*565OPINION
TOCI, Judge.Timothy Roosevelt Boles (“defendant”) appeals his convictions for eighteen felony offenses involving four separate victims and several counts of burglary, kidnapping, sexual assault, sexual abuse, sexual conduct with a minor, and child molestation. He raises two issues:
1) Was defendant’s July 3, 1991, arrest unlawful, thereby rendering defendant’s statements to the Phoenix police inadmissible at trial?
2) Did the trial court commit fundamental error by allowing the state to introduce expert testimony regarding an autorad “match” between defendant’s DNA and the DNA in the samples recovered from two of the four victims?
We summarize our holdings as follows: First, defendant’s statements to the police were admissible in evidence because he was lawfully arrested. Second, the trial court properly admitted evidence that defendant’s DNA autorad matched the DNA autorad of the samples recovered from two of the victims. Nevertheless, the court committed fundamental error by allowing the state’s experts to express opinions that implied that the autorad match positively identified the defendant. The court’s error, we conclude, tainted all the jury’s verdicts. Accordingly, we reverse defendant’s convictions and remand for a retrial.
I. FACTS AND PROCEDURAL HISTORY
A. General Facts
The charges in this case relate to a series of sexual assaults committed over a period of three years within neighboring apartment complexes in the area of 15th Avenue and Glendale Avenue in Phoenix, Arizona. Three of the four female victims (victim one, victim two, and victim four) were under fourteen years of age when they were assaulted. The other victim (victim three) was twenty years old. Defendant resided in the same general neighborhood as the victims.
The assaults occurred under the following circumstances. Each attack occurred early in the morning hours in the same general geographic area. In each case the assailant attempted to hide his identity from the victim in one way or another. Three of the victims testified that the assailant wore some type of glove. Two of the victims (victims one and four) were taken from their apartments while they slept and then assaulted. A third victim (victim two) was assaulted after she was taken from her apartment laundry to a different location. The older victim (victim three) was assaulted in her own bedroom, while the attacker held a knife to her neck.
The victims involved in the first thirteen counts of the indictment (victims one, two, and three) gave similar descriptions of the assailant: a young, black man with a slim, muscular build, short black hair, and no facial hair. Victim four could only state that her attacker was a black man with black hair. Victims one, two, and four, but not victim three, were shown photo lineups with defendant’s photo. None of the victims could identify defendant from the photo array. At trial, although victims one and two identified defendant as their attacker, victims three and four could not.
Two police officers, arriving within minutes of a report that victim one had been abducted from her apartment, observed a black man running in the alley behind victim one’s apartment. He was not wearing a shirt, had on blue shorts, and was wearing glasses. They briefly stopped him, but released him because the earlier report indicated the assailant was white. Later, when the officers interviewed victim one, she described the rapist as a black man wearing glasses, fingerless gloves, cut-off jeans, and no shirt. In 1991, after defendant’s arrest, the officers identified defendant from a photo line-up as the man they had stopped in 1988.
Victim three was assaulted at about 3:40 a.m. in her apartment at 1717 West Glendale Avenue. During the struggle, the attacker covered the victim’s face with a shirt and gagged her with a sock. The victim observed that the man was wearing no clothing other than a fingerless glove on his right hand. That same morning, at about 3:30 *566a.m., a newspaper delivery man saw a slender black man in a neighboring apartment complex at 1609 West Glendale Avenue, the site of victim one’s assault. The man was wearing black spandex shorts, a shirt, and a blue and white painter’s cap with the bill turned up.
Later, at about 4:00 a.m., the delivery man approached an apartment complex at 17th Avenue and Glendale Avenue and again saw the same mam The man was then running toward the apartments at 1609 West Glendale Avenue, where victim one had been assaulted in 1988. Several months earlier, the delivery man had observed the same black man, wearing spandex shorts and a T-shirt, loitering around the same apartments.
During April 1991, the Phoenix police maintained a surveillance operation in the area surrounding 15th Avenue and Glendale Avenue. On several occasions in the early morning, some of the .surveillance officers saw a man fitting the assault suspect’s description, wearing running clothes. He was running through the surrounding apartment complexes and occasionally looking into individual apartment windows. On each of these mornings, the officers lost track of the suspect before they could follow him.
On July 3,1991, at approximately 4:00 a.m. at the corner of 15th Avenue and Glendale Avenue, two of the surveillance officers again saw the suspect. This time they were able to stop him as he ran through the breezeway of the apartment complex at 1609 West Glendale Avenue. The suspect was carrying a single sock in his hand and was wearing Asics brand running shoes. The suspect gave his name as Timothy Boles. One of the detaining officers identified Boles as the. same person the officer had seen looking into apartment windows on April 11, 1991. Another surveillance officer identified defendant as the man he had seen peering into apartment windows on April 16, 1991.
Boles was arrested for trespassing and taken to the police station for questioning. He denied that he was involved in any of the sexual assaults or that he had engaged in peering into apartment windows. He also denied owning any of the types of running attire described by the assault victims, the officers, and the newspaper delivery man. He further denied that he was an early morning jogger. After giving the police blood and hair samples, defendant was released. Later, the Arizona Department of Public Safety (“DPS”) crime lab extracted a DNA sample from defendant’s blood.
