United States v. Mendonca

22-826-cr
United States of America v. Mendonca

UNITED STATES COURT OF APPEALS
FOR THE SECOND CIRCUIT

August Term, 2022

Argued: June 26, 2023 Decided: December 6, 2023

Docket No. 22-826-cr

UNITED STATES OF AMERICA,

Appellee,

— v. —

ANTHONY CHRISTOPHER MENDONCA,

Defendant-Appellant.

B e f o r e:

LYNCH, LOHIER and KAHN, Circuit Judges.

Defendant-Appellant Anthony Christopher Mendonca appeals his
conviction following a jury trial in the Eastern District of New York (Cogan, J.) on
one count of possessing child pornography. On appeal, Mendonca presents two
separate challenges. The first targets the exclusion of the public from substantial
portions of his jury selection, which was constrained by stringent pandemic-era
restrictions; the other targets the admission at trial of inculpatory statements that
Mendonca argues were coerced by law enforcement’s suggestion that he had
“failed” a polygraph exam – not just as to his possession of child pornography,
but as to the “hands on abuse of kids” – and associated threats and promises.
Crucially, neither matter was properly preserved below. Thus, although we are
troubled by aspects of both challenges, because neither can withstand our
exacting standards governing plain-error review, we AFFIRM the judgment of
the district court.

Judge LOHIER files an opinion concurring in the judgment and in Part II of
the majority opinion.

GENNY NGAI, Assistant United States Attorney (Kevin Trowel,
Marietou Diouf, Assistant United States Attorneys, on
the brief), for Breon Peace, United States Attorney for the
Eastern District of New York, Brooklyn, NY, for Appellee.

SARAH BAUMGARTEL, Federal Defenders of New York, Inc.,
Appeals Bureau, New York, NY, for Defendant-Appellant.

GERARD E. LYNCH, Circuit Judge:

Defendant-Appellant Anthony Christopher Mendonca (“Mendonca”)

appeals his conviction by a jury in the United States District Court for the Eastern

District of New York (Brian M. Cogan, J.) on one count of possession of child

pornography. On appeal, he presents two separate constitutional challenges, one

2
centering on the exclusion of the public from large portions of the jury selection

process, the other on the admission of inculpatory statements he made to

investigators after he was told he had “failed” a polygraph exam – not just as to

the possession of child pornography, but also as to the “hands on abuse of kids”

– a tactic he argues was so coercive that, in conjunction with other similar

behavior, it rendered his subsequent inculpatory statements involuntary.

Although both challenges raise serious concerns, neither was properly

preserved below, and we are thus constrained by the exacting standards that

govern our review of arguments raised for the first time on appeal. Guided by

those standards, we AFFIRM the district court’s judgment.

BACKGROUND

I. The Government’s Investigation of Mendonca

The events that led to Mendonca’s arrest began in spring 2018, when

federal and New York investigators determined that an online user with a

particular IP address had been accessing child pornography using peer-to-peer

networks. Investigators were eventually able to link that IP address to

Mendonca’s home in East Flatbush, Brooklyn, and at 6:00 a.m. on November 20,

2018, they executed a search warrant for that home. One of the Homeland

3
Security Investigations (“HSI”) agents who participated in the search would later

testify at trial that when he and his team first arrived that morning, they could

see Mendonca through a second-floor window, but by the time they entered they

found him in the basement.

During their search, agents seized approximately 141 electronic devices,

including several computers, external hard drives, and internal hard drives,

mainly from a workspace in that basement that only Mendonca used, including

“maybe 20 to 30” devices in a desk drawer stuffed “almost completely full” with

hard drives. App’x 597-98. Among the internal hard drives in that drawer was

one containing over 18,000 images of child pornography. No child pornography

was found on any other device. Agents conducting the search, however, did not

discover any logbooks or other means of identifying which devices, including the

drive with the child pornography, belonged to whom. At the time, both

Mendonca and his wife (the only other person living in the home) were

employed as information technology (“IT”) professionals, and as part of his

work, Mendonca would sometimes bring colleagues’ devices home to work on

them.

While in the Mendoncas’ home, HSI agents interviewed Mendonca, and

4
eventually brought him to a nearby New York City Police Department (“NYPD”)

precinct for a polygraph examination and additional questioning. Those

interviews – and the pretrial litigation they inspired – take center stage in this

appeal and are discussed in greater detail below. During the precinct interview,

Mendonca admitted to downloading and possessing child pornography, and was

promptly arrested.

II. Mendonca’s Trial

In November 2018, Mendonca was charged in the Eastern District of New

York (“E.D.N.Y.”) with a single count of possession of child pornography in

violation of 18 U.S.C. §§ 2252(a)(4)(B) and 2252(b)(2). The matter was assigned to

Judge Cogan, who referred all pretrial matters to Magistrate Judge Sanket J.

Bulsara. Trial and surrounding proceedings unfolded over three days in early

June 2021, at a time when trials in E.D.N.Y. were subject to stringent pandemic-

era protocols. Jury selection took place on June 7, 2021 before Magistrate Judge

Cheryl L. Pollak. Those proceedings are also central to this appeal and are

discussed in greater detail below.

Judge Cogan presided over the trial proper beginning on June 8, 2021. At

trial, the jury was instructed on the following four elements of the charged child

5
pornography offense: (1) “that the defendant knowingly possessed matter

containing one or more visual depictions”; (2) a nexus to interstate commerce; (3)

“that the production of the visual depiction involved a minor” (and the depiction

portrays that minor) “engaging in sexually explicit conduct”; and (4) “that the

defendant knew that the production of the visual depiction involved a minor”

(and that the depiction portrays that minor) “engaging in sexually explicit

conduct.” App’x 884. The second and third elements were undisputed.

The trial thus focused primarily on whether Mendonca knew that he was

in possession of child pornography. On that point, the government called three

HSI agents who were involved in the search of Mendonca’s home, the review of

the electronic devices seized there, and/or Mendonca’s interrogation. It also

introduced videotaped excerpts of Mendonca’s inculpatory statements during his

precinct interview.1 At closing, the government urged the jury to heed the

“overwhelming” forensic evidence, id. at 834, which it argued dovetailed with

key details from Mendonca’s interview statements, including his admission to

1
The rest of the government’s case-in-chief consisted of several of the offending
images themselves along with the testimony of two other HSI agents concerning
the origin of some of the images, including that they were produced outside of
New York, and that their subjects were indeed minors.

6
using a file sharing service called Newshosting to download child pornography

on a timetable that aligned with the forensic evidence of activity from the

computer in his basement workspace.

Mendonca did not put on a defense. At closing, his counsel argued that

“[w]hen the agents showed up at the Mendonca residence, they had already

made up their minds” about who was responsible for the child pornography. Id.

at 839. They then “fed” that predetermined narrative to Mendonca who, “worn

out” from “hours of pressuring and questioning,” eventually “fed it back to

them.” Id. at 840. Counsel also queried why, even though both Mendonca and his

wife worked in IT, and even though Mendonca’s workspace was a “tornado” of

devices belonging to many people, investigators “never bothered to ask” which

devices belonged to whom, or to undertake other forensic work to “find out

where those devices came from.” Id. at 846-48.

The case was submitted to the jury on June 9, 2021. Deliberations lasted

less than three hours. During that time, the jury conveyed a single note to the

court, requesting to review the admitted excerpts from the recordings of

Mendonca’s interrogation. It was permitted to do so, and it returned a

unanimous guilty verdict shortly thereafter.

7
 In April 2022, Mendonca was sentenced to 30 months’ imprisonment, to be

followed by a five-year term of supervised release. He is currently serving that

sentence and is projected to be released around July 2024. Following the

imposition of sentence, Mendonca filed this appeal.

DISCUSSION

Mendonca appeals two discrete aspects of his conviction: (1) the process by

which his jury was selected, and (2) the admission at trial of inculpatory

statements that he argues were coerced. We address each challenge in turn.

I. Public Trial Right

Mendonca first argues that his Sixth Amendment right to a public trial was

violated by Magistrate Judge Pollak’s failure, amid a cloud of pandemic-era

logistical constraints, to livestream substantial portions of jury selection to the

courtroom designated for the public. Although Mendonca’s concerns are not

without some merit, they were not raised below, and are therefore subject to

plain-error review. His arguments on appeal do not satisfy the demanding

standard for such review.

A. Mendonca’s Jury Selection

On Thursday, June 3, 2021, Judge Pollak held a telephone conference with

8
the parties to discuss jury selection, which was to begin the following Monday.

The impetus for the conference, she explained, was to “run through the selection

procedure with you because it is a little bit different than it was before the

pandemic.” App’x 123. To start, because of social distancing requirements, the

court would only be able to screen 35 potential jurors at a time: one batch in the

morning and then (at least) another 35 in the afternoon. Second, the process

would straddle multiple courtrooms, with the court addressing the “basics” of

voir dire with all 35 prospective jurors in the courthouse’s larger “ceremonial”

courtroom, before, in Judge Pollak’s initial vision, breaking into smaller groups

who would be shuffled between courtrooms for additional questioning and for-

cause dismissals. Id. at 125.

Judge Pollak then raised a remaining sticking point that would complicate

that initial plan: sensitive questions “that we would normally handle at sidebar,”

but that could not be so handled in light of the court’s stringent social distancing

requirements. Id. at 129. The parties offered various ideas. The government

proposed tracking jurors’ threshold responses to those questions as the court

proceeded, and then “at the end” reconvening to “bring out each juror

individually and do a sidebar on everything so it’s much more efficient.” Id.

9
Judge Pollak noted with approval that this approach “does avoid having them

have to come up to sidebar and hovering with the court reporter there and all the

attorneys . . . because obviously we are not permitted to ask jurors if they are

vaccinated.” Id. at 131. Although defense counsel was “fine” with that proposal,

he noted his concern that such a lengthy gap between question and response

might make jurors less likely to answer honestly. Id. at 132. The government

offered that another alternative would be “to have them come out individually,”

but noted that “[i]t seems that would take quite a bit of time.” Id. Defense counsel

agreed: “Yeah, I think if it moves smoothly I think it’s a good system and we can

always adjust if it’s not working.” Id.

Reflecting on those remarks, Judge Pollak floated the prospect of

“winnow[ing] out all of the potentially problematic questions.” Id. at 133. In other

words, she explained, she could confine her initial questioning in the ceremonial

courtroom to less fraught subject matters – e.g., “has anyone here served on a

jury before, when did you serve,” etc. – and then going “one by one” in a smaller

courtroom, 2E, for “the more sensitive case specific questions.” Id. Both parties

expressed their approval, and the conversation moved on to other voir dire

details.

