Available via license: CC BY-NC 4.0
Content may be subject to copyright.
Accepted Manuscript
© Sleep Research Society 2021. Published by Oxford University Press on behalf of the Sleep
Research Society.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-
NonCommercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-
commercial re-use, distribution, and reproduction in any medium, provided the original work is
properly cited. For commercial re-use, please contact journals.permissions@oup.com
Sleep and Interrogation: Does Losing Sleep Impact Criminal History Disclosure?
Zlatan Krizan
Anthony J. Miller
Christian A. Meissner
Iowa State University
Zlatan Krizan, Ph.D.
Department of Psychology
Iowa State University
(515)294-1975
zkrizan@iastate.edu
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Abstract
Study Objectives: Despite centuries of using sleep deprivation to interrogate, there is
virtually no scientific evidence on how sleep shapes behavior within interrogation settings.
To evaluate the impact of sleeplessness on subjects‟ behavior during investigative interviews,
an experimental study examined the impact of sleep restriction on disclosure of past illegal
behavior. Methods: Healthy participants from a university community (N=143) either
maintained or curbed their sleep (up to 4 hours a night) across two days with sleep monitored
via actigraphy. They were then asked to disclose past illegal acts and interviewed about them.
Next, they were re-interviewed following an example of a detailed memory account (model
statement). Disclosures were blindly coded for quantity and quality by two independent
raters. Results: Sleep-restricted individuals reported similar offenses, but less information
during their disclosure with slightly less precision. Model statement increased disclosure but
did not reduce the inhibiting impact of sleep loss. Mediation analysis confirmed the causal
role of sleep as responsible for experimental differences in amount of information, and
subjects‟ reports suggested impaired motivation to recall information played a role.
Conclusions: The findings suggest that even moderate sleep loss can inhibit criminal
disclosure during interviews, point to motivational factors as responsible, and suggest
investigators should be cautious when interrogating sleepy subjects.
Keywords: Sleep, Interviewing, Interrogation, Disclosure, Model Statement
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Statement of Significance
Thousands of investigative interviews are conducted every day by public safety, law-
enforcement, and military organizations. Historically, sleep disruption has been used in these
contexts as a tool to compel disclosure or confessions. Despite frequently imposed sleep
disruption on detainees and commonly experienced sleep loss among interview subjects such
as victims or witnesses, there is no direct scientific evidence on how sleep shapes intelligence
disclosures during investigative interviews. This experimental study evaluated how moderate
self-imposed sleep-restriction impacted criminal history disclosure during a laboratory
interview about past illegal acts, finding that sleep-deprived subjects provided substantively
less information with a trend for them to report less motivation to recall information. Asking
for a second disclosure (following an example of the desired level of detail) increased the
amount of information provided overall, but did not reduce the suppressive impact of sleep.
The findings carry direct implications for science and practice of investigative interviewing
and point to the importance of sleep for long-term memory retrieval.
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Introduction
Historical and journalistic accounts reveal persistent and ongoing use of sleep
deprivation as a means of “breaking the resistance” of uncooperative interrogation subjects.1
Khalid Sheikh Mohammed (KSM), who planned the 2001 attacks on the United States, was
detained and subjected to 132 hours of continuous sleep deprivation.2 Even when not
imposed on detainees, sleep disruption is endemic to custodial environments across military
and law-enforcement sectors.3 Sleep disruption is also inevitable among various subjects of
investigative interviews, as police often interview suspects, victims, and witnesses at night or
following traumatic events that disrupt sleep.4-5 As a result, a significant number of
investigative interviews are certain to involve sleep-deprived subjects.
The lengthy interrogations themselves can produce sleep disruption, and may be
intentionally aimed at wearing down suspects. A survey of more than 600 law-enforcement
investigators indicated their longest interrogation to last 5 hours, on average, ranging up to 72
hours.4 Inducing tiredness alongside other types of discomfort is common to accusatorial
interview approaches frequently taught to investigators, which contributes to lengthy
interviews.6-8 Military investigators have also reported intentional deprivation of sleep among
detainees in service of ensuring compliance.9
The prevalence of sleep deprivation as a tactic to obtain information is based on the
premise that disrupting sleep will ultimately increase the amount of accurate and useful
information that a subject is otherwise unwilling to provide.10 While sleep deprivation has
also been involved in coercive approaches aimed to ensure compliance or force
confessions,8,9 the key question for investigators and those collecting human intelligence
involves efficacy of the tactic and the value of the information obtained. Although there is
little direct scientific evidence regarding these assumptions, behavioral evidence that bears on
this question suggests that sleep-deprived interview subjects should be less likely to provide
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
reliable human intelligence. First, sleep-deprived interview subjects may be more cognitively
impaired, with problems in sustaining attention on the task at hand, recalling relevant
information, or understanding situational demands.11-13 Second, sleep-deprived subjects may
be more socially withdrawn, namely less attentive to social cues, less motivated to meet
interview demands, and more socially avoidant or disfluent.14-16 Finally, sleep-deprived
interview subjects may be more emotionally distressed, with increased stress responses,
anxiety, anger, fatigue, and behavioral unpredictability.17-19 These observations are
inconsistent with the premise that sleep deprivation helps increase (accurate) disclosure, even
if subjects show apparent behavioral compliance with interrogators (e.g., falsely confess).20
In this vein, research targeting the role of sleep in investigative settings suggests that
