dissenting.
The judgment of the Court of Appeals should be affirmed without remand, except perhaps in light of Almanza v. State, 686 S.W.2d 157 (Tex.Cr.App.1985). Govan v. State, 682 S.W.2d 567 (Tex.Cr.App.1985) should be overruled with regard to the language approved on page 3 of the majority opinion:
[T]he State argues that such error [failure to apply the law of vicarious culpability to the facts] may be harmless if the evidence clearly supports appellant’s guilt as a principal actor. We agree.
A jury should never be authorized to convict on a theory not supported by evidence. To the extent that Cumbie v. State, 578 S.W.2d 732 (Tex.Cr.App.1979), and other cases support this proposition, they were not overruled by Almanza. Al-manza merely held that such defects do not produce a reversal automatically without an inquiry regarding the degree of harm, if any.
When there is evidence that the criminal acts alleged in the indictment were committed by a person other than the accused, but that the accused is nevertheless criminally responsible for those acts, it is of course not erroneous for the trial judge to authorize conviction of the accused upon this theory. But unless the jury is authorized to do so by an instruction which specifically applies the law of vicarious culpability to the facts, expressly requiring the jury to find, among other things, that some person other than the accused committed the offense alleged, the jury charge is objectionable. The majority admits as much by concluding that it was error for the trial judge to refuse an application paragraph on the law of parties in the instant cause.1
The only remaining inquiry is that required by Almanza. Since the error was sufficiently preserved, according to the majority’s analysis, appellant’s conviction must be reversed if “some harm” has been demonstrated.
If we hold that the jury was authorized by evidence to convict the appellant as a nonprimary culpable party, we cannot simply assume without knowing that it did not do so, regardless of whether we think the evidence was legally sufficient to support his conviction as a primary actor. The jury, as sole factfinder, was free to conclude from the evidence that appellant was not a primary actor, even if the evidence was also sufficient in contemplation of law to support a contrary conclusion. Thus, as an appellate tribunal, we are not at liberty to assume that the jury in this cause did not convict the appellant under a theory of vicarious culpability.
In order to find the error harmless it would, therefore, be necessary to conclude not only that the evidence was sufficient to support conviction as a primary actor, but also that the jury did in fact convict upon this theory. And this we do not and cannot know. Therefore, if for purposes of argument we assume rather that the jury did not convict appellant as a primary actor, can we say that he did not suffer “some harm” from the concededly erroneous failure of the trial judge to apply the law of vicarious culpability to the facts of this case, admitting as we must that the evidence would have been adequate to raise the issue and to support an affirmative finding by the jury upon it? Govan auto*683matically answers this question in the affirmative. Almanza forbids automatic answers.
In the face of a timely and specific objection, such as the one interposed in this case, a showing of “some harm” has been demonstrated because there is a logical possibility, supported by evidence, that the jury convicted appellant upon a theory of vicarious culpability without being expressly informed of the facts it must find before doing so. The rule of Govan precludes a finding of harm in such context.
I would, therefore, overrule Govan to the extent that it holds a failure to apply the law of vicarious culpability to the facts is automatically harmless whenever there is sufficient evidence to convict the accused as a primary actor. Moreover, to remand this case for reconsideration by the Court of Appeals would be wasteful of judicial resources. But in any event I would affirm the Fifth Court of Appeals or, at most, vacate and remand for reconsideration in light of Almanza only.
In this connection, I agree with the majority’s characterization of Almanza in footnote 2, except as follows. First, although Almanza speaks of “actual harm,” I do not understand the Court’s opinion in that case to suggest that a reliable prediction of a jury’s actual reasoning can ever be made. Second, I do not understand Almanza to imply that an examination of the pleadings, evidence, and jury argument is to be given greater deference in the assessment of harmfulness than the jury charge itself.
Having said that, my disagreement with some of Almanza’s language is a matter of record. See, Hayes v. State (Tex.Cr.App. No. 955-82, delivered March 12, 1986, Teague, J., dissenting). I am convinced that harmfulness can never rationally be made to depend upon the probability of a jury returning a different verdict absent charging error. The data necessary to make such predictions are simply not available to this Court. It is far better in my view, and more in keeping with the structure of our legal system, to insist upon legally acceptable jury instructions and leave factfinding to the triers of fact.
Nevertheless, while there may be a certain tension between my “logical possibility” test, articulated in Hayes, and a literal reading of certain language in Almanza, I do not regard the two as patently inconsistent. Indeed, it is clear to me, and should be equally clear to anyone who has read this Court’s opinions applying Almanza, that no uniformly comprehensible method is at work. This only goes to show that Almanza is in need of some refinement. My opinion in Hayes was an effort in this direction. To my knowledge, no other member of this Court has suggested an alternative methodology which can be intelligibly applied in practice.
