OPINION ON STATE'S PETITION FOR DISCRETIONARY REVIEW

BENAVIDES, Judge.

Appellant was convicted of conspiracy to commit arson and his punishment assessed at fifteen years confinement in the penitentiary. The San Antonio Court of Appeals reversed and remanded the cause for a new trial because the jury had not been instructed (1) that "[a] conviction cannot be had upon the testimony of an accomplice unless corroborated by other evidence tending to connect the defendant with the offense committed[,]" and (2) that appellant's codefendant was an accomplice as a matter of law. Art. 38.14, V.A.C.C.P. Although appellant made no request for such an instruction at trial, the Court of Appeals held that omission of this charge was egregiously harmful, denying appellant a fair and impartial trial because the State had adduced no corroborating evidence. Saunders v. State, 794 S.W.2d 91 (Tex.App. — San Antonio 1990). We granted the State's petition for discretionary review to decide whether the Court of Appeals correctly applied the harm analysis required by Almanza v. State, 686 S.W.2d 157 (Tex.Cr.App. 1984).

I.

The San Antonio Court took the law to be that omission of a jury instruction requiring corroboration of accomplice witness testimony, even if not requested by the defendant, usually constitutes reversible error only when such testimony is essential to the State's case because (1) the witness is in fact an accomplice and there is no evidence to corroborate his testimony, (2) the nonaccomplice evidence adduced at trial is insufficient as a matter of law to support conviction, or (3) the witness provides testimony which is the only corroboration for another accomplice witness. Otherwise, the Court opined, omission of the instruction is generally harmless.

We agree that this is a reasonably accurate reading of our holding in Gonzales v. State, 441 S.W.2d 539, 542 (Tex.Cr.App 1969), and that we have held the rule to be viable even after our decision in Almanza. See Burns v. State, 703 S.W.2d 649, 651 (Tex.Cr.App. 1985). However, we are now persuaded that a rigorous application of the Gonzales test is inconsistent with the general principles underlying Article 36.19 of the Code of Criminal Procedure, even in a case such as this where the evidence would plainly be insufficient absent the accomplice testimony.

Much of the impetus for our holding in Almanza was a perceived need to avoid the tyranny of hard and fast rules. Over the years prior to that decision, slow accretion of the case law had produced a catalog of jury-charge errors which were automatically considered fundamental and reversible by this Court without regard to their actual impact on the fairness or reliability of the specific trial in which they occurred. See Cumbie v. State, 578 S.W.2d 732 (Tex.Cr.App. 1979). Yet, under the statute law, appellate courts were admonished not to set aside criminal convictions on account of error in the trial court's charge to the jury unless the error "was calculated to injure the rights of defendant, or unless it appears from the record that the defendant has not had a fair and impartial trial." Art. 36.19, V.A.C.C.P.

Historical investigation of this statute revealed that the legislative intent, as envisioned by this Court's checkered interpretation of it, was that the degree of harm necessary for reversal should depend both upon procedural default of the appellant and upon the extent to which the outcome of trial was actually affected. Accordingly, we held in Almanza that,

. . . one reason to reverse for error in the charge arises if the error, having been properly objected to at trial, is harmful and therefore "calculated to injure the rights of the [sic] defendant." An independent basis for reversal arises if the error, even though not timely objected to, is so egregious and creates such harm that it deprives the accused of a "fair and impartial trial."

686 S.W.2d at 172 (Emphasis added). In this context, "error" meant, among other things, failure of the trial judge to charge the jury concerning some "law applicable to the case[.]" See Art. 36.14, V.A.C.C.P. The issue of harm was seen mainly as an empirical question and, therefore, primarily contingent on the individual peculiarities of each trial, considered as a whole.

But the approach of an opinion like Gonzales is antithetical to this method of review, if for no other reason, because it institutionalizes a discrete universe of facts in which error of a particular kind is automatically held to be harmful. Thus, while the circumstances calling for reversal under Gonzales do identify some fairly compelling examples of potentially egregious error, we are unwilling after Almanza to hold that such cases must invariably be reversed, just as we are unwilling to hold after Almanza that no other combination of circumstances may be found sufficiently harmful for reversal. If our recent precedents teach anything, it is that all harmless error applications, including that prescribed by Almanza, are essentially empirical inquiries concerning the effect of flaws and mistakes on the particular strengths and weaknesses of individual cases. A formal taxonomy of reversible errors has no place in such a scheme. Only the methodology is rule-governed.

