Boggs, Justice.After a jury trial, appellant Mellecia Spencer was convicted of one count of driving under the influence of alcohol (less safe) and one count of possession of an open container. She appealed the judgment of conviction and sentence only as to the conviction for DUI. The Court of Appeals affirmed her conviction in Spencer v. State, 337 Ga. App. 360 (787 SE2d 320) (2016) (Spencer I). We granted this petition for certiorari to consider whether the Court of Appeals erred in holding that the trial court properly admitted a police officer’s testimony correlating the results of a horizontal gaze nystagmus (“HGN”) test with a numeric blood alcohol content or “BAC”. Because this testimony was admitted without a sufficient foundation having been laid under Harper v. State, 249 Ga. 519 (292 SE2d 389) (1982), we reverse the judgment of conviction and sentence with respect to the DUI.The underlying facts are laid out in detail in Spencer I. In brief, Spencer was stopped for a nonworking headlight, and the investigating officer noted her slurred speech, an odor of alcoholic beverage, a wristband from a bar, and a plastic cup in the center console that appeared to contain an alcoholic drink. The officer administered the HGN test to Spencer, who exhibited 4 out of 6 “clues” indicating impairment.At trial, the officer was questioned by the solicitor regarding the HGNtest:Q: Just based on your training and experience, have you noticed a correlation between four out of six clues on the HGN test that you perform in the field and a blood alcohol or breath alcohol content that would be in a person’s system?A: Based on my training and my experience, four out of six clues generally indicates a blood alcohol level equal to or greater than a .08.Spencer objected to this testimony and, after a lengthy colloquy, the trialcourt overruled the objection. The solicitor questioned the officer again[1]:Q: Is there a correlation between what you saw on the HGN test and, in general, a person’s alcohol content level on their system?A: Based off my training and my experience, generally there is.Q: And what is that correlation?A: Generally, the four out of six clues indicates an alcohol concentration equal to or greater than a .08.A jury found Spencer guilty, the trial court denied her motion for new trial, and she appealed, contending inter alia that the trial court erred in allowing this testimony. The Court of Appeals affirmed her conviction, holding:[I]t is true that an arresting officer’s testimony identifying a specific numeric blood alcohol content based solely on a defendant’s HGN results should be excluded. But the officer here did not give such testimony. Rather, he testified that in performing the test he looks for up to six clues in a subject’s eyes, that observing four or more clues indicates impairment due to alcohol, and that four out of six clues generally indicates a blood alcohol level equal to or greater than .08.(Citations and punctuation omitted.) Spencer I, 337 Ga. App. at 360-361 (1). The court held that, because the officer did not identify a specific blood alcohol level for Spencer, but merely testified that a finding of four out of six clues generally exceeds the impairing level of .08, the trial court did not err in allowing the testimony. Id. at 361 (1). We granted certiorari to consider whether the Court of Appeals erred in so holding.Our decision in Harper, supra, guides a trial court’s determination of whether a scientific principle or technique is competent evidence in a criminal case:It is proper for the trial judge to decide whether the procedure or technique in question has reached a scientific stage of verifiable certainty, or in the words of Professor Irving Younger, whether the procedure “rests upon the laws of nature.” The trial court may make this determination from evidence presented to it at trial by the parties; in this regard expert testimony may be of value. Or the trial court may base its determination on exhibits, treatises or the rationale of cases in other jurisdictions. The significant point is that the trial court makes this determination based on the evidence available to him rather than by simply calculating the consensus in the scientific community.(Citations and footnote omitted.) 249 Ga. at 525-526 (1). And “[o]nce a procedure has been recognized in a substantial number of courts, a trial judge may judicially notice, without receiving evidence, that the procedure has been established with verifiable certainty, or that it rests upon the laws of nature.” Id. at 526 (1).[2]It is generally accepted that the HGN test “has reached a state of verifiable certainty in the scientific community and is admissible as a basis upon which an officer can determine that a driver was impaired by alcohol.” Hawkins v. State, 223 Ga. App. 34, 38 (1) (476 SE2d 803) (1996). And here, the Court of Appeals relied upon its prior holding in Parker v. State, 307 Ga. App. 61, 64 (2) (704 SE2d 438) (2010), that “a score of four out of six clues on an HGN test constitutes evidence of impairment.” (Emphasis supplied). Spencer I at 361 (1). But whether the HGN test may properly be used as evidence that a driver is impaired by alcohol is not the same question as whether the HGN test has been established as an indicator of either a specific number or a numeric range of blood alcohol content.In Bravo v. State, 304 Ga. App. 243 (696 SE2d 79) (2010), our Court of Appeals addressed this distinction, noting that its earlier decision in Webb v. State, 277 Ga. App. 355 (626 SE2d 545) (2006), had correctly framed the question:[w]e do not wish to imply that a trial court must always admit numerical evidence of a defendant’s blood alcohol content adduced by an HGN test. The HGN test is a procedure that has reached a state of verifiable certainty in the scientific community and is admissible as a basis upon which an officer can determine that a driver was impaired by alcohol. It may be an open question, however, whether the HGN test has reached a state of verifiable certainty in the scientific community as a basis for determining the numerical level of a driver’s blood alcohol level.