dissenting. At first blush this does not appear to be a case of broad significance. The majority has neither announced a new principle of law nor clarified an old one. Yet, this case troubles me as it joins many recent state and federal cases as part of a slow, deliberate movement to significantly reduce the protections provided by the Fourth Amendment. The danger in today’s opinion is that it does not declare a departure from settled law — a departure whose merits can be vigorously debated by the bar and the public. Rather, like many other decisions here and elsewhere, it announces the court’s adherence to the precepts of the Fourth Amendment while it quietly declines to honor them. Thus, the accepted rules governing searches and seizures are not openly challenged and changed, but are subtly weakened with each passing case.
I see two readily identifiable problems with the majority opinion: the structure of its constitutional analysis and its conclusion that the alleged constitutional error was harmless beyond a reasonable doubt.
A
Since the United States Supreme Court formally recognized application of the harmless error doctrine to errors involving constitutional rights in Chapman v. California (1967), 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705, cases in which harmless error has been argued by the state have generally been analyzed by first considering the alleged constitutional error and, second, if an error has been found, deciding whether the error was harmless. See, e.g., Arizona v. Fulminante (1991), 499 U.S. -, 111 S.Ct. 1246, 113 L.Ed.2d 302. While this has not always been the case, see, e.g., State v. Tabasko (1970), 22 Ohio St.2d 36, 51 O.O.2d 64, 257 N.E.2d 744, I believe that this form of analysis places the Constitution in the proper perspective. The majority does great harm to the Fourth Amendment by elevating application of the harmless error doctrine over analysis of the alleged constitutional error. By focusing “exclusively on the proper application of the harmless error doctrine,” the majority avoids confronting the Fourth Amendment. I believe that in criminal cases in which a constitutional violation is alleged, the court’s first duty always is to determine whether there has indeed been a constitutional error. If the court determines that there has been an error, it may proceed, in certain *488cases,2 to inquire whether that error was harmless.
This may seem only to be an insignificant matter of emphasis. I see, however, three very real problems with the court’s limited focus on harmless error. First, the majority implicitly holds that there has been a constitutional violation without dealing squarely with the issue. I believe that it is improper for this court ever to assume that the state has violated either the Ohio or the United States Constitution. State law enforcement officials are entitled to be told unequivocally whether the police conduct at issue in a given case is constitutionally permissible. Second, if there has been a constitutional violation it is important for the court to state specifically what it was and to explain why it occurred. The decisions of this court are the law of this state. The bench and bar must follow them in arguing and deciding future cases, those decisions concerning the Fourth Amendment define the parameters of Ohioans’ reasonable expectations of privacy, and law enforcement officials are guided by these decisions in developing and carrying out their practices and policies. Third, and most important, the harmless error doctrine is but a narrow exception to the exclusionary rule and it should be treated as such. To give harmless error analysis top billing in a Fourth Amendment case is to trivialize the alleged constitutional error. The court’s treatment of this case seems to indicate, as United States Supreme Court Justice Robert Jackson warned, “that Fourth Amendment freedoms are tacitly marked as secondary rights, to be relegated to a deferred position.” Brinegar v. United States (1949), 338 U.S. 160, 180, 69 S.Ct. 1302, 1313, 93 L.Ed. 1879, 1893 (Jackson, J., dissenting).
In this case, the warrantless search of Brown’s van clearly violated the Fourth Amendment. “It remains a ‘cardinal principle that “searches conducted outside the judicial process, without prior approval by judge or magistrate, are per se unreasonable under the Fourth Amendment — subject only to a few specifically established and well-delineated exceptions.” ’ ” California v. Acevedo (1991), 500 U.S. -, -, 111 S.Ct. 1982, 1991, 114 L.Ed.2d 619, 634. The state argues that the inventory search and the inevitable discovery exceptions to the warrant requirement are applicable in this case. I disagree.