After police issued a warrant for defendant’s arrest, they learned defendant had moved to Illinois. In conjunction with the Cook County Illinois Sheriffs Office, Phoenix police arrested defendant and searched his Illinois residence. The police found a pair of blue running shorts, a pair of black spandex pants, two painter’s caps, a red bandanna, and a photograph of defendant in his running apparel wearing gloves.
When arrested, defendant was questioned by a Cook County investigator. Although unaware of the details of the assaults, the investigator knew of the DNA test results. When the investigator told defendant about the DNA evidence, defendant responded that he was not concerned, explaining that he had read that DNA testing was not known to be accurate.
When defendant was originally interrogated in Phoenix, he had been provided no details of the assaults except for their location. Nevertheless, when questioned in Illinois, defendant told the Cook County investigator that the victims were white, young teenagers. Later, defendant recanted his earlier admissions, stating that he was unsure whether the victims were white and that he merely assumed that they were young girls. Defendant also claimed that Phoenix Police Detective Townsend told him in July of 1991 that some of the victims, while asleep in their apartments, were abducted and raped elsewhere.
At trial, the state presented an analysis of the physical evidence. A serologist testified that the pubic hairs retrieved from victims two and three, and the blood retrieved from victim two’s leg, were consistent with defendant’s hair and blood type. The serologist also testified to the consistency between defendant’s Asics shoes and the partial shoe print left on the counter under the window where the assailant entered victim four’s *567apartment. The state called two experts to testify that defendant’s DNA matched that of the samples taken from victims two and three.
B. DNA Profiling
1. The Process1
The deoxyribonucleic acid (“DNA”) molecule has a double stranded structure called a double helix, which resembles a spiral ladder. The molecule is composed of three billion base pairs of four different chemicals. The particular order or pattern of these base pairs dictates genetic characteristics.
The base pairs or rungs of the ladder are composed of four substances called adenine (A), cytosine (C), guanine (G), and thymine (T). In the middle of each rung the substances bond, but A will only bond with T and C will only bond with G. Thus, the only possible combinations that can result from the bonding of the rungs of the DNA ladder are A-T, T-A, C-G, and G-C. When the DNA molecule is separated, and the base pairs are divided, they always return to the same configuration that existed before division.
Because 99% of the molecule is the same for all humans, DNA profiling focuses on those areas of the DNA molecule where there is significant differentiation of the base pair pattern.2 These areas of differentiation are called “polymorphic,” and base pair patterns in polymorphic areas are called “alleles.” A “match” is said to exist when the rungs in a part of the DNA ladder from the crime scene specimen have the same chemical sequence as the rungs in a part of the ladder in the DNA specimen taken from the defendant.
The matching process involves the use of two distinct disciplines: molecular biology and population genetics. The biological component of the test utilizes a process called Restriction Fragment Length Polymorphism (“RFLP”). In that process, the specific alleles in the polymorphic areas of the DNA molecule are isolated, photographed, and measured.
The first step of RFLP entails extracting DNA from both the evidence found at the crime scene and from samples provided by the suspect. The DNA is “cut” with enzymes that recognize certain base pair sequences. These “molecular scissors” migrate along the partial DNA strand and cut it only where these certain polymorphic chemical base pair sequences occur.
The cut fragments are placed in a gel to which an electrical current is applied. The process, known as gel electrophoresis, causes the large pieces to remain at one end of the gel and the smaller fragments to move to the other. Because gel consistency may vary and thus cause a difference in the speed with which the fragments move through the gel, a defendant’s sample and the comparison sample are run on the same gel but in different tracks or lanes.
After the DNA fragments are sorted by size, they are transferred to a nylon membrane, which is easier to handle than the gel. During this stage, the DNA fragments are split lengthwise along each base pair so that the base pairs are separated into two strands. Next, radioactive probe markers designed to match or complement the single-stranded alleles are applied to the membrane. The use of several probes is necessary because it is common for two or more individuals to have the same alleles even in a polymorphic area. It is less common, however, for two individuals to share several alleles as identified by four or five different probes.
The nylon membrane is then placed against a piece of X-ray film so that the radioactive probes can expose the film where all the tagged alleles are located. This picture is called an autoradiograph, or autorad, and resembles the bar code on grocery packages.- It is also widely known as the “DNA fingerprint.”
*568The last step of the RFLP process involves the “match” process. This involves a comparison of the alleles or bar codes on the crime scene sample with the autorad from the defendant’s sample. The two autorads are examined to determine if the dark bands or bar codes from the two samples are similar in size, relative position, and color intensity. A “match” is declared if the two samples are comparable within a certain percentage margin of error.
If a match exists, scientists then use population genetics to determine the probability that a randomly selected person, if tested, would have the same DNA profile as that left at the scene. Population geneticists rely heavily on a data base comprised, of RFLP results of a sample population. The data base serves two functions. First, it is supposed to guarantee that the target alleles are truly polymorphic. If an allele appearing on the autorad is shared by everyone, it imparts no information about the defendant. Second, the database is thought to provide information about the frequency with which specific • alleles appear in the population. Whether the data base does this is the subject of scientific controversy. Streich, 658 A.2d at 47-48; see State v. Bible, 175 Ariz. 549, 584, 858 P.2d 1152, 1187 (1993), cert. denied, — U.S. -, 114 S.Ct. 1578, 128 L.Ed.2d 221 (1994); State v. Hummert, 183 Ariz. 484, 488, 905 P.2d 493, 497 (App.1994).