10
 At the end of the conference, Judge Pollak apprised the parties that “you

should be aware that there is a room [that] will be set aside . . . for the public and

the press. And they will be able to hear everything that goes on during the jury

selection . . . including the questioning in 2E.” Id. at 143 (emphasis added).

That turned out not to be entirely accurate. When the parties reconvened in

the ceremonial courtroom on June 7, Judge Pollak briefly recapped, in broad

strokes, the jury selection plan, and then called in the first group of prospective

jurors. After outlining various COVID protocols and orienting the venire on jury

service in general, she laid out the multi-courtroom protocol she intended to

employ. That plan largely aligned with what the parties had agreed to a few days

earlier: the court would begin with “general questions” of all 35 prospective

jurors in the ceremonial courtroom; jurors who had answers requiring follow-up

would simply stand up and provide their juror number; the court would then

circle back to those issues in the private one-on-one questioning in courtroom 2E.

But Judge Pollak’s next instruction was a conspicuous departure from the

original plan:

[D]uring any private questioning the live stream, which
is being used from this courtroom into a separate room
for the public to observe the proceedings, that live stream

11
 will be disconnected during the private questioning and
the only people who will hear your answers are those in
the courtroom now; the parties, the lawyers, the judge
and court personnel, okay? All right.

Id. at 164 (emphasis added). The same instruction was given, almost verbatim,

during the afternoon session. But although it directly contradicted what Judge

Pollak had said previously, no one objected during either session.

Unencumbered by any objection, Judge Pollak forged ahead with her first

set of questions in the ceremonial courtroom. These included, among other

similar inquiries, whether prospective jurors anticipated having difficulty

following basic instructions about the presumption of innocence and jurors’ duty

to apply the law as instructed, whether any had prior familiarity with the case or

the people involved, and whether any jurors had any physical or sensory issues,

language issues, or pre-paid travel plans that would pose a hardship should they

be selected to the panel.2 Very few jurors responded in the affirmative; those who

did were asked to identify their juror numbers and then to be seated. Judge

Pollak next asked, among other similar inquiries, whether any prospective jurors

had served on a jury before, had experience (or loved ones with experience) in or

2
In all relevant respects, the afternoon session unfolded much the same as the
morning session.

12
around the criminal justice system, or had worked in fields related to computers,

software programming, or IT. Here, whenever a juror said yes, Judge Pollak

immediately followed up with detailed questioning in front of the rest of the

venire. She then closed out that portion of the proceedings by asking if any jurors

would have difficulty rendering a verdict based solely on the evidence presented

at trial without regard for outside beliefs or ideas, or if any so far felt they would

be inclined to favor one side or the other. When two jurors in the morning

session stood in response to that last question, Judge Pollak recorded their juror

numbers and asked them to return to their seats. At that point, she announced

that “we are going to retire one by one to Courtroom 2E” for individual

questioning. Id. at 201.

Those one-by-one proceedings in courtroom 2E – which, true to Judge

Pollak’s instruction that morning, were not livestreamed – covered a range of

topics, including various claims of hardship (medical appointments, travel plans,

employment issues, etc.) and follow-up from earlier unelaborated affirmative

responses in the ceremonial courtroom (relatives in law enforcement, prior

litigation history, etc.). Jurors were also asked a variety of additional questions,

including whether they or their loved ones had ever been subject to criminal

13
investigation or prosecution, accused of child pornography or similar offenses, or

connected to child pornography in some other way (via groups that investigate

child pornography, provide services to victims, etc.). Several jurors were excused

for cause based on their answers during that portion of the proceeding.

Next, Judge Pollak put to each juror who had not yet been excused a suite

of basic biographical questions, including (1) their place of residence, (2) duration

of residence, (3) whether they rent or own, (4) whether they live with others, (5)

marital/parental status and history, and where applicable, details about their

spouse and/or children, (6) employment status and details, (7) education history,

(8) hobbies, (9) whether they regularly read newspapers or magazines or use the

internet, and (10) whether they watch TV and what they like to watch. Finally,

after asking each juror whether they felt they could not be fair and impartial,

Judge Pollak gave the parties the chance to request additional inquiry.

At no point, morning or afternoon, did any party object to the fact that

none of this was being livestreamed.

After all prospective jurors had completed their one-on-one questioning in

courtroom 2E, those who remained were directed to return to the ceremonial

courtroom for peremptory challenges. Eventually, the final panel was seated, the

14
rest of the venire was dismissed, last-minute instructions were given, and court

was adjourned for the evening. The next morning, June 8, 2021, Judge Cogan

gave his own preliminary instructions to the final panel, and trial began.

B. The Public Trial Right

The Sixth Amendment guarantees criminal defendants “the right to a . . .

public trial.” U.S. Const. amend. VI; see United States v. Gupta, 699 F.3d 682, 686

(2d Cir. 2012) (“[W]hile this right derives from both the First and Sixth

Amendments, it is the latter that supports a defendant’s public trial right . . . .”),

citing Presley v. Georgia, 558 U.S. 209, 212 (2010). That right “unquestionably

extends to voir dire,” Gupta, 699 F.3d at 687, and can abide exceptions “only for

good cause shown,” Press-Enterprise Co. v. Superior Ct. of California, Riverside

County, 464 U.S. 501, 505 (1984). Under the Supreme Court’s ruling in Waller v.

Georgia, 467 U.S. 39, 48 (1984), “[t]rial courts must – before excluding the public

from any stage of a criminal trial – satisfy themselves” of four requirements: (1)

the party seeking closure must “advance an overriding interest that is likely to be

prejudiced”; (2) the closure “must be no broader than necessary to protect that

interest”; (3) the court must consider “reasonable alternatives” to closure; and (4)

the court must make “specific findings” sufficient to support the closure. Gupta,

15
699 F.3d at 687 (emphasis in Gupta) (internal quotation marks and citations

omitted). “In other words, if a court intends to exclude the public from a criminal

proceeding, it must first analyze the Waller factors . . . .” Id. (emphasis in original).

There are a number of doctrinal carveouts to that rule. For one thing, the

standard is somewhat less rigorous for so-called “partial” closures, which require

only a “substantial” rather than an “overriding” interest to justify the closure.

United States v. Smith, 426 F.3d 567, 571 (2d Cir. 2005).3 But even for closures

deemed partial, the other Waller requirements persist: the closure must still be no

broader than necessary, and the trial court must justify the partial closure with

sufficient findings, having considered reasonable alternatives. United States v.

Laurent, 33 F.4th 63, 95 (2d Cir. 2022).

Separately, although a violation of the public trial right is a “structural

error” and therefore “not subject to harmless-error review,” United States v.

Gomez, 705 F.3d 68, 74 (2d Cir. 2013), we have recognized a narrow “triviality”

3
In determining whether a closure is sufficiently narrow to qualify as partial, we
look to factors like “its duration, whether the public can learn what transpired
while the trial was closed (e.g. through transcripts), whether the evidence was
essential, and whether selected members of the public were barred from the
courtroom, or whether all spectators were excluded.” Smith, 426 F.3d at 571
(internal quotation marks omitted).

16
exception that, while “very different from a harmless error inquiry,” likewise

safeguards convictions from vacatur under certain circumstances, Peterson v.

Williams, 85 F.3d 39, 42 (2d Cir. 1996); see Gupta, 699 F.3d at 688 (emphasizing

“the doctrine’s narrow application”). In particular, the triviality inquiry looks to

“whether the actions of the court and the effect that they had on the conduct of

the trial deprived the defendant – whether otherwise innocent or guilty – of the

protections conferred by the Sixth Amendment.” Peterson, 85 F.3d at 42.4 Unlike

harmless-error review, it “does not dismiss a defendant’s claim on the grounds

that the defendant was guilty anyway or that he did not suffer prejudice or

specific injury.” Id. (internal quotation marks omitted).

C. Standard of Review

Although the structural nature of a public trial violation removes it from

the ambit of harmless-error review, where that violation is not objected to below,

4
To that end, we consider “whether the closure subverts the values the drafters
of the Sixth Amendment sought to protect: 1) to ensure a fair trial; 2) to remind
the prosecutor and judge of their responsibility to the accused and the
importance of their functions; 3) to encourage witnesses to come forward; and 4)
to discourage perjury.” Smith v. Hollins, 448 F.3d 533, 540 (2d Cir. 2006) (internal
quotation marks omitted). If it does not, we will conclude that the exclusion did
not “implicate the Sixth Amendment public trial guarantee.” Id. (internal
quotation marks omitted).

17
we still “review the claim for plain error” under Rule 52(b) of the Federal Rules of

Criminal Procedure. Laurent, 33 F.4th at 95-96 (emphasis added); see Gomez, 705

F.3d at 74 (“Whether an error can be found harmless . . . is simply a different

question from whether it can be subjected to plain-error review.” (internal

quotation marks omitted)). Under the standard set forth by the Supreme Court in

in United States v. Olano, before we may correct an error forfeited by a party’s

failure to object below, “there must be (1) error, (2) that is plain, and (3) that

affects substantial rights”; then, once those initial conditions are met, “an

appellate court may then exercise its discretion to notice a forfeited error, but

only if (4) the error seriously affects the fairness, integrity, or public reputation of

judicial proceedings.” Gomez, 705 F.3d at 75 (internal quotation marks and

alterations omitted), citing Olano, 507 U.S. 725, 732-36 (1993). It is the defendant’s

burden to establish all four requirements. Greer v. United States, 141 S. Ct. 2090,

2097 (2021); see Gomez, 705 F.3d at 75 (“Meeting all four prongs of this test is

difficult, as it should be.” (internal quotation marks omitted)).

D. Plain-Error Analysis

That exacting standard of review controls here.

With respect to the first element, the district court demonstrably erred

18
when it decided not to livestream large portions of the jury selection process to

the separate courtroom reserved for the public. That is true regardless of the

extent of the closure. Partial or not – and we do not have occasion to decide

whether it is appropriate to characterize as partial the decision to shut all

members of the public out from essentially all of the individualized voir dire,

sensitive and non-sensitive questions alike – our precedents make clear that the

extent of the closure affects only the requisite degree of the interest advanced

under the first Waller factor. It does not eliminate that factor altogether, nor any

of the other Waller obligations; it does not excuse the court from ensuring that the

closure is no broader than necessary, from considering alternatives, from making

specific findings that justify the closure, Laurent, 33 F.4th at 95, nor from doing all

of those things “before excluding the public from any stage of a criminal trial,”

Gupta, 699 F.3d at 687 (emphasis in original) (internal quotation marks omitted).