To be valid, an inventory search of a vehicle “must be conducted in good faith and in accordance with reasonable standardized procedure(s) or established routine.” State v. Hathman (1992), 65 Ohio St.3d 403, 604 N.E.2d 743, paragraph one of the syllabus. Further, even if an inventory search of a vehicle is valid, a closed container found in the vehicle “may only be opened as *489part of the inventory process if there is in existence a standardized policy or practice specifically governing the opening of such containers.” Id., paragraph two of the syllabus. The purpose of these rules is to limit the discretion of individual police officers to ensure that inventory searches are not used as “a purposeful and general means of discovering evidence of crime.” Colorado v. Bertine (1987), 479 U.S. 367, 376, 107 S.Ct. 738, 743, 93 L.Ed.2d 739, 749 (Blackmun, J., concurring). To satisfy the requirements of the Constitution, a standardized procedure or routine must limit an officer’s discretion in two ways. “First, it must limit the officer’s discretion regarding whether to search a seized vehicle. * * * Second, [it] must limit an officer’s discretion regarding the scope of an inventory search * * *.” (Citations omitted; emphasis sic.) United States v. Salmon (C.A.3, 1991), 944 F.2d 1106, 1120.
In this case, the state concedes that the Canton police did not have a written policy governing the opening of closed containers during inventory searches. Moreover, there is no evidence in the record that the Canton police had anything more than a very vague inventory policy which made the arresting officers responsible for the contents of vehicles when a driver was arrested. One of the officers who arrested Brown and searched the van, in fact, testified that he was aware of no policy “concerning inventory and contents of automobiles or vans.” To the extent that any policy regarding inventory searches did exist, it did not adequately limit the officers’ discretion regarding their decision to search the van or the scope of that search.
The state also contends that the evidence was admissible under the inevitable discovery exception to the exclusionary rule because the police would have eventually inventoried the car and discovered the evidence. Under the inevitable discovery exception, “illegally obtained evidence is properly admitted in a trial court proceeding once it is established that the evidence would have been ultimately or inevitably discovered during the course of a lawful investigation.” State v. Perkins (1985), 18 Ohio St.3d 193, 18 OBR 259, 480 N.E.2d 763, paragraph one of the syllabus.
I believe that because the police did not have a mandatory, clearly articulated inventory search policy, the inevitable discovery exception cannot apply. In the absence of a standard police policy requiring an inventory search and defining the parameters of the search, it was never inevitable that the search in question here would occur. Without a standardized practice or established routine, individual police officers have the discretion to decide whether to conduct an inventory search. The factors the police take into account on their own in deciding whether to inventory a car are not subject to any objective judicial evaluation. The basic principle of the inventory search exception is *490that an articulated policy removes the discretion of the police from the determination whether and how thoroughly to search an impounded vehicle. The search of Brown’s van and the black leather bag found in the van was “inevitable” only because the police later testified that they inevitably would have searched them. In fact, the search would have been inevitable only if a standardized policy required the police to inventory the vehicle and the contents of all of the closed containers in the vehicle. Because the record shows that no such policy existed, I cannot conclude that any of the evidence found in the van was admissible under the inevitable discovery exception.
Because the search of Brown’s van does not fall within any of the exceptions to the warrant requirement, I would hold that the evidence found in the van was illegally seized and improperly admitted at trial.
B
In Chapman v. California (1967), 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705, the United States Supreme Court formally recognized the application of the harmless error rule to constitutional errors. The.rule announced by the court was 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.” (Emphasis added.) Id. at 24, 87 S.Ct. at 828, 17 L.Ed.2d at 710-711. In Chambers v. Maroney (1970), 399 U.S. 42, 90 S.Ct. 1975, 26 L.Ed.2d 419, the court affirmed that the admission of evidence obtained in violation of the Fourth Amendment can be subject to harmless error analysis.