2. DNA Identification of Defendant
Here, the DPS lab used RFLP to compare defendant’s DNA sample with the DNA samples procured from victims two and three.3 Autorads containing defendant’s DNA and autorads containing the DNA obtained from the victims were compared over five probes. The DPS criminalist who performed the test visually inspected the autorads and found that defendant’s autorad matched the other autorads over all five probes. The criminalist then, with the aid of a computer, measured the autorads and concluded that defendant’s autorad matched the other autorads over four probes. The fifth probe was too faint to measure precisely.
C. The Frye Hearing
After a hearing at which experts for the state and the defense testified, the trial court ruled that the scientific techniques used by the DPS lab for declaring a match satisfied the Frye4 test. Thus, the court allowed the state to introduce expert testimony of the autorad match between defendant’s DNA and the DNA found on the clothing of victims one and three. The trial court ruled, however, that the random match probability , statistics would not be admitted at trial because the state had failed to show general acceptance of such statistics in the scientific community.
D. Verdicts and Sentence
The jury returned guilty verdicts on all eighteen counts, and the trial court sentenced defendant to aggravated, consecutive prison terms on all counts. In total, defendant was sentenced to 333 years in prison. Defendant was awarded credit for 454 days of presentence incarceration, to be applied to his sentence on count one. This appeal followed.
II. DISCUSSION
A Illegal Arrest
Defendant contends that his statements to Detective Townsend denying that he owned any of the jogging attire described by the victims should have been suppressed. He asserts that these statements were the fruit of his illegal arrest on July 3, 1991. Defendant, however, did not challenge the legality of his arrest in the trial court. Thus, our analysis is limited to a review for fundamental error. See State v. Thomas, 130 Ariz. 432, 435-36, 636 P.2d 1214, 1217-18 (1981).
*569On July 3, 1991, defendant was arrested for trespassing. Ariz.Rev.Stat.Ann. (“A.R.S.”) section 13-1502(A)(1) (1989), defines third-degree trespass as:
1. Knowingly entering or remaining unlawfully on any real property after ... reasonable notice prohibiting entry.
A.R.S. section 13-1504(A)(2) defines first-degree trespass as:
2. Entering any residential yard and, without lawful authority, looking into the residential structure thereon in reckless disregard of infringing on the inhabitant’s right of privacy.
In reviewing the legality of a suspect’s arrest, we first determine what facts and circumstances the police were aware of at the time of arrest. See State v. Lawson, 144 Ariz. 547, 553, 698 P.2d 1266, 1272 (1985). If those facts and circumstances gave the police reasonable cause to believe that the suspect was committing or had committed an offense, the arrest was lawful. Id. Even if the arresting officer lacked personal knowledge of all the facts, the arrest was lawful if probable cause existed from the collective knowledge of all the officers involved in the investigation. Id.
Having reviewed the record, we conclude that defendant’s arrest was lawful. On separate occasions in April of 1991, two officers observed defendant running between the apartment complexes in the middle of the night, looking into individual apartment windows. These observations gave the police probable cause to arrest defendant for first-degree trespass.
Furthermore, on July 3, the arresting officer and other surveillance officers observed defendant leaving the back yards of some private residences on the north side of Glendale Avenue. The arresting officer himself stopped defendant in an apartment complex. posted “no trespassing.” And, the officer knew that defendant did not live in that complex. On these additional facts, the police had probable cause to arrest defendant for third-degree trespass. Consequently, defendant’s statements to Detective Townsend were admissible at trial.
B. Admission of DNA Evidence
1. The Scope of Bible
Bible is the first Arizona Supreme Court case to consider the admissibility of DNA evidence. The preliminary issue in Bible was which admissibility standard to apply to scientific evidence: the Frye test or the test in Daubert v. Merrell Dow Pharmaceuticals, Inc., — U.S. -, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993)? In Daubert, the United States Supreme Court concluded that the Federal Rules of Evidence passed by Congress superseded the Frye test. — U.S. at -, 113 S.Ct. at 2794. The Court applied a more lenient test—one that requires a court to weigh several factors, including the Frye test, to ascertain the relevancy and reliability of the evidence. — U.S. at-, 113 S.Ct. at 2796-97. The Bible court, without rejecting Daubert, applied the Frye test because Arizona traditionally has followed it. 175 Ariz. at 580, 858 P.2d at 1183.
The substantive issue in Bible was whether the random match probability statistics derived by the product rule were admissible in evidence. Noting widespread disagreement among scientists concerning the reliability of the product rule, the court held that random match probability statistics based on that rule do not satisfy the Frye test. They are, therefore, inadmissible. Bible, 175 Ariz. at 587, 858 P.2d at 1190. Bible, however, “uncouples” the statistical probability component of the DNA analysis from the RFLP or autorad match evidence. See id.