after about two days without sleep, individuals can become more vulnerable to leading
questions, more likely to change their answers when contradicted, and show more confidence
in false memories.21-23 To examine confessions, one experiment compared students who were
typically rested to those who stayed up all night in their willingness to sign a (false) statement
acknowledging they broke an important rule of the study (all were innocent) – 50% of sleep-
deprived individuals signed a false statement on the first request, compared with only 18% of
their better-rested counterparts.24 Another investigation examined circadian factors, finding
that individuals were more likely to admit to wrongdoing when time-of-day was mismatched
with their chronotype, such that evening-oriented individuals confessed to more offenses in
the morning than at night, while morning-oriented individuals admitted to more offenses at
night than in the morning.25 These findings suggest that total sleep deprivation (24 hours or
more) or circadian misalignment could render individuals more responsive to social
pressures, namely susceptible to social contamination of memory and false confessions. In
support of this conclusion, a recent analysis of 600 Palestinian ex-detainees revealed that
those who reported significant sleep deprivation during detention produced a higher number
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
of confessions with either true or false statements, questioning the efficacy of sleep loss in
extracting the truth. Moreover, the higher number of confessions did not result in more
convictions or longer prison sentences, further questioning the utility of sleep loss as means
to ensure prosecution.3
However, this evidence does not address whether the amount and detail of
information individuals are likely to provide during investigative interviews is affected by
sleep loss. Instead of confessions, most investigative interviews seek to obtain information or
intelligence that is detailed, relevant, and can be corroborated. In words of one military
investigator, “If the bad guy does not want to admit he is bad but wants to share reliable
information in order to try and convince me he is a good guy, then it would be foolish of me
to focus on his guilt”.9 In everyday law-enforcement contexts, investigative interviews
frequently involve reluctant witnesses, embarrassed victims, and confidential informants,
where the goal is always obtaining information instead of a confession. Even co-operative
interviews can involve lost sleep, and losing information in such cases can exert a significant
influence on investigations. Regardless of its veracity, obtaining more intelligence (i.e.,
details, names) is also critical to evaluating credibility of information provided by interview
subjects as well as detecting deception.1 For example, it is easier to spot a false alibi when
accounts provided by individuals need to include a lot of detail and context, because liars
usually produce fewer details.26
Finally, it is not clear whether more common, real-world doses of sleep loss impact
disclosure and behavior during investigative interviews. Because total sleep deprivation has
been the exclusive focus of the few experimental studies addressing interrogation behavior, to
what extent those effects carry over to the more commonly experienced partial sleep
restriction is unknown. Given most interviews occur with subjects that have had at least some
sleep, it is critical to examine the impact of more ecologically-valid levels of sleep loss.
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Study Purpose
The purpose of the present experiment was thus to examine the causal impact of
sleep-restriction on the disclosure of past criminal behavior within an interview setting. To
our knowledge this is the first experiment to examine the impact of sleep loss on gathering
human intelligence about past crimes. To this end, it tested the impact of at-home sleep
restriction among 143 members of a university community on the quantity and quality of
information disclosed during a criminal history interview with an experimenter. While the
participants did not face legal risk, they nevertheless discussed embarrassing or guilt-
inducing illegal behaviors – important obstacles to disclosure among actual victims,
witnesses, and suspects. Following their initial disclosure, interview subjects also listened to
a „model statement‟, a detailed memory account of an unrelated event known to increase
information yield.27 Subjects then provided information about this event again. As mentioned
earlier, eliciting more details during an interview is critical both for amassing facts as well as
for distinguishing deceptive from honest accounts. In this case, a model statement
intervention was used as it might reduce any deleterious influence of sleep restriction, namely
differences in information yield due to sleep loss could be ameliorated following exposure to
a detailed model statement. Alternatively, sleep loss could exert a more pervasive effect,
affecting the amount of information similarly across initial and post-model-statement
disclosures.
To evaluate intelligence yield from interview subjects, blind raters listened to audio
recordings, counted actionable pieces of intelligence (who, what, where, when, why, how)
and rated their level of quality (detail and precision). Moreover, participants self-reported on
a variety of interview experiences. Our primary hypothesis was that sleep-restricted
individuals would disclose less information than their more rested counterparts, given that
even moderate sleep loss increases cognitive impairment, social withdrawal, and emotional
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
distress, reactions likely to interfere with efficacious interviewing and provision of human
intelligence11-19. Note that all interviews focused on events that occurred prior to the study;
hence, any effects of sleep restriction can only be attributed to how people recall or report
episodic memories, and not how well they encoded or consolidated such memories. While
there was no access to ground truth, random assignment eliminated any confounds with
motivated distortion or pre-existing memory differences across conditions. Finally, any effect
of the model statement on disclosure can only be attributed to motivational or practice
factors, as cognitive competency (e.g., working memory) should not be affected by a simple
instruction to repeat retrieval. As a result, should the model statement qualify the impact of
sleep restriction on disclosure (e.g., compensate for the hypothesized lower disclosure),
motivational rather than competency processes would be implicated.
