concurring. I agree with the rescript reversing the judgment of the Appellate Court and ordering a new trial, but I disagree with the view of the majority that “a per se rule of automatic reversal more properly vindicates the denial of the defendant’s fundamental constitutional right to assistance of counsel guaranteed by the sixth amendment.” Because the decision of the United States Supreme Court in Geders v. United States, 425 U.S. 80, 96 S. Ct. 1330, 47 L. Ed. 2d 592 (1976), leaves unresolved the issue of whether automatic reversal is required where a court prohibits communication between attorney and client during a trial recess of brief duration, I would follow the opinion of the United States Court of Appeals for the Second Circuit in United States v. DiLapi, 651 F.2d 140 (2d Cir. 1981), cert. denied, 455 U.S. 938, 102 S. Ct. 1427, 71 *604L. Ed. 2d 648 (1982), which rejected the per se reversal rule and applied a harmless error analysis where the communication ban was in effect only for a recess of five minutes. Such an analysis requires, however, that the state demonstrate that the error in barring attorney-client communication was harmless beyond a reasonable doubt. I disagree with the conclusion of the Appellate Court in applying such an analysis that the state has sustained its burden under this standard, which is applicable to errors of constitutional dimension.
It is a difficult proposition to maintain that a criminal conviction must be reversed because of a trial court error that, beyond a reasonable doubt, has resulted in no prejudice to a defendant and thus is harmless. One possible purpose of such an automatic reversal rule, however, maybe to advance values beyond those involved in the case under consideration in a manner not adequately accomplished by mere appellate recognition of an error. See Vasquez v. Hillery, 474 U.S. 254, 262, 106 S. Ct. 617, 88 L. Ed. 2d 598 (1986); Rose v. Mitchell, 443 U.S. 545, 578, 99 S. Ct. 2993, 61 L. Ed. 2d 739 (1979) (Stewart, J., concurring). The United States Supreme Court, however, has never articulated such a basis for its declaration that “there are some constitutional rights so basic to a fair trial that their infraction can never be treated as harmless error . . . .” Chapman v. California, 386 U.S. 18, 23, 87 S. Ct. 824, 17 L. Ed. 2d 705, reh. denied, 386 U.S. 987, 87 S. Ct. 1283, 18 L. Ed. 2d 241 (1967). Presumably, therefore, those infringements of a defendant’s constitutional rights in a criminal case where harmless error analysis is deemed inappropriate involve situations where the constitutional violation must be deemed harmful under all circumstances or where it is of such a nature that ascertainment of its harmfulness is impossible or so difficult as not to warrant the *605endeavor. Such cases are rare exceptions to “the principle that an otherwise valid conviction should not be set aside if the reviewing court may confidently say, on the whole record, that the constitutional error was harmless beyond a reasonable doubt.” Delaware v. Van Arsdall, 475 U.S. 673, 681, 106 S. Ct. 1431, 89 L. Ed. 2d 674 (1986).
Although the denial of the assistance of counsel at a trial has been placed within the narrow category of constitutional violations requiring automatic reversal, it does not necessarily follow that every erroneous interference with attorney-client communication, however brief or inconsequential, is inevitably harmful. The unexpressed concern of courts that have indulged this nonsequitur rather than engage in a harmfulness evaluation is that there often is nothing in the record from which a reviewing court can determine the effect of an erroneous ban on communication. Consequently, it is assumed that resort to speculation in such cases is likely in estimating the prejudicial impact of the error. The requirement under harmless error analysis that the state demonstrate harmlessness beyond a reasonable doubt, however, should dispel such a fear. Where the appellate record contains nothing to support a conclusion that the communication ban is harmless beyond a reasonable doubt, under that doctrine a reversal is required.
For a similar reason the concern expressed in the majority opinion that a harmless error analysis may somehow trespass upon the attorney-client privilege is unwarranted. The state has the burden of providing a record from which harmlessness can be found, while the defendant and his counsel may remain silent, except for making a bare objection, which the majority opinion incongruously requires to preserve this constitutional error claim. See State v. Evans, 165 Conn. 61, 70, 327 A.2d 576 (1973). Even if something more were *606required to protest a communication embargo, it would not be necessary for defense counsel to reveal his purpose in wishing to confer with his client. Furthermore, the attorney-client privilege protects communications that have taken place and is not at all applicable to those that have not occurred because they were forbidden.
