CONNOR, Justice.

After a jury trial in the district court, appellants were convicted of illegal commercial fishing. The alleged offense occurred on August 21, 1967, in a portion of Red Bluff Bay closed to commercial salmon fishing. The sentences imposed consisted of substantial fines, suspension of appellants' commercial fishing licenses, and confiscation of the proceeds from a quantity of fish which appellants had caught.

This appeal is brought on the grounds that the district court committed prejudicial error in admitting evidence of an experiment which had been conducted by personnel of the Alaska Department of Fish and Game and that the superior court erred in affirming the conviction.

The major issue before the trial court was whether the vessel WHIDBY, in command of appellant Love, had made a legal set with its seine outside the closed area, and had then drifted into the closed area of Red Bluff Bay, or whether, as charged by the state, a set was made or had not reached completion at a time when appellants were inside the closed area.

THE EVIDENTIARY SETTING

Because of the nature of the issue presented, it is necessary to set forth the major facts testified to at trial.

It should be explained that Red Bluff Bay is a narrow bay on Baranof Island, that it extends inward for a distance of about four miles, with a narrows about one-tenth mile wide which is located about one and one-half miles from the head of the bay. In the narrows there are markers delimiting the area which is closed to fishing at all times, even when the outer portion of the bay is open to commercial fishing pursuant to regulations.

The Department of Fish and Game stream guards who testified at trial were stationed almost a mile inside the markers and could not actually see the markers from their camp.

At the trial the stream guards, Peter Nord and James R. Jones, testified for the prosecution. Both were college students first employed for the summer by the department on July 24, 1967. Both were inexperienced in the art of purse seining, and, though they had received some oral instruction, neither had attended the stream guard training school operated by the department. On August 18, 1967, they were assigned to a stakeout of the area of Red Bluff Bay closed to commercial salmon fishing.

Both Nord and Jones testified that they observed the WHIDBY, as well as other fishing vessels, enter the closed area of the bay on the evenings of August 19 and August 20, and anchor for the night. Neither observed the WHIDBY engage in fishing on either occasion, although they kept the vessel under close scrutiny.

They did not again sight the WHIDBY in closed waters until August 21, which was the opening day for fishing outside the markers. At about 5:00 p.m. the WHIDBY cruised around the closed waters for about an hour, and then went back out through the narrows to the legal waters beyond. The witnesses stated that they saw the WHIDBY again in the closed area that evening at about 9:00 p.m., proceeding toward the head of the bay. Jones said that this time the vessel seemed to be under full power. The vessel was out of view for about 20 minutes while he went to his camp to obtain certain physical possessions. When he next saw the WHIDBY, it had drifted closer to where he was standing. At this time the stream guards heard the outboard motor of a skiff start, and they saw the skiff make what they described as a rapid circle and rejoin the WHIDBY.

After observing this circling pattern, both witnesses heard clanging noises, shouting, and engines operating. After listening to the noises for some time, Jones proceeded in his kayak to the WHIDBY and placed appellants under arrest. At the time the WHIDBY was boarded, about half the seine was aboard and all of the purse rings were on board. Jones had not observed any wind while he was ashore, but noticed gusts of wind when he boarded the WHIDBY. When the arrest was made, appellant Love denied the charges and said that they had drifted into the closed area because of bad wind and tide. After the arrest the crew finished securing the fish, which were later sold to a buyer, with the proceeds impounded for possible confiscation by the state. The set resulted in the capture of 8,763 salmon, weighing about 43,647 pounds.

Each appellant and William F. Love, Jr., son of the appellant, testified that they had entered the closed area of the bay on August 19 and 20 for the purpose of investigating the numbers and whereabouts of salmon in the area and to ascertain their movements. Appellants Love and Snyder testified that their investigations revealed a large number of salmon and that the salmon tended to drift in and out of the closed area with the tide. They testified that on the opening day, August 21, 1967, they scouted around the bay until the other vessels which had come there to fish had departed for lack of finding any significant numbers of fish. They testified that they went into the closed waters of the bay about 5:00 p.m., at which time they saw a large number of salmon. The WHIDBY then headed out past the markers to wait for the fish to drift out with the last stages of the ebb current. They testified that at about 8:30 or 9:00 p.m., while the WHIDBY was approximately 100 yards outside the markers, the large school of salmon was spotted drifting past the markers toward them. It was then that they made the set. This was just about at the turn of the tide, as it began to flood. The height of the tide was about nine feet above datum plane. Each appellant stated that the purse rings were brought aboard the vessel while they were still in legal waters. Thereafter, they testified, the WHIDBY drifted through the narrows into the closed area.