None of that happened here, and the government does not attempt to argue

otherwise.5 And that was surely error. See Gupta, 699 F.3d at 678-

5
Instead, the government offers several responses that rather miss the point.
First, it argues that the closure was justified by a substantial reason, which, true
or not, is immaterial where the court made no such finding on the record, to say
nothing of meeting Waller’s other demands. Next, the government insists that
“[t]he magistrate judge’s individual questioning of the prospective jurors in

19
88 (holding that “despite not making any Waller findings, the district court

intentionally excluded the public from the courtroom for the entirety of voir

dire” and therefore, “[o]n these facts alone, the closure was unjustified”); see also

courtroom 2E was the functional equivalent of a sidebar conference,” and
therefore either not a closure at all, or too trivial to “subvert any of the values
derived from Waller.” Appellee’s Br. 43, 51 (internal quotation marks omitted)
Both of those arguments fail for the same reason: the proceedings in courtroom
2E went well beyond what could be reasonably characterized as typical of a
sidebar. The court conducted essentially all of its individualized voir dire in
courtroom 2E outside the view of the public. Once the livestream was
discontinued, the court did not confine itself to follow-up on potentially sensitive
matters broached in the ceremonial courtroom, nor even to sensitive questions
reasonably likely to invite a sidebar if answered in the affirmative (much less to
situations where there had actually been an affirmative answer to one of those
threshold questions). We are aware of no authority supporting the notion that
standard and unremarkable voir dire questions like “[w]here do you live,” “[t]ell
me about your education,” and “[d]o you watch TV at all,” e.g., App’x 215-17, are
typically shielded from public view absent an “affirmative request” for privacy
because the question touches on “deeply personal matters that [a juror] has
legitimate reasons for keeping out of the public domain.” Press-Enterprise, 464
U.S. at 511. Finally, the government compares the closure in this case to one we
deemed covered by our triviality carveout in Gibbons v. Savage, 555 F.3d 112, 121
(2d Cir. 2009). But in that case, “nothing of significance happened” during the
closure: “The judge read the indictment, asked questions of a few jurors, and
provided administrative details on what the jurors should expect if chosen.” Id.;
see also United States v. Shipp, No. 21-1284-CR, 2022 WL 16543193, at *1 (2d Cir.
Oct. 31, 2022) (summary order). The same can hardly be said of a closure that
encompassed not just sensitive responses normally heard at sidebar, but the
lion’s share of the jury selection process and nearly all of the individualized voir
dire, for-cause dismissals, and sensitive questioning – including most of the
innocuous threshold questioning that might prompt a sensitive sidebar.

20
Presley, 558 U.S. at 214 (“[T]rial courts are required to consider alternatives to

closure even when they are not offered by the parties.”).

Turning to the second Olano requirement, we conclude that the error was

“plain,” for many of the same reasons. 507 U.S. at 734. Waller and Press-Enterprise

are long-standing precedents, and the government does not dispute that there is

a substantial body of circuit case law addressing their application to jury

selection as a significant part of a criminal trial right to which the public trial

right attaches. See, e.g., Gupta, 699 F.3d at 687.

But whatever the phrase “plain error” might otherwise suggest, the

existence of (1) “error” that is (2) “plain” is only half the battle. See United States v.

Lewis, 424 F.3d 239, 246-47 (2d Cir. 2005). There are still two more steps: “[E]ven

though the error be plain, it must also [3] affect substantial rights,” and even then

we must still determine (4) “whether the forfeited error seriously affects the

fairness, integrity or public reputation of judicial proceedings.” Johnson v. United

States, 520 U.S. 461, 468-69 (1997) (internal quotation marks and alterations

omitted), citing Olano, 507 U.S. at 732-37.

Turning briefly to the “substantial rights” requirement, although that third

Olano step ordinarily tests whether the error “affected the outcome of the district

21
court proceedings,” United States v. Ragonese, 47 F.4th 106, 110 (2d Cir. 2022)

(internal quotation marks omitted), the Supreme Court has repeatedly

“reserv[ed] the question whether ‘structural errors’ automatically satisfy the

third ‘plain error’ criterion,”6 United States v. Marcus, 560 U.S. 258, 263 (2010)

6
There is some reason to think not. To start, plain error is “simply a different
question” from harmless error, Gomez, 705 F.3d 74, and protects different
interests. “[T]he harmless-error doctrine is essential to preserve the principle that
the central purpose of a criminal trial is to decide the factual question of the
defendant’s guilt or innocence, and promotes public respect for the criminal
process by focusing on the underlying fairness of the trial rather than on the
virtually inevitable presence of immaterial error.” Arizona v. Fulminante, 499 U.S.
279, 308 (1991) (internal quotation marks omitted). In contrast, plain-error review
“encourage[s] timely objections and reduce[s] wasteful reversals by demanding
strenuous exertion to get relief for unpreserved error.” United States v. Dominguez
Benitez, 542 U.S. 74, 82 (2004). It is therefore not farfetched to imagine that
structural error would play different roles in each setting. In the context of
harmless error, moreover, the “structural” umbrella really covers two different
categories of error: (1) those so “obviously” problematic that it is all-but
impossible for them not to have prejudiced the defendant, such as the total
deprivation of the right to counsel throughout trial, Fulminante, 499 U.S. at 309, or
the “denial of an impartial judge,” United States v. Sedillo, 10 F. App’x 690, 691
(10th Cir. 2001) (unpublished disposition), citing Liteky v. United States, 510 U.S.
540, 555 (1994) (the “high degree of favoritism or antagonism” required to
amount to such an error “make[s] fair judgment impossible”); and (2) those
errors – like a public trial right violation – whose effects are “often difficult” for
appellate courts to measure, and that are unlikely to be outcome-determinative in
deciding guilt or innocence, but that nonetheless ordinarily warrant reversal
because they destabilize the very “framework within which the trial proceeds,”
United States v. Marcus, 560 U.S. 258, 263 (2010) (internal quotation marks and
alteration omitted). But in the context of plain error, where even the defendant
does not believe, in the moment, that an error is worth objecting to, it is

22
(collecting cases); see United States v. Eldridge, 2 F.4th 27, 38 (2d Cir. 2021)

(acknowledging that the Supreme Court has “noted the possibility that certain

errors, termed structural errors, might affect substantial rights regardless of their

actual impact on an appellant’s trial,” but declining to reach the issue), rev’d on

other grounds, Eldridge v. United States, 142 S. Ct. 2863 (2022), quoting Marcus, 560

U.S. at 263.

We need not wade into those waters, however. Even assuming the third

requirement was met here, Mendonca’s challenge founders on the fourth and

final plain-error prong. See Johnson, 520 U.S. at 469 (rejecting a plain-error

challenge where “even assuming that the [error] affected substantial rights, it

does not meet the final [plain-error] requirement” (internal quotation marks and

alterations omitted)); see also United States v. Hougen, 76 F.4th 805, 813 (9th Cir.

reasonable to wonder whether our inability to retrospectively measure any
prejudice should matter. Put bluntly: if the defendant did not fear any prejudice,
why should we? See United States v. Nelson, 277 F.3d 164, 206 (2d Cir. 2002) (“[W]e
do not imply that all claims of structural error (of error that requires automatic
reversal rather than harmless error review) are unwaivable.”), citing Freytag v.
Comm’r, 501 U.S. 868, 896 (1991) (Scalia, J., concurring in the judgment) (“Must a
judgment already rendered be set aside because of an alleged structural error to
which the losing party did not properly object? There is no reason in principle
why that should always be so.”). There may well be a compelling answer to that
question, but like a long line of courts before us, we need not search for it today.

23
2023) (rejecting, on the fourth plain-error prong, a pandemic-era public trial right

challenge, rebuffing the defendant’s argument that the court was “preclude[d]”

from conducting “an individualized fourth prong analysis in cases implicating a

structural error”).

Courts have yet to define with much particularity what precisely it means

to “seriously affect[] the fairness, integrity or public reputation of judicial

proceedings.” Johnson, 520 U.S. at 469 (internal quotation marks omitted). That is

not entirely by accident; the Supreme Court has explained that “[t]he fourth

prong is meant to be applied on a case-specific and fact-intensive basis.”7 Puckett

v. United States, 556 U.S. 129, 142 (2009); see Rosales-Mireles v. United States, 138 S.

Ct. 1897, 1906 (2018) (rejecting another Circuit’s formulation of the fourth prong

as “unduly restrictive”); United States v. Young, 470 U.S. 1, 16, (1985) (“[W]hen

addressing plain error, a reviewing court cannot properly evaluate a case except

by viewing such a claim against the entire record . . . [because] each case

necessarily turns on its own facts.” (internal quotation marks and citations

7
Separately, we have also observed – and this is almost as true today as it was in
1997 – that “[t]his requirement is not discussed in many of our prior cases,
because a court need not reach the issue unless Olano’s first three conditions are
satisfied.” United States v. Knoll, 116 F.3d 994, 1001 (2d Cir. 1997) (internal citation
omitted).

24
omitted)). It has also underscored that even in circumstances where “the integrity

of the system may be called into question,” there may be key “countervailing

factors in particular cases” that nonetheless excuse the unpreserved error.

Puckett, 556 U.S. at 142-43.

That guidance resonates here. In this case, there are indeed conspicuous

countervailing circumstances. First, this trial was held relatively early in the

pandemic,8 when the legal community in New York was still sorting out how to

accomplish justice under the many new constraints imposed to keep jurors,

litigants, and court staff safe. While a defendant’s rights of course do not

disappear during a pandemic, that context is vital to our understanding of the

fourth plain-error prong. This challenge would not be before us had the district

court not been compelled to juggle multiple courtrooms, multiple lines of

questioning, and multiple batches of jurors as it labored to satisfy tried-and-true

voir dire norms in circumstances that were anything but normal. Tellingly, we

have excused other public trial violations under the fourth prong where the error

similarly arose from the district court’s understandable – though flawed –

8
Judge Cogan confided to the parties at the end of trial that this was in fact his
very “first trial back since the pandemic.” App’x 893. Judge Pollak had presided
over at least one prior pandemic-era jury selection.

25
response to logistical constraints and other environmental concerns. See, e.g.,

Gomez, 705 F.3d at 75-76 (excluding defendant’s family members during voir dire

“cannot be viewed as [an error] that affected the fairness, integrity, or public

reputation of judicial proceedings” where it was “clear from [the] inquiry by the

district judge, when the prospective jurors were about to be brought to the

courtroom, that the judge anticipated having at least some of the persons already

seated leave because, as the record reveals, there were not enough seats for all of

the venirepersons”); United States v. Killingbeck, 616 F. App’x 14, 16 (2d Cir. 2015)

(summary order) (claimed Sixth Amendment error of “restricting the display of

trial exhibits [containing child pornography] to the courtroom audience,” in

addition to not being plain, “did not seriously affect the fairness, integrity, or

public reputation of the trial”).