Thus, Bible neither precludes evidence of a match nor bars all testimony about its significance. The court stated that, in absence of probability evidence, testimony of a match was relevant to establish one fact: “that the blood on Defendant’s shirt could have come from the victim (but not that it necessarily and conclusively did come from the victim).” Id.
The reason that it cannot be said that two matching DNA samples are conclusive of identity is simply this: population scientists cannot agree on how probable it is that a randomly selected person from the general population, if tested, would have the same DNA profile as that of the sample left at the *570crime scene. See id. at 585, 858 P.2d at 1188. For that reason, if two DNA fragments match, one can only say that they might have a common source. On the other hand, if two fragments do not match, no statistical probability determination is made. In that event, the concerns about the probability of a random match vanish; the samples cannot come from the same person. Id. at 582 n. 20, 858 P.2d at 1185 n. 20.
Furthermore, although Bible barred random match probability statistics derived by the product rule, it did not close the door on all testimony concerning the significance of a DNA match. The court stated that “evidence of a match, accompanied by evidence that a match means that it is possible or probable that the two samples came from the same individual, could be admissible.” Id. The court could not “foresee what explanatory evidence might be available to militate in favor of admissibility of the match evidence independent of the probability calculations.” Id. (emphasis added).
2. Resolution of This Case
a. Match Testimony
Defendant asks this court to determine the following issue: does the inadmissibility of random match probability statistics require the preclusion of evidence of a “match?”
We stated in Hummert that a match is “meaningless” without a valid estimate of the likelihood of a random match. 183 Ariz. at 488, 905 P.2d at 497. This simply means that an autorad match itself is not positive evidence of identification. But one should not infer that Hummert deems such a match irrelevant. Indeed, in Hummert, we stated that “Bible supports the conclusion that experts can properly testify that the declaration of a DNA match means that defendant could not be excluded as the contributor of the sample.” Id. at 489, 905 P.2d at 498.
Thus, Hummert gave the trial court the authority to admit, after a Frye hearing, evidence of the autorad match. Here, defendant’s expert conceded, and the trial court found, that the scientific techniques (the RFLP process) used by the DPS lab in declaring a match satisfied the Frye test. Moreover, although defendant’s expert testified concerning the potential for human error in testing DNA, he did not even suggest that the DPS lab erred in any way. Thus, the trial court correctly ruled that the state could offer evidence of the autorad match.
At trial, both of the state’s experts declared a match between defendant’s DNA sample and the DNA samples retrieved from the victims’ clothing. One of the experts also explained the meaning of a match:
When two DNA samples match, that means that they are consistent with having come from the same individual. But it doesn’t necessarily mean they come from the same individual. It means that if you have a suspect’s DNA and evidence, it means you cannot exclude the suspect as a source of that evidence if they do match.
In light of Bible and Hummert, this testimony was admissible. Thus, had the prosecution elicited no more testimony concerning the significance of the DNA autorad match, no error would have been made. The prosecution, however, went beyond the boundary defined in Hummert and elicited opinion testimony that effectively nullified the testimony quoted above.
b. Opinion Testimony
Here, based on the autorad match between defendant’s DNA sample and that taken from the victims, the state’s experts offered opinions to the effect that it was highly unlikely that someone other than defendant was the source of both samples. This testimony presents us with the following issue: in the absence of admissible random match probability calculations that meet the Frye test, or some other explanatory evidence independent of the random match probability calculations, may the trial court allow unqualified expert opinions that imply that an autorad match is positive evidence of defendant’s identification? Because defendant did not object to the testimony about the significance of the match, our analysis of this issue is limited to a review for fundamental error. See Thomas, 130 Ariz. at 435-36, 636 P.2d at 1217-18.
*571The testimony given here is similar to the testimony given by the state’s expert in Hummert. There, the expert testified that the possibility of a random match was “rare.” 183 Ariz. at 490, 905 P.2d at 499. Another expert testified that “one can, by carefully choosing particular parts of the DNA that vary a lot between people, uniquely identify every person with just a sample of each person’s DNA.” Id. Both experts further asserted that if two samples match over three probes, the samples must have come either from the same individual or from identical twins. Id.
This court found such testimony to be inadmissible and highly prejudicial. We reasoned that, by “informing the jury that a match over three probes was a ‘rare’ event or that a DNA match could ‘uniquely identify* an individual, the experts effectively conveyed to the jury, at least implicitly, the random match probability statistics____” Id. We further reasoned that “cross-examination of the experts about the validity of their assumptions ... was practically impossible without resorting to the precluded random match probability statistics.” Id.
On the other hand, a recent opinion of this court faced nearly the same issue and reached.the opposite result. State v. Bogan, 183 Ariz. 506, 905 P.2d 515 (App.1995). There, a woman was murdered near a cluster of palo verde trees. Two seed pods from a palo verde tree were found in the bed of the defendant’s truck. A biology professor, using the “RAPD”5 technique for testing DNA, compared the DNA of the seed pods found in the defendant’s truck with the DNA of the twelve trees near the murder scene. The seed pod DNA matched the DNA of one of the trees (“PV-30”). 183 Ariz. at 508, 905 P.2d at 517. The professor then tested sixteen additional trees throughout Maricopa County to estimate the probability of a random match.