Appellate courts have taken three basic approaches to harmless error analysis in Fourth Amendment cases: (1) some have emphasized the nature and relevance of the tainted evidence; (2) some have emphasized the strength of the untainted evidence; and (3) most have attempted to compare the probable effect of the tainted evidence with the strength of the untainted evidence. 4 LaFave, Search and Seizure, A Treatise on the Fourth Amendment (2 Ed.1987) 531-537. I believe the third approach to be the best because it is designed to assess pragmatically the impact of the tainted evidence on the trial as a whole. The Illinois Supreme Court explained this method of analysis:
“In considering whether constitutional error constitutes harmless error beyond a reasonable doubt * * * it is not enough that the erroneously admitted evidence be considered merely cumulative or that there be other evidence in the record sufficient to support the conviction. * * * The inquiry of a court of review should not be as to the amount of untainted evidence as compared to the amount of tainted evidence. The focus should rather be upon the character and quality of the illegally obtained evidence as it *491relates to the other evidence bearing on the same issue and the court should appraise the possible impact upon the jury of the wrongfully obtained evidence.” (Emphasis added.) People v. Black (1972), 52 Ill.2d 544, 555, 288 N.E.2d 376, 383.
It is important to remember that even using this approach, an appellate court can never be completely certain that a given error was harmless. Professor Stephen Saltzburg, an authority on evidence and trial practice, made this observation:
“[W]hen an evidentiary error occurs in the course of a trial, it disturbs [defense counsel’s] delicately balanced decision-making process. The abnegation of a particular rule upon which the defense relied may inflict more damage than initially appears. A meritorious line of defense may be dropped, an important witness held back, or entire strategies abandoned even though they should prevail. The impact of the error upon the defendant’s case may be amplified by the fact that because the error may be held harmless few lawyers will themselves attempt to depend or advise clients to depend on the appellate court’s setting the record straight. It is much more likely that trial strategy will change to accommodate rulings of the trial court, however erroneous.” Saltzburg, The Harm of Harmless Error (1973), 59 Va.L.Rev. 988, 990.
In light of the great difficulty in accurately determining the effect of erroneously admitted evidence, appellate courts must be careful to apply the harmless error doctrine only in the clearest cases. Harmless error should only be found when the state’s case is airtight even without the tainted evidence and the tainted evidence cannot have materially influenced the jury. “[W]e must recognize that harmless-error rules can work very unfair and mischievous results when, for example, highly important and persuasive evidence, or argument, though legally forbidden, finds its way into a trial in which the question of guilt or innocence is a close one.” Chapman, supra, at 22, 87 S.Ct. at 827, 17 L.Ed.2d at 710.
In this case, a comparison of the probable effect of the tainted evidence and the strength of the untainted evidence leads me to the firm conclusion that the admission of the tainted evidence was not harmless beyond a reasonable doubt. Brown was charged with drug abuse, which required the state to prove that he knowingly obtained, possessed, or used cocaine. The state sought to prove only that Brown obtained and possessed cocaine. The evidence presented by the state can be briefly summarized: The police officers who arrested Brown testified that they saw him throw something to the ground as they approached him. The officers searched the area and discovered a small zip lock bag containing crack cocaine. The officers then searched *492the van Brown had been driving and found a black leather pouch which contained a plastic film canister, fourteen zip-lock bags like the one containing the crack, and a trace amount of cocaine. The bag containing the crack and the contents of the black leather pouch were introduced as evidence at trial.
Brown’s defense was that the crack cocaine was not his. He testified that he did not throw the bag containing the crack to the ground as the police approached him, that the bag was never in his possession, and that he had never used crack cocaine. To support his argument, his attorney stressed to the jury that the state did not initially request the police lab to check the bag containing the crack for fingerprints, nor did the lab find any fingerprints when it ultimately examined the bag.
In this debate over whether the crack cocaine belonged to Brown the physical evidence seized from Brown’s van is absolutely crucial to the state’s case. Without the black leather pouch and the fourteen plastic bags, the state’s evidence directly implicating Brown is limited to the oral testimony of the two arresting officers. Without the evidence from the van, the jury would have been presented with conflicting oral testimony: the officers’ word against Brown’s word (coupled with the fact that there were no fingerprints on the bag containing the crack).
