¶ 16. dissenting. The State and defendant both agree that the trooper in this case was unqualified to speculate that defendant’s blood-alcohol level was over .10 at the time he administered the horizontal-gaze nystagmus (HGN) test. And the majority accepts for the sake of discussion that it was error for the court to allow the trooper to testify to potentially unreliable HGN test results on account of the trooper’s admitted failure to follow the appropriate protocol for administering the test to someone "with a head injury. Notwithstanding these two errors — *8one conceded by the State and one assumed for the sake of argument — the majority concludes that the errors were harmless and thus affirms defendant’s conviction. In doing so, I believe that the majority misapplies the concept of “harmless error” and conducts its own analysis of the evidence, rather than considering whether, beyond a reasonable doubt, the jury’s verdict was unaffected by the improperly admitted evidence.
¶ 17. I note at the outset that defendant’s objections to the trooper’s testimony extend far beyond the trooper’s purported quantification of defendant’s BAC on the basis of the HGN test, and far beyond the use of the test even though the trooper failed to evaluate the impact of defendant’s head injury, as required by the applicable protocol. Defendant’s deeper argument is that, while the trooper was competent to explain what protocols he followed in the field, how he administered the test, and what he observed while performing the test, he was not competent to interpret the results of the HGN test for the jury, including testifying that certain observations in the testing are indicative of intoxication. That’s because, defendant argues, in contrast to the walk-and-turn and one-leg-stand field-dexterity exercises, the fink between the jerks in a subject’s eyeball as he or she follows a moving finger across the horizon and his or her intoxication is based on scientific principles that lay jurors generally do not understand. Defendant argues that the HGN test is a scientific technique requiring a scientific foundation for admission, and cites decisions from twenty states supporting this view. I infer that the majority’s harmless-error analysis thus extends beyond the trooper’s testimony as to the specific BAC he speculated defendant had at the time of the HGN test to the trooper’s testimony concerning the HGN test in its entirety.3
*9¶ 18. Under the harmless-error standard,4 we may find an error harmless only if we can state a belief that the error was harmless beyond a reasonable doubt. State v. Carter, 164 Vt. 545, 553-57, 674 A.2d 1258, 1263-66 (1996). An error in admitting evidence cannot be considered harmless if “ ‘there is a reasonable possibility that the evidence complained of might have contributed to the conviction.’ ” Id. at 553, 674 A.2d at 1264 (quoting Fahy v. Connecticut, 375 U.S. 85, 86-87 (1963)). The burden is on “the beneficiary of the error to prove it was not harmful,” id. at 553-54, 674 A.2d at 1264, and “we must approach harmless error arguments cautiously,” id. at 556, 674 A.2d at 1266.
¶ 19. There is no dispute in this case that defendant drank two or three five-ounce glasses of wine while eating a large meal with a neighbor over some period of time prior to 7:00 p.m. on the night in question. That evidence in itself is not particularly damning. The neighbor, a former bartender, confirmed that he did not observe anything prior to his departure around 7:00 p.m. that indicated to him that defendant was intoxicated. There is no dispute that defendant drove her car around 8:40 p.m. There is no dispute that around 9:00 p.m., after she drove, but shortly before she was subjected to field-sobriety tests, defendant drank from a bottle of 100-proof Southern Comfort, although the amount that she drank at that time is in dispute. And there is no dispute that defendant’s BAC, as measured by a blood test, was .160 at 11:35 p.m. The critical question for the jury was whether around 8:40 p.m., when she last operated a motor vehicle, and before she consumed the liquor, defendant was intoxicated.
¶ 20. Nobody saw defendant when she was actually driving. The jury was left to infer from defendant’s conduct before and after driving, and from expert testimony of her likely level of intoxication, what her condition was at the time she drove. The value of most of that evidence is greatly compromised by her consumption of the undetermined amount of liquor after she drove. For example, the trooper’s testimony that he smelled the odor of intoxicants on defendant is of little import, since defendant *10undisputedly drank at least two ounces of Southern Comfort shortly before the trooper pulled over the car in which she was a passenger.
¶21. In this context, the jury heard from two experts who analyzed her BAC as of 11:35 p.m., and, using “relation-back” calculations, opined as to her likely level of intoxication as of 8:40 p.m. They made disparate assumptions in their calculations based on conflicting evidence of how much Southern Comfort defendant had consumed. The State’s expert assumed that defendant had consumed two ounces — an assumption supported by neighbor’s estimate that the one-inch drop in the level that he observed in the bottle corresponded to two ounces. Based on this expert’s testimony, defendant’s BAC at the time of operation would have been .136 — a level that would support a finding of driving under the influence. That expert conceded that if defendant had consumed four ounces of liquor at neighbor’s house after she drove, her BAC at the time of operation would have been closer to .068 — below the legal limit.
¶22. Defendant’s expert measured how many ounces are contained in an inch of liquor in a 750-milliliter, or a fifth, Southern Comfort bottle. She testified that a one-inch drop in the liquor in that bottle corresponds to four or five ounces, as opposed to neighbor’s two-ounce estimate. Defendant’s expert thus assumed that defendant had consumed four or five ounces after she drove, and calculated a likely BAC at the time of operation as .067 or .033, respectively.