Appellants pointed in their testimony to two factors which caused them to drift into the closed area. First, after the purse rings were brought aboard, the seine was pulled up to a boom about 20 feet above the deck of the vessel, and the wind, which they estimated was gusting at 15 to 25 miles per hour toward the head of the bay, blew upon the hoisted seine, which acted like a sail to push them into the closed area. Second, after the rings were aboard, there remained a large bag of fish extending 24 to 30 feet below the water's surface and upon which the current of the incoming tide and the motion of the fish themselves, which tend to swim with the tide, exerted a force toward the head of the bay.

Appellants testified that when they realized that the WHIDBY was drifting into the closed area, an anchor was cast overboard. However, the anchor caught in the seine, and only when this was later discovered was the anchor cleared. It finally caught hold. They stated that the clanging noises heard by the stream guards were caused by several factors. The purse rings, once aboard, are stacked on a pin located on the deck, called a hairpin, and this stacking procedure causes clanging noises. However, the process of gathering in the balance of the seine requires that the rings be taken off the hairpin. They are then pulled through a block and allowed to come down again; whereupon they are put on what is called a ring pin. These procedures also cause clanging noises. This process goes on during and until the last of the seine is back on the stern of the vessel, having gone through the block and back onto the stern. Other similar noises are caused by winches, blocks, and engines.

It was claimed that the circling pattern observed by the stream guards was not the making of a set, because the skiff with an outboard motor was not sufficiently powerful to pull the seine in a circle, i.e., starting from the vessel and returning to it again. Appellants testified that the purpose of sending the skiff out was to prevent the seine's corks from being pulled under the water by the weight of the salmon in the seine.

The dissenting opinion states that there was no dispute as to the time the set was made, that stakeout witness Jones saw the circling pattern of the small boat immediately after 8:30 p.m., and that stakeout witness Nord testified that at about 9:00 p.m. appellants' boat came near them and made a circle, which he thought to be a set. From this it is urged that appellants failed to explain how the WHIDBY could have drifted one mile without the passage of appreciable time, and that this probably accounts for the disbelief of their testimony by the jury. We find the record far from clear about the times at which various phases of the operation either occurred or were observed.

Mr. Nord said that the WHIDBY came into the closed area "around nine" in the evening. He then said that before he saw the vessel, he heard noises from it. It is not clear whether he heard noises at about 9:00 p.m. or observed the vessel then. But when he saw the WHIDBY, it was coming down somewhere in the middle of the bay. It then came over toward the camp until it was close to shore. How much time elapsed between when Nord first heard or saw the WHIDBY and when it arrived opposite the camp is not clear from the record.

The witness Jones first testified that he saw the WHIDBY at 8:30 or 9:00 p.m. At one point on cross-examination he testified that when the WHIDBY was in front of the camp it could have been 9:30 p.m. In a written statement made shortly after the incident he said that at approximately 9:00 p.m. the WHIDBY entered the closed area of the bay. He testified that he then went to his camp, that it took about 20 minutes to get there, that he was at the camp for about 15 to 20 minutes, and that it took about 10 minutes to get back to where he was previously located. While he was at the camp the WHIDBY was out of sight, but he heard what sounded like an outboard motorboat. Returning to his previous location from the camp, he saw the circling pattern of the outboard motorboat that he believed to be a set of the purse seine. Later, in his redirect examination, he referred to the time of first observing the WHIDBY by saying, "It was closer to 9:30."

Appellant Snyder testified that the time of making the set outside the closed area was "close to low tide and it seems to me like somewheres, approximately 8:30, which could certainly be off a few minutes."

William F. Love, Jr., and appellant Kapron testified that it was about two hours between the making of the set and when Jones came aboard. Appellant Kapron testified that the vessel drifted after making the set for "an hour, two hours maybe."