Second, we give considerable weight to Judge Pollak’s deliberate efforts to

foster a collaborative environment for the court and the parties to figure out how

best to complete a fair voir dire under challenging and unusual circumstances.

While, for purposes of the first plain-error prong, she fell short of her duty under

Waller to explore alternatives to closure on the record, she was open to, and

indeed eagerly solicited and incorporated, input from the parties about how best

26
to conduct voir dire. The transcript of the pretrial conference demonstrates that

Judge Pollak was guided by a genuine aim to collaboratively navigate difficult

and novel circumstances towards a just solution – including by ensuring that the

parties were empowered to (and often did) provide feedback at every turn. As

the Supreme Court has noted, “the public legitimacy of our justice system relies

on procedures that are neutral, accurate, consistent, trustworthy, and fair, and

that provide opportunities for error correction.” Rosales-Mireles, 138 S. Ct. at 1908

(internal quotation marks omitted). Judge Pollak’s clear effort to satisfy that

aspiration does much to rebut any claim that her actions brought the legal system

into “disrepute.”

Granted, the final procedure differed from what the parties had agreed to,

and from what Judge Pollak had previewed, days earlier. That was undoubtedly

error. But whether that departure resulted from a changed mind or was purely

inadvertent, Mendonca had every opportunity to object – opportunities, in fact,

since the court employed the same procedures in the morning and in the

afternoon, after a lunch break – and elected not to. Had he done so, “the trial

court might well have adopted an alternative” or, at the very least, “been alerted

to announce the findings . . . as to the necessity for, the needed breadth of, and

27
the possible alternatives to, any exclusion.” Gomez, 705 F.3d at 75-76. Or, likely

enough, Judge Pollak would have realized that she had already announced a

different plan to the parties, and simply corrected the mistake. See Puckett, 556

U.S. at 134 (“[T]he timely raising of claims and objections . . . gives the district

court the opportunity to consider and resolve them.”). Neither party has

disputed that other trials in E.D.N.Y. around this time proceeded without raising

public trial concerns. See, e.g., United States v. Shipp, No. 21-1284-CR, 2022 WL

16543193, at *1 (2d Cir. Oct. 31, 2022) (summary order). That fact, along with the

judge’s earlier assumption that the livestream would be operational throughout

voir dire, suggest that there was no technological or logistical barrier to

livestreaming any portion of voir dire the court wished to livestream.9 All that

was missing was an objection.

We have remarked in the past that reversal “serves little purpose when the

defendant makes no more than a perfunctory effort to prevent the closure,” let

alone no effort at all, and that “to the extent that a public trial serves the broader

public interest in fair trials designed to end in just judgments, that purpose is best

9
Nor indeed did Judge Pollak ever imply that the reason she was deviating from
her earlier plan was because of any such constraint. She gave no explanation at
all – in no small part because no objection ever occasioned one.

28
served by requiring a defendant to press his argument in a forceful and

persuasive way rather than by awarding a windfall” to one who chooses not to

do so. Brown v. Kuhlmann, 142 F.3d 529, 542 (2d Cir. 1998) (internal quotation

marks, citations, and alterations omitted); see Gomez, 705 F.3d at 74 (“Due regard

generally for the public nature of the judicial process does not require disregard

of the solid demands of the fair administration of justice in favor of a party who,

at the appropriate time and acting under advice of counsel, saw no disregard of a

right, but raises an abstract claim only as an afterthought on appeal.” (emphasis

omitted)), quoting Levine v. United States, 362 U.S. 610, 619-20 (1960).10 That

perspective does not require us to affirm every time any defendant has failed to

object to any error, but we find it instructive under the present circumstances.

For all of those reasons, we think “no miscarriage of justice will result here

if we do not notice the error,” Johnson, 520 U.S. at 470 (internal quotation marks

omitted), and we therefore decline to vacate Mendonca’s conviction on that basis.

10
In the absence of a contemporaneous objection, moreover, we have no record
as to whether any member of the public ever was present or attempted to enter
the room reserved for the public at any time while the livestream was cut off, or
for that matter while it was on.

29
 II. Voluntariness of Mendonca’s Statements

Mendonca’s second challenge to his conviction attacks the voluntariness of

the inculpatory statements he made during his police interrogation that were

introduced against him at trial. Once again, however, Mendonca failed to

preserve the issue below, and thus while we are troubled by some of the police

tactics employed in this case, we discern no plain error in the district court’s

decision to admit those statements.

A. Mendonca’s Interrogation

The following account of Mendonca’s interrogation is undisputed. It is

drawn from the transcript and video recordings of the interrogation; as explained

below, there was no evidentiary hearing regarding Mendonca’s motion to

suppress the statements.

Mendonca was first questioned by HSI Special Agents Krista Cousins and

Luanne Walter while his home was being searched. In the course of that initial

80-minute home interview, which was recorded, Mendonca was advised of, and

agreed to waive, his Miranda rights. Afterwards, he voluntarily accompanied

police to the nearby 63rd NYPD Precinct, where he was questioned for the better

part of four hours. Throughout that precinct interview, which was also recorded,

30
he was periodically offered water, bathroom breaks, and other accommodations.

The precinct interview began around 8:50 a.m. with questioning by NYPD

Detective Anthony Santilli. After some preliminary discussion, re-reading (and

re-waiving) of Mendonca’s Miranda rights, and small talk, Santilli broached the

prospect of Mendonca taking a polygraph exam. He assured Mendonca that the

exam was “100% voluntary” and that Mendonca was free to stop, and leave, at

any time; he was not under arrest. App’x 1083-84. “Right now,” he continued,

“you’re the only person that knows whether or not [] you did this,” but “[i]f you

take this test in about two hours we’ll both know.” Id. at 1085. He advised

Mendonca that “[i]f you did do this, you should refuse the test,” but “[i]f you

didn’t do it, it’s an easy way for me to eliminate you as a suspect.” Id. Mendonca

agreed to take the polygraph exam and signed a consent form to that effect.

After further questioning on an eclectic mix of subjects – from Mendonca’s

background and education, to whether his wife could have been responsible for

downloading the child pornography at their home,11 to whether Mendonca

11
Santilli would revisit this subject later in the interview, countering Mendonca’s
continued resistance to the child pornography allegations by asking, “It’s your
wife’s then? So . . . we’ll arrest your wife once we got that. . . . Are you kidding
me? . . . Like, you’re gonna blame your wife.” App’x 1151-52. Mendonca retorted
that he was blaming neither his wife nor himself: “It’s not our pornography.” Id.

31
“ha[d] ever watched that TV show Mindhunters,” id. at 1119-20 – Santilli began

priming Mendonca for the exam. He regaled Mendonca with an extended

overview of the polygraph technology and underlying science before eventually

connecting Mendonca to “the instruments,” explaining that “[b]asically the pad

you’re sitting on . . . detects motion from your waist down” and two other

devices attached to his chest “detect motion from your waist up.” Id. at 1110.

Fingertip sensors would detect sweating, blood pressure, and heart rate, “like [in]

a doctor’s office.” Id. at 1112. In other ways, the exam might recall a “school test”

– though, unlike in school, where a score of “97” might earn top marks, here

“you need 100 to pass.” Id. at 1123.

Finally, before proceeding with the exam, Santilli conducted some

preliminary diagnostics. He asked Mendonca to circle numbers on a piece of

paper, close his eyes, be still, and answer yes or no questions about the numbers

he had circled. Based on that process, Santilli immediately announced that he had

determined that the exam would be effective on Mendonca.

The exam itself involved, in Santilli’s characterization, three different

categories of questions: (1) “control” questions; (2) “incident” questions, targeting

at 1152.

32
whether Mendonca did the specific thing he was accused of; and (3) “profile

character” questions, which, according to Santilli, “if you pass, you know, I can

competently say that, listen, not only did he not do it [but] he’s not even the type

of person that would have done it.” Id. at 1115. The actual exam questions

included, among others, the following:

• Is today Sunday?
• Do you intend to . . . answer every test question
truthfully?
• During the first 42 years of your life have you ever
been attracted to someone too young for you?
• As an adult have you ever had sexual contact with a
minor in any way?
• In your entire life can you remember ever doing
anything to a female you[r] age without her
permission?
• During the first 42 years of your life did you ever lie
to avoid responsibility for something you did?
• Did you ever knowingly possess child pornography?
• Have you done anything to try and deliberately beat
this test?

Id. at 1129-31. Mendonca completed the exam multiple times, denying all

wrongdoing. At one point, Santilli accused Mendonca of “moving around, . . .

you’re doing something with your body. . . . You’re trying to play games. I can

see it . . . .” Id. at 1132. After the third run-through of the questions, Santilli left

33
the room for about ten minutes to “go grade this and give you some feedback.”

Id. at 1142.

Immediately upon his return, he announced, “O.K. There is no doubt that

you’re lying.” Id. at 1143. And “you didn’t just fail the child pornography test

question. You failed, uh, hands on abuse of kids. Having sexual contact with a

minor.” Id. Much of the ensuing questioning played on that theme. To start,

Santilli offered that he did not believe the agents “really care about the child

pornography” at all, but rather had targeted Mendonca because he fit the second,

and more problematic, of two common child pornography archetypes: (1) cases

where “we just knock on the door and say, hey listen, man, uh, you’ve been

downloading some stuff you really shouldn’t be downloading” and “[w]e’ll try

to get you therapy,” as opposed to (2) cases where the suspect is believed to pose

an ongoing, direct risk to children. Id. at 1144-45. Mendonca, Santilli confided, fell

into the second bucket because he worked in IT at a school building that housed

both elementary and middle school students. And that left Santilli concerned:

“[Are] you the type of person to take a brand new kid off the street, rape them

and murder them and throw their body in your back yard? Are you?” Id. at 1147.

Those worries, Santilli claimed, flowed from his own experiences on the force.

34
“I’ve had cases where the individual literally had taken a kid off the street . . .

and basically put his hands down her pants and then ran off.” Id. at 1150-51.

Mendonca repeatedly denied abusing children.