The trial court held a Frye hearing. The experts agreed that RAPD was a generally accepted technology and that the test results were reliable. The experts differed, however, on the probability statistics. The professor placed the odds of a random match at about one in a million. The defense expert gave a more conservative estimate of one in 136,000. The court ruled that the statistical evidence did not satisfy the Frye test.
Nevertheless, finding no dispute concerning RAPD, the trial court allowed the professor to testify at trial about the significance of the DNA match. Id. at 511, 905 P.2d at 520. The professor stated that the seed pod DNA samples and the DNA sample from PV-30 “were identical” and that the “two samples ... most likely did come from [PV-30].” Id.
The Bogan court, deciding not to follow Hummert, held that, because the trial court found that RAPD is generally accepted, the professor’s opinion testimony was admissible. The court pointed out that experts are allowed to declare matches between shoe tracks, tire tracks, fingerprints, paint smears, hair and fiber samples, bullet marks, and bite marks, without addressing the statistical possibility of a random match. Id. at 510-514, 905 P.2d at 519-23. The court reasoned that DNA evidence should not be treated differently than these other types of scientific evidence. Id. The court concluded that an expert’s opinion should not be suppressed “on the basis of a remote statistical possibility that the opinion might be incorrect.” Id. at 514, 905 P.2d at 523.
In our view, Bogan conflicts with Bible. By referring to the “remote statistical possibility that the [expert’s] opinion might be incorrect,” the Bogan court employs probability statistics to support the opinion testimony. Bible, however, leaves open the admissibility of expert opinion testimony about the frequency of a random match only where such testimony is “independent of the probability calculations.” Bible, 175 Ariz. at 587, 858 P.2d at 1190.
Additionally, the Bogan majority relied on the admissibility of other types of scientific evidence—evidence that is distinguishable from DNA evidence. First, the *572“comparative” scientific evidence that the Bogan court cited in admitting the state’s DNA testimony—shoe prints, bite marks, fiber, hair, paint, and bullet markings—usually is admissible without a Frye hearing. See State v. Richards, 166 Ariz. 576, 578, 804 P.2d 109, 111 (App.1990). Furthermore, the jury itself, by looking directly at photographs or models, is able to make a comparison without relying totally on scientific interpretation. Id.
More important, no one seriously challenges the reliability of the scientific principles behind “matches” of shoe prints, bite marks, fiber, hair, paint, and bullet markings. The scientific theory used as a basis for expert testimony about such comparisons has long been generally accepted by the scientific community. For example, in Arizona, fingerprint identification has been held to be scientifically reliable since 1921. See Moon v. State, 22 Ariz. 418, 198 P. 288 (1921).
On the other hand, the reliability of the scientific principles underlying expert testimony about the odds against a random DNA match is, at least for the moment, disputed by the scientific community. For this reason, Bible held that the statistical component of a DNA match is inadmissible because it does not meet the Frye test.6 Recognizing this important distinction, the concurring opinion in Bogan rejected the majority’s analysis.
Turning to this case, we first consider the testimony of the state’s experts. Both experts testified that they had never seen or heard of two unrelated individuals whose DNA autorad matched the sample over five probes.7 The state’s primary expert, Richard Hallick, went so far as to say that to find such a match would require a sample size equal to or greater than the world population. Hallick also testified, “[H]ad such an event occurred, it would certainly be published in a journal that I would see.”
One expert further testified:
Now, of course you can find individuals, two individuals, that will have the same pattern with only one probe. But the general expectation is that as more and more probes are used, that if two DNAs are from different individuals, you will identify differences in their DNA Because such highly variable regions are being analyzed, it is the expectation in this case that you can exclude an individual from being a source of DNA if that individual was indeed not the source of the DNA
(Emphasis added.) On redirect examination, the prosecutor and this same expert engaged in the following dialogue:
Q. [Defense counsel] asked you about whether it is theoretically possible to find two unrelated individuals who match over four or five probes. Do you recall that questioning?
A. Yes.
Q. Is it fair to say that before we would be able to find those two unrelated persons, we would have to basically test everyone in the United States, perhaps the world, to even find that type of situation? A. The sample size could be that large or larger.
The record of the Frye hearing, however, does not provide any support for these opinions. At the commencement of the Frye hearing, the state conceded that there was no generally accepted scientific method of calculating the frequency of a random match. Furthermore, although Bible recognizes that *573an autorad match may be accompanied by explanatory evidence “that it is possible or probable that the two samples came from the same individual,” such evidence must be “independent” of the probability statistics. See Bible, 175 Ariz. at 587, 858 P.2d at 1190. The state offered no evidence independent of the probability statistics to support the opinion of its experts.8
The conclusion is inescapable that the state’s experts, in one way or another, necessarily assumed the validity of the inadmissible probability statistics. Otherwise, no foundation existed for the experts’ testimony that defendant was the likely source of the DNA samples taken from the victims. Such testimony strongly implies, without any underlying generally accepted scientific consensus, that a person unrelated to defendant, picked at random from the general population, will never have the same DNA profile as defendant. The testimony of the state’s experts thus had far more significance than “simply failing to exclude the defendant.”