Appellant Love testified that low water according to the tide table was at 8:42 p.m., that the tidal current was barely moving outward when he saw the school of fish, but that he did not look at a clock when he commenced the set. He used the tide table for the Sitka station as a reference. Nothing appears in the record about the differentiation between the time of low tide at Sitka and Red Bluff Bay, which are on the west and east sides of Baranof Island, respectively. Nor is there anything in the record about the variability of time of tide on any given day, as compared with the predictions of the tide tables, nor is there anything in the record about the differences between time of tide and the time that the flood current commences into Red Bluff Bay.

Appellant Love testified that after the set was completed the tidal current turned. Later, on cross-examination, he testified that the set was made between 8:30 and 9:00 p.m. but closer to 8:30, and that Jones boarded the WHIDBY to make the arrest at about 10:00.

Thus, we conclude that all of the testimony is consistent with a considerable elapsed time during which the vessel could have drifted between making the set and its arrival at the head of the bay.

THE EXPERIMENTAL EVIDENCE

Near the close of the state's case the prosecution introduced testimony of an experiment conducted by Frank W. Sharp, District Protection Officer of the Department of Fish and Game, the experiment having been conducted on September 3, 1967. The purpose of the experiment was to show that the WHIDBY would not have drifted into the closed area of the bay on August 21. This testimony was admitted over the objection of the appellants, and it is upon the denial of the motion to exclude this evidence that this appeal is based.

On September 3, Mr. Sharp took the patrol boat HARLEQUIN to Red Bluff Bay. The HARLEQUIN was brought up to the markers which designated the closed area, the engines were put in neutral, and the vessel drifted away from the closed area into the outer portions of Red Bluff Bay. The test was repeated with the same results. At the time of the experiment the tide was about 9.2 feet above datum plane, and it was about two hours after the turn of the tide while it was on the flood. It was established by examination of the witness that the HARLEQUIN was a 50-foot craft with a low silhouette and a four-foot draft. In comparison, the WHIDBY was a 40- to 45-foot fishing vessel with a five- to six-foot draft and a substantial amount of gear above the deck. The HARLEQUIN had no seine hoisted above the deck and no seine full of fish submerged below the water. The wind during the experiment was estimated as gusting at five to six miles per hour into the bay, although at times during the experiment there was no wind at all. Mr. Sharp testified that he thought the reason for the HARLEQUIN's outward drift was the fresh water flow from a waterfall inside the closed area which created an outgoing current in the bay at all times. However, no hydrographic evidence on this or any other characteristic of current flow in Red Bluff Bay was presented. It is noteworthy that Mr. Sharp did not claim qualifications as an expert on the hydrographic characteristics of Red Bluff Bay nor even as an expert general mariner.

Counsel for appellants moved to exclude the experimental evidence on the ground that the conditions of the experiment were essentially dissimilar to those on the night of the alleged offense. He said,

"About the only similarity, apparently, is that the tide was approximately the same but we don't have any gear, fish in the water, we've got different vessels, different drafts, different wind conditions, and I can't see any value that this testimony would serve except possibly to confuse the jury."

The district court overruled the objection, stating that the differences and circumstances went to the weight of the testimony rather than to its admissibility, and admitted the evidence.

The superior court affirmed the judgments of conviction. The court reasoned that even though the testimony of Sharp should not have been admitted, there was more than sufficient other evidence upon which the jury might have based their verdicts.

ADMISSIBILITY OF THE EXPERIMENTAL EVIDENCE

There are no Alaska cases dealing directly with the admissibility of the results of an experiment. But there is little quarrel about the general outlines of the rule. It is stated almost universally that for such evidence to be admissible, it should have been developed under conditions substantially similar to those surrounding the event in issue. The rule of substantial similarity of conditions does not require an identity of conditions but only that degree of similarity which will insure that the results of the experiment are probative. In some cases a high degree of similarity may not be attainable, yet the evidence nevertheless may be enlightening to the jury.

As with other forms of circumstantial evidence, the trial judge may, in his discretion, exclude the experimental evidence after a determination that the probative value of the experimental evidence is outweighed by the possibility of prejudice, confusion of the issues or undue consumption of time. This discretion is, however, dependent upon a showing of substantial similarity of conditions by the proponent of the evidence.

As the court observed in Tuite v. Union Pac. Stages, Inc., 204 Or. 565, 284 P.2d 333 (1955), quoting from Leonard v. Southern Pac. Co., 21 Or. 555, 28 P. 887 (1892):

"The principle is that at best it is within the discretion of the court to admit any testimony whatever about experiments or similar occurrences. But in any event the conditions must appear to be substantially the same. It is not within the discretion of the court to admit evidence about experiments, unless the conditions are substantially alike." 284 P.2d, at 345.

