dissenting.
I agree with the majority’s resolution of all but one of the many issues raised- in this appeal. The one exception ■ is the judge’s decision to allow plaintiffs to present evidence contrary to their Rule 36 admissions. The judge abused his discretion by applying the wrong legal standard toresolve this pivotal issue incorrectly. What the majority downplays as a “technical error” allowed plaintiffs to ambush Glatt at trial on a critical issue. I would reverse and remand for a new trial on that basis.
Plaintiffs offered two theories to hold defendant Glatt liable for the fire. One was that one or more plastic filters in the granulator were cracked, and that cracks allowed metal inside the filter to contact other metal parts so as to cause a spark that started the fire. Defendant Glatt served' requests for admissions under Federal Rule of Civil Procedure 36. Three asked plaintiffs to admit in essence that they had no evidence that filters were cracked at the time of the fire.
Plaintiffs’ responses were convoluted, but the bottom line was that plaintiffs admitted they had not discovered any evidence that one or more of the filters or filter housings were cracked at the time of the fire. Cf. McCann v. Mangialardi, 337 F.3d 782, 788 (7th Cir. 2003) (district court erred by failing to give preclusive effect to plaintiffs Rule 36 admission that he had no evidence that defendant withheld exculpatory evidence; admission was fatal to plaintiffs claim).
At trial, however, the judge allowed plaintiffs to contradict their admissions on this critical issue. The majority errs by claiming “there was no contradiction” because the admissions were “not unqualified.” Ante at 571. Read as a whole, the majority argues, Tate & Lyle indicated in its admissions that it had indirect evidence of cracked filters at the time of the fire. The admissions should be read as a whole, and to that end I have reproduced them in full as an appendix.1
Glatt’s requests for admissions tried to learn in three different ways whether plaintiffs had any evidence that cracked filters were in the granulator, at the time of the September 17, 2009 fire. The first two requests (Nos. 7 and 8) focused on whether plaintiffs had “observe[d]” or were “informed” about cracked filters after Dynacoil’s filter repairs in August 2009. Although plaintiffs “denied” those requests, they went on to say: “it is admitted T&L is not presently aware of anyone that observed cracked, chipped, or poorly fitting filters in the Granulator following work performed by Dynacoil in August or September of 2009."
Despite this admission, plaintiffs’ expert witness testified at trial that he had “been told that there were some 16 of these damaged or cracked filters, in fact, in the machine before the fire.” SA 61. The dis*573trict court excused the jury and acknowledged that although the expert had said the broken filters were in the granulator “before the fire, it certainly came across as at the time of the fire so that’s why I excused the jury to make sure that this didn’t become worse than it needs to be.” SA 66-67. Outside the presence of the jury, plaintiffs’ counsel admitted that he had told the expert that 16 broken filters were inside the granulator on September 1st. SA 65. The expert also said: “What I have been told is this, that on September 1st, if I remember that date correctly, I may be off a day or two, that they had put in there those.broken filters. And from September 1st to the 17th, and for a good number of days in fact, the machine had not been in operation.” SA 69-70.
Making matters worse, one of plaintiffs’ employees volunteered in cross-examination that he reinserted cracked filters into the granulator after the repairs:
Q: Okay. Well, isn’t it true that you had a conversation with Mr. Wenrich sometime before the fire where Mr. Wenrich had asked you about the filter issue, and you had told him that you weren’t aware of any filters that have been damaged since the repair was done three weeks ago?
A: Yes, so that, that does say that. But we had cracked filters that we had been using to fill the granulator. [Note that this was volunteered and non-responsive.]
Q: Well, you had a whole—this is the first time I’m hearing this so—you had a full stock of filters in your possession, did you not?
A: Yep.
Q: So, you didn’t use those stock filters to replace the cracked ones then?
A: We used some of them, but we continued using some of the cracked ones also because I didn’t want to keep cracking more filters.
Dkt. No. 299 at 140-41 (emphases added). This testimony directly contradicted Admissions 7 and 8.
Glatt’s third request (No. 9) was even broader than the first two. It sought admission that plaintiffs “have not discovered any evidence that one or more of the Granulator filters” was cracked at the time of the fire on September 17,2009, (Emphasis added.) Tate & Lyle’s first sentence in response said: “It admits the allegations in Paragraph 9 as it pertains to cracked or chipped filters or housing,” It continued by noting that it had “not discovered direct evidence” of cracked filters-, but it speculated that “it wás possible” some filter elements were under stress after the repairs and could have broken again.
Again, this admission was directly contradicted by plaintiffs’ witness at trial:
Q: Okay. My question, though, is on September 17th of 2009 and after Dyna-coil had gone and done this retrofit program and had modified these filter housings, my understanding is that there is no evidence whatsoever that there was any cracked filters inside the granulator; isn’t that true?
A: No. That is not true.
Q: So, you know there were cracked filters inside the granulator áfterwards? A: Yeah, while we were running, yes.
Q: Okay. So, you’re saying that on September 17th of 2009, there were cracked filters inside the granulator?