It is not difficult to envision situations in which the record may demonstrate harmlessness in accordance with the constitutional standard. For example, if the prohibition against communication with a defendant has been limited to a brief recess occurring during cross-examination and the additional cross-examination following the recess has been clearly insignificant, there would ordinarily be a sufficient basis to find the standard to have been satisfied. Indeed, those courts that restrict automatic reversal to cases in which defense counsel have raised objection to a communication ban, as does the majority opinion, appear to engage in a limited harmless error analysis, reasoning that the absence of an objection indicates that counsel had no desire to discuss anything with his client. Stubbs v. Bordenkircher, 689 F.2d 1205, 1207 (4th Cir. 1982), cert. denied, 461 U.S. 907, 103 S. Ct. 1879, 76 L. Ed. 2d 810 (1983); Bailey v. Redman, 657 F.2d 21, 24 (3d Cir. 1981), cert. denied, 454 U.S. 1153, 102 S. Ct. 1024, 71 L. Ed. 2d 310, reh. denied, 455 U.S. 995, 102 S. Ct. 1624, 71 L. Ed. 2d 857 (1982). The emphasis placed by the court’s opinion in Geders itself upon the fact that the communications ban had continued for a period of seventeen hours, including an overnight recess, and must, therefore, have interfered with the normal discussion between attorney and client of the progress of the case also indicates that harmfulness was a factor in determining whether a deprivation of the right to counsel had occurred. Because I am reluctant to expand the category of constitutional errors for which automatic reversal is mandated, and authoritative prece*607dent remains indecisive upon the subject, I choose not to join in the majority opinion.
I agree with the result, nevertheless, because the state has not proved that the “sequestration” order of the trial court was harmless beyond a reasonable doubt. In my judgment, the Appellate Court erred in its application of that standard to the circumstances of this case. Although the Appellate Court construed the trial court’s order as effective during only the twenty-one minute recess following its issuance, the order may well have been viewed by the defendant or his counsel as precluding all discussion between them until completion of his testimony. No limitation was placed on the duration of the order and it followed immediately the issuance of a sequestration order for all defense witnesses, presumably effective throughout the remainder of the trial. The court in granting the motion relating to the defendant specifically ordered that “the witness is to be sequestered,” although the state had requested only that the defendant be ordered “not to talk to his counsel since [the state’s attorney was] in the middle of cross-examination.”
The cross-examination following the recess consisted of a single question to which the court sustained an objection, although the defendant’s negative answer was never stricken. If the testimony of the defendant had then terminated, I would agree with the Appellate Court that the record sufficiently supported a conclusion that the prohibition of communication between the defendant and his client had not resulted in any harm. There was, however, additional questioning on redirect concerning the circumstances of the defendant’s arrest, intended to cast doubt upon the identification of the defendant by the police. An extensive recross-examination followed in which the state’s attorney attempted to impeach the defendant’s version of the arrest as well as his credibility. During this exami*608nation the court remarked to defense counsel upon the defendant’s habit of answering a question after his counsel had objected thereto, “[s]o whatever comes out, that’s his problem.” At counsel’s response that he had not had the opportunity to talk to the defendant during the recess when he “could have reminded him at that time,” the court conceded the accuracy of that observation.
The recross-examination was followed by additional redirect examination and then by further recross-examination and another brief redirect examination before court adjourned for the day. On the following day the defendant was recalled for redirect examination and was again cross-examined.
Even if the communication ban had clearly been in effect only during the twenty-one minute recess, as the Appellate Court assumed, a highly significant portion of the defendant’s testimony followed the recess. This testimony might well have been affected by the absence of any opportunity for the guidance of counsel during the recess. The court’s remark about the defendant’s propensity to answer questions before his counsel’s objection had been ruled upon, which had been evident before the recess, indicates that proper coaching during the recess would have helped the defendant’s later testimony. When a defendant testifies at his trial, his credibility is critical and the value of suggestions of counsel during a recess after he has observed his client testify cannot be minimized. There is nothing in the record that I can perceive to support the conclusion of the Appellate Court that the state has sustained its burden of proving the error harmless beyond a reasonable doubt.
Accordingly, I concur in the result.