Accordingly, whether we search for "some error" preserved by objection at trial or for "egregious error" urged for the first time on appeal, our approach to an assessment of its harmful impact is the same.

In both situations the actual degree of harm must be assayed in light of the entire jury charge, the state of the evidence, including the contested issues and weight of probative evidence, the argument of counsel and any other relevant information revealed by the record of the trial as a whole.

Almanza, 686 S.W.2d at 171. Because the tenor of Gonzales encourages compartmentalizing common factual scenarios for purposes of analyzing harm, it is contrary to the basic methodology prescribed by Almanza and should no longer be considered vital to the criminal jurisprudence of Texas. We, therefore, sustain the State's fourth ground for review, which contends that the Court of Appeals erred in applying the Gonzales test in the present context.

With these principles in mind, we turn now to a review of the trial record in the cause before us.

II.

Appellant's house in Sabinal burned down while he was vacationing with his family in New Mexico. Firefighters thought the circumstances suspicious because no one was home at the time, and they reported the incident to arson investigators. Within a few days, representatives of appellant's insurance carrier and members of the local fire department carefully studied the charred remains and removed carpet samples for laboratory analysis. One such sample, taken from a hallway near the water heater closet, showed evidence of kerosene, which had been spilled there either in modest amount within the past six months or in substantial amount more than six months before the fire. In the opinion of the insurance investigator, the fire was set deliberately at the site of the kerosene spill.

Local law enforcement officers arranged for a further inspection of the property by arson investigators from San Antonio, but before the investigators arrived, appellant made a controlled burn of the remaining structure under fire department supervision, ostensibly for reasons of safety. This second fire was the subject of intensely conflicting testimony at trial. Local law enforcement officers recalled insisting that appellant delay clearing the site until after further inspection. But the appellant, and others present at the time of the supervised fire, did not remember any such admonishment. Rather, it seemed clear to them that appellant was left free to raze the remains at his option. In any case, the San Antonio team was unable to reach an opinion regarding cause of the first fire on account of the second.

According to one witness, however, appellant himself was responsible for the fire that destroyed his home. James Woodley, a life-long friend, testified that appellant had earlier asked him to set the fire, but later hired another person to do it. Woodley did agree, however, to safeguard some relatively inexpensive items of personal property for appellant which, as a result, were not destroyed. For his small part in this transaction, Woodley was also indicted for conspiracy to commit arson, although the charges were later dismissed in exchange for his testimony at trial. The hypothetical torch man, presumably a third member of the conspiracy, never surfaced.

In spite of Woodley's professed agreement to keep some of appellant's things and to remove others from the house before it burned, virtually all of appellant's valuable chattels were destroyed in the fire, including a collection of arrow heads worth up to several hundred thousand dollars. Only a few firearms, left for repair with a gunsmith, and a fragment of appellant's arrow head collection from a nearby guest house survived.

To corroborate Woodley's testimony and generally to strengthen its case of a criminal conspiracy, the State also sought to prove that appellant was a man in financial difficulty whose only available remedy short of bankruptcy and destitution was to defraud his insurance carrier into paying more for his house and its contents than they were worth. To this end, it was disclosed at trial that the burned structure had recently been insured for about one hundred thousand dollars more than its replacement cost, leading the State to contend that appellant had destroyed his own home for profit. But undisputed testimony also demonstrated that the house, because of its antiquity, had an actual fair market value much greater than the cost of rebuilding it, and that the increased coverage was in fact recommended by appellant's new insurance carrier for that very reason.

It was also shown that appellant, a successful breeder and trainer of thoroughbred horses, typically borrowed large sums of money to finance his operation. When his house burned down, he owed the bank about two hundred thousand dollars on a note consolidating several earlier loans. This circumstance led the State to suggest at trial that appellant's profit motive in the destruction of his own home arose from a condition of personal and professional insolvency. But undisputed testimony established that appellant's obligation to the bank was not unusually large, that the bank was satisfied with the condition of his account, and that his business prior to the fire had been profitable.

Further testimony revealed that, during his career, appellant was twice involved in insurance claims arising from the unexpected death of promising horses born at his stable or in which he had a financial interest. It was the State's contention at trial that appellant profitted inordinately from these claims and, in any event, that his experience in such matters somehow rendered it more likely that he would defraud his insurance carrier. But the testimony of witnesses made it clear that the sudden death of livestock is an ordinary occupational risk in appellant's business, that he did not suffer such losses in suspiciously high numbers, and that insurance proceeds on the claims were not expected to exceed, and actually fell short of, the profits reasonably to be anticipated had the horses lived a normal lifespan.