Neither does the casting of the expert’s opinion in terms of personal experience serve to comply with Bible. In Hummert, the state argued, and the trial court agreed, that the expert testimony was permissible under Bible “because the experts couched their opinions in terms of their personal experience as scientists.” 188 Ariz. at 490 n. 2, 905 P.2d at 499 n. 2. We found this approach to be “nothing more than a subtle evasion of Frye.” Id. We concluded, “If such testimony were admissible, any expert could express an opinion regarding matters not generally accepted by the relevant scientific community merely by framing it in terms of his or her own observations.” Id. We therefore held that the expert’s testimony was inadmissible under Bible. We reach the same result here for the same reasons.
The dissent, relying on Bogan, argues that “statistical evidence is collateral to, rather than foundational for, otherwise admissible opinion testimony of DNA experts.” According to the dissent, the scientific community’s general acceptance of RFLP provides the foundation for expert opinion testimony that a match between a defendant’s DNA and an unknown DNA sample means that the defendant in fact was the source of the sample. In our view, there are several flaws in this reasoning.
We begin with the dissent’s assertion that the DNA match itself proves that the semen obtained from the crime scene came from defendant. This argument overlooks the distinction between the two components of DNA testing: the RFLP or autorad match component, and the statistical probability component. Bible, 175 Ariz. at 581-82, 858 P.2d at 1184-85. The RFLP or autorad match is simply a visual and computer comparison of the alleles or bar codes on the defendant’s autorad with those on the autorad prepared from the crime scene DNA sample. A “match” is declared if the two samples are comparable within a certain margin of error.
Nevertheless, Bible tells us that an RFLP or autorad match simply means “the samples could have come from the same individual.” 175 Ariz. at 581, 858 P.2d at 1184. Thus, the RFLP or autorad match by itself could not have provided the foundation for the experts’ opinion testimony in this ease; only the second component of DNA testing—the random match probability calculations—permit one to pinpoint the actual source of an unknown DNA sample. Id. at 582, 858 P.2d at 1185. But here the state conceded and the trial court therefore ruled that the probability calculations did not meet the Frye test. Thus, in this case, neither component of DNA testing can provide the foundation for the state’s experts’ opinion testimony that the effect of the match was to positively identify defendant.
*574Neither do we agree with the dissent’s proposition that statistical evidence is always collateral to expert opinion testimony. That is clearly not true where the expert opinion claims a scientific basis. See Bible, 175 Ariz. at 578, 858 P.2d at 1181. Where the opinion rests on a scientific basis, “there must be a demonstrable, objective procedure for reaching the opinion and qualified persons who can either duplicate the result or criticize the means by which it was reached, drawing their own conclusions from the underlying facts.” United States v. Baller, 519 F.2d 463, 466 (4th Cir.), cert. denied, 423 U.S. 1019, 96 S.Ct. 456, 46 L.Ed.2d 391 (1975).
Here, the state’s experts’ identification of defendant as the likely source of the DNA samples found on the victims obviously claimed a scientific basis. Otherwise, the opinions would seem to have had little probative value. Furthermore, the opinions rendered by the experts relied on what they had learned from journals and heard from others about the frequency of a random match. Accordingly, the Frye standard is clearly applicable to the purported scientific techniques and principles underlying the opinions of the state’s experts.
The dissent also argues that an expert’s opinion need not be accompanied by guarantees of trustworthiness because the expert is subject to cross-examination and rebuttal. Again, we disagree. Where Frye is applicable, opinion testimony not based on generally accepted principles is inadmissible regardless of the fact that the expert rendering the opinion is subject to cross-examination. See State ex rel. Collins v. Superior Court, 132 Ariz. 180, 199, 644 P.2d 1266, 1285 (1982); Baller, 519 F.2d 463, 466. This rule is particularly important here. To effectively cross-examine the state’s experts the defendant would have had to inquire into the probability statistics. Such cross-examination would result in introducing in evidence the very statistics defendant wished to exclude.
Finally, the dissent likens DNA evidence to blood-typing statistics, which are derived from the product rule and have long been admissible at trial. We respectfully disagree with this analogy. DNA random match probability calculations are different from blood-typing statistics. First, unlike DNA random match probability evidence, blood typing evidence does not purport to make a positive identification of a defendant. Furthermore, the data base for blood-typing is reliable because it is large and comprehensive—many people in the general population have been blood typed. There is little disagreement in the scientific community as to the percentage of a particular blood type existing in the general population.
On the other hand, DNA testing is still in its infancy, and the DNA data bases around the country are small. Consequently, scientists do not agree that the DNA data bases accurately reflect the DNA characteristics of the general population. We do not doubt that as the DNA data base grows the scientific community will reach agreement that the data base reflects the characteristics of the general population. But until such agreement is reached, random match probability statistics derived from the product rule are inadmissible. See State ex rel. Collins, 132 Ariz. at 199, 644 P.2d at 1285.