Methods
Participants
Due to lack of prior research to suggest effect size, we aimed to capture at least a
moderate effect of d = .42, the average effect size in experimental behavioral research (from a
meta-analysis of 177 randomly selected experiments).28 A power analysis using G*Power 3.1
for a two-tailed t-test suggested 140 participants would be sufficient to ensure .80 power for
detecting this difference. Furthermore, for a two-factor mixed design targeting a within-
between-subjects interaction with 160 participants would be sufficient to ensure .80 power
for detecting even a small interaction effect approximating d = .20, assuming .60 correlation
between repeated observations and .05 level of significance.29 Findings are interpreted with a
focus on effect sizes, confidence intervals, and practical differences, rather than arbitrary p-
value thresholds for rejecting implausible null hypotheses.30
A total of 143 participants from a large Midwestern university and the surrounding
community were recruited to participate in a study on “sleep restriction and interviewing in
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
healthy adults.” They were randomly assigned (using a randomly generated list of two values
each representing one condition) to 2 × 2 mixed design with sleep restriction manipulated
between-subjects, and a within subject manipulation where a free recall interview was
followed by introduction of the model statement and a second recall attempt. Prior to
participating in the study, all interested parties were screened over the telephone. To
minimize risks from participating in the study, individuals who self-reported a diagnosis of
(1) sleep, mental, or a physical illness, (2) overnight work, or (3) sleeping less than 6 hours a
night on average were not invited to participate. All participants provided signed informed
consent which informed them that any disclosures will be confidential and will not result in
legal risk. The sample was 63% male, had an average age of 21.81 (SD = 4.16, range 18–69),
and 71% of participants identified as white. All research procedures were approved both by
the Iowa State University Office for Responsible Research and the Federal Bureau of
Investigation‟s Internal Review Board, with research conducted according to relevant
guidelines and regulations. The analyses were not formally preregistered. Relevant research
materials, coding instructions, stimuli, and de-identified data can be found on
https://osf.io/yqa4f/.
Procedures and Measures
Participants came to the lab twice. During their initial session they signed informed
consents, were randomly-assigned to a sleep restricted or a control condition, and then
completed several survey measures on personality and typical sleep patterns, as well as
several unrelated computer tasks (either an emotion suppression task or an empathy task that
required identifying basic emotions, subjects of separate research inquires). These measures
were delivered via the Qualtrics Survey® platform; they lasted around 30 minutes and are not
discussed further.
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Based on random assignment, the experimenter then created a “target” sleep schedule
for each participant. If the participant was in the control condition, the experimenter
instructed the participant to adhere to their expected sleep and wake times for those particular
days. First, we asked each participant to report their expected bed time and rise time for the
next two days. If the participant was in the sleep-restriction condition, the experimenter
provided a new target sleep time two hours later than originally expected by the participant,
as well as a new wake time two hours earlier than originally expected by the participant.
Participants in the sleep-restricted condition were instructed to adhere to these new sleep and
wake times as best they could and to avoid naps, alcohol, caffeine, or psychoactive drugs.
This procedure encouraged participants to remove up to eight hours from their typical sleep
across two days and has been found effective in prior research.18 To help participants wake-
up at their designated wake times, they received an automated phone call at their scheduled
wake up time. Participants were compensated with $50, and better adherence to the assigned
sleep-wakes schedule yielded compliance bonuses ($5 for each hour awake assigned between
10pm and 6am).
After establishing their schedule with the experimenter, participants received the
Actiwatch Spectrum Pro (by Phillips Respironics), which they wore on their non-dominant
wrist until their return to the laboratory. This device recorded movement at 30 second epochs
to estimate sleep–wake state. Although sleep–wake state as recorded by the Actiwatch highly
converges with polysomnographic recordings, all recordings were visually inspected by the
second author to increase accuracy.31 If necessary, sleep–wake recordings were manually
adjusted on the basis of participant-entered sleep and wake times (using Actiwatch buttons),
alongside recorded light or movement (by the second author).32 Daytime naps were not
included in sleep calculations, but were extremely rare. Actiwatch data on sleep duration
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
served as a manipulation check and were used as an intervening factor in a causal mediation
analysis.
Criminal History Disclosure
Following the two nights at home, participants returned to the laboratory and returned
the Actiwatch. Using an established paradigm,33 participants were then given an opportunity
to disclose and discuss past illegal behaviors. First, individuals were escorted to a private
room and provided with a sheet of paper (without their name) that listed 20 criminal
behaviors in roughly ascending order of legal severity (ranging from transporting fireworks
and trespassing, to shoplifting and driving under the influence; see34 and Supplemental
Online Materials). They were asked to indicate whether they had engaged in each of these by
checking „yes‟ or „no‟. The experimenter then obtained the information sheet from the
participants and initiated a brief investigative interview.
The interview room was of moderate-size, and involved both the experimenter and a
camera facing the participants across a larger table. The experimenter then assured the
participant of confidentiality and asked the participant to recall a specific time that they
committed the most severe crime they disclosed (the highest number the participant checked
on the sheet), asking “would you be willing to talk to me about that time?” If the participants
declined (or did not admit to any illegal behaviors on the list), they were asked if they were
willing to discuss a time when they wronged someone (even if not breaking the law).
If the participant agreed, they were asked to describe “a specific instance when you
did this behavior” and then “to provide as many details as you can remember about the
event”. Participants then described this event to the experimenter while being video recorded
(though only audio was stored and used for coding purposes). All experimenters were female
(with the exception of one session), wore white lab coats throughout the session, and were
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
trained to read from a script during the interview procedures while maintaining a neutral, yet
serious, demeanor.
After providing their initial disclosure of the event, participants were told that prior
research has shown that providing a model of the level of detail required was helpful to
interviewees. They were then asked to listen to a “model statement,” an audio recording of a
man recounting his experience at the Iowa State Fair in prodigious detail (adapted from27).
Participants were then prompted to recount their story again, with a focus on providing a
higher level of detail as illustrated by the model statement they just heard. No additional
prompts were used.
