attorneys generally attacked
the decision in a 1976 Supreme Court brief. See supra note 133 and accompanying text.
n147. Letter from Darrel W. Stephens, Executive Director, to Patrick A. Trueman, Office of Liaison
Services, U.S. Dept. of Justice (Feb. 11, 1987) (on file with author). When asked "do you support efforts of the
Justice Department to have the Supreme Court reverse the Miranda decision?," seven answered yes, eight
preferred modification to reversal, and six answered no. Id. The pollster also reported that "two of the answers
recorded as yes, also fit the category of "prefer modification to reversal,' according to the comments they added
after initially answering." Id.
n148. Systematic surveys suggest also that confessions are important to law enforcement. See Stephens et
al., supra note 142, at 422 (Table VI); Witt, supra note 26, at 324 (Table I); Yale Project, supra note 23, at 1592
n.195.
n149. American Bar Assoc. Comm. on Criminal Justice in a Free Soc'y, supra note 5, at 9.
n150. The police respondents were asked whether they agreed with the following three propositions: First,
"Miranda warnings and motions to suppress are effective." Id. at 33. To what end Miranda was "effective" was
left undefined; thus this proposition would likely draw assent from even the strongest critics of Miranda
(perhaps especially the strongest critics). In any event, it is clear that assessing Miranda's effects by looking at
suppression motions dramatically understates Miranda's costs. See Cassell, supra note 1, at 391-94 (noting
problems with suppression rate analysis). Second, "Miranda warnings routinely prevent police from detecting or
solving crimes." American Bar Assoc. Comm. on Criminal Justice in a Free Soc'y, supra note 5, at 33. To my
knowledge, no one alleges that Miranda makes "detecting" crimes more difficult. While "solving" crimes is
much closer to the mark, it too fails to accurately capture all of Miranda's harmful effects. See Cassell, supra
note 1, at 397-98 (discussing screening of cases after police arrest a person, i.e., after police "solve" a crime).
Moreover, one could agree with my 3.8% Miranda cost figure and still conclude that Miranda does not
"routinely" affect cases. Third, "Problems with Miranda warnings routinely result in prosecutorial refusal to
accept cases or plea bargains." American Bar Assoc. Comm. on Criminal Justice in a Free Soc'y, supra note 5, at
9. Again, it is rarely problems with the Miranda warnings that lead to refusal to accept cases or plea bargains
(much less "routinely" do so); rather, it is the failure to obtain confessions in the first place that is our concern.

Page 31
90 Nw. U.L. Rev. 1084, *1124

See Cassell, supra note 1, at 393-94. For further criticisms of the survey, see Fred E. Inbau, A Flawed Survey
Regarding Miranda by an American Bar Association "Special Committee," 23 Prosecutor 7 (1990).
n151. For example, on the question regarding the effectiveness of Miranda suppression motions, on a scale
of 1 to 5 (1 being "strongly disagree" and 5 being "strongly agree") the mean police score was 3.39. Barbara E.
Smith, Telephone Survey of Criminal Justice Practitioners, Final Report to the American Bar Association
Criminal Justice Section 30 (Table 18) (1987) [hereinafter Smith, Telephone Survey], reprinted in Craig M.
Bradley, The Failure of the Criminal Procedure Revolution 43 (1993). This suggests that less than half of the
sample agreed with the proposition by giving a score of 4 or 5.
n152. Richard A. Leo, Police Interrogation and Social Control, 3 Soc. & Legal Stud. 93, 114 (1994);
Wermiel, supra note 132, at 72 ("We have literally a whole generation of police officers who have never known
police work except under Miranda.") (quoting a police representative).
n153. See, e.g., Leo, supra note 36, at 397 (noting that Miranda has contributed to improved perception of
police professionalism).
n154. See Smith, Telephone Survey, supra note 151, at 43 (Table 25) (most prosecutors estimate that 1% to
5% of their cases are dismissed due to Miranda problems). Drawing any conclusions from the ABA's study is once again - made perilous by the peculiar phrasing of the question. Prosecutors were apparently asked to
estimate the "percentage of cases dismissed due to Miranda problems" and may have been restricted to the
choices "none," "1-5 percent," "6-10 percent," or "11 percent or more." See id. at 43 (Table 25). Thus, it is
unclear exactly where a prosecutor would respond with a dismissal estimate corresponding to Schulhofer's
0.78%. But despite that glitch, it is quite likely that any cost estimate is far too low because it rests on cases
dismissed after filing, not cases never filed because of Miranda problems. The survey taker reported the
following: "Many prosecutors told us that [dismissed cases were] ... not a problem, because they do not file
cases in the first place where there is a Miranda ... problem." Id. at 42; see also supra note 150 (elaborating on
this point).
n155. Bradley, supra note 151, at 44. In discussing the study, Bradley misplaced a decimal in his
extrapolation. He concluded that the "disturbingly high" numbers of criminals released by Miranda might be as
high as 15,000. Id. But that number was derived not by taking 5% of the total number of criminals (as suggested
by the prosecutor's survey), but by taking only 0.5% of all criminals. See id. at 43 (estimating that number of lost
cases from Miranda for 3,000,000 crimes per year would be 15,000, 0.5% of the total). Thus, Bradley was
disturbed by an even smaller percentage of criminals, 0.5%, than indicated by Schulhofer's 0.78% figure.
n156. See Cassell, supra note 1, at 440-46.
n157. See Cassell & Hayman, supra note 35, at 907 (table 14); see also id. at 907-16 (reporting other data on
the importance of confessions).
n158. Thomas, supra note 35, at 939. Thomas noted that "the O.J. Simpson case is but one more example of
how a circumstantial case, even one that appeared overwhelming, can be undermined." Id.