Nevertheless, the state asks us to determine whether its random match probability statistics satisfy the Daubert rule.9 This we cannot do. The Frye test is the law in Arizona. Although it is true that the supreme court stated in Bible that in the future it would decide whether to adopt the Daubert test, 175 Ariz. at 580, 858 P.2d at 1183, until the supreme court does so, we are bound, to follow Frye.
Our holding, then, is narrow. We hold that, in absence of a foundation consisting of random match probability statistics that satisfy the Frye test or some other generally accepted scientific principle independent of the probability statistics, *575opinion testimony of the kind discussed above is not admissible.10 Without such foundational testimony, evidence about the significance of a DNA match is limited to testimony that the DNA tests did not exclude the defendant as a suspect.
c. Fundamental Error Analysis
One issue remains: was the trial court’s admission of the expert testimony discussed above fundamental error? An error is fundamental if it disrupts “the foundation of the case or takes from the defendant a right essential to his defense.” Thomas, 130 Ariz. at 435, 636 P.2d at 1217 (quoting State v. Gamble, 111 Ariz. 25, 26, 523 P.2d 53, 54 (1974)). Put simply, an error is fundamental if it denied the defendant a fair trial. Id. at 435-36, 636 P.2d at 1217-18. In making this determination, we “examine the prejudicial nature of unobjected-to error in light of the entire record.” Bible, 175 Ariz. at 572, 858 P.2d at 1175. Here, we find that the inadmissible testimony was fundamentally prejudicial to the defense.
First, by avoiding the application of the Frye test to the testimony of the state’s experts, the state was able to immunize its experts from an effective cross-examination by defendant. The frequency of a random match depends wholly on the disputed validity of the product rule and its underlying assumptions. To effectively attack the state’s experts’ testimony about the conclusive effect of a five probe autorad match, defendant had to challenge the product rule and the assumptions on which it is based. See Hummert, 183 Ariz. at 489, 905 P.2d at 498.
The trial court, however, ruled at the Frye hearing that any cross-examination about probability 'statistics would “open the door” to the state’s introduction of such statistics. Thus, defendant was placed in a dilemma. He could waive his challenge to the state’s experts’ testimony about the conclusive nature of a five probe autorad match. Or, through his cross-examination, he could risk admission of the very statistics he wished to dispute—statistics that are not otherwise reliable enough to be admitted.
Faced with this choice, defense counsel asked no questions about the product rule or its assumptions. Thus, the jury was left with the testimony of the state’s experts implying a conclusive match between defendant’s DNA and the samples recovered from the victims. This was arguably even more damaging to defendant than had the trial court admitted the random match probability statistics barred by Bible. See Hummert, 183 Ariz. at 489, 905 P.2d at 498.
Neither did the testimony of defendant’s DNA expert dispel the impression that a match on five probes was virtually conclusive. Although defendant’s expert testified that a four probe match in unrelated persons was not significant unless the data base was large, no testimony was elicited from him about population genetics or the validity of the assumptions underlying the methods for deriving random match probability statistics. Thus, the jury never heard any testimony about the scientific controversy surrounding the derivation of probability statistics. The picture presented to the jury was not that of a “hotly disputed scientific controversy,” as in Bible, but rather one with “the aura of infallibility surrounding an unchallenged scientific theory.” Bible, 175 Ariz. at 588, 858 P.2d at 1191.
Furthermore, defendant’s sole defense was that he was mistakenly identified. Defendant had an alibi for the 1991 charges. His' wife testified that she gave birth to their child in early 1991 and that defendant did not run in the mornings during that period. Although two of the victims identified defendant in court, none of the victims identified *576defendant in photo line-ups shown to them earlier. In fact, the Phoenix Police Department case agent conceded that without the DNA evidence the state “did not have anything other than some circumstantial evidence.” Thus, the improper opinion testimony of the state’s experts to the effect that they had never seen or heard of a match like this in unrelated persons, effectively destroyed defendant’s mistaken identification defense.
Finally, during closing argument, the prosecutor improperly characterized the DNA evidence as being conclusive of guilt. The prosecutor argued:.
How many probes do we have to do in order to come up with identity or a statement as to the identity of an individual?
Although the scientists will say, well, it’s theoretically possible that two unrelated individuals could match over five probes or four probes, no one has ever seen it ... And so it’s an incredible tool, it’s a good tool, for identifying individuals.
And when you know that the defendant has committed the sexual assaults, you know that beyond any reasonable doubt because you know the power of DNA from listening to the experts.
The above argument demonstrates how DNA evidence can be distorted. The prosecutor’s argument has been called “prosecutor’s fallacy.” State v. Bloom, 516 N.W.2d 159,163 (Minn.1994). It is the “the fallacious equation of random match probability with the probability that the defendant is the perpetrator of the crime.” Id. As our supreme court stated in Bible, “although the random match probability may factor into the guilt probability calculation, the opposite is not true. Nor are the formulae for determining the two different probabilities the same.” Bible, 175 Ariz. at 582 n. 18, 858 P.2d at 1185 n. 18.
In sum, the state was able to portray the DNA evidence as virtually infallible and nearly conclusive of guilt. We conclude, therefore, that the trial court’s admission of the experts’ opinion testimony denied defendant a fair trial. We also conclude that the inadmissible testimony tainted the entire trial.