That hardly satisfied Santilli. “[I]f I’m asking you if you ever had sexual

contact with a kid how am I supposed to believe that when you’re lying about”

possessing child pornography. Id. at 1146. Bemoaning that Mendonca was

“[l]ying to my face” even after he had “bombed” the polygraph exam – “[D]on’t

play me like I’m a fucking idiot. . . when you keep lying to me, you disrespect

me,” id. at 1147-48, 1154 – Santilli urged that he was “looking for your

cooperation to make sure . . . that there’s no children that you’ve . . . raped or

murdered.” Id. at 1154-55.

Moreover, he cautioned, that cooperation would affect his

recommendation to prosecutors. “[S]hould my recommendation to the district

attorney be that he’s an asshole who probably raped and murdered a bunch of

kids?” Id. at 1155. Or should his recommendation be that Mendonca was still at

an early stage where “i[f] we intervene and make sure that they get help at this

stage before it progresses to the point where they pull a kid off the street . . . [o]r

while you have some one-on-one time with a kid in a classroom[?]” Id. He

35
repeatedly threatened that Mendonca’s lack of candor as to the child

pornography accusations would trigger larger concerns for police and prosecutor

alike, which could in turn affect Mendonca’s punishment for possessing child

pornography: “[I]f you’re not being honest about [child pornography], then I’m

automatically jumping [to the conclusion] that something happened with the kids

at school. And then . . . when I talk to the district attorney, my recommendation

isn’t going to be, hey, this is a guy that has a problem[,] that can’t control it, that

doesn’t know how to get the help, maybe he should be put in some sort of

therapy instead of, you know, criminal charges or jail time.” Id. at 1156.

Mendonca continued to deny all wrongdoing, and eventually a frustrated Santilli

stepped out of the room shortly before 11 a.m.

About 15 minutes later, HSI Agent Cousins entered and took the lead on a

variety of topics. These initially included Mendonca’s home, his electronic

devices, his personal and professional life, his religion, and his practice of

downloading feature films and adult pornography online. When the

conversation turned to adult pornography tastes, Santilli, who had been

observing quietly, chimed in with his own pornography preferences, prodding

Mendonca to disclose whether he was “into like people peeing on each other”

36
and whether he knew what “Lolita” means. Id. at 1187-89.

Temperatures soon began to rise again. Both Santilli and Cousins

continued to press Mendonca for some time, insisting that they knew he was

lying, that they knew that he knew that he had child pornography, and either

that they wanted to help him avoid getting to the point where he would actually

hurt children, or, alternatively, that they suspected he already had. Santilli

reiterated that “[s]omeone who is found with a terabyte of child pornography on

their hard drive is probably someone that has a problem with touching kids.” Id.

at 1193. “[I]f we can’t feel confident that you’ve realized you have a problem with

it, how can I comfortably let you out into the world?” Id. Accordingly, once

again, while he might prefer to recommend “court mandated therapy” for

Mendonca, “I can’t give that recommendation to the district attorney, if you’re

not being a hundred percent honest with us.” Id. at 1194.

By this point, Mendonca was beginning to waver. “I don’t think I’m . . .

sexually attracted to kids,” he pondered, but “[h]ow do I determine if that’s part

of who I am, that I don’t know about. Is that something in me?” Id. at 1199. Soon

thereafter, he offered that although he “d[id]n’t think” he needed such help, he

was “willing . . . to open up to a counselor . . . and see if there’s something that

37
I’m hiding [from] myself.” Id. at 1206. Cousins responded that “this is your

chance to talk about you, and talk about why, and talk about . . . if you’re sorry.”

Id. at 1207. “If you don’t want to do this anymore, this is your chance to say that

because you get one chance with us, right?” Id. Moments later, she repeated that

“[w]e’re doing this for you because this is your only chance to talk.” Id. at 1208.

At last, Mendonca began to confess, albeit in fits and bursts. He first

admitted to “downloading a lot of videos and put[ting] them on the hard drive.

But . . . I don’t think they were kids of some sort.” Id. Moments later, though, he

acknowledged that he had downloaded “maybe 10, 20” videos containing child

pornography, “and pictures too.” Id. at 1211. He described those files as depicting

“a young person, like a young girl, a young man, or . . . images of like a child on a

beach or . . . running around in a house, or something like that.” Id. He

speculated that the children depicted were “maybe 6 to 10 years old,” and that

they were mostly young girls – answering “[y]eah” when asked whether

“females” were his preference – but “you’d probably find like a boy or two in

there somewhere.” Id. at 1211-12. He explained that he had downloaded the files

using a program called Newshosting, discussing at length the mechanics of that

process and (without being prompted) providing his username and password.

38
 Cousins and Santilli then circled back to the prospect of Mendonca

“hurt[ing] a child.” Id. at 1237. First, Cousins asked whether Mendonca viewed

these videos “[b]ecause you’re a terrible person and you want to see those

children hurt or do you view these videos because you can’t help but be sexually

attracted, because you do masturbate because it’s something that you can’t help.”

Id. at 1238. “It’s one or the other,” she would later reiterate. Id. at 1248. When

Mendonca denied masturbating to the videos, she quipped, “So you just want to

see children hurt.” Id. at 1238. Mendonca denied both that he was sexually

attracted to the children and that he enjoyed watching children get hurt. Around

the same time, asked whether he thought that those who create child

pornography feel ashamed, Mendonca answered, “I mean I would feel

ashamed,” which prompted Santilli to ask, “What happened? Where did it

happen?” Id. at 1242. Mendonca responded, “What? The downloading?” Id.

Santilli clarified, “With the kids. Taking the videos. Taking a picture.” Id. When

Mendonca denied that he had personally created child pornography, Santilli

remarked, “The way you’re talking it sounds like you did.” Id. at 1243.

As the clock ticked closer to 1 p.m., Cousins reminded Mendonca that “this

is your time” to say anything else “to help yourself.” Id. at 1244-45. After locking

39
down one more unambiguous confession – Mendonca’s admission that he had

been downloading child pornography for “[m]aybe about two years or so,” id. at

1246 – the interviewers asked Mendonca if he would be willing to write and sign

a statement in his own words. That statement would in turn be sent to

prosecutors, who would then decide how to proceed “based on our

recommendation, based on your statement, based on the facts.” Id. at 1250.

Mendonca consented, and produced a handwritten statement in which he

admitted to “us[ing] a site call[ed] Newshosting, to download[] Games, movies,

books, vidos [sic]” and to “download some videos to an external hard Drive and

keep it.” Id. at 933. The statement continued: “I very sorry [sic]. . . . I need help to

on this [sic], and what to do. I Never take pic [sic] or videos [of] kids in any way,

and will never do that. Thank [sic] for understanding.” Id.

B. Mendonca’s Pretrial Motions

Nearly two years before trial, in September 2019, Mendonca moved to

suppress all the inculpatory statements from the precinct interrogation, arguing

that under the totality of the circumstances, both his Miranda waivers and

subsequent statements had been coerced. He based that motion on three specific

pre-polygraph exam tactics: (1) Santilli’s misrepresentation that Mendonca

40
“could prove his innocence by taking the exam,” which was “a promise [Santilli]

knew could not be fulfilled”; (2) Santilli’s explanation that if Mendonca was

indeed innocent, it would be in his interest to take the polygraph exam, which

Mendonca characterized as “impermissibl[e] . . . legal advice”; and (3) Santilli’s

suggestion that if Mendonca “chose not to speak,” police “would assume his

guilt,” which Mendonca argued invalidated his prior Miranda waiver. Dkt. No. 26

at 3-5. The motion did not address the post-exam conduct that has become

central to this appeal.

The motion was referred to Magistrate Judge Bulsara, who held oral

argument in December 2019.12 Opting to rely on the recording and transcript of

the interrogation, Judge Bulsara did not hold an evidentiary hearing and did not

otherwise invite testimony from Santilli, Cousins, or others. Consequently, there

is no sworn testimony regarding, among other things, Santilli’s experience and

training in administering polygraph tests or interpreting their results, whether

Santilli’s characterizations of the polygraph results to Mendonca accurately

12
Although at the time of oral argument, Judge Bulsara had apparently not yet
had the opportunity to review the full interview recording or transcript, the
parties eventually provided the court with a recording and transcript of the
entire interview before the report and recommendation issued.

41
reflected his subjective opinion as an expert polygraph examiner, or for that

matter whether Mendonca had actually been connected to a properly functioning

polygraph. Nor does the record contain any raw polygraph data.13

In February 2020, Judge Bulsara issued a report and recommendation

(“R&R”) concluding that both Mendonca’s Miranda waivers and subsequent

statements were voluntary and recommending that the suppression motion be

denied. Although the R&R contained a conspicuous warning that, pursuant to

Rule 59(b)(2) of the Federal Rules of Criminal Procedure, objections must be filed

within 14 days, Mendonca never objected. A few weeks later, Judge Cogan

adopted Judge Bulsara’s recommendations in a brief text order:

Defendant has filed no objections to [the] Report and
Recommendation. I agree with Magistrate Judge
Bulsara’s well-reasoned decision that defendant’s
statements were made voluntarily and without

13
The parties made a few vague representations at oral argument as to their
understanding of what polygraph records exist, and to what extent they were
provided to the defense before trial. Mendonca also represented in one of his
submissions to the district court that the government had provided some raw
polygraph data to the defense during discovery. But none of that is in the record,
and regardless, these representations do little to advance our understanding of
the basic question left hanging by the record: whether the statement that (at least
in Santilli’s opinion) Mendonca failed a polygraph exam not just with respect to
child pornography, but also with respect to the “hands on abuse of kids,” was a
fabrication.

42
 coercion. Therefore, defendant’s [] motion to suppress is
denied.

Dkt. Entry dated Feb. 19, 2020.

Over a year later, as trial approached, the government moved in limine to

introduce portions of the videotaped precinct interrogation. Mendonca countered

that if any inculpatory statements were admitted, the court should play the entire

recording (subject to some redactions) “in the interest of context and

completeness.” Dkt. No. 56 at 3. “[I]n addition to several hours of coaxing Mr.

Mendonca into making statements,” he urged, “the interrogation in this case

involved an especially atypical and coercive feature: an NYPD detective

administered a polygraph examination to Mr. Mendonca, and then told Mr.

Mendonca he had failed the polygraph as a means of continuing the

interrogation.” Id. Judge Cogan primarily sided with the government but

permitted Mendonca to introduce a few excerpts to show “the nature of the

agents’ questioning and the environment of the interrogation.” Dkt. No. 75 at 2.