With these cautions in mind, several points should be made about my assessment of the instant cause. Although I have looked principally to the jury charge, consistent with my opinion in Hayes, I have not simply ignored the statement of facts and arguments of counsel. With regard to the former, I find, as does the majority, that evidence was adduced raising an issue of vicarious culpability. Thus, there was a factual, not merely a theoretical, basis upon which the jury in this cause might have convicted the appellant as a nonpri-mary culpable party. Had there been no evidence to this effect, it would have been erroneous for the trial judge to authorize conviction upon this theory, either abstractly or by means of a complete application paragraph. But that is not the issue. In the present context, it is enough to observe that the question of vicarious culpability arose during trial, not by gratuitous submission of a jury question unsupported by evidence, but by the evidence itself.
Secondly, the majority fails to discuss, not only those portions of the statement of facts pertinent to the inquiry required by Almanza, but any relevant portion of the jury argument either. If this Court continues to insist that the entire record should be examined for harm, surely it is reasonable to expect that relevant portions will be *684identified in its opinion and that the significance of such portions will be revealed.
With specific regard to the jury argument, it is true that vicarious culpability was not strenuously urged by the State as a distinct theory upon which the jury might predicate a finding of guilt. In spite of my belief that the probable impact of argument by counsel on jury deliberations cannot fairly be gauged without rank speculation, I will concede that juries as a whole are probably somewhat less likely to convict upon a theory not urged by either party than upon a theory strenuously urged. However, without speculating, I am unable to say how much less likely, if at all, it was in this case. Accordingly, I am inclined to afford the circumstance of jury argument little weight in my harm analysis. If I could gamer a majority of this Court, I would abandon the emphasis on jury argument altogether in determining whether there has been “some harm” under the test prescribed by Almanza. Correct jury instructions are, among other things, designed precisely to eliminate the need for speculation by appellate courts.
Be that as it may, the record in the instant cause, taken as a whole, does not demonstrate that the erroneous jury instructions were harmless, nor does the majority disclose which parts of the record should be deemed relevant in this connection, or why. True, I have looked to the charge itself, but what is wrong with that? Almanza does not prohibit my doing so, and the charge is, after all, most important to resolve the issue. The majority, on the other hand, evidently assigns no importance to the fact that the jury was expressly authorized to convict appellant without being required specifically to find each element necessary for a determination of vicarious culpability.
I have also looked to the evidence and found in the record some support for a proper jury instruction on vicarious culpability. I do not take this to be an indication of harmlessness.
I have looked to the jury argument. There is nothing in it strongly suggesting that the jurors convicted upon a theory of vicarious culpability. But there is also nothing in it strongly suggesting that they did not. In any case, the jury argument does not militate in favor of finding harmlessness, and the majority itself does not claim otherwise.
I am astonished, in light of this record, that the majority represents the Govan rule to be based on “realistic probabilities.” There is, on the contrary, nothing in its analysis which could honestly be called either realistic or probabilistic. The majority instead contends that a jury acts irrationally whenever it finds an accused guilty as a nonprimary culpable party in the face of evidence not legally insufficient to convict him as the primary actor. This proposition is outrageous! Soon this Court will be saying that a jury is irrational when it convicts of a lesser included offense upon evidence not legally insufficient to convict of the greater, or some other such nonsense. This kind of reasoning is backwards, takes no account of the proper role alternative jury instructions serve in our system of jurisprudence, and proclaims the demise of law in the factfinding process.
I recognize, of course, that this Court has ordered this cause remanded to the Court of Appeals for a harm analysis, which perhaps explains the absence of any harm analysis in its own opinion. This would at least make some sense to me if the case were remanded for reconsideration in light of Almanza. But instead, the question of harm has been remanded for consideration in light of the Govan rule only. To iterate the blasphemy, the Govan rule, succinctly stated, provides that “[w]here the evidence clearly supports a defendant’s guilt as a principal actor, any error of the trial court in charging on the law of parties is harmless.” See, p. 675. This is a “per se rule” and does not indicate a need for remand. Accordingly, the Court of Appeals can be expected simply to apply the rule in a straightforward fashion without regard to the sufficiency of evidence to support conviction, particularly since the appellant did not complain of evidentiary *685insufficiency on appeal. The majority’s action today perpetuates a rule which is at odds both with logic and with Almanza. And it ill serves the Dallas Court of Appeals, which will inevitably be left “groping in the dark for a light switch.”
In view of Almanza, I therefore urge the Dallas Court on remand to clearly state its reasons for finding the error in this case harmless, if it does so — to lay on paper the evidence which affirmatively demonstrates no prejudice to the appellant here. Perhaps the effort alone will prove illuminating, since it will surely confirm the futility of such an enterprise. But even if it does not, at least bench and bar will have been given the analysis promised by Almanza, and not the wrongheaded “automatic af-firmance” of Govan.
. I am compelled to disagree with the conclusion that Presiding Judge Onion reaches in the dissenting opinion that he files, that the application paragraph in the charge on the law of parties was sufficiently specific to meet appellant’s objection. To have met the appellant’s objection, it would have been necessary for the trial judge to apply specifically what he charged the jury in the abstract on the law of parties. His failure to do so constitutes error.