Another statement of this rule is found in Ft. Worth Denver Ry. v. Williams, 375 S.W.2d 279 (Tex. 1964):

"Before it can be said that dissimilarity of conditions merely goes to the weight and hence presents a jury question, the judge must be able to say with some degree of certainty, that confusion will not occur and such dissimilarities as are disclosed are capable of explanation so as to be readily understood." 375 S.W.2d, at 282.

In other words, if the differences of condition can be explained, so that the effect of those differences upon the experiment can be evaluated rationally, the judge may exercise his discretion and admit the evidence, for it can be helpful to the jury. But the judge cannot use his discretion to decide that despite a plain lack of substantial similarity in conditions he will, nevertheless, admit the evidence. In cases concerning the admissibility of experimental evidence, the foundation for admissibility should be scrutinized closely to determine whether the conditions surrounding the experiment were substantially similar to those of the alleged occurrence.

In applying the test of substantial similarity, the trial court should be guided by the following principles: Are the dissimilarities likely to distort the results of the experiment to the degree that the evidence is not relevant? Can the dissimilarities be adjusted for or explained so that their effect on the results of the experiment can be understood by the jury? In this connection the court must consider the purpose of the experiment and the degree to which the matter under experiment is a subject of precise science. Absolute certainty is not required if the experiment would be considered valid by persons skilled or knowledgeable in the field which the experiment concerns.

This determination of whether substantial differences exist may not always be capable of a mechanical solution, but the same may be said about most trial court evidentiary determinations that employ notions of relevance or materiality. Frequently common sense provides a good guide to whether a factor entering into an evidentiary determination is substantial or merely unimportant.

Applying the test to the present case, it is apparent that the district court committed error in admitting the testimony about the experiment. The differences between the conditions of the experiment and those surrounding the occurrences on the night in question are so great that the offered testimony did not fulfill the requirement of basic relevancy. The similarity of tide levels and wind direction are not enough to meet the test. The failure of the experiment to test the current below the surface, the effect of the wind on the hoisted seine, the effect of the possibly stronger wind, and particularly the effect of the large number of fish in the submerged portion of the seine render the results too speculative to be useful in a court of law. It is no help that there was prosecution testimony that the wind conditions were similar on both occasions. The record reveals only controversy on this point. Moreover, it is noteworthy that no expert testimony was offered which would explain or evaluate the dissimilarities between the experiment and the actual event.

Nor is the rule of substantial similarity any less demanding because the evidence is claimed by its proponent to have been introduced only for the purpose of rebutting the anticipated testimony of the other side. The reason it is unacceptable is given in Sewell v. MacRae, 52 Wn.2d 103, 323 P.2d 236 (1958), in which the court said:

"The experiment did not enlighten the jury as to whose testimony was more reliable, in that it disregarded the evidence adduced by the appellant. The result of an experiment which rests upon the testimony of the party offering it, and disregards the conflicting evidence adduced by the opposing party, has no more probative value than the oral testimony upon which it is based." 323 P.2d, at 239.

In the present case the evidence of the experiment was materially relevant only to rebut the appellants' testimony which stated that certain natural elements combined to cause them to drift into the closed area of the bay. This gave the prosecution an additional benefit because it could argue, under the maxim falsus in uno, falsus in omnibus, that the testimony of appellants should be disbelieved in its entirety. It does not matter whether the prosecution did in fact so argue, or to what degree. The jury itself could apply the same reasoning, unaided by the prosecutor. But neither the experiment nor the testimony about it accounted for the critical, missing elements. Accordingly, it had no probative value and could only mislead the jury.

THE HARMLESS ERROR QUESTION

The appellee urges that the admission of the experimental evidence, if error, was harmless. In Daniels v. State, 388 P.2d 813 (Alaska 1964), this court, in a criminal case, applied the harmless error rule to the admission of evidence which was irrelevant, saying,

"[E]ven if that part of her testimony objected to below had been stricken there is no reasonable probability that a fairminded jury would have arrived at a different verdict. We find no prejudicial error on this point." 388 P.2d at 816.