In the end, scenes from the precinct that were played for the jury included:

(1) a five-minute clip beginning at timestamp 8:48 a.m. that shows Santilli

bringing Mendonca water, reading him his Miranda rights, and asking some

43
preliminary questions, GX 206-E; (2) a one-minute excerpt at 10:44 a.m. that

shows Santilli warning Mendonca “don’t play me like I’m a fucking idiot,” GX

206-K; (3) a nine-second clip at 10:52 a.m. that shows Santilli telling Mendonca

that his concern was to ensure “that there’s no children that you’ve raped or

murdered,” GX 206-L; (4) a 16-minute clip beginning at 10:56 a.m. that shows

Santilli storming angrily out of the room, leaving Mendonca alone for nearly 15

minutes, and then Cousins entering with Santilli to begin the next stage of the

interview, GX 206-G; (5) a 16-second clip at 11:29 a.m. that shows Mendonca

confirming that his wife does not use his external hard drives, GX 206-H; (6) a

26-minute excerpt beginning at 12:06 p.m., containing the first wave of

confessions and the detailed discussion of Newshosting, GX 206-I; and (7) the

final 25 minutes of the interview, beginning at 12:47 p.m., containing more

incriminating statements as well as the period when Mendonca was left alone to

write out his statement. GX 206-J. The jury was also shown the handwritten

statement itself. Most of the interactions at the heart of this appeal were not

shown to the jury.

C. Voluntariness Standards

Challenges to the voluntariness of a confession are based on two

44
overlapping constitutional provisions: (1) due process protections under the Fifth

(or Fourteenth) Amendment, and (2) the Fifth Amendment privilege against

self-incrimination. Dickerson v. United States, 530 U.S. 428, 433 (2000). We need not

disaggregate the two: “As a general matter, courts’ descriptions of statements as

‘compelled’ (invoking the text of the Self-Incrimination Clause) and/or

‘involuntary’ (invoking, arguably, the Due Process Clause) are often used

interchangeably and the words often treated synonymously.” United States v.

Allen, 864 F.3d 63, 82-83 n.84 (2d Cir. 2017).14 And “[a]lthough this principle and

14
The privilege against self-incrimination also creates the affirmative right to be
warned of one’s “Miranda rights.” Although Mendonca has at times invoked
Miranda (and although the two strains of Fifth Amendment authority have
intermingled over the years), this appeal is a classic voluntariness challenge.
Mendonca does not argue, for example, that the Miranda warnings he received
were deficient; despite the government’s fixation on that question, Mendonca has
disavowed any Miranda argument on appeal. See Oral Argument 7:43-8:11. In any
event, “[t]he existence of a knowing and voluntary [Miranda] waiver does not . . .
guarantee that all subsequent statements were voluntarily made,” and we
therefore “cannot ‘dispense with the voluntariness inquiry’ simply because we
determine that a defendant’s waiver was valid.” In re Terrorist Bombings of U.S.
Embassies in East Africa, 552 F.3d 177, 211-12 (2d Cir. 2008), quoting Dickerson, 530
U.S. at 444. A letter-perfect recitation of the classic Miranda warnings and a freely
given waiver of the right not to respond to questions would not render a
subsequent confession voluntary if, for example, in between the waiver and the
confession, officers tortured the subject. The same principle applies to other
forms of coercion. See, e.g., United States v. Anderson, 929 F.2d 96, 98-102 (2d Cir.
1991) (although it was “quite clear” that Miranda warnings had been “adequate,”
agent’s “coercive” suggestion “that if [the suspect] asked for a lawyer it would

45
its elaboration have been deemed to be due variously to the Due Process clause

and the Fifth Amendment right against self-incrimination, the modern test for

voluntariness is well established and multi-faceted.” United States v.

Orlandez-Gamboa, 320 F.3d 328, 332 (2d Cir. 2003) (internal citations omitted).

The test is indeed multi-faceted. There is “no talismanic definition of

‘voluntariness,’ mechanically applicable to the host of situations where the

question has arisen.” Schneckloth v. Bustamonte, 412 U.S. 218, 224 (1973). Broadly,

we have recognized that “voluntary” means “the product of a free and deliberate

choice rather than intimidation, coercion, or deception,” United States v. Plugh,

648 F.3d 118, 127 (2d Cir. 2011) (internal quotation marks omitted), and thus that

a confession is not given voluntarily when “obtained under circumstances that

overbear the defendant’s will at the time it is given,” United States v. Anderson,

929 F.2d 96, 99 (2d Cir. 1991).

In practice, “[w]e look at the totality of circumstances surrounding a

Miranda waiver and any subsequent statements to determine knowledge and

voluntariness.” United States v. Taylor, 745 F.3d 15, 23 (2d Cir. 2014). Those

permanently preclude him from cooperating with the police” nonetheless
rendered involuntary confessions “made immediately after the warnings were
given”).

46
circumstances generally fall into three categories: “(1) the characteristics of the

accused, (2) the conditions of interrogation, and (3) the conduct of law

enforcement officials.” United States v. Haak, 884 F.3d 400, 409 (2d Cir. 2018),

quoting Green v. Scully, 850 F.2d 894, 901 (2d Cir. 1988) (“No single criterion

controls whether an accused’s confession is voluntary . . . .”). Mendonca’s

challenge focuses almost entirely on the third category, which the Supreme Court

has positioned as both a necessary and (potentially) sufficient condition to a

finding of involuntariness. See Colorado v. Connelly, 479 U.S. 157, 167 (1986) (“We

hold that coercive police activity is a necessary predicate to the finding that a

confession is not ‘voluntary’ within the meaning of the Due Process

Clause . . . .”); Haynes v. Washington, 373 U.S. 503, 504 n.1, 513-15 (1963) (holding

that a confession was involuntary based on police conduct, even though the

defendant made “no claim that he was physically abused, deprived of food or

rest, or subjected to uninterrupted questioning for prolonged periods”).

Ultimately, the government bears the burden to establish “by a

preponderance of the evidence that a suspect waived his Miranda rights, and that

his confession is truly the product of free choice” and thus admissible at trial.

Anderson, 929 F.2d at 99; see Taylor, 745 F.3d at 23.

47
 D. Standard of Review

Ordinarily, “[w]e review the factual findings underpinning the district

court’s voluntariness determination for clear error while subjecting the ultimate

conclusion that a defendant’s statements were voluntarily [made] to de novo

review.” United States v. Siddiqui, 699 F.3d 690, 707 (2d Cir. 2012). In this case,

however, the parties dispute the proper standard of review based on

(1) Mendonca’s failure to raise issues key to his voluntariness challenge on

appeal in his suppression motion below, coupled with (2) his failure to object to

Judge Bulsara’s R&R. We agree with the government that the particular challenge

Mendonca advances on appeal is doubly encumbered by these dual procedural

failures.

We begin with Mendonca’s shifting rationales. On appeal, Mendonca

argues in substantial part that his interrogators coerced his confession by

misleadingly declaring, after he had taken his polygraph exam, that his test

results suggested he had “committed ‘hands on abuse’ of kids and ‘sexual contact

with a minor,’” and then “quickly extrapolat[ing] this to evidence that Mendonca

had physically abused children at the school where he worked” – in other words,

by cowing him with the kind of “false evidence” that has been “implicated in the

48
vast majority of documented police-induced false confessions.” Appellant’s Br.

53-54 (internal quotation marks omitted). He also argues that after his polygraph

exam, and based on his purported results, his interrogators improperly

threatened that if he “did not confess to possessing child pornography, the

detective would have to tell ‘the district attorney’ that Mendonca was ‘an asshole

who probably raped and murdered a bunch of kids.’” Id. at 57, quoting App’x

1155. Similarly, he argues that after his polygraph exam, he was hoodwinked into

believing that this interrogation was “his only chance” not just “to admit that he

viewed child pornography” but also “to avoid accusations” of the even more

serious crimes purportedly flagged by the polygraph results. Id. at 55-57 (internal

quotation marks omitted).

None of that was argued below. Mendonca’s suppression motion

exclusively addressed the pre-exam tactics Santilli used to coax Mendonca into

taking the test in the first place.15 “It is a well-established general rule that an

15
Mendonca did eventually set his sights on post-exam conduct before the
district court, but only in his response to the government’s bid to introduce his
confessions at trial. Even in that context, though, he did not argue that the cited
conduct rendered his statements involuntary as a matter of law and thus
excludable; rather, he simply urged that if the jury was to see the confessions, it
should in fairness see “the complete picture” – including the details of Santilli
and Cousins’s post-exam conduct – so that jurors could themselves “assess the

49
appellate court will not consider an issue raised for the first time on appeal.”

United States v. Gershman, 31 F.4th 80, 95 (2d Cir. 2022) (internal quotation marks

and alteration omitted); see Allianz Ins. Co. v. Lerner, 416 F.3d 109, 114 (2d Cir.

2005) (declining to entertain new arguments raised on appeal where the party

had raised different arguments in support of the same general legal claim below).

“This rule is not an absolute bar to raising new issues on appeal; the general rule

is disregarded when we think it necessary to remedy an obvious injustice.”

United States v. Stillwell, 986 F.3d 196, 200 (2d Cir. 2021). When we exercise our

discretion to consider a forfeited argument, however, we typically review such

matters only for plain error. See, e.g., United States v. Edwards, 342 F.3d 168, 179

(2d Cir. 2003) (reviewing a challenge to the admissibility of a statement under

Rule 404(b) for plain error); see also, e.g., United States v. Stuckey, 317 F. App’x 48,

50 (2d Cir. 2009) (summary order) (same, for the admission of allegedly

Miranda-deficient statements).

Then, compounding that omission, Mendonca also failed to object to Judge

Bulsara’s R&R rejecting his arguments and recommending that the suppression

motion be denied. Rule 59(b) of the Federal Rules of Criminal Procedure provides

voluntariness and truthfulness of the inculpatory statements.” Dkt. No. 56 at 3.

50
that failure to timely object to a magistrate’s recommendation “waives a party’s

right to review.” Fed. R. Crim. P. 59(b)(2). In this case, the R&R provided a

14-day window for Mendonca to object. He never did.

To be sure, there is ample authority from this Court, and others, holding

that because a Rule 59 waiver is “non-jurisdictional,” it “can be excused in the

interests of justice.” United States v. Ballares, 317 F. App’x 36, 38 (2d Cir. 2008)

(summary order); see United States v. Romano, No. 15-992-CR, 2022 WL 402394, at

*4 (2d Cir. Feb. 10, 2022) (summary order) (same), citing United States v. Male Juv.