Audio recordings were edited to create separate sets of (1) initial disclosures and (2)
post-model-statement disclosures. These two sets of recordings were provided to two raters
not previously involved in the study and blind to experimental conditions as well as
hypotheses. Each rater coded all of the disclosures in a random order on both quantity and
quality of information (per33). To capture the quantity of information, each rater counted the
number of verifiable details (potentially falsifiable by independent evidence) obtained during
the disclosure, namely who, what, where, when, why, and how (thus ranging 0-6). Even if
multiple people or locations may have been involved, specifying only one of the relevant
elements was sufficient to count the presence of the detail. To capture quality of information,
raters evaluated the precision of each detail by assigning values between 1 (not explicit, have
to infer) and 3 (additional context or specificity). For example, for time of the event
(“when”), raters indicated whether only season or time of day was reported (1), month or
hour was reported (2), or the exact date and time were reported (3). Failing to report a detail
automatically resulted in a score of „0‟. Inter-rater agreement was evaluated for both quantity
and quality of information by calculating agreement correlations across both raters‟ sets of
judgments (number and precision of details provided, respectively) separately for those taken
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
before and after the model statement. Overall agreement rates for the number of details
provided averaged 89.2% (ranging from 77.2% to 98.7%). The rated number of details
provided by the two independent raters correlated .71 and .70 across the initial and model
statement disclosures, respectively. The level of precision correlated between .47 and .74
across various six details for initial disclosures, and between .43 and .66 for repeat
disclosures. There was thus acceptable to strong agreement between the raters in
distinguishing one event account from another35. The precision of “what” and “how”,
arguably the most complex details, yielded the least agreement. Because independent sets of
raters evaluated pre- and post-model statement accounts, it was not possible to evaluate the
number of novel details provided the second time.
Interview Reactions
Following the recorded interview, participants completed computerized survey
measures regarding their interview experience. These included (1) perceived treatment by the
interviewer (e.g., professionalism), (2) impressions of the interviewer (e.g., likability), (3)
subjects‟ level of co-operation (e.g., how much they disclosed and how much they resisted),
(4) motivation and effort when disclosing information, (5) meta-cognition regarding the event
(e.g., how strong they felt their memory of the event was, did they recall verbatim or gist
content), and (6) emotions about the event (e.g., guilt). All items were rated on 7-point
Likert-type scales (see Supplemental Online Materials for exact wording of all items).
Participants also indicated their perceived disclosure, by reporting “How much information
do you think you provided to the interviewer,” rated on a 1 (None at all) to 7 (A lot) Likert-
type scale. Finally, participants were then debriefed, offered transportation, and paid for their
participation.
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Results
Data Exclusions and Manipulation Checks
Data were first inspected for completeness. Across all participants, 8 were removed
because they did not disclose any illegal or antisocial behavior to be interviewed about, and
17 additional participants were removed because the interview recording wasn‟t available due
to equipment failure, refusal to discuss the event, or unexpected errors (e.g., discussing an
event different from one reported on the checklist, or lack of condition information in one
case). This resulted in the final sample size of 118 participants (58 in restricted, and 60 in the
control condition, with 11 and 12 participants removed from each condition, respectively, but
note 4 with no condition data). Note a slightly higher number for analyses involving self-
reported data not contingent on video recordings or lower for analyses involving model
statement disclosures. All remaining participants were included in the experimental analyses
regardless of their (varying) compliance with the sleep schedule, as to minimize data loss and
avoid endogenous selection biases in estimates of experimental effects (see).36
Sleep Restriction Effectiveness
Examination of actigraphic recordings indicated that participants in the control
condition slept an average of 422 and 406 minutes on the first and second nights,
respectively, whereas those who restricted slept an average of 269 and 292 minutes on the
first and second nights, respectively. This yielded an average nightly sleep duration of 6.9
hours for the control condition (4.4 to 10.2 hours), and 4.6 hours (2.4 to 7.8 hours) for the
restriction condition, an overall difference of 4.4 hours, t (118) = -10.18, p < .001, 95% CI (-
317, -214 minutes), d = -1.87. In brief, the average sleep-restricted participants lost little
more than half a night of sleep across the two days of the study. As would be expected given
increased sleep pressure among the restricted participants, they also took fewer minutes to
fall asleep (Sleep Onset Latency, M = 15.1; SD = 14.8) than did control participants (M=24.0;
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
SD = 20.2), t(113) =-2.67, p = .009, d = .50. Sleep-restricted participants spent slightly fewer
minutes awake after sleep onset (Wake-After-Sleep-Onset, M = 30.4; SD = 34.3) than did
control participants (M=45.1; SD = 45.8), t(113) =-1.93, p = .055, d = .36.
Finally, there were no systematic difference in timings of interview sessions, with the
average timing of interviews for the restricted participants at 12:40pm (SD = 2:30 hours) and
for control participants at 1:13pm (SD = 2:11 hours), t(115) = -1.27, p = .207. Sessions
occurred between 8:30am and 5:00pm.
Types of Disclosed Offenses
Absolute frequencies of specific offenses selected for interviews (reported across
experimental conditions) appear in Figure 1. Sixteen out of 20 listed offenses were selected
for interviews at least once. Using alcohol before 21 years old (#4), experimenting with
illegal drugs (#11), and driving under the influence (#19) were the most common. Overall,
the patterns of criminal history disclosure were nearly identical across the conditions and
typical for the sampled population (young adults).