Page 32
90 Nw. U.L. Rev. 1084, *1124

n159. Schulhofer, supra note 2, at 543.
n160. Id. at 541.
n161. Cassell, supra note 1, at 433. Professor Schulhofer reduces my confession necessity figure by, in large
part, including data from the New Haven study. As I explained initially, the authors of that study labeled a
confession in a case as "not important" even when no plausible investigative alternatives existed and the
evidence, besides the confession, was sufficient only to take the case to trial but not to obtain a conviction. See
Cassell, supra note 1, at 427 & n.235 (discussing Yale Project, supra note 23, at 1583). Professor Schulhofer
believes that this categorization is appropriate given that uncertainties always inhere in judgments about such
things. Schulhofer, supra note 2, at 541 n.172. But even putting the best face on it, the New Haven classification
can only be regarded as bizarre. The authors admitted that they were using standards that "were legal, i.e., taken
from case-book cases, not empirical." Yale Project, supra note 23, at 1582 n.167. Under those "legal" standards,
everything was swept into the interrogation "unnecessary" category where the evidence fell between the low
standard of surviving a directed verdict to the high standard of "conviction would almost certainly result absent a
strong defense by the suspect." See id. at 1582. This is a far cry from making a reasonable empirical judgment
about when a confession is "necessary" for a conviction.
n162. Id. at 544.
n163. See id. at 541-44.
n164. See Cassell, supra note 1, at 445 n.340.
n165. See Schulhofer, supra note 2, at 543 (assessing cases in which "prosecutors can use many sorts of
leverage to obtain guilty pleas in some cases counted as "lost' " but failing to assess parallel possibility of cases
in which defendants use their additional leverage - because of the lack of a confession - to obtain dismissals in
some of the cases counted as won). This omission is particularly glaring, since I outlined several reasons for
believing that the estimates of confession necessity might be too low. For example, the Pittsburgh study (relied
upon by both Schulhofer and me to determine when confessions were necessary for conviction) assumed
"cooperation by witnesses." Cassell, supra note 1, at 424 (citing Seeburger & Wettick, supra note 26, at 14). In
today's world of witness intimidation and citizen skepticism about the criminal justice system, such an
assumption will inevitably overstate the strength of the prosecution's case. See id. at 424 n.218 (citing study
proving many robbery cases are dropped because of some sort of witness problem); Paul G. Cassell, Balancing
the Scales of Justice: The Case for and the Effects of Utah's Victim's Rights Amendment, 1994 Utah L. Rev.
1373, 1410-11 (collecting sources on witness and victim intimidation).
n166. Stephen J. Schulhofer, Plea Bargaining as Disaster, 101 Yale L.J. 1979, 1984 (1992).
n167. See Cassell & Hayman, supra note 35, at 914 (table 16) (finding dismissal rate of 4.8% in cases when
questioning was successful versus a dismissal rate of 28.3% when questioning was unsuccessful); David W.
Neubauer, Confessions in Prairie City: Some Causes and Effects, 65 J. Crim. L. & Criminology 103 (1974)
(Tables 4 & 5) (finding a dismissal rate of 0% in property cases with a confession versus a dismissal rate of 13%
without a confession and not finding a significant difference in the dismissal rate in nonproperty cases
controlling for effect of trials).