(95-CR-1074), 121 F.3d 34, 38 (2d Cir. 1997) (holding, in a pre-Rule 5916 case, that

“failure to object timely to a magistrate judge’s report may operate as a waiver of

any further judicial review of the decision, as long as the parties receive clear

notice of the consequences of their failure to object”).17 But Mendonca has not

16
Rule 59 was promulgated in 2005, and amended in 2009 to extend the default
objection window from 10 to 14 days. See Fed. R. Crim. P. 59 advisory
committee’s notes to 2005 adoption and 2009 amendment.
17
Accord United States v. Broadnax, 691 F. App’x 604, 605 (11th Cir. 2017)
(unpublished opinion); United States v. Merrett, 9 F.4th 713, 716 (8th Cir. 2021)
(similar, for failure to object under Rule 59(a)), citing Thomas v. Arn, 474 U.S. 140,
155 (1985) (holding, in a pre-Rule 59 case, that because the failure to object to a
magistrate’s report implicates a “nonjurisdictional waiver provision, the Court of
Appeals may excuse the default in the interests of justice”).

51
alerted us to (nor have we found) any authority for the proposition that “the

interests of justice” extend to situations where a defendant not only failed to

object to a magistrate’s R&R, but where the heart of the defendant’s appellate

case was never presented to the magistrate to begin with.

There is good reason for that. Timely raising issues and objections below,

once again, “gives the district court the opportunity to consider and resolve

them.” Puckett, 556 U.S. at 134 (adding that the district court “is ordinarily in the

best position to determine the relevant facts and adjudicate the dispute”). In this

case, Mendonca’s failure to raise his concerns below has left key evidentiary

holes in all the places where, as discussed further below, a properly alerted

district court might have been able to develop a record that could have

illuminated (or refuted) the police trickery central to Mendonca’s coercion claims

– in particular, whether the polygraph was just a prop to playact a suspect into

confessing, or whether there was indeed fire beneath all the smoke and mirrors.18

18
We do not suggest that polygraph examinations are scientifically valid tools of
“lie detection.” The issue here is not whether such an exam would be “proof” (let
alone admissible evidence at trial) one way or the other as to Mendonca’s guilt. In
this case, no one sought to introduce evidence of the polygraph exam at trial. The
question, rather, is whether the purported results of the examination were
presented to Mendonca in good faith, were exaggerated, or were altogether
fabricated in an effort to trick Mendonca into confessing.

52
Instead, we are left to guess. It is difficult to fathom how the “interests of justice”

require us to effectively imagine into existence, for a defendant’s benefit, a record

that does not exist precisely because of that defendant’s failure to occasion its

creation.

Consequently, absent any other persuasive reason to exercise our

discretion to consider arguments not raised below, we are left to review the

district court’s decision to admit Mendonca’s inculpatory statements for plain

error. Cf. Spence v. Superintendent, Great Meadow Corr. Facility, 219 F.3d 162, 174

(2d Cir. 2000) (explaining that our discretion to review an unobjected-to

magistrate recommendation “is exercised based on, among other factors, whether

the defaulted argument has substantial merit or, put otherwise, whether the

magistrate judge committed plain error in ruling against the defaulting party”).

E. Plain-Error Analysis

In analyzing that question, we are guided by the same plain-error

standards that governed our review of Mendonca’s public trial right challenge.

Of particular import for this issue is the second plain-error requirement: “For an

error to be plain it must be ‘clear or obvious, rather than subject to reasonable

dispute.’” United States v. Montague, 67 F.4th 520, 534 (2d Cir. 2023), quoting

53
Marcus, 560 U.S. at 262.

As lofty as that bar is for any issue, it is especially daunting for a

voluntariness challenge. Not only is a voluntariness determination intensely

“fact-specific,” Tankleff v. Senkowski, 135 F.3d 235, 245 (2d Cir. 1998), but “what

was adequate in one case to produce an involuntary confession does not establish

that the same result has been created in a different, but somewhat similar set of

circumstances,” Green, 850 F.2d at 902. Thus, in a context where “[n]o single

criterion controls,” where we cannot decide voluntariness until “after careful

evaluation of the totality of the surrounding circumstances,” Green, 850 F.3d at

901, and where we have recognized that similar circumstances can and should

yield divergent results, it is all the more difficult for any set of facts to clear the

plain-error bar.

Despite the concerns we lay out below, this case is ultimately no exception.

The most charitable formulation of Mendonca’s challenge is that he was lied to

about the nature of the evidence against him, and based in large part on that lie,

misleadingly threatened on the one hand with punishment for a more serious

crime he was never actually suspected of committing, and misleadingly

promised leniency on the other. Even assuming for a moment that this case offers

54
a unique blend of those elements that might, de novo, invite the extraordinary

remedy of a vacated conviction, on plain-error review, that uniqueness is

precisely the problem. We have consistently held that “a confession is not

involuntary merely because the suspect was promised leniency if he cooperated

with law enforcement officials,” or threatened with harsh legal penalties if he did

not. United States v. Ruggles, 70 F.3d 262, 265 (2d Cir. 1995) (internal quotation

marks omitted); see also, e.g., United States v. Gaines, 295 F.3d 293, 299 (2d Cir.

2002) (“[V]ague promises of leniency for cooperation are just one factor to be

weighed in the overall calculus and generally will not, without more, warrant a

finding of coercion.”); Haak, 884 F.3d at 412 (“[T]here is nothing improper in

police truthfully telling a defendant that he will be prosecuted to the full extent of

the law if he chooses not to cooperate.”).

We are similarly constrained by a substantial volume of authority that

police conduct that is “false, misleading, or intended to trick and cajole the

defendant into confessing does not necessarily render the confession

involuntary.” Haak, 884 F.3d at 409 (internal quotation marks omitted); see also,

e.g., Mara v. Rilling, 921 F.3d 48, 80 (2d Cir. 2019) (holding that officers’

“misrepresentations about the strength of the evidence against [the suspect]” did

55
not render subsequent confessions involuntary). Even on de novo review, those

general principles would pose a formidable obstacle to a finding of

involuntariness, and any such finding would have to be rooted in the highly

specific facts of this case.

Finally, we are constrained by the record itself. Here, Mendonca’s failure to

preserve the issue below looms large. A central premise powering his appeal is

that the voluntariness of his confessions was undermined by Santilli “falsely

informing Mendonca that he had failed the polygraph” not just with respect to

child pornography, but also with respect to “‘hands on abuse of kids’ and ‘sexual

contact with a minor.’” Appellant’s Br. 52, quoting App’x 1143. That premise not

only stands as an independent claim of error, but also underlies other key

components of this appeal – for example, the oft-repeated interrogation refrain

that if Mendonca did not admit to possessing child pornography, Santilli would

have no choice but to disclose to prosecutors that Mendonca was also suspected

of personally abusing children, and to recommend that he be punished

accordingly.

But the record contains no finding that Santilli’s claim was false, and no

evidence from which a fact-finder could assess its accuracy. Apart from his

56
representations to Mendonca, which one might reasonably be reticent to trust,

there is no evidence in the record of Mendonca’s actual polygraph results, or of

what Santilli subjectively believed those results meant, or indeed whether the

purported polygraph exam was actually a real polygraph exam at all. All of that

would bear on any assessment of whether Mendonca was coercively duped into

confessing, and we are in the dark as to all of it.

And crucially, we likely would not have been in the dark had Mendonca

flagged these issues below, either at the outset or via an objection to the R&R. An

argument or objection focused on the post-polygraph interrogation might have

occasioned the district or magistrate judge to schedule an evidentiary hearing, to

review exhibits and affidavits illuminating the polygraph results, to invite

testimony from Santilli and others, and to otherwise develop a record that would

have shed light on what exactly happened, thus permitting an evaluation of

whether the interrogators’ behavior went beyond the kinds of deception, threats,

and promises (or combinations thereof) that courts have held to be permissible.

Instead, as a direct result of Mendonca’s choices below, we have none of

that.

57
 Admittedly, were we reviewing Mendonca’s challenge de novo, or if the

record were further developed, things might well look different. We agree with

Mendonca that his case presents a troubling blend of ingredients that, at the very

least, imperil voluntariness. The first is Santilli’s determined pre-exam attempt to

position the polygraph as an unassailable investigative tool that would be key to

determining Mendonca’s fate. Not only did Santilli imply that the test results,

alone, could determine whether the police continued to consider Mendonca a

suspect,19 but he also foreboded that unlike in school, “here you need 100 to

pass.” Id. at 1123. The thrust of those pre-exam tactics was surely to cement in

Mendonca the sense that his fortunes were tied to his polygraph performance,

and to imply that because only a guilty person would refuse the test, Mendonca’s

refusal to participate would be taken as incriminating evidence.

Then, as soon as Santilli was in a position to characterize (honestly or not)

19
The focus of Mendonca’s initial suppression motion, which he to some extent
reprises here, was on these efforts to encourage him to take the polygraph exam:
in particular, Santilli’s assurance that passing the exam would “eliminate”
Mendonca as a suspect, coupled with what Mendonca has characterized as
improper (and bad) legal advice that Mendonca should not take the test if guilty,
but should absolutely take it if innocent. Like the district court, we disagree that
these pre-exam tactics were sufficient on their own to render his subsequent
statements involuntary. See Haak, 884 F.3d at 409. But that does not disqualify
them from factoring into a larger set of circumstances that threaten voluntariness.

58
Mendonca’s polygraph performance, he was able to pour on additional threats

and promises. The instant Santilli returned from (purportedly) reviewing

Mendonca’s polygraph results, the detective – at first on his own, and later in

tandem with Cousins – began the first of what would become many variations on

the theme that Mendonca was suspected not just of a child pornography offense,

but also of actually assaulting children himself, and that his candor with respect

to the former was directly tied to whether the latter would play any role in his

prosecution or punishment.20 There is little evidence that any of that was true.

20
The government maintains that the interviewers’ focus on potential abuse of
children was “reasonabl[e] in light of the defendant’s job at an elementary
school.” Appellee’s Br. 68. That may be true in the abstract. But it is also clear
from the transcript that their end game here was not to uncover abuse committed
by Mendonca, but rather to cajole him into admitting to the child pornography
offense they actually thought he had committed. There is vanishingly little in the
record that even hints at any investigation into whether Mendonca had
personally assaulted children – a response one might reasonably expect from
investigators who were truly concerned about that risk in its own right and not
simply seizing upon an opportunity for leverage. Similarly, there is some tension
between Santilli’s representation that the polygraph results indicated that
Mendonca was lying about whether he had physically abused children and his
statement that unless Mendonca admitted to possessing child pornography,
Santilli would assume he was also a hands-on abuser. If the polygraph results
really did reflect that Mendonca had already physically abused children, why
would a confession to possessing child pornography warrant a recommendation
that Mendonca be treated as an “early-stage” child pornography offender who
could be prevented from spiraling into a hands-on abuser through therapy?