Althoujgh no DNA evidence supported the counts relating to victims, one and four, the state bootstrapped those counts onto the counts supported by the DNA evidence. Throughout the trial, the state emphasized the similarities among the four assaults. Then, during closing argument, the prosecutor vigorously argued that the offenses were so similar that the same person must have committed all of them. Before enumerating the similarities, the prosecutor explained:
Who committed the sexual offenses against these victims? We know from the facts and the evidence ... that the defendant ... committed the offenses against all four of these girls and how do we know that? There’s a latin word called modus operan-. di, M.O. you hear it, and basically what that means is a way that someone does something, a way they usually have a characteristic—a plan or a scheme that happens regularly. And it’s almost like the defendant’s signature on each of the faces of these girls. Because when he rapes them, he leaves his own little mark, and it’s similar in each of the four cases.
The prosecutor then argued, based on the inadmissible expert testimony, that the DNA match proved that defendant committed the offenses against victims two and three. Afterward, the prosecutor tied the counts unsupported by DNA evidence to the other counts:
Go back to the [victim one] case and the [victim four] case and look at the similarities between these two cases and the [victim two and three] cases and you will know that the same rapist that committed these offenses committed the rapes against these two girls.
Thus, the state indirectly used expert testimony based on inadmissible DNA probability statistics relating to victims two and three to prove that defendant committed the assaults on victims one and four. Because the evidence against defendant on the latter counts was not overwhelming, we cannot say beyond *577a reasonable doubt that the jury verdicts on such counts were uninfluenced by the inadmissible expert- testimony.
III. CONCLUSION
We hold that the trial court properly admitted in evidence defendant’s statements to the Phoenix police and the expert testimony of the DNA autorad match. We further hold, however, that the trial court committed fundamental error by allowing the state’s experts to give opinion testimony about the significance of the DNA autorad match. We find no other fundamental error in the record. Accordingly, we reverse defendant’s convictions and remand for a retrial consistent with this opinion.
LANKFORD, P.J., concurs.. Our descriptions of DNA and DNA testing are primarily taken from State v. Streich, — Vt. -, 658 A.2d 38 (1995).
. Technology does not exist to compare the entire DNA molecule from defendant with the crime scene sample. If all of a DNA molecule could be removed from the nucleus of a cell it would be approximately 6 feet long.
. The police investigating the rape extracted DNA from a semen stain on victim two's underwear and a small amount of dried blood that was stuck to victim two's calf. The DPS lab also extracted DNA from the semen found on a shirt in victim three's apartment.
. Frye v. United States, 293 F. 1013 (D.C.Cir. 1923). Under Frye, the court determines if a scientific principle used as a basis for expert testimony is generally accepted in the relevant scientific community. See Bible, 175 Ariz. at 578, 858 P.2d at 1181.
. Randomly amplified polymorphic DNA.
. Bible found that evidence of random match frequency failed to meet the Frye standard for several reasons. First, the statistics offered in Bible were based on the product rule, which requires a large enough data base to be statistically significant. 175 Ariz. at 583, 858 P.2d at 1186. Additionally, the product rule is based on the assumption that each band on the autorad represents a DNA segment that is independent of the other bands. If this assumption is incorrect, the results of the product rule "may be incorrect by a substantial margin.” Id. Another relevant assumption upon which the product rule is based is the absence of population substructuring in the data base. The conclusion that population sub-structuring does not occur even within racial classifications assumes that the general population freely migrates and mates in a totally random fashion. Id. The validity of this assumption and its impact on the statistical accuracy of the data base has divided the scientific community. Id. at 584, 858 P.2d at 1187.
. The state elicited similar conclusions from the defense expert.
. The dissent asserts that admission of the expert opinion testimony is supported by this passage from Bible: "evidence of a match, accompanied by evidence that a match means that it is possible or probable that two samples came from the same source, could be admissible.” 175 Ariz. at 587, 858 P.2d at 1190. This language must be read in context with Bible's immediately following statement that such explanatory evidence must be “independent of the probability calculations.” Id. Thus, the scientific principle or technique underlying the opinion testimony of the state’s experts would not have been the type of evidence that Bible contemplated might be admissible to explain the significance of an auto-rad match.
. Recent cases are split on whether the product rule satisfies the Daubert test. Compare Streich, 658 A.2d at 48, with Taylor v. State, 889 P.2d 319, 336-37 (Okla.Crim.App.1995).
. In State v. Johnson, 183 Ariz. 623, 905 P.2d 1002 (App.1995), Division Two of this court held that the ceiling principle—another method of deriving random match probability statistics—has gained general acceptance. Johnson, however, is not pertinent to our disposition of this appeal. Here, the state conceded that no method of calculating random match probability statistics has gained general acceptance in the scientific community. Thus, the state did not offer evidence of the ceiling principle at the Frye hearing and did not rely on the ceiling principle as the foundation for its experts’ opinions on the significance of a DNA autorad match. Furthermore, because the trial court generally precluded defendant from cross-examining the state’s witnesses on population statistics, file defendant had no opportunity at trial to challenge the reliability of the ceiling principle.