The Impact of Sleep Restriction on Disclosure
To examine differences in information yield as a function of sleep-restriction and the
model statement, raters‟ aggregated scores for quantity of information (the number of details
disclosed) and quality of information (averaged ratings of precision) were each submitted to a
2 (sleep-restriction, manipulated between-subjects) by 2 (model-statement, manipulated
within subjects) mixed ANOVA. The results appear in Figure 2, while descriptive statistics
and correlations between the key measured variables appear in Table 1.
Compared to the number of details disclosed by participants who followed their
typical sleep schedule (M = 5.00 details, SD = 1.04), sleep-restricted participants disclosed
fewer details on average (M = 4.67 details, SD = 1.04), F (1, 109) = 4.49, p = .036, d = .32,
95% CI [-.63, .11]. In an absolute sense, sleep-restricted participants provided 7% less
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
information (across both disclosures) than their more rested counterparts. Participants also
disclosed more information following the model statement (M = 5.18; SD = .91) than before
(M = 4.50; SD = 1.04), F (1, 109) = 40.97, p < .001, d = .70, 95% CI [-.97, -.42]. In an
absolute sense, participants provided 15% more information following the model statement
than during their initial disclosure. There was little evidence that administering the model
statement moderated the impact of sleep restriction, Finteraction (1, 109) = .341, p = .561. As
evident by individual means, the impact of sleep restriction on information quantity was only
slightly higher before the model statement (7% less information) than after it (5% less
information). Overall, these differences have clear practical significance. If an investigator
interviewed ten people who lost a similar amount of sleep as participants in this study
(relative to ten more rested people), she could expect 3 fewer pieces of critical information
about the case in terms of who, where, when, what, why, or how the crime was committed.
Similarly, a repeated disclosure following a clearer standard of desired detail (i.e., the model
statement) could lead to 6 or more pieces of information relative to the original disclosure
(regardless of sleep restriction). To explore whether particular pieces of information may be
more affected, we evaluated frequency of given types of detail across the conditions. These
non-parametric tests (not corrected for multiple comparisons) suggested sleep-restricted
individuals reported fewer „when‟ details in their original disclosures (p = .053) and fewer
„where‟ details in the second disclosures (p = .029, complete results appear in the
Supplementary Materials).
Compared to precision of information disclosed by participants who followed their
typical sleep schedule (M = 1.37, SD = .33), sleep-restricted participants did not provide
substantially less precise information, on average (M = 1.31, SD = .38), F (1, 109) = 1.00, p =
.32, d = .17, 95% CI [-.59, .17]. Participants disclosed appreciably more precise information
following the model statement (M = 1.48; SD = .40) than before (M = 1.21; SD = .31), F (1,
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
109) =58.85, p < .001, d = .77, [-1.0, -.46]. As with quantity of information, there was little
evidence that the model statement qualified any impact of sleep restriction, Finteraction (1, 109)
= .119, p = .730. When contrasted with the effect on the number of details, the precision of
those details was thus less sensitive to sleep restriction, but slightly more sensitive to repeated
disclosure following the model statement.
In contrast, when asked for their own subjective assessment of how much information
they provided to the interviewer overall, sleep-restricted participants tended to reportt
providing slightly more information (M = 5.41, SD = 1.17), than those in the control
condition (M = 5.12, SD = 1.17), t (116) = 1.38, p = .17, d = .25, 95% CI (-.13, .72), although
this difference did not reach significance. At the same time, there was some evidence that
participants invested less effort during their disclosure, with tendencies to report less
motivation to remember (d = .35, p = .06) and more effort needed to remember the event (d =
-.29, p = .12). Taken together, while not reaching conventional standards of statistical
significance, the similar differences across several queries suggest that sleep loss may have
hampered the motivation to remember actual details or made it more difficult, which may
have contributed to less disclosure. In this vein, subjects‟ reports of impaired motivation and
effortfulness did exhibit correlations both with sleep duration and quality of details disclosed
in the hypothesized directions (see Supplemental Online Materials for all differences in self-
reports and associated confidence intervals).
Sleep Duration as a Causal Mediator of Experimental Effects on Disclosure
To directly examine whether sleep duration was responsible for the observed
experimental differences in quantity of disclosure, we tested a causal mediational chain with
experimental condition as the independent variable, actigraphically-recorded sleep duration
as the mediating variable, and quantity of information during the initial disclosure as the
dependent variable (using PROCESS v.3 MACRO for SPSS that utilizes 5,000 bootstrap
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
samples, Hayes, 2018).37 The results of this analysis appear in Figure 3 and indicate a
singular role of sleep duration in explaining the experimental effect of sleep restriction on the
initial number of details (R2 = .07, p = .03). Specifically, there was no significant direct effect
on disclosure but only a substantial indirect effect, with sleep duration mediating the impact
of experimental manipulation on the number of details reported, IE = .43, 95% CI (.09, .78).
This provides strong evidence that restricted sleep, rather than reactions to the sleep-
restriction protocol independent of lost sleep (e.g., frustration about changing daily routine),
is responsible for inhibited disclosure during investigative interviews the next day.
Discussion
These findings are the first to address whether sleep loss causally influences
intelligence yield from investigative interview subjects. By imposing moderate and common
levels of sleep loss (losing a half-night of sleep over two days), the findings speak to levels of
sleep disruption that professional investigators often encounter in their interview subjects.