59
 These circumstances resemble those that imperiled the determination of

voluntariness in Green, perhaps the closest reference point this Circuit has to

offer. 850 F.2d at 903. In that case, even though there was nothing in “the

characteristics of the accused or the conditions of interrogation” that gave us

pause,21 we nonetheless remarked that the voluntariness of the defendant’s

confession was “not free from doubt” based on “troubling” police conduct alone.

21
That is generally true here as well. Mendonca is an educated professional who
appeared perfectly cogent and responsive throughout his interactions with
police. Nor is there indication of physical mistreatment. Santilli and Cousins
repeatedly offered Mendonca water, bathroom breaks, and the like. Still,
although Mendonca’s challenge is primarily reliant upon the third voluntariness
factor (“the conduct of law enforcement officials”) alone, rather than “(1) the
characteristics of the accused“ or “(2) the conditions of interrogation,” Haak, 884
F.3d at 409 (internal quotation marks omitted), it is worth acknowledging – as we
have in the past – that his inexperience with the mechanics of the criminal justice
system left him more vulnerable to police bluster and deception than other
suspects may have been in his place. Compare Ruggles, 70 F.3d at 265
(emphasizing the “extensive criminal record” that had made the defendant
“familiar with police questioning”), with United States v. Young, 964 F.3d 938, 946
(10th Cir. 2020) (explaining that the defendant’s past state arrests “do not
convince us that [the defendant] could withstand the coercion created by [the FBI
agent’s] legal misrepresentations and promises of leniency” based on a fabricated
link between the agent and the judge presiding over the defendant’s federal
case). It is also relevant that throughout the entirety of his interrogation,
Mendonca was not represented by counsel. See Orlandez-Gamboa, 320 F.3d at 333
(acknowledging that the defendant’s representation by counsel was relevant to
whether statements induced by promises of leniency undermined the
voluntariness of those statements).

60
Id. (describing “the issue of voluntariness” as “a close one”). In Green, the

“chicanery” included (1) “detectives’ misrepresentations that they already had

enough evidence for [the defendant’s] arrest”; (2) “false promises of help which

held out the prospect of more lenient treatment”; and (3) a detective’s “reference

to the electric chair.” Id. Ultimately, we deemed the confession voluntary, relying

heavily on two mitigating circumstances: the fact that one interviewer’s improper

tactics were later walked back by a second interviewer, and a statement from the

suspect himself that the reason he confessed was not to curry leniency, but rather

that “he was afraid that what he had done to the victims in a blackout would be

something he was going to do to his own family – maybe even his mother.” Id. at

903-04. This case features no analogous mitigating factors and a number of

analogous aggravating factors, including (1) similar (possible) misrepresentations

about the evidence against Mendonca, (2) similar threats and promises founded

on those (possible) misrepresentations, and (3) a persistent cloud of potential

criminal liability for the sexual assault of children – assuredly among the most

pernicious accusations that can be dangled in front of a suspect.

The circumstances here – assuming for the sake of argument that the

polygraph test results were misrepresented – also resemble those that other

61
Circuits have determined to be sufficiently coercive to render a confession

involuntary. See, e.g., United States v. Lopez, 437 F.3d 1059, 1061, 1065 (10th Cir.

2006) (holding that a confession was involuntary based on agents’ insinuations

that “the gun residue test they had conducted on Lopez earlier in the day had

produced positive results, even though the agents had not actually yet received

any test results,” along with threats to “prove Lopez’s mother was a liar if she

tried to corroborate Lopez’s alibi” and promises of extreme sentencing leniency

in exchange for a confession); Tobias v. Arteaga, 996 F.3d 571, 582-83 (9th Cir. 2021)

(“[A]nyone in Tobias’s shoes would have understood that Detective Arteaga

considered anything less than a confession to the murder a ‘lie,’ because all of

Tobias’s attempts to deny the accusation were met with statements such as

‘you’re full of shit’ . . . [and thus] in the context of this particular interrogation,

any threat that Tobias would be treated harshly for ‘lying’ was no different than a

threat that he would be treated harshly for refusing to cooperate by confessing –

exactly the type of threat we have previously held to be categorically

impermissible.”). On de novo review, those comparators would raise significant

concerns.

62
 But we are not on de novo review. And, in spite of our concerns, we cannot

conclude that the district court was “clear[ly] or obvious[ly]” wrong to admit

Mendonca’s inculpatory statements on the record presented. Montague, 67 F.4th

at 534 (internal quotation marks omitted). To that end, even Green, the Second

Circuit precedent arguably most helpful to Mendonca’s case, morphs from a de

novo sword into a plain-error shield: the Green panel of course ultimately deemed

the confession at issue there voluntary, notwithstanding its analogous concerns.

See 850 F.2d at 904. And though Green may reasonably be distinguished by the

mitigating factors that resolved that panel’s “doubt[s]” in favor of voluntariness,

id. at 903, a court might also reasonably construe Green as evidence that even the

many red flags raised here do not necessarily render a statement involuntary.

In other words, even assuming the circumstances here might warrant

reversal on de novo review, we simply cannot confidently declare that the matter

is no longer “subject to reasonable dispute.” Montague, 67 F.4th at 534 (internal

quotation marks omitted). Therefore, the district court did not plainly err in

conforming to the general thrust of that authority and admitting Mendonca’s

inculpatory statements at trial.

63
 CONCLUSION

We have considered Mendonca’s other arguments and conclude that they

are without merit. Thus, for the foregoing reasons, we AFFIRM the judgment of

the district court.

64
LOHIER, Circuit Judge, concurring in part and concurring in the judgment:

I concur in the judgment and in Part II of the majority opinion but write

separately because I somewhat disagree with the majority’s discussion in Part I

about the public trial right and structural as opposed to trial error.

Mere trial errors are discrete mistakes that “occur[] during the presentation

of the case to the jury” and are subject to harmless error review. Arizona v.

Fulminante, 499 U.S. 279, 307–08 (1991). The “defining feature of a structural

error,” by contrast, “is that it affects the framework within which the trial

proceeds, rather than being simply an error in the trial process itself.” Weaver v.

Massachusetts, 582 U.S. 286, 295 (2017) (cleaned up). A structural error thus may

consist of “a special category of forfeited errors that can be corrected regardless

of their effect on the outcome.” United States v. Olano, 507 U.S. 725, 735 (1993).

As several sister Circuits have pointed out, “[t]he third requisite of plain error

review” — the error’s effect on a defendant’s substantial rights — “is necessarily

met where the error at issue is structural.” United States v. Becerra, 939 F.3d 995,

1005 (9th Cir. 2019); see United States v. Negrón-Sostre, 790 F.3d 295, 305–06 (1st

Cir. 2015); United States v. Simmons, 11 F.4th 239, 269 (4th Cir. 2021); United States

v. McAllister, 693 F.3d 572, 582 n.5 (6th Cir. 2012); United States v. Janis, 898 F.3d

1
847, 851 n.2 (8th Cir. 2018); United States v. Clark, 981 F.3d 1154, 1169 & n.4 (10th

Cir. 2020). And in United States v. Feliciano, 223 F.3d 102 (2d Cir. 2000), our Court

explained that “[e]rrors are properly categorized as structural only if they so

fundamentally undermine the fairness or the validity of the trial that they require

voiding its result regardless of identifiable prejudice,” id. at 111 (quotation marks

omitted). Structural error “should be presumed prejudicial if the defendant

cannot make a specific showing of prejudice.” Olano, 507 U.S. at 735. In that

regard, our review of structural errors differs substantially from plain error

review for trial errors, which requires a clear and obvious error that has a

demonstrated rather than presumed effect on substantial rights.

In the context of the Sixth Amendment right to a public trial, presuming

prejudice makes sense: “[I]t would be, in most cases, virtually impossible for a

defendant to demonstrate that the absence of family members or friends from his

trial affected its result,” and the public trial right “could well become a right in

name only if the defendant were required to show prejudice.” Carson v. Fischer,

421 F.3d 83, 95 (2d Cir. 2005) (quotation marks omitted).

But should we presume that structural errors also “seriously affect[] the

fairness, integrity, or public reputation of judicial proceedings,” Johnson v. United

2
States, 520 U.S. 461, 467 (1997) (quotation marks omitted) — the fourth and final

prong of plain error review? At least in the context of the public trial right, in my

view, the answer is yes. “[I]f the searchlight of a public trial serves as a restraint

against the abuse of judicial power, then extinguishing it through an improper

closure surely threatens to erode the integrity and public reputation of the trial.”

Daniel Levitas, Comment, Scaling Waller: How Courts Have Eroded the Sixth

Amendment Public Trial Right, 59 Emory L.J. 493, 528 (2009) (quotation marks

omitted). The Ninth Circuit has followed that reasoning and developed a rule

that structural errors invariably “satisfy the third and fourth prongs of plain error

review.” United States v. Ramirez-Ramirez, 45 F.4th 1103, 1109 (9th Cir. 2022)

(emphasis added). I agree with the majority that such a categorical rule goes too

far. See Majority Op. [25–26] (citing Puckett v. United States, 556 U.S. 129, 142

(2009)). It is enough to say that structural errors typically compel the reversal of a

conviction whether or not harm is shown.

But not in every case. In rare cases, “countervailing factors” may “satisfy

[us] that the fairness, integrity, and public reputation of the proceedings will be

preserved absent correction.” Rosales-Mireles v. United States, 138 S. Ct. 1897, 1909

(2018). In my view, this appeal involves the very rare case where the

3
presumption that a structural error impugns the judicial process is rebutted.

During jury selection, the District Court faced obvious and unprecedented health

and safety restrictions arising from the pandemic and grappled with

extraordinary logistical challenges in picking a jury. Nothing could have

prepared the court for the thorny set of decisions and careful safety measures

that the pandemic demanded. Even in hindsight, it’s fair to say that any risk of

harm to the perception of fairness, integrity, and reputation of our court

proceedings caused by limiting public access to jury selection during a deadly

pandemic in New York City was not as grave as the real risk of harm the

pandemic itself posed to potential jurors and other court personnel. “Sincere

efforts to accommodate the extraordinary practical concerns raised by the

pandemic — even if later revealed to be legally erroneous — simply do not

impair the perceived fairness or ‘public reputation’ of judicial proceedings in the

same way that such measures undoubtedly would if undertaken during ordinary

times.” United States v. Mayo, No. 21-10181, 2023 WL 5198767, at *2 (9th Cir. Aug.

14, 2023) (quoting Olano, 507 U.S. at 736).

For this reason I concur.

4