(concurring separately):
Justice Stewart’s dissenting opinion states that “a majority of the Court has never decided whether the unanimity instruction was correct or harmless error if incorrect.” I think that he is wrong. Chief Justice Hall and Justice Howe considered the issue and concluded that the instruction was correct because of their (minority) view on the requirements of the unanimity standard.1 Justice Stewart agreed with this conclusion, albeit on different grounds. He concluded that una*223nimity was required but that the instruction was adequate to convey this to the jury.2 Justice Zimmerman and I concluded that the instruction was incorrect because it did not inform the jury that unanimity was required.3 Therefore, each member of this court considered whether the instruction was correct, and a majority of the court held that it was.
In a footnote, Justice Stewart noted in dicta his view that the instruction would be harmless even if it were erroneous.4 According to his holding, however, the instruction was correct. No other justice joined in his harmless error dicta. Chief Justice Hall and Justice Howe presumably refrained because they found no error in the instruction under the standard they adopted for jury unanimity. I concluded that the instruction was “entirely inadequate” to inform the jury that it was required to decide unanimously on a specific aggravating factor to convict Tillman of first degree murder. State v. Tillman, 750 P.2d 546, 588 (Utah 1987) (Durham, J., concurring and dissenting). I did not consider whether the failure to instruct the jury properly on the aggravating circumstances could have been harmless, however, because I concluded that there was no basis for a first degree murder conviction at all under the merger doctrine. Id. Justice Zimmerman joined in that view. Id. at 591 (Zimmerman, J., concurring and dissenting).
The end result is that three justices joined in a holding predicated on the correctness of the instruction.5 Justice Stewart continues in this habeas review to subscribe to that holding, but inexplicably proceeds to conduct a harmless error review of the giving of the instruction. As he says in his opinion, “[I]t is necessary ... to address on the merits in this habeas proceeding the issue of whether the failure to give a correct instruction on unanimity was harmless error.” I do not understand how he can legitimately analyze the “failure to give a correct instruction” when he and two other justices held that the instruction was correct on direct appeal.6
I did not agree then, and do not agree now, with the dicta in his footnote and the analysis in his current opinion that the error was harmless in the guilt phase of the trial. Finally, I do not agree with his finding in this opinion that the error was harmful as to the penalty phase. I am not persuaded by his argument that the failure to require unanimity in the guilt phase affected the penalty phase. Even if the guilt phase instruction might have caused confusion in the penalty phase, it was clearly corrected by instruction 6 at the penalty phase, which informed the jury that “you may consider as aggravating circumstances only those circumstances listed above which you, as a jury, unanimously found to exist beyond a reasonable doubt.” (Emphasis added.) But I do not treat the question of harmfulness in any further detail, since neither Justice Stewart nor the court has ever held that the instruction was error in the first place; I regard the issue *224as purely hypothetical. I note, however, that because I do not join Justice Stewart’s conclusion that the instruction was harmless in the guilt phase or harmful in the penalty phase, it is not our disagreement over the holding in the direct appeal which decides this case.
I have not changed my views on this conviction and penalty since the direct appeal, but I regard myself bound by the majority holding in that opinion and therefore concur in the majority opinion’s affir-mance herein.
. See State v. Tillman, 750 P.2d 546, 562-68 (Utah 1987) (plurality opinion by Hall, C.J.); id. at 582 (Howe, J., concurring in the result on this point).
. See id. at 577-79 (Stewart, Assoc. C J., concurring and concurring in the result).
. See id. at 585-88 (Durham, J., concurring and dissenting); id. at 591 (Zimmerman, J., concurring and dissenting).
. Id. at 580 n. 1 (Stewart, Assoc. C.J., concurring and concurring in the result).
. Justice Stewart appears to assume that when a majority of the court agrees on a conclusion but they do so under different theories, there is no binding majority holding. This is incorrect. On direct appeal, three justices of this court concluded that the instruction was correct. The only way for the question to come properly before us for disposition now would be for Justice Stewart, Justice Howe, or Chief Justice Hall to change his view and determine that the instruction was in fact error. Chief Justice Hall and Justice Howe might do so by applying the majority holding on unanimity from the direct appeal to the language of the instruction, an analysis they did not do at that time, and concluding that the instruction was erroneous. Justice Stewart could only do so by abandoning his earlier holding that the instruction was correct. Justice Zimmerman and I are not free to ignore the holding unless one of the majority who created it departs from it.
.Contrary to the suggestion in Justice Stewart’s opinion, what this court has previously held in the same case is not a mere "nicety." Rather, it goes to the core of the legitimacy of the exercise of judicial power.