The findings indicated that losing even 4-5 hours of sleep over two days suppressed the
amount of information that subjects provided, both initially (7% less) and during a repeat
disclosure (5% less). A causal mediation analysis confirmed the role of shortened sleep
duration as responsible for these experimental differences. While these percentages should be
taken only as estimates given the inherent error in counting pieces of information from
subjective ratings, they still speak to substantive consequences of sleep. When extrapolated to
numerous interviews that interrogators conduct in the field, these differences imply that
multiple pieces of information about crimes (who, what, where, when, why, how) could be
lost when interviewing sleepy subjects. For example, across 10 interviews (frequently
necessary in homicide investigations), those who lost 5 hours of sleep would together provide
around 5 fewer verifiable details about who, where, when, what, and how (relative to those
who did not lose sleep). These differences were not attributable to being treated differently by
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
the interviewers. Of note, when queried about their interviews, the sleepier subjects tended to
report providing more information (when they objectively provided less). While not reaching
statistical significance, this results suggests participants did not have insight that sleep loss
actually inhibited their disclosures .
Why did sleep loss inhibit criminal history disclosure? Subjects‟ reports of their
interview experiences did suggest they liked and felt slightly more comfortable with the
interviewers when sleep-restricted, although they did not indicate they were treated
differently (see Supplemental Online Materials for results). As these differences were slight
and theoretically inconsistent with providing less information (people disclose more to those
they like),38 they are unlikely explanations for differences due to sleep restriction. Note there
were no systematic differences in how participants experienced or felt about the events nor
the types of crimes they disclosed (See Supplemental Online Materials for results). Finally,
note that disclosures about the same event from the same participants were coded
independently for pre- and post-model statement accounts; as a result, it is not clear how
many of the details provided the second time were new, and whether sleep restriction
impacted generation of novel details.
Of note, sleep-restricted individuals exhibited trends toward being less motivated to
recall information and finding disclosures more effortful, although these differences did not
reach conventional standards of statistical significance. These patterns suggest that increased
fatigue due to sleep loss may be one factor worth exploring as the reasons for less disclosure.
The model statement increased the amount of information generally, but it did not
significantly reduce the deleterious impact of sleep loss. Note that the level of sleep loss was
relatively modest in this study, so future research should assess the impact of more severe
doses of sleep loss on information yield, which could exert larger effects. Moreover, subjects
of night-time interviews (after midnight) could be affected by circadian misalignment in a
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
similar way, with such interviews extremely common.4,25 With that said, moderate sleep loss
in the context of fact-finding interviews examined here is much more common in everyday
law-enforcement than lengthy custodial interviews.
There are important limitations to consider as well. The study did not quite reach the
target sample size. More critically, interview subjects in this study did not face severe legal
consequences as actual suspects would, so whether disclosure would be similarly affected by
sleep when the stakes are very high is unclear. In fact, the long passage of time in most cases
would mean that the likely consequences would have already occurred. Nevertheless, most
investigative interviews involve witnesses, victims, or informants, rather than suspects
themselves. In such cases the barriers to disclosures are not necessarily legal, but rather
psychological or social (e.g., guilt, shame, or embarrassment, as in this study). As the power
and status and power of the interviewer were relatively low in the current study, it is not clear
whether the current dynamics will translate into more high-steak interrogations or those with
authority figures. Additionally, the interviewers in this study were not professionals or trained
law-enforcement officers. Furthermore, there was no access to ground truth; it is unknown
whether individuals actually committed the offenses they acknowledged. However, there are
strong pressures against admitting wrong-doing, so it is unlikely that individuals from this
population are fabricating serious crimes they did not commit. It is more likely that they
failed to disclose offenses. Regardless, random assignment would have equalized such
distorting influences across experimental conditions. Note that goal in some interrogations
has been to induce confessions regardless of the truth, but the present data does not speak to
that particular purpose. The study did not
Finally, the exact reasons for less disclosure following sleep loss are only suggested
by the data. On one hand, self-reported fatigue and impaired motivation were somewhat
related to both sleep loss and lower intelligence yield, but the findings were exploratory.
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Future research should focus directly on mechanisms that underlie these effects. Note that
different doses of sleep loss and different interviewing contexts may involve quite different
pathways. For example, the interviews in this study were relatively brief. Surveys of law-
enforcement suggest that interviews often last multiple hours,4 which is bound to increase
fatigue and may exacerbate the impact of sleep loss.
Conclusions
In sum, sleep loss among investigative interview subjects may be a substantive and
persistent influence on the amount of information that investigators collect from suspects,
victims, and witnesses. As each of these populations are likely to experience sleep disruption
during crimes (e.g., due to trauma or need to evade authorities), investigators should consider
the sleep history of their interview subjects. Anecdotal reports also suggest that interrogators
often have to wake up suspects in custody in order to interview them.39 In one New York case
a just-awoken robbery suspect was interviewed for nearly 3 hours, even if according to the
interviewing detective himself the suspect “seemed like he was dozing off, and we had to
stress to him did he understand what was going on”.40 This further highlights the need to
understand how sleep loss impacts interview subjects, as well as how sleep inertia
(grogginess upon awakening) may impact disclosure. While immediate debriefing or
interrogation is advised in many cases (because the subject may be unavailable later, because
social influence may contaminate memory, or because others may dissuade subjects from
reporting), delaying interviews may be advised in certain circumstances to off-load any sleep
pressure and aid memory or effort investment. Similar recommendations have been suggested
in cases of interviews with law-enforcement following officer-involved shootings, but there
are no universal standards.41 Policy makers should also consider institutional constraints that
may impact the extent to which interview subjects are sleep-deprived, especially when in
custodial environments.
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Deposit of Material in a Data Repository
All data described in this report, as well as all materials used in this investigation are
publicly available in the Open Science Framework repository and can be accessed at:
//osf.io/yqa4f/.
