concurring in part, and dissenting in part:
I agree with much of the majority opinion. However, I would affirm in part, and reverse in part.
In circumstances where there was uncon-troverted evidence of a history of beatings with belts and extension cords that left visible and permanent sears, there is hardly dispute on the merits of the findings of neglect and abuse. Nonetheless, as the trial continued, some of the questions and comments of the presiding judge became *80more hostile to the mother of the children. Indeed the judge, himself, recognized this as a problem. Thus, in keeping with the Canons of judicial conduct, serious questions of judicial impartiality were introduced into the proceedings.
Generally, in cases of this nature, alleged judicial bias must have arisen from an extrajudicial source and rest on some basis other than what the judge learned from participation in the case. In re Bell, 373 A.2d 232, 233 (D.C.1977). This jurisdiction has a practice, embodied in Super.Ct.Civ.R. 63-1, which requires a party alleging bias to file, in good faith, an affidavit asserting the basis for the claim prior to the proceeding. When this is appropriately done, such judge shall proceed no further.
In this instance, appellant responds that, prior to trial, there was no known identifiable personal basis for seeking recusal. It was only during the proceedings, that the judge’s behavior caused concern. This is an important point of focus. Although the requirement of judicial impartiality, and the appearance of same, applies to all proceedings, we must consider whether the assertion of a conflict of interest on the part of a judge is essentially no different — for purposes of appellate review — than intemperate judicial behavior. Of course, our recu-sal rule, and the decisions in Liljeberg v. Health Servs. Acquisition Corp., 486 U.S. 847, 108 S.Ct. 2194, 100 L.Ed.2d 855 (1988), and Scott v. United States, 559 A.2d 745 (D.C.1989) (en banc), specifically address the former circumstance. Thus, the Lilje-berg analysis for harmless error looks only to the (1) risk of injustice to the litigants, (2) the risk of injustice in other cases, (3) and the risk of undermining public confidence in the judiciary.
Recognizing that the requirements of judicial impartiality, and the appearance of impartiality, apply with equal force to the instant case, it is my view that the nature of the present assertion — that the judge’s intemperate courtroom behavior reflected at least the appearance of partiality— makes our review more akin to the usual review for harmless error. That is, we should assuredly include in the consideration of factors the weight of the evidence.
In a context, as here, where the findings of abuse and neglect are virtually beyond dispute, but where the challenge to the disposition decisions do raise unanswerable questions regarding at least the appearance of judicial impartiality, it is my view that we should remand the case for only a new disposition hearing. In doing so, I believe we would be entirely faithful to the concerns and protections afforded by the requirement of judicial impartiality, and the appearance of it, but we legitimately need not order a second “trial” regarding the underlying events. The thrust of the challenge in this case is not so much to reliti-gate the treatment of the children, but rather to get a balanced or fair hearing on the future course of action for all the persons involved. The children, of course, are not litigants, but merely passive participants in a procedure designed to recognize everyone’s rights, but also to serve their best interest. In this unusual situation, I think we can prudently remedy the problem by a remand for a new disposition hearing.
The majority opinion, at note 12, states that the dissent overlooks the fact that the primary issue concerns the appearance of bias. Starting with the premise that not every error of this nature requires a per se reversal, I suggest that our true difference turns on the appropriate scope of harmless error review in the present circumstances.