A: Yes.
Dkt. No. 299 at 137.
Despite these contradictions, plaintiffs never moved under Rule 36(b) to .withdraw or amend their admissions. When Glatt objected based on plaintiffs’ Rule 36 admissions, the judge told Glatt it should simply “argue the weight of the evidence.” *574The judge explained that in allowing plaintiffs to offer the evidence contradicting their Rule 36 admissions, he was relying on David v. Caterpillar, Inc., 324 F.3d 851, 856-57 (7th Cir. 2003).
That was an abuse of discretion because the judge applied the wrong law, the wrong legal standard. See, e.g., Cooter & Gell v. Hartmarx Corp., 496 U.S. 384, 402, 110 S.Ct. 2447, 110 L.Ed.2d 359 (1990); Ervin v. OS Restaurant Services, Inc., 632 F.3d 971, 976 (7th Cir. 2011). The David case dealt with parties’ more general duty to supplement discovery responses under Rule 26(e) and sanctions under Rule 37. By relying on David and handling this issue as he did, the judge failed to appreciate: (1) the effect of a Rule 36 admission; (2) the procedure for withdrawing such an admission; and (3) the different standard that applies under Rule 36(b). See 8B Charles Alan Wright & Arthur R. Miller, Federal Practice and Procedure § 2264 at 400 (3d ed. 2010) (“To the extent that there is an inconsistency, the specific provision of Rule 36(b) should control over the general provision of Rule 26(e)”; withdrawal or amendment of Rule 36 admission must satisfy requirements of Rule 36(b)).
First, the judge overlooked the most important difference between other discovery responses and Rule 36 admissions: admitted matters are not merely evidence but are “conclusively established.” Rule 36(b); Banos v. City of Chicago, 398 F.3d 889, 892 (7th Cir. 2005) (affirming denial of leave to withdraw admissions). Advisory committee notes show that clarifying just this point was a chief purpose of the 1970 amendments to Rule 36(b).
Second, if a party wants to withdraw or amend a Rule 36 admission, Rule 36(b) expressly requires a “motion.” No motion was filed here. See Kalis v. Colgate-Palmolive Co., 231 F.3d 1049, 1059 (7th Cir. 2000) (affirming decision based on admissions where responding party never filed a motion to withdraw); American Automobile Ass’n v. AAA Legal Clinic, 930 F.2d 1117, 1120 (5th Cir. 1991) (reversing admission of evidence contradicting Rule 36 admission without motion to withdraw or amend admissions); 999 v. C.I.T. Corp., 776 F.2d 866, 869 (9th Cir. 1985) (affirming refusal to allow evidence at trial to contradict Rule 36 admission). And as the AAA ' and 999 courts made clear, trial is ordinarily far too late even for a formal motion to withdraw or amend.
Third, even if we overlook the lack of a motion, the substantive standard for withdrawal or amendment under Rule 36(b) is much more demanding than the David standard the trial judge applied here. Rule 36(b) provides:
Subject to Rule 16(e) [endorsing importance of final pretrial order that may be modified only to prevent “manifest injustice”], the court may permit withdrawal or amendment if it would promote the presentation of the merits of the action and if the court is not persuaded that it would prejudice the requesting party in maintaining or defending the action on the merits.
Admissions are not merely evidence to weigh. They are judicial admissions. They cannot be contradicted by the responding party without the court’s permission to withdraw or amend them. E.g., United States v. Kasuboski, 834 F.2d 1345, 1350 (7th Cir. 1987) (“Unless the party securing an admission can depend on its binding effect, he [or she] cannot safely avoid the expense of preparing to prove the very matters on which he [or she] has secured the admission, and the purpose of the rule is defeated.”), quoting 1970 Fed. R. Civ. P. 36 advisory committee note; Airco Industrial Gases, Inc. v. Teamsters, 850 F.2d 1028, 1037 (3d Cir. 1988) (reversing district court’s failure to give Rule 36 admission *575binding effect; “This admission is not merely another layer of evidence, upon which the district court can superimpose its own assessment of weight and validity. It is, to the contrary, an unassailable statement of fact that narrows the triable issues in the case.”); Williams v. City of Dothan, 818 F.2d 755, 762 (11th Cir. 1987) (reversing district court’s refusal to recognize Rule 36 admission); Brook Village North Associates v. General Electric Co., 686 F.2d 66 (1st Cir. 1982) (reversing district judge’s decision to permit evidence contradicting Rule 36 admissions); see generally 8B Wright & Miller, Federal Practice and Procedure § 2264 (3d ed. 2010).
Under the correct Rule 36(b) standard, allowing withdrawal or amendment in this case would have been an abuse of discretion. Defendant Glatt had been relying on the Rule 36 admissions for two years. Plaintiffs waited until at least the eve of trial to try to avoid their admissions (though without filing a Rule 36(b) motion admitting what they were doing). The correct standard here should have required a showing of “manifest injustice” consistent with Rule 16(e). That certainly was not shown in light of plaintiffs’ gamesmanship regarding the admissions.