Financial Disclosure
This research was supported by the Federal Bureau of Investigation Award
#15F06718C000253 awarded to the first and last authors.
sNon-Financial Disclosure
Authors do not report any non-financial conflicts of interest.
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
References
1. Vrij A, Meissner CA, Kassin SM, Morgan III A, Fisher, RP, Kleinman, SM. Psychological
perspectives on interrogation. Perspectives on Psych Sci, 2017;12: 927-955. doi:
10.1177/1745691617706515
2. DeFraia D. Harrowing cables detail how the CIA tortured accused 9/11 mastermind Khalid
Sheikh Mohammed, jeopardizing the case against him. The Intercept Website.
https://theintercept.com/2019/09/11/khalid-sheikh-mohammed-torture-cia/. Published
September 11, 2019. Accessed January 3rd, 2020.
3. Sehwail M, Perez-Sales P, Rasras KMA, Sehwail W, Guasch A, Galan A. Sleep
deprivation does not work: Epidemiology, impacts and outcomes of incidental and systematic
sleep deprivation in a sample of Palestinian detainees. Torture Journal, 2019; 29. Accessed
January 3rd, 2020.
4. Kassin SM, Leo RA, Meissner CA, et al. Police interviewing and interrogation: A self-
report survey of police practices and beliefs. Law & Human Behavior, 2007;31: 381-400.
Accessed January 3rd, 2020.
5. Kleim B, Wysokowski J, Schmid N, Seifritz E, Rasch B. Effects of sleep after
experimental trauma on intrusive emotional memories. SLEEP, 2016; 39 2125.
DOI: 10.5665/sleep.6310
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
6. Kozinski W. "The reid interrogation technique and false confessions: A time for change,"
Seattle Jour for Soc Justice, 2018;16 (2). https://link.springer.com/article/10.1007/s42822-
019-00025-8Accessed on January 3rd, 2020.
7. Meissner CA, Kelly CE, Woestehoff SA. Improving the effectiveness of suspect
interrogations. Annual Rev of Law & Soc Sci, 2015; 11: 211-233.
doi: 10.1146/annurev-lawsocsci-120814-121657
8. Stuart GL. Innocent until interrogated: The true story of the Buddhist temple massacre.
Tucson, AZ: The University of Arizona Press, 2010.
https://uapress.arizona.edu/book/innocent-until-interrogated. Accessed October 30, 2020.
9. Semel MD. Military interrogations: Best practices and beliefs. Perspectives on Terrorism,
2013; 7(2): 39-61.
https://www.researchgate.net/publication/303151335_Military_Interrogations_Best_Practices
_and_Beliefs. Accessed October 30, 2020.
10. Rodriguez J, Harlow B. Hard measures: How aggressive cia actions after 9/11 saves
american lives. Threshold Editions; 2013.
11. Lim J, Dinges DF. A meta-analysis of the impact of short-term sleep deprivation on
cognitive variables. Psych Bulletin, 2010; 136(3): 375–389. https://doi.org/10.1037/a0018883
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
12. Payne JD, Stickgold R, Swanberg K, Kensinger EA. Sleep preferentially enhances
memory for emotional components of scenes. Psych sci, 2008; 19(8): 781–788.
https://doi.org/10.1111/j.1467-9280.2008.02157.x
13. Whitney P, Hinson J, Jackson M, & Van Dongen HPA. Feedback Blunting: Total Sleep
Deprivation Impairs Decision Making that Requires Updating Based on Feedback. Sleep,
2015; 38(5); 745-754. doi: 10.5665/sleep.4668
14. Ben-Simon E, Walker M. Sleep loss causes social withdrawal and loneliness. Nat Comm,
2018; 9(1). DOI: 10.1038/s41467-018-05377-0
15. Engle-Friedman M, Young SG. Sleep‟s role in effortful performance and
sociability. In : Z. Krizan ed. Sleep, Personality, and Social Behavior, New
York: Springer-Nature. 2019; (63-82).
16. Harrison Y, Horne JA. Sleep deprivation affects speech. Sleep: Journal of Sleep
Research & Sleep Medicine, 1997; 20(10): 871–877. https://doi.org/10.1093/sleep/20.10.871
17. Babson KA, Trainor CD, Feldner MT, Blumenthal H. A test of the effects of acute sleep
deprivation on general and specific self-reported anxiety and depressive symptoms: an
experimental extension. Jour of behav therapy and exp psychiatry, 2010; 41(3): 297–303.
https://doi.org/10.1016/j.jbtep.2010.02.008
18. Krizan Z, Hisler G. The essential role of sleep in self-regulation. In: K. D. Vohs & R. F.
Baumeister ed. Handbook of Self-Regulation (3rd ed.). New York, NY: Wiley. 2016.
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
19. Plicher JJ, Huffcutt AI. Effects of sleep deprivation on performance: A meta-analysis.
Sleep, 1996; 19(4): 318-326. https://doi.org/10.1093/sleep/19.4.318
20. O‟Mara S. Why Torture Doesn’t Work: The Neuroscience of Interrogation. Boston, MA:
Harvard University Press. 2015.
21. Blagrove M. Effects of length of sleep deprivation on interrogative suggestibility. Jour of
Exp Psych: Applied, 1996; 2(1): 48–59. 10.1037/1076-898X.2.1.48
22. Blagrove M, Akhurst T. Effects of sleep loss on confidence-accuracy relationships for
reasoning and eyewitness memory. Journal of Experimental Psychology: Applied, 2000; 6:
59-73. DOI: 10.1037//1076-898x.6.1.59
23. Calvillo DP, Parong JA, Peralta B, Ocampo D, Van Gundy R. Sleep increases
susceptibility to the misinformation effect. Applied Cogn Psych, 2016; 30: 1061-1067.