The State also made much of the fact that some personal property had been removed from the house before it burned, suggesting that appellant consciously saved those items from destruction. Certainly, the prospective rescue of irreplaceable, unique, or especially valuable belongings does intimate foreknowledge of coming disaster. But, in this case, the surviving property was not shown to be of such kind. Indeed, the evidence conclusively established that moveable effects of far greater value and distinction were actually consumed by the fire. When taken in combination with generally plausible explanations for the removal of other personalty, the inference that appellant effected a preemptive rescue of some undistinguished property is not at all compelling.

Finally, it was the State's contention that appellant deliberately impeded investigation of the fire by burning down the damaged building before law enforcement investigators from San Antonio could examine it. Assuredly, the record will support a conclusion that appellant consciously disregarded orders to leave the site intact for several days, although the evidence is far more probative of misunderstanding than of disobedience. But, even assuming that he purposely brought down the remaining structure without permission, the inference that he sought thereby to prevent effective police investigation because conscious of his own guilt remains weak.

A day-long examination of the site in his presence by a professional, experienced fire investigator had just been completed. No testimony in the case suggests that the San Antonio investigators would have conducted tests or made examinations of a kind different than those already performed. As appellant intended to live on the property in the undamaged guest house, he was justifiably concerned about the danger to himself and his family from the unstable structure remaining nearby. Under such circumstances, and without additional evidence of criminal motive, a conclusion that he disobeyed strict orders not to clear the site because he thought further investigation might incriminate him is tenuous at best.

III.

Under Almanza, omission of an unrequested jury instruction "applicable to the case" calls for a new trial only when the defendant was greatly disadvantaged thereby. This degree of harm, sufficiently serious to be called "egregious," is present whenever a reviewing court finds that the case for conviction or punishment was actually made clearly and significantly more persuasive by the error.

In the instant cause, the State's most incriminating evidence against appellant was the testimony of an accomplice. Our law requires that such testimony be corroborated by evidence connecting appellant with the offense before conviction is warranted. Failure to inform the jury of this requirement makes it possible for rational jurors to convict even absent corroboration which they find convincing. But, if the omission is not made known to the trial judge in time to correct his error, appellate review must inquire whether jurors would have found the corroborating evidence so unconvincing in fact as to render the State's overall case for conviction clearly and significantly less persuasive.

We think that, even when viewed in a light favorable to conviction, the evidence taken as a whole does not have a strong tendency to connect appellant with the commission of arson or with a criminal conspiracy. Proof of a large debt, for example, does not suggest a motive for fraud unless it somehow increases the probability of the actor's insolvency. Likewise, augmented insurance coverage should not be taken to establish preparation for fraud without an apparent absence of other plausible reasons for the policy revision in question.

In the instance cause, not only is neither inference well supported by evidence, but each is contradicted by other testimony. Rational jurors may not utterly disregard undisputed evidence without a sensible basis for thinking it unreliable any more than they may simply assume a critical part of the proof without evidence having an inclination to confirm it. Here, the circumstances of debt and increased insurance coverage are just not germane to the case unless coupled with some evidence from which personal insolvency and misevaluation of the insured property can rationally be inferred.

What remains of the State's case is a rather weak inference that appellant attempted to impede the criminal investigation by intentionally destroying evidence, and that he consciously saved a few items of personal property from the fire by contriving to have them stored elsewhere at the time. But because the testimony was conflicting on these points and the inculpatory inferences tenuous in any case, it is clear to us that rational jurors would certainly have found the State's case significantly less persuasive had they been told that Woodley's testimony could not be accepted without corroboration.

For these reasons, we think the Court of Appeals's basic analysis in this cause was acceptable under Almanza, although it relied rather heavily on an opinion which is now defunct. While we may disagree with the lower court that no evidence, apart from accomplice testimony, tended to connect appellant with the charged offense, we nevertheless find such evidence to have been exceedingly weak. Accordingly, we agree that failure to instruct the jury of the need to corroborate Woodley's testimony was critical to the outcome in this cause and effectively denied appellant a fair trial.

The State's remaining grounds for review are, therefore, overruled. The judgment of the Court of Appeals is affirmed.

MILLER and WHITE, JJ., dissent.

BAIRD and MALONEY, JJ., dissent to Part I and concur in the result reached in Part III.