In a mistaken effort to minimize the prejudice to Glatt, the majority writes: “And two months after Tate had submitted its very limited admissions, it supplemented them by presenting evidence that no fewer than 16 filters installed in the Gran-ulator before the fire had been damaged.” Ante at 571. The majority is referring to plaintiffs’ mere production of one email as part of a much larger document production, without comment or explanation. That is not a fair basis for excusing plaintiffs’ or the court’s error. Merely producing an unexplained document does not signal to the requesting party that the admitting party intends to withdraw or amend its admissions. That’s why Rule 36(b) requires a motion—to give fair notice of an intent to change positions from the formal admission.
No trial is perfect, and I do not vote lightly to retry a case like this. But this was not a minor technical error. Glatt was entitled to rely on the Rule 36 admissions. It was prejudiced at trial by the trial court’s use of the wrong legal standard to nullify those admissions at trial. By treating the trial court’s error in this case as a “technical error,” the majority fails to grasp the unfair advantage that plaintiffs gained by these tactics. By tolerating these tactics, the majority opinion will muddy the waters unnecessarily in this branch of federal civil practice. And by undermining justified reliance on Rule 36 admissions, the majority opinion is also likely to reduce the benefits of using Rule 36, which will tend to raise the costs of civil litigation. For these reasons, I hope trial judges in this circuit and beyond will respond to this odd outlier of a decision by resisting any temptation to follow the shortcuts tolerated in this case and by turning square corners in applying Rule 36(b). I respectfully dissent.
Appendix
Plaintiff Tate & Lyle Americas LLC Response to Defendant Glatt Air Techniques Inc.’s Request to Admit. Separate Appendix for Defendant-Appellant at 128-29.
7. After the August 2009 filter repairs to the Granulator were completed to the time of the September 17, 2009 tire, you did not observe any cracked, chipped, or poorly fitting filters in the Granulator.
RESPONSE: It denies the allegations in Paragraph 7. T&L is not aware of any filter repairs performed. If this Request was intended to ask about complaints after the granulator filter housing repair efforts were made, it is admitted T&L is not presently aware of anyone that observed *576cracked, chipped, or poorly fitting filters in the Granulator following work performed by Dynacoil in.August-or September of 2009. Answering further, Reinard Ortloff advised David O’Connor, William Wenrich, Paul Buser, Steve .Sirabian, Ulrich Walter, on July 80, 2009, that it was possible some filter elements had stress on the support in the GFG and if that was the case there was a high risk that new filter elements would break- again. If this Request was intended to ask about observations of cracked,, chipped or poorly fitting filters, after the granulator filter housing repair efforts were made, it is admitted it is not currently aware of any such observations'. 8. After the August 2009 filter repairs to the Granulator were completed to the time of the September 17, 2009 fire,- no one informed you that one or more of the Granulator filters o[r] filter housings were cracked, chipped, or poorly fitting.
RESPONSE: It denies the allegations in Paragraph 8. T&L is not aware of any filter repairs performed. T&L is not presently aware of anyone that observed cracked, chipped, or poorly fitting filters in the Granulator following work performed by Dynacoil in August or September of 2009. Answering further, Reinard Ortloff advised David O’Connor, William Wenrich, Paul Buser, Steve Sirabian, Ulrich Walter, on July 30, 2009, that it was possible some filter elements had stress on the support in the GFG and if that was the case there was a high -risk that hew filter elements would break again. The filters provided with the Granulator, -even if not cracked, chipped or poorly fitting, had exposed metal in the frame. If this Request was intended to ask about information given of cracked, chipped or poorly fitting filters, after the granulator filter housing repair efforts were made, it is admitted it is not currently aware of any such information given.
9. You have not discovered any evidence that one or more of the Granulator filters or filter housings [w]as cracked, chipped or poorly fitting at the time the September 17,2009 fire started. ’ •
RESPONSE: It admits the allegations in Paragraph 9 as it pertains to cracked or chipped filters or housing. T&L has not discovered direct evidence that one or more of the Granulator filters or filter housings was cracked or chipped at the time the fire started. For further answer, Reinard Ortloff advised David O’Connor, William Wenrich, Paul Buser, Steve Sira-bian, Ulrich Walter, on July 30, 2009, that it was possible some filter elements had stress on the support in the GFG and if that was the case there was a high risk that new filter elements, would break again. Answering further, and denying the allegations as it relates to poorly fitting filters and filter housing, following the fire, additional work needed to be done to the filter housing because filters were poorly fitting which would suggest the filter housing and filters remained poorly fitting even at the time of the fire.
. While parts of the responses were convoluted and as the majority says ‘‘.not unqualified,” the qualifications were not relevant to the ' way in which plaintiffs' contradicted their admissions. The "qualifications” in the admissions speculated that the “new” filters installed after the August 2009 repairs could have . experienced stress and cracked. Plaintiffs offered evidence at trial, however, that the old cracked filters were reinstalled in the granulator after the repairs.