On other occasions this court has reversed or affirmed evidentiary rulings as being either harmless or prejudicial, but without discussing at any length the standards governing harmless error in criminal cases. In a recent case this court applied the constitutional harmless error test which was announced in Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), to an error claimed on other than constitutional grounds.

The present case permits a closer look at the principles which should guide decision in applying the harmless error rule. The subject is one of great importance, not only because it pervades the appellate process but because it goes to the vitals of our entire legal system. This is because the system must be judged not only by what it says, but by what it does.

A formal statement of legal rules is of little value unless we know the methods by which those rules operate in practice. What matters pragmatically is only the portion of the formal rule that survives after the judicial apparatus has done its work.

The application of the harmless error rule in a given case is a broad act of judgment, with all that the term implies. It is not easy to express in mechanistic verbal formulae a rule comprehending the many factors which motivate that act of judgment. The interplay of impression and analysis, the experience and legal philosophy of the judge, the necessity to balance between competing interests, and a detailed consideration of the actualities in each case, all contribute inevitably to the result.

Still it is possible to set bounds by which decision can be approximated. The borderlines may sometimes be blurred, but even those may become more distinct when enough case-by-case determinations permit a trend of decisional law to be observed.

The harmless error doctrine originated in part because of the rule, at common law, that a reversal of a criminal conviction on appeal required the discharge of the prisoner, who could not be retried. The gist of the earliest rule was that if, from the record, the reviewing court had no reasonable doubt that guilt was proved by proper evidence, the reception of improper evidence would not result in reversal. But if it were otherwise, reversal was required. This was supplanted in the early Nineteenth Century by the Exchequer rule which, with rare exceptions, gave the defeated party a right to a new trial for any erroneous ruling by the trial court. The Exchequer rule gained rapid ascendancy in the United States and was still thriving here long after it had been reformed in England. The apotheosis of the rule by so many of our courts resulted in shocking injustice in many instances. Presented by the possibility of reversal for the most trifling errors, the rule became a tool for the crafty, the rich oppressors, and the hypertechnicians of the law. The stultifying effects of the rule created a condition approaching a crisis in judicial administration, and brought forward a strong movement for change. Eventually reforms were sought and obtained after a long struggle.

All of our jurisdictions today employ standards which attempt to distinguish between errors which are merely technical and those which inflict harm by affecting the substantial rights of a party. The difficulties of applying the harmless error rule are summarized ably by Professor Wright:

"Unhappily the appellate courts, on whom the duty falls to enforce contemporary standards of fairness, are not vouchsafed the luxury of abstract decisions, and they are constantly forced to make the most agonizing kind of choice between the public interest in maintaining fair trials and the public interest in putting criminals behind bars. The dilemma is not resolved by elegant language about protecting the pure fabric of criminal prosecution against the `ghostly phantom of the innocent man falsely convicted'; neither is it resolved by a doctrinaire insistence on freeing the guilty in every case in which the prosecutor has overreached or the trial court has lapsed." 3 C. Wright, Federal Practice and Procedure, 354 (1969) (footnotes omitted).

Some of the guides for decision can be stated more aptly in the negative than otherwise. For example, it is not the function of an appellate tribunal, working from the dead record, to make the determination of guilt or innocence. Bollenbach v. United States, 326 U.S. 607, 66 S.Ct. 402, 90 L.Ed. 350 (1946). That issue is for the jury. People v. Mleczko, 298 N.Y. 153, 81 N.E.2d 65 (1948). The test is not whether, with the erroneous matter elided from the record, there would be enough evidence to support a conviction. It is not for us to speculate on the outcome at a retrial, absent the erroneous matter. The pivotal question is what the error might have meant to the jury. Our function is to consider not how the error would have affected us if we had tried the case, but how it may have affected a jury of reasonable laymen. It is the impact on their minds which is critical in determining whether an error impaired or affected the substantial interest of the defendant in having a fair trial.

A keen analysis of the harmless error rule is found in Kotteakos v. United States, 328 U.S. 750, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946). The majority, speaking through Mr. Justice Rutledge, stated the ultimate test as follows:

"But if one cannot say, with fair assurance, after pondering all that happened without stripping the erroneous action from the whole, that the judgment was not substantially swayed by the error, it is impossible to conclude that substantial rights were not affected. The inquiry cannot be merely whether there was enough to support the result, apart from the phase affected by the error. It is rather, even so, whether the error itself had substantial influence. If so, or if one is left in grave doubt, the conviction cannot stand." 328 U.S., at 765, 66 S.Ct., at 1248.