(Citations, punctuation, and footnote omitted; emphasis in original.) Bravo, supra, 304 Ga. App. at 247 (1). After noting that numerous jurisdictions have concluded that HGN tests are “not admissible to quantify a specific BAC,” id. at 247 n.13, the Court of Appeals concluded that the trial court erred in admitting a police officer’s testimony that he “estimated Bravo’s BAC was 0.25 grams based on a mathematical calculation,” because the evidence “[fell] short of establishing that the method at issue has reached a scientific stage of verifiable certainty.” Id. at 249 (1).Additionally, Bravo cites the Court of Appeals’ decision in Kirkland v. State, 253 Ga. App. 414 (559 SE2d 161) (2002), for the proposition that “field sobriety tests, including the HGN, are admissible to show that a detainee’s BAC exceeds a particular impairing level.” (Citation and footnote omitted.) Bravo, supra, 304 Ga. App. at 248 (1). Spencer I likewise relies upon this statement. 337 Ga. App. at 361 (1). Kirkland, however, was convicted under OCGA § 406-391 (a) (5), DUI per se, rather than OCGA § 40-6-391 (a) (1), DUI less safe. Kirkland consented to a breath test, and the two samples taken showed a BAC of 0.124 grams and 0.127 grams. Id. at 415. And the officer was asked, “[D]o you have an opinion based on what you observed in the HGN and what you read on the Intoximeter, do you have an opinion as to how they correlate with one another.” (Punctuation omitted.) Id. Ultimately, the Court of Appeals relied upon the breath test to affirm the conviction, observing that the breath test results “provide independent support for the jury’s determination that Kirkland was guilty of DUI by having an alcohol concentration of 0.10 grams or more, in violation of OCGA § 40-6-391 (a) (5).” Id. at 416-417. Because of these unusual facts, this decision provides no support for avoiding the application of Harper. To the extent that Kirkland could be relied upon to permit an estimate of blood alcohol content based on HGN alone, without sufficient evidence to establish the scientific validity or reliability of any correlation as required by our holding in this case, it is disapproved.Here, the Court of Appeals erroneously relied upon Kirkland and Parker, supra, without addressing its holding in Bravo that the use of an HGN test to identify a specific numeric BAC has not “been recognized in court as reaching the requisite scientific stage of verifiable certainty,” id. at 248 (1), that “the State failed to establish the scientific validity and reliability of the procedure at issue,” and that the admission of the officer’s testimony was therefore an abuse of discretion. Id. at 249 (1). It did not examine the reasoning of the many jurisdictions that have rejected the use of the HGN test for this purpose. Nor did it address the caution in Bravo that, “the Harper decision requires recognition of a procedure or technique in a substantial number of courts before a trial court may judicially notice, without receiving evidence, that the procedure has been established with verifiable certainty, or that it rests upon the laws of nature.” (Citations and punctuation omitted.) Id. at 248 n. 16. It avoided all of these pertinent issues by creating an indiscernible distinction between the officer’s testimony in Bravo and the testimony given here.We see no valid distinction to be made between testimony that a driver’s blood alcohol content is “estimated” at 0.25 grams, as in Bravo, and the testimony presented here that the results of Spencer’s HGN test generally indicate a blood alcohol content “equal to or greater than .08.”[3] In either case, only a single number was presented to the jury, and that number established a numeric blood alcohol content exceeding the per se legal limit. In effect, these “linguistic gymnastics,” as noted by the Missouri Court of Appeals in State v. Rose[4], enable the State to present to the jury a blood alcohol content that is conclusive as to the driver’s intoxication, without satisfying the required evidentiary standard. This is improper. Before any such evidence may be admitted, the proponent must satisfy the requirements established by Harper.Here, the State failed to meet that standard, presenting less supporting evidence than that found insufficient in Bravo. On cross-examination, the officer testified that his knowledge of the HGN test was based on his participation in police training totalling approximately two weeks, and that he had no medical, physiological, or other specialist training. While the officer testified, over objection, that the test was “scientific” because his training “has shown [him] that there’s a correlation between the clues observed in this evaluation and blood alcohol content. There’s a direct connection between the two of them,” and that “several studies” supported this, he did not identify the studies and they were not admitted into evidence.[5] No scientific or medical testimony was presented at trial.We conclude that the evidence presented by the State in this case was insufficient to establish the scientific validity or reliability of any correlation between a particular number of clues on an HGN test and a numeric blood alcohol content, whether a specific percentage or “equal to or greater than” a specific percentage. The trial court therefore abused its discretion in admitting this evidence. In light of the repeated questioning regarding the offending evidence, as well as testimony that Spencer was not stopped for unsafe or erratic driving, that the officer acknowledged on cross-examination that she did not exhibit many of the usual signs of intoxication, that Spencer had had recent surgery, and that Spencer presented evidence that she was not less safe to drive, we cannot say that this error was harmless, and we therefore reverse Spencer’s conviction for DUI (less safe).Judgment reversed in part. All the Justices concur.