Page 33
90 Nw. U.L. Rev. 1084, *1124

n168. See Cassell, supra note 1, at 442-44 (recounting data from "Prairie City," "Middle America," New
Haven, and Salt Lake City).
n169. The matrix is described and used in the New Haven study, see Yale Project, supra note 23, at 1583,
and is adopted in the Seaside City study, see Witt, supra note 26, at 324. Schulhofer relies on both of these
studies for his estimate of interrogation necessity. See Schulhofer, supra note 2, at 541 n.172.
n170. See Yale Project, supra note 23, at 1583 (Table 19) (emphasis added) (intersection of "none" and
"some" in amount of evidence columns and "available and probably adequate" row in investigative alternatives
column).
n171. See Cassell, supra note 1, at 468-69 (collecting evidence that some police agencies are now " "awash
in a deluge of violence' " (citation omitted)).
n172. Schulhofer, supra note 2, at 543.
n173. Id.
n174. See Cassell, supra note 1, at 446 n.346 (citing, inter alia, Stephen J. Schulhofer, Due Process of
Sentencing, 128 U. Pa. L. Rev. 733 (1980)).
n175. Id.
n176. Schulhofer, supra note 2, at 544.
n177. See Cassell, supra note1, at 444-45.
n178. See id. at 445 n.339 (recounting data from "Middle America," six scattered cities, the San Francisco
Bay Area, and "Prairie City").
n179. See supra notes 128-46 and accompanying text.
n180. See Cassell, supra note 1, at 448-49 & n.358 (54% of the variance explained by population).
n181. See id. at 419-21; Cassell & Hayman, supra note 35, at 876-80.
n182. Cassell, supra note 1, at 421.
n183. See supra note 154 and accompanying text; see also Cassell, supra note 33.

Page 34
90 Nw. U.L. Rev. 1084, *1124

n184. See supra note 154 and accompanying text.
n185. Cassell, supra note 1, at 486-99.
n186. Schulhofer, supra note 2, at 548.
n187. Yale Kamisar, The "Police Practice" Phases of the Criminal Process Revolution and the Three Phases
of the Burger Court, in The Burger Years 143, 150 (Herman Schwartz ed., 1987).
n188. Moran v. Burbine, 475 U.S. 412, 433 n.4 (1986).
n189. See Schulhofer, supra note 2, at 553.
n190. See Schulhofer, Reconsidering Miranda, supra note 28, at 436-55.
n191. Schulhofer does offer new discussion of Withrow v. Williams, 113 S. Ct. 1745 (1993), which refused
to extend the equitable rule of Stone v. Powell, 428 U.S. 465 (1976), to Miranda claims, largely on prudential
grounds. See Schulhofer, supra note 2, at 554. However, while Withrow says much about the breadth of the
equitable doctrine of Stone, it says little about the issue of concern here - the constitutional contours of Miranda.
That Withrow worked no sea change in current doctrine is made clear by the later decision of Davis v. United
States, 114 S. Ct. 2350 (1994), in which the Court restated the proposition that the Miranda questioning cut-off
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." Davis, 114 S. Ct. at 2355. Also perhaps new (or at least different
from some other defenses of Miranda) is Schulhofer's candid admission that Miranda constituted "a radical
break with prior precedent." Schulhofer, supra note 2, at 552.
n192. Grano, supra note 110; see also Joseph D. Grano, Miranda's Constitutional Difficulties: A Reply to
Professor Schulhofer, 55 U. Chi. L. Rev. 174 (1988).
n193. See Cassell, supra note 1, at 472 n.498.
n194. And perhaps in the view of others. See Michael Chertoff, Chopping Miranda Down to Size, 93 Mich.
L. Rev. 1713 passim (1995) (praising Grano's analysis on this point); Gerald Caplan, Book Review, 40 Wayne
L. Rev. 279 (1993) (same); Daniel Collins, Book Review, 1995 Pub. Interest L. Rev. 185 (same); cf. George C.
Thomas III, An Assault on the Temple of Miranda, 85 J. Crim. L. & Criminology 807 (1995) (reviewing Grano's
book and finding some of Grano's arguments hard to answer).
n195. Schulhofer, supra note 2, at 553.
n196. Grano, supra note 110, at 183.
n197. Id. at 198. For other arguments refuted by Grano, compare, e.g., Schulhofer, supra note 2, at 550