This is the test which we must employ. Any approach more permissive would allow the conviction of persons for criminal offenses by means which do not measure up to established legal standards. Our quest is not justice according to one's personal belief about the desirable result in a given case, but justice according to law.

As the New York Court of Appeals has put it, speaking through Fuld, J.:

"It cannot be overemphasized that our legal system is concerned as much with the integrity of the judicial process as with the issue of guilt or innocence. The constitutional and statutory safeguards provided for one accused of crime are to be applied in all cases. The worst criminal, the most culpable individual, is as much entitled to the benefit of a rule of law as the most blameless member of society. To disregard violation of the rule because there is proof in the record to persuade us of a defendant's guilt would but lead to erosion of the rule and endanger the rights of even those who are innocent." People v. Donovan (1964), 13 N.Y.2d 148, 243 N.Y.S.2d 841, at 845, 193 N.E.2d 628, at 631.

The test we employ today is less rigorous than that employed by the United States Supreme Court in cases where constitutional rights are affected by evidentiary determinations in state courts. In those cases we are bound by the rule of Chapman v. California, supra, that before a federal constitutional error can be held harmless, "the court must be able to declare a belief that it was harmless beyond a reasonable doubt." 386 U.S., at 24, 87 S.Ct., at 829. As to ordinary errors, not based on a constitutional claim, we read Chapman as permitting us to employ a different approach. The announced standards of Kotteakos v. United States, supra, which we adopt today for ordinary criminal trial errors, were applied in that case to federal ordinary errors and not those based on a claim under the Constitution of the United States. It is not clear whether the Chapman standard in the future will be applied to all errors in federal criminal trials. But for our own purposes the Kotteakos principles are, in our opinion, a proper guide where the Chapman standard does not come into play.

A case similar to the one before us is Tippit v. State, 332 P.2d 222 (Okl.Cr. 1958), in which an experiment had been conducted to demonstrate that the victim of an automobile accident could have been knocked or dragged a certain distance by the automobile driven by the defendant. A quantity of wet sand was placed in a burlap bag, together with a partially inflated truck tire innertube, which was placed in a vertical position, in an attempt to simulate the body of the victim as it was hit by the automobile. The court held that the admission of this experiment was error, as there was so little similarity between the conditions of the experiment and the facts of the case itself. The court deemed the admission of this evidence prejudicial error, as the testimony concerning it could not have done anything other than create confusion and prejudice against the defendant.

We cannot fairly say that the experimental evidence did not appreciably affect the jury's verdict against appellants. The admission of the evidence was prejudicial error.

INAPPLICABILITY OF CHAPMAN v. CALIFORNIA

One reasonably might ask why we have not adopted the Chapman test as the one to be applied to errors not affecting federal constitutional rights. There is nothing grossly wrong with the rule laid down in Chapman when viewed against the purposes which brought that rule into being and which indicate the appropriate boundaries of its operation. But to adopt that rule without a critical inquiry into its underlying basis would not, in our opinion, fulfill the obligation of independent, informed judgment which rests upon us as the judicial officers who are expected to determine questions of Alaska law.

The fundamental rule applied by the United States Supreme Court over many years was that of automatic reversal for the introduction of constitutionally prohibited evidence or the commission of other trial court errors which impaired federal constitutional rights. The rule of automatic reversal has been used in cases concerning coerced confessions, unconstitutional presumptions, prejudicial publicity, speedy trial, discriminatory jury selection, unconstitutional state statutes, and denial of counsel, to name only a few. In some of these cases the constitutional breach may have played little or no role in the conviction. Between Bram v. United States, 168 U.S. 532, 18 S.Ct. 183, 42 L.Ed. 568 (1897), and the decision in Chapman in 1967 the court used the harmless error principle in treating constitutional claims on only a few occasions, and then only in trivial or rather special situations. It was reasonably believed by many observers that the harmless error doctrine might be totally irrelevant to claims of error based on a constitutional right or privilege.