When the evidence found in Brown’s van is before the jury, however, the entire complexion of the case changes. The oral testimony of the arresting officers is buttressed by physical evidence which links Brown to the bag containing the crack. The importance of this evidence was not missed by the prosecutor. In his final point to the jury during the state’s closing argument, he sarcastically responded to Brown’s defense by expressly raising the relationship between the bag containing the crack and the bag found in Brown’s van:
“And I suppose you know it’s a coincidence that the bag that contains this crack cocaine this bag is identical identical to the bags that Mr. Brown had in his pouch that he admitted he owned. * * * For some reason they picked Bergen Brown out of all the residents of Canton, Ohio and I came in here and said boy Bergen was standing there and just matter of fact we found this bag laying [sic] there so we’re going to say that he dropped it and coincidentally this bag is identical to the bag that Mr. Brown has in his pouch that he admits is his. * * *
“Those are the facts. That’s what the officer saw and I ask you how much credibility[,] how much credibility can you give to a guy that sits here and tells you that he doesn’t use the cocaine when there’s cocaine in his pouch that he takes to work everyday. The same pouch that contains the baggies and they’re identical to the baggie in State’s Exhibit 1.”
*493The evidence found in the van was highly relevant, extremely strong, and not cumulative. The evidence was relevant because it directly countered Brown’s defense that the bag containing the crack was not his and because the state used it to directly challenge Brown’s credibility as a witness. The evidence was strong because it was physical evidence which Brown admitted belonged to him. And the evidence was not cumulative because no other physical evidence introduced by the state connected Brown to the crack.
Contrary to the majority’s view, I do not find the remaining evidence to be “overwhelming.” With the exception of the evidence found in the van, the state’s case against Brown relied entirely on oral testimony by the two arresting officers that they saw Brown drop the drugs. This testimony was based on a fleeting glimpse at night.3
Brown’s conviction depended on the jury’s resolution of the credibility of the witnesses. I cannot say that this determination was not influenced by the physical evidence illegally seized from Brown’s van and erroneously admitted by the trial court. “There is thus at least ‘a reasonable possibility that the evidence complained of might have contributed to the conviction.’ ” Stoner v. California (1964), 376 U.S. 483, 490, 84 S.Ct. 889, 893, 11 L.Ed.2d 856, 861, fn. 8 (quoting Fahy v. Connecticut [1963], 375 U.S. 85, 86, 84 S.Ct. 229, 230, 11 L.Ed.2d 171, 173).
C
Overzealous use of the harmless error doctrine profoundly weakens the deterrent effect of the Fourth Amendment. If the state feels that it can introduce illegally obtained evidence with impunity it will see little reason to avoid infringing on our Fourth Amendment rights during criminal investigations.
Because the warrantless search of Brown’s van did not fall within any of the exceptions to the warrant requirement, I would hold that the evidence discovered was erroneously admitted at trial. Because that evidence bore *494directly and influentially on material determinations made by the jury, I would hold that the error was not harmless beyond a reasonable doubt. The judgment of the court of appeals should be affirmed. I respectfully dissent.
Sweeney and H. Brown, JJ., concur in the foregoing dissenting opinion.. “[T]here are some constitutional rights so basic to a fair trial that their infraction can never be treated as harmless error * * Chapman, supra, at 23, 87 S.Ct. at 827-828, 17 L.Ed.2d at 710 (citing examples from United States Supreme Court cases preceding Chapman).
. The case cited by the majority in support of its conclusion is factually distinguishable. I believe that the oral testimony in State v. Tabasko (1970), 22 Ohio St.2d 36, 51 O.O.2d 64, 257 N.E.2d 744, was substantially more compelling than the testimony in this case for two reasons: First, the three witnesses in Tabasko were unbiased — they were private citizens with no interest in the outcome of the trial. Second, in Tabasko the factual issue before the jury was whether the witnesses themselves had been permitted to use narcotics in the defendant’s home. The witnesses’ testimony was based on their own actions, not on their perceptions of the defendant’s actions. In contrast, the police officers who arrested Brown cannot be considered completely unbiased — they were active participants in the arrest and conviction of the defendant. Moreover, their crucial testimony did not recount their own actions, but rather was based entirely on their perceptions of Brown’s actions.