Page 35
90 Nw. U.L. Rev. 1084, *1124

(arguing that Griffin v. California, 380 U.S. 609 (1965), established a general principle that pressure on a
suspect equates with compulsion) with Grano, supra note 110, at 138 ("Schulhofer's argument ... overlooks that
Griffin involved that aspect of Fifth Amendment doctrine that governs the trial ... [and] the Court has since
limited Griffin to the trial context."). Compare also Schulhofer, supra, note 2, at 550 (wondering whether
waivers of Miranda rights include waivers of "protections against torture") with Grano, supra note 110, at 142
(criticizing Miranda because "while the notion of waiving a right of silence is intelligible, the notion of waiving
a right not to be compelled, especially when compel is a synonym for coerce, is not"). Schulhofer has responded
elsewhere to some of Grano's arguments. See Schulhofer, Reconsidering Miranda, supra note 28.
n198. Schulhofer, supra note 2, at 550.
n199. Akhil R. Amar & Renee B. Lettow, Fifth Amendment First Principles, 93 Mich. L. Rev. 857, 861
(1995); see also infra note 204 (discussing how Amar & Lettow's theory of the Fifth Amendment confirms the
constitutionality of my proposed alternative). Yale Kamisar wrote an extended response to Amar and Lettow,
but conceded that there are strange tensions in current doctrine. See Yale Kamisar, On the "Fruits" of Miranda
Violations, Coerced Confessions, and Compelled Testimony, 93 Mich. L. Rev. 929 passim (1995); cf. Akhil R.
Amar & Renee B. Lettow, Self-Incrimination and the Constitution: A Brief Rejoinder to Professor Kamisar, 93
Mich. L. Rev. 1101 (1995) (replying to Kamisar).
n200. Schulhofer, supra note 2, at 556.
n201. Id. at 558.
n202. New York v. Quarles, 467 U.S. 649, 655-56 (1984). Contrary to Schulhofer's intimations, Quarles is
routinely understood as allowing not only questioning without explicit waivers but also after suspects' attempts
to stop questioning. See, e.g., United States v. Mobley, 40 F.3d 688, 693 (4th Cir. 1994) (allowing custodial
questioning after invocation of right to counsel under "public safety" exception), cert. denied 115 S. Ct. 2005
(1995); United States v. DeSantis, 870 F.2d 536, 540-41 (9th Cir. 1989) (same); Trice v. United States, 662 A.2d
891, 894-95 (D.C. 1995) (same).
n203. Quarles, 467 U.S. at 658 n.7.
n204. Id. at 654-55 & n.5. Quarles thus stands squarely in the path of Schulhofer's proposed reading of the
Court's prophylaxis language as meaning "only that the specifics of the Miranda approach can be modified or
replaced if a state provides equivalent protection to the suspect." Schulhofer, supra note 2, at 554. The Court
required no such "equivalent protection" in Quarles.
n205. Cassell, supra note 1, at 477-78 (internal quotation omitted).
n206. Stephen J. Schulhofer, Confessions and the Court, 79 Mich. L. Rev. 865, 880-84 (1981).
n207. Id. at 892.