In Fahy v. Connecticut, 375 U.S. 85, 84 S.Ct. 229, 11 L.Ed.2d 171 (1963), the court granted certiorari to determine whether state harmless error rules could be applied in cases where illegally seized evidence had been introduced contrary to the prohibitions of Mapp v. Ohio, 367 U.S. 643, 81 S.Ct. 1684, 6 L.Ed.2d 1081 (1961). But the court did not reach that issue. It found that the state court had misapplied its own harmless error rule and ostensibly reversed on that ground. Fahy v. Connecticut, supra, 375 U.S., at 91-92, 84 S.Ct. 229.

Finally, in the Chapman case the court made clear that some errors violative of constitutional rights can, in the setting of a particular case, be "so unimportant and insignificant that they may, consistent with the Federal Constitution, be deemed harmless, not requiring automatic reversal of the conviction." 386 U.S., at 22, 87 S.Ct., at 827. But before a constitutional error may be declared harmless there are two major hurdles to be crossed. First, the beneficiary of the error has the burden to show that it was harmless. 386 U.S., at 24, 87 S.Ct., at 824. Second, the court passing upon it "must be able to declare a belief that it was harmless beyond a reasonable doubt." 386 U.S., at 24, 87 S.Ct., at 828. Applying this test to improper prosecutorial comment on the failure of the accused to testify, outlawed by Griffin v. California, 380 U.S. 609, 85 S.Ct. 1229, 14 L.Ed.2d 106 (1965), the court found the error to be harmful and necessarily reversible.

We cannot tell from a careful reading of the majority opinion in Chapman what will be the future evolutions and consequences of the test there announced, as there is much that is not stated in the Chapman majority opinion. Even the rationale for enforcing a federally created standard of constitutional harmless error is not explored in detail. As the reflective dissent of Mr. Justice Harlan points out, there are many reasons the majority may have had for employing the rule, but it did not state them. Perhaps the majority was concerned that state-fashioned harmless error rules might be used as instruments to dilute and evade federal constitutional guarantees, particularly the new doctrines made applicable to the states by Fourteenth Amendment interpretations during the last decade. But this we do not know because we are not told.

In sum, Chapman tells us that although harmless error can be applied to constitutional guarantees, this will be permitted only sparingly and under strict standards.

By contrast, the errors we must review touch upon a host of matters which have nothing to do with the effectuation of federal constitutional rights, or have only a tenuous connection with those rights. The errors with which we must deal are considerable in their variety. They may concern matters of our local procedure and rules of evidence; they may employ notions of discretion, convenience, and the historical evolution of rules; and they may present problems peculiar to the physical, economic and social circumstances of Alaska. The range of problems is so broad that we are unwilling to bind ourselves to a rule fashioned for a different purpose, the future interpretations of which rule are largely unpredictable. We feel that more flexibility is needed than is to be found in a test stated in the stark terms of reasonable doubt.

The second troubling feature of the Chapman test is its very brevity. The mere statement of a formula is of no great value unless we know something about the principles by which it will be applied. It is true that we have selected a simple statement to summarize our own approach to harmless error: whether we can fairly say that the error did not appreciably affect the jury's verdict. But we have also stated some of the principles by which we hope to reach decision and we have indicated some of the perimeters within which the harmless error rule is to be applied.

We have consciously used the terms "fair assurance" and "substantially swayed" interchangeably with "fairly say" and "appreciably affect." Any meaningful difference between those sets of terms and the thought expressed becomes a subject more appropriate for investigation by semanticists and metaphysicians than by us. If anything, the use of those slightly different terms to express a single standard illustrates our view that the application of the harmless error rule brings into play a group of principles and approaches to decision which cannot be imprisoned in any one terse phrase or formula. It is often illusory to attempt, within an apparently simple statement of a rule, to comprehend a set of principles which cannot be confined to rigidity in their application. Of course, legal principles seem to carry within themselves an innate reductive tendency by which statements of specific rules, or sets of rules, eventually replace the principles themselves. The law could not function well without this method. But we should never forget that,

"The nadir of mechanical jurisprudence is reached when conceptions are used, not as premises from which to reason, but as ultimate solutions. So used, they cease to be conceptions and become empty words." Pound, "Mechanical Jurisprudence," 8 Colum.L.Rev. 605 (1908), at 620-621.

So it is that we have tried to retain some elasticity in our statement of the avenues by which harmless error will be determined in our future decisions. We have also tried to avoid an approach so standardless as to become a potential instrument of judicial whimsy.

The superior court's affirmance is reversed with directions to remand to the district court for a new trial.