(concurring as to the result). Deck v Missouri (544 US 622 [2005]) addresses what the Constitution requires to rectify the due process violation occurring “where a [trial] court, without adequate justification, orders the defendant to wear shackles that will be seen by the jury ” (id. at 635 [emphasis supplied]). In contrast to Deck (see id. at 634), there is no indication that the leg irons defendant Cruz was forced to wear during his trial were, in fact, visible to the jury and I do not believe that this case comes within Deck’s analytical framework by reason of the circumstance that there may exist some residual, unnegated possibility that defendant’s shackles were seen or that their existence was inferred (see majority mem at 944-945). Thus, while it is clear that People v Clyde (18 NY3d 145 [2011] [decided today]), involving the visible, indeed blatant, use of shackles without record justification, was necessarily decided in accordance with Deck, it is not similarly clear that the analytic course charted in Deck must be followed here. The question posed on this appeal — one that did not have to be addressed in Clyde — is what the appellate response should be when a criminal defendant has, without adequate record justification, been forced to wear a physical restraint during trial, but there is absent from the record any indication that the restraint was actually perceived by the jury. This is an inquiry to which I believe we have already spoken as a matter of state law in People v Buchanan (13 NY3d 1, 4 [2009]).
In Buchanan, as here, there was no indication that the jury had become aware of the restraint, a stun belt, which was concealed by Buchanan’s outer clothes. And, the People argued as they do here that “Deck is distinguishable because the stun belt here was not visible to the jury” (Buchanan, 13 NY3d at 4). We *946replied that it was not necessary to decide whether there had been a constitutional due process violation since “as a matter of New York law . . . it is unacceptable to make a stun belt a routine adjunct of every murder trial, without a specifically identified security reason” (Buchanan, 13 NY3d at 4). While shackles are different from stun belts, the principal underlying rationale for requiring judicial justification for restraining a defendant at trial is the same. A defendant is presumed innocent and must be treated by the State, to the extent possible, in a way that is consonant with that presumption. As we said in People v Roman, “[a] defendant is presumed innocent and he is entitled to appear in court with the dignity and the self-respect of a free and innocent man” (35 NY2d 978, 979 [1975]). If a stun belt may not, under New York law, be a routine adjunct of a murder trial, neither may shackles be a routine adjunct of a trial for burglary.
We did not engage in harmless error analysis in Buchanan, but rather reversed, apparently, solely by reason of the error in requiring the restraint without case specific record justification. While we did not explain this summary approach, there is much to recommend it.
Although the mere hypothesis that an unjustified shackling has affected a verdict — and that is all there is here — is not enough to sustain a Deck analysis, since the essential premise of requiring the People to prove beyond a reasonable doubt that the shackling did not contribute to the verdict is that there has been prejudice from the jury’s actual perception of the defendant in restraints, the concern that seems to underlie the Court’s expansion of Deck's applicability is entirely justified. The introduction of a physical restraint into a criminal trial unavoidably risks its discovery, either through mischance or because the elaborate measures taken to keep the restraint from the jury’s awareness will themselves disclose to the jury that there is something about the defendant’s appearance that is being kept from them; as the Court has noted, the misgivings of defendant’s attorney as to the efficacy of the bunting used to shield the shackles from the jury’s eyes — particularly when it was used only at the defense table — were not misplaced. These risks to the fairness of the proceeding are not ones that should be routinely undertaken. Nor should other likely consequences of employing restraining devices on a defendant at trial — the burdening of representational rights and the right to be present at all material stages of the trial and, indeed, to testify in one’s behalf — be minimized, even though the manner in which they *947affect a verdict may be indirect and exceedingly difficult to account for accurately in a harmless error analysis. For these reasons it would seem prudent to attach a definite consequence to a trial court’s failure to make the findings essential to justify the use of physical restraints — a consequence that does not depend upon the limitations and vagaries of harmless error analysis. That is what we did in Buchanan. Given the very basic interests at stake, and the ease with which they can properly be afforded the judicial consideration they are due, there should be a clear rule that the failure to make a record to justify restraining a defendant at trial will necessitate a new trial. We observed in Buchanan that “a trial court has broad discretion in deciding whether a restraint is necessary for courtroom security” (13 NY3d at 4) and that a formal inquiry was not required so long as the requisite record was made. This is not an onerous condition of a decidedly onerous imposition. There is no reason why it should not be met.
Judges Graffeo, Read, Smith and Pigott concur in memorandum; Chief Judge Lippman concurs in result in an opinion in which Judges Ciparick and Jones concur.
Order reversed, etc.