Page 36
90 Nw. U.L. Rev. 1084, *1124

n208. Id. at 882.
n209. See Kamisar, supra note 199, at 934 n.19 ("Many commentators - and I am one of them - consider
[the failure to require recording of interrogation] a serious weakness in Miranda.").
n210. Amar & Lettow, supra note 199, at 873. While they have in mind a model of in-court depositions, id.
at 908, they explain that other "more relaxed schemes are also compatible with our approach," id. at 909. Indeed,
they outline a modified warning regime as an alternative to Miranda in which a suspect would be affirmatively
told that his silence might be used against him. Id.
n211. Kamisar, supra note 199, at 933-34. Although Kamisar's preference would be to reinforce the
Miranda rules with an electronic recording requirement, he reports that if, in the years immediately preceding
Miranda he were offered the choice between (1) my proposed system and (2) the Miranda doctrine, which, at
least as it has been interpreted by the Burger and Rehnquist Courts, allows the police to obtain waivers of a
suspect's rights without the presence of any disinterested observer and without any objective record of the
"waiver" transaction (even when a tape recording is feasible), he would have chosen my system. Letter from
Yale Kamisar, Univ. of Michigan, to Paul Cassell, Univ. of Utah (Sept. 11, 1995) (on file with author)
[hereinafter Letter from Kamisar to Cassell]. As he noted the year before the Miranda case, "in the long run, no
statute, court rule or court decision pertaining to warnings or waivers will suffice - for the same reason that the
flood of appellate opinions on "involuntary' confessions has not sufficed - until police interrogation is stripped of
its "most unique feature ... its characteristic secrecy.' " Yale Kamisar, Equal Justice in the Gatehouses and
Mansions of American Criminal Procedure, in Crim. Just. in Our Time 85-86 (A.E. Dick Howard ed., 1965)
(quoting Bernard Weisberg, Police Interrogation of Arrested Persons: A Skeptical View, in Police Power and
Individual Freedom 153, 179 (Claude Sowle ed., 1962)). But Kamisar believes that by the early 1960s the
"voluntariness"-"totality of the circumstances" test was on its way to becoming a formidable restriction on police
interrogation methods, see Yale Kamisar, Remembering the "Old World" of Criminal Procedure: A Reply to
Professor Grano, 23 U. Mich. J.L. Ref. 537, 572-75 (1990). Kamisar also believes that this development would
have accelerated greatly once judges were able to listen to tape recordings of the actual interrogation (from the
very beginning to the end) rather than base their opinions on "sanitized" police versions of what occurred in the
interrogation room. Letter from Kamisar to Cassell, supra.
n212. Schulhofer, supra note 2, at 557.
n213. See, e.g., Model Code of Pre-Arraignment Procedure 130.4 (Proposed Official Draft 1975); see also
Harold Rothwax, Guilty: The Collapse of the Criminal Justice System 66-87, 237 (1996) (criticizing Miranda
and arguing for videotaping as a substitute).
n214. Schulhofer, supra note 2, at 557. Professor Schulhofer implies that I believe unconstitutional coercion
must involve physical brutality. Id. at 550-51. But in my earlier article I noted specifically that "coercive police
questioning can involve not only police brutality but also other techniques as well." Cassell, supra note 1, at 475.
n215. See Schulhofer, supra note 2, at 560 (concluding that "Miranda assuredly was a compromise, because
its rules permit some compelling pressures to continue ...."); Schulhofer, supra note 206, at 880-81 (concluding
that "when the suspect ... remains swept from familiar surroundings surrounded by antagonistic forces, and
particularly when the all-important warnings are delivered by those same antagonistic forces, it is hard to see
how the intimidation can be reduced very much") (internal quotations omitted).