¶ 23. The trooper who pulled over the car in which defendant was a passenger was also allowed to testify about the field-sobriety tests that he administered to defendant shortly after the time she last drove her car. The trooper testified that he had been involved in approximately forty to fifty DUI investigations, and had administered the standardized field-sobriety exercises approximately one hundred times. He described completing two week-long trainings at police academies focused on administering the field-sobriety exercises, and he provided a detailed explanation of the various components, including the HGN test. In particular, he explained that for the HGN test, he asks the subject to follow his fingertip with his or her eyes as he moves it from about twelve to fifteen inches from the bridge of their nose to the side. He explained that a nystagmus, or involuntary jerkiness of the eyes, before the eyes hit a 45-degree angle indicates a BAC level above *11.10. He also described six clues evaluated in the test, and said that if four or more clues are present, the subject fails the test, implying that the subject is intoxicated. Based on his administration of the test to defendant, he testified that she was impaired, and “that she was over a .10.” These statements purportedly relating a subject’s performance on the test to a specific BAC, and purporting to do so specifically with respect to defendant, are the ones the State concedes were improper.
¶ 24. The issue before this Court is not whether we would have, as factfinders, considered the trooper’s testimony to be superfluous, or whether we would have concluded that defendant was intoxicated at the time that she drove even without the testimony. It isn’t whether a juror could have viewed the testimony in that way. The question is whether, given the record as a whole, we can conclude beyond a reasonable doubt that there is no “ ‘reasonable possibility that the evidence complained of might have contributed to the conviction.’ ” Carter, 164 Vt. at 553, 674 A.2d at 1264 (quoting Fahy, 375 U.S. at 86-87). I cannot fathom how the majority can answer that question in the affirmative.
¶ 25. The majority rightly notes that the jury was likely forced to rely on estimates of the defendant’s BAC at the time of operation to decide the case, since it had no testimony of direct observations of defendant while she was driving, and before she drank the Southern Comfort. But the suggestion that there is no reasonable possibility that the trooper’s testimony affected the jury’s analysis is impossible to square with this record. The jury was faced with disparate expert opinions about defendant’s level of intoxication when she was last driving, driven in large part by divergent assumptions concerning how much she drank. The amount she drank was itself unknowable, and the jury heard competing estimates of the amount from different witnesses. In the face of this unknown, the jury also heard testimony from a state trooper that when he administered the field-sobriety tests — very shortly after she drank the liquor — she had a BAC of at least .10. This testimony concerned her supposed BAC at an instant very close in time to her last driving. It didn’t rely on relation-back analysis, and a jury could reasonably conclude that it did not fully reflect the liquor she had consumed only moments before. It is entirely consistent with the State’s expert’s relation-back testimony and associated assumptions, and is in conflict with defendant’s expert’s opinion. And it came cloaked with a veneer of *12specialized training and knowledge from a trained law enforcement officer who has administered the test “approximately one hundred times.”
¶26. The majority postulates that the jurors credited the neighbor’s estimate of how much liquor defendant drank after she drove, and thus credited the associated relation-back testimony, such that the trooper’s testimony was irrelevant to their calculus. That’s certainly one possible story of what happened in the jury room, or in individual jurors’ minds. But it seems at least as plausible that in trying to figure out how much defendant actually drank and defendant’s level of intoxication when she last drove, at least some jurors were influenced by evidence from an officer of the law that effectively contradicted the testimony of defendant’s expert. The officer’s testimony was not tangential to the case. In its closing argument, the State specifically pointed to the HGN test, the trooper’s training and experience with the test, and his conclusion that the test showed that defendant was impaired by intoxicants. This was an important piece of evidence the State relied upon in making its case. I cannot understand how the majority can conclude that there is “no reasonable possibility” that this testimony might have contributed to the conviction.
Because I conclude that reversal is required on the basis of the more specific testimony purporting to quantify defendant’s BAC, which the State concedes was improper, I do not address defendant’s broader objection to the court’s admission of the trooper’s testimony concerning the interpretation of the HGN test. Defendant raises a substantial issue that merits more extensive consideration in an appropriate case. Likewise, I do not address the merits of defendant’s argument, accepted by the majority for the sake of discussion, that it was error for the court to allow the trooper to testify to potentially unreliable HGN test results on account of the trooper’s admitted failure to follow the appropriate protocol for administering the test to someone with a head injury. As set forth above, I strongly reject the notion that if this testimony was admitted in error, the error was harmless. But because I believe the trooper’s purported quantification of the BAC level in *9this case was itself sufficient to support reversal, I need not evaluate the merits of this claim.
Because defendant preserved her objections to the trooper’s testimony by conducting voir dire, and arguing to the court that the trooper had no ability to explain the scientific principles underlying his testimony concerning the HGN test, I conclude that this is not a plain error review case.