Accessed on October 30th, 2020
24. Frenda SJ, Berkowitz SR, Lofuts ER, Fenn KA. Sleep deprivation and false confessions.
Proc of National Aca of Sci, 2016; 113(8). DOI: 10.1073/pnas.1521518113
25. Scherr KC, Miller JC, Kassin SM. "Midnight confessions": The effect of chronotype
asynchrony on admissions of wrongdoing. Basic and Applied Social Psychology, 2014;
36(4):321-328. https://doi.org/10.1080/01973533.2014.917974
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
27. Ewens S, Vrij A, Leal S, et al. Using the model statement to elicit information and cues to
deceit from native speakers, non-native speakers and those talking through an interpreter. App
Cogn Psych, 2016; 30: 854–862. doi:10.1002/acp.3270
28. Schäfer T, Schwarz MA. The meaningfulness of effect sizes in psychological research:
Differences between sub-disciplines and the impact of potential Biases. Frontiers in
Psych, 2019; 10:813. doi: 10.3389/fpsyg.2019.00813.
29. Faul F, Erdfelder E, Lang A. et al. G*Power 3: A flexible statistical power analysis
program for the social, behavioral, and biomedical sciences. Behavior Research
Methods, 2007; 39: 175–191. https://doi.org/10.3758/BF03193146
30. McShane BB, Gal D, Gelman A, Robert C, Tackett JL. Abandon statistical
significance, The American Stat, 2019; 73: 235-245, DOI: 10.1080/00031305.2018.1527253
31. Marino M, Li Y, Rueschman MN, et al. Measuring sleep: Accuracy, sensitivity, and
specificity of wrist actigraphy compared to polysomnography. Sleep, 2013; 36: 1747-1755.
doi: 10.5665/sleep.3142
32. Chow CM, Wong SN, Shin M, et al. Defining the rest interval associated with the main
sleep period in actigraph scoring. Nat and sci of sleep, 2016; 8: 321–328.
https://doi.org/10.2147/NSS.S114969
33. Dianiska RE, Swanner JK, Brimbal L, Meissner CA. Conceptual priming and context
reinstatement: A test of direct and indirect interview techniques. Law & Human
Behavior, 2019; 43: 131-143. doi: 10.1037/lhb0000323
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
34. Madon S, Yang Y, Smalarz L, Guyll M, Scherr KC. How factors present during the
immediate interrogation situation produce short-sighted confession decisions. Law and
Human Behavior, 2013; 37: 60-74. http://dx.doi.org/10.1037/lhb0000011
35. Oh, I, Wang, G, Mount, MK. Validity of Observer Ratings of the Five-Factor Model of
Personality Traits: A Meta-Analysis. Journal of Applied Psychology, 2011; 96: 762-773.
https://doi.org/10.1037/0021-9010.79.2.272
36. Elwert F, Winship C. Endogenous selection bias: The problem of conditioning on a
collider variable. Annual Review of Sociology, 2014; 40: 31-53.
https://doi.org/10.1146/annurev-soc-071913-043455
37. Hayes AF. Introduction to Mediation, Moderation, and Conditional Process
Analysis.(2nd Ed.). New York: The Guilford Press; 2018.
38. Brimbal L, Dianiska RE, Swanner JK, Meissner CA. Enhancing cooperation and
disclosure by manipulating affiliation and developing rapport in investigative interviews.
Psych, Public Policy, & Law, 2019; 25:107-115. doi: 10.1037/law0000193
39. Jones M. Personal interview with Det. Matthew Jones. March 15, 2020
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
40. Floyd J. Interrogation of sleeping suspect causes reversal of conviction. John Floyd
Website. https://www.johntfloyd.com/conviction-reversed-miranda/. Published March 12,
2014. Accessed April 24, 2020.
41. Potts J. Enhanced interviewing techniques to improve memory recall. Police Foundation
Website.https://www.policefoundation.org/improved-police-legitimacy-through-cognitive-
interviewing-methods-the-challenges-of-memory-recall-post-traumatic-event/. Accessed on
March 11th, 2020.
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Figure Captions
Figure 1. Distribution of Interview-Targeted Offenses as a Function of Sleep-Restriction
Figure 2. Quantity (top) and Quality (bottom) of Disclosed Information as Function of Sleep
Restriction and the Model Statement Instructions
Figure 3. Sleep duration as a causal mediator of the impact of sleep-restriction on the number
of details initially disclosed
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Table 1. Correlations and Descriptive Statistics for all Dependent Variables
1.
2.
3.
4.
5.
M
SD
1. Sleep Duration
--
690m
187m
2. Initial Details
.24*
--
4.5
1.0
3. Model-Statement Details
.11
.35***
--
5.2
.92
4. Initial Precision
.16
.72***
.36***
--
1.2
.33
5. Model-Statement Precision
.04
.27**
.77***
.47***
--
1.5
.40
6. Subjective Disclosure
-.13
.26**
.08
.26**
.10
5.3
1.2
Note. * p < .05; ** p < .01 *** p < .001. N = 106-119
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Figure 1
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Figure 2
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
Accepted Manuscript
Figure 3
Downloaded from https://academic.oup.com/sleep/advance-article/doi/10.1093/sleep/zsab124/6276397 by guest on 16 May 2021