Page 37
90 Nw. U.L. Rev. 1084, *1124

n216. Schulhofer, supra note 2, at 557.
n217. Id. at 559; cf. Schulhofer, supra note 206, at 877-78 (criticizing book by Yale Kamisar for tending "to
reinforce impressions often conveyed elsewhere that Miranda marked the death of the due process test and that,
at least for the time being, it remains buried. [Under Miranda] careful attention to the voluntariness issues
remains an imperative, though sometimes overlooked, obligation of court and counsel."). Of course, even after a
waiver, if a suspect asks for a lawyer or for questioning to stop, the Miranda questioning cutoff rules apply.
However, the limited empirical evidence on this point suggests that such post-waiver invocations are relatively
unusual. See infra note 230.
n218. Fred P. Graham, The Self-Inflicted Wound 316 (1970) (quoting Inbau and Reid manual). Indeed,
some have crafted an argument that, by keeping the details of police interrogation free from judicial scrutiny,
Miranda has allowed police to develop more subtle and coercive psychological tactics than Inbau and Reid could
describe in the 1960s. See Richard A. Leo, From Coercion to Deception: The Changing Nature of Police
Interrogation in America, 18 Crime, L. & Soc. Change 35, 54 (1992) (arguing that "police power in interrogation
has become more subtle, more indirect, more manipulative, and more likely to be effectuated through
psychological processes").
n219. Schulhofer, supra note 2, at 557.
n220. See Fred E. Inbau et al., Criminal Interrogation and Confessions 216-19 (3d ed. 1986) (explaining
that "the vast majority of criminal offenders are reluctant to confess and must be psychologically persuaded to
do so" and then discussing "permissible interrogation tactics and techniques" without reference to Miranda
rules). See generally Stephanie Simon, Getting Suspects to Confess: How Far Can Police Go?, L.A. Times, Sept.
21, 1995, at A1 (discussing lack of clear guidance to police officers on permissible interrogation tactics and
concluding that the case law offers "a confusing, and sometimes contradictory, jumble of rules").
n221. Cassell, supra note 1, at 486-89; accord Schulhofer, supra note 2, at 556 ("No doubt a videotaped
record would often prevent police abuse and manipulation of the swearing contest.").
n222. Cassell, supra note 1, at 488-89.
n223. Richard Jerome, Suspect Confessions, N.Y. Times Mag., Aug. 13, 1995, at 28, 30.
n224. Schulhofer, supra note 2, at 556.
n225. See id. at 556-57.
n226. See Cassell, supra note 1, at 495 n.623 (reviewing available evidence).
n227. To be sure, Miranda may confer some modest benefit on suspects even after a waiver through its
questioning cut-off rules. See id. at 432 n.267. But the benefits of videotaping would be more than adequate
compensation.

Page 38
90 Nw. U.L. Rev. 1084, *1124

n228. Miranda's costs appear to be concentrated in this group of suspects. See id., at 492-96.
n229. See Cassell & Hayman, supra note 35, at 859-60 (finding that of suspects questioned, 16.3% invoked
initially and 3.9% during questioning); Leo, supra note 36, at 262 (finding that of suspects questioned, 19.8%
invoked initially and 1.1% invoked during questioning).
n230. Thomas, supra note 194, at 819.
n231. 384 U.S. at 533-34 (White, J., dissenting). Professor Thomas discussed the problem exposed by
Justice White at length and concededly without effective response. Thomas, supra note 194, at 819-27.
n232. Grano, supra note 110, at 219 (quoting Caplan, supra note 110, at 1469); see also William T. Pizzi,
The Privilege Against Self-Incrimination in a Rescue Situation, 76 J. Crim. L. & Criminology 567, 595 (1985)
(noting "the folly of approaching the scope of the Fifth Amendment solely from the defendant's point of view
while totally ignoring ... the purpose and function of the police conduct").
n233. See Schulhofer, supra note 2, at 557 (contending that such cases are a "suitable illustration of the
difference between the Miranda safeguards and those that Cassell would put in their place"). The available
empirical evidence suggests that Schulhofer is talking here about less than 4% of all suspects. See supra note
229.
n234. Under current doctrine, if the suspect says the wrong magic words, "I might want a lawyer,"
questioning can continue without a videotape recording. Davis v. United States, 114 S. Ct. 2350 (1994). Also
under current doctrine, if the suspect says the magic words, "I absolutely want to invoke my right to remain
silent," questioning can continue at a later point, again without a videotape recording. See Michigan v. Mosley,
423 U.S. 96 (1975).
n235. Phillip E. Johnson, A Statutory Replacement for the Miranda Doctrine, 24 Am. Crim. L. Rev. 303,
310 (1987).
n236. See id.
n237. See, e.g., Minnick v. Mississippi, 498 U.S. 146, 151 (1990) (concluding that benefits from Miranda's
questioning cut-off rule "have been thought to outweigh the burdens [it] ... 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") (internal quotation omitted).
n238. The public's elected representatives have expressed their view that Miranda's prophylactic rules are
unjustified. See Davis v. United States, 114 S. Ct. 2350, 2358 (1994) (Scalia, J., concurring) (arguing 18 U.S.C.
3501 "reflects the people's assessment of the proper balance to be struck between concern for persons
interrogated in custody and the needs of effective law enforcement"); see also supra notes 111-12 and
accompanying text (discussing 3501).