OPINION

I. FACTUAL AND PROCEDURAL BACKGROUND

Danny Haynes was arrested for driving while intoxicated. Following his arrest, Haynes submitted to an “Intoximeter 3000” breath analysis, which chemically tests for the presence of alcohol. See AS 28.35.031 (“A person who operates or drives a motor vehicle in this state ... shall be considered to have given consent to a chemical test or tests of the person’s breath.... ”). The In-toximeter 3000 breath test produced a reading of .106 grams of alcohol per 210 liters of breath. The Intoximeter 3000 has a recognized margin of error of .01 grams per 210 liters of breath. Therefore, Haynes’ actual breach alcohol content, as measured by the Intoximeter 3000, may have been as high as .116 or as low as .096.

An alcohol level of .10 grams or more per 210 liters of the person’s breath is required for the Department of Public Safety (the Department) to revoke a driver’s license. AS 28.15.165(c); AS 28.35.030(a)(2). Applied in Haynes’ favor, the .01 margin of error would equate to an actual result of .096 grams, a reading below the statutory minimum of .10 required for revocation. The hearing officer considered the margin of error inherent in Haynes’ .106 test result, but declined to apply it in Haynes’ favor. Instead, the hearing officer relied on the arresting officers’ testimony that Haynes’ personal appearance and behavior indicated that he was intoxicated, to find it more probable than not that Haynes’ breath alcohol content was .10 or higher at the time of the test. Consequently, the Department revoked Haynes’ license. The superior court affirmed the Department’s revocation order, and this appeal followed.

II. DISCUSSION

A. Standard of Review

We review the hearing officer’s decision to revoke Haynes’ driver’s license independent of the superior court’s decision, as the superior court was acting as an intermediate court of appeal. Jager v. State, 537 P.2d 1100, 1106 (Alaska 1975); State v. Marathon Oil Co., 528 P.2d 293, 298 (Alaska 1974). Because the issues presented in this appeal are purely questions of law, we are not bound by the lower court’s decision. Rather, we will “adopt the rule of law that is most persuasive in light of precedent, reason, and policy. Guin v. Ha, 591 P.2d 1281, 1284 n. 6 (Alaska 1979).

B. Consideration of Other Factors

Alaska Statute 28.15.165(c) provides that the Department of Public Safety may revoke a person’s license if “a chemical test under AS 28.85.031(a) produced a result described in AS 28.35.030(a)(2).” AS 28.15.-165(c) (emphasis added). Alaska Statute 28.-35.030(a)(2) provides that a person commits the crime of driving while intoxicated (DWI) if a breath analysis reveals that the person’s breath sample contains .10 grams or more of alcohol per 210 liters of the person’s breath. AS 28.35.030(a)(2). Alaska Statute 28.15.-165(c) does not provide for consideration of other factors or circumstances, such as the appearance and behavior of the individual, in determining whether the person’s breath did, in fact, contain the requisite level of alcohol.1 AS 28.15.165(c). Therefore, the hearing offi*755cer should not have considered evidence regarding Haynes’ appearance and behavior in determining whether his actual breath alcohol level was over .10 grams. The officers’ testimony regarding Haynes’ appearance and behavior does not provide support for license revocation pursuant to AS 28.15.165(c).2

C. Margin of Error

The legislature has the power to require the revocation of a driver’s license on the basis of a particular test result or reading, despite its inherent margin of error, when the legislature expressly considers that margin and deems it sufficiently negligible such that it may be disregarded.3 In such circumstances, the test result is considered tolerably inaccurate, and, therefore, the Department may revoke a license on the basis of the test result without regard to the test’s margin of error.

In Barcott v. Department of Public Safety, 741 P.2d 226 (Alaska 1987), we addressed the issue of whether the hearing officer must consider the inherent margin of error in a chemical analysis designed to test the presence of alcohol in a person’s breath.4 In the course of our analysis, we examined how courts in other jurisdictions interpreted their own DWI statutes with regard to the issue of inherent margin of error in a chemical blood/ breath alcohol test. Id. at 229. Essentially, the courts’ analyses hinged on whether the particular court interpreted its jurisdictional DWI statute to create an offense upon a test reading in excess of their statutory limit or upon an actual level of alcohol in excess of the limit. Courts that interpret their DWI statutes to create an offense upon a test reading in excess of the statutory limit presume that the legislature considered the inherent risk of error in the chemical analysis and found it to be tolerably inaccurate; thus, the courts did not require the fact finder to consider the inherent margin of error of a particular testing device.5 See State v. Rucker, 297 A.2d 400, 402-03 (Del.Super.Ct. 1972); Nugent v. Iowa Dep’t of Tramp., 390 N.W.2d 125, 128 (Iowa 1986); Schildgen v. Comm’r of Pub. Safety, 363 N.W.2d 800, 801 (Minn.App. 1985); State v. Lentini, 240 N.J.Super. 330, 573 A2d 464, 466-67 (N.J.Super.Ct.App.Div. 1990); Slagle v. State, 570 S.W.2d 916, 919 (Tex.Crim.App. 1978). In contrast, courts that interpret their DWI statute to create an offense upon an actual level of alcohol do not presume that their legislature considered the inherent margin of error of a chemical test; thus, those courts require the fact finder to consider the inherent margin of error before rendering a decision. See State v. Boehmer, 1 Haw.App. 44, 613 P.2d 916, 918-19 (1980); State v. Bjornsen, 201 Neb. 709, 271 N.W.2d 839, 840 (1978); State v. Prestier, 7 Ohio Misc.2d 36, 455 N.E.2d 24, 27 (Ohio Mun.Ct. 1982); State v. Keller, 36 WashApp. 110, 672 P.2d 412, 414 (1983).

In Barcott, we adopted the latter reasoning and held that there was no evidence or indication that the Alaska Legislature considered the margin of error inherent to the Intoximeter 3000. Id. at 230. The legislature did not specifically approve the Department’s use of the Intoximeter 3000 test, but rather authorized the Department to approve satisfactory techniques, methods, and standards of performing the analysis. AS 28.35.-*756038(d); Barcott, 741 P.2d at 230.6 There is no indication that the legislature considered the .01 margin of error inherent to the Intox-imeter 3000 in setting the legal limit at .10 grams per 210 liters of the person’s breath. Therefore, we do not interpret AS 28.35.-030(a)(2) as creating an offense for violation of the statutory .10 grams per 210 liters of breath, without regard to the margin of error inherent to the particular testing device utilized. In Alaska, a driver commits a DWI offense when his or her actual alcohol level exceeds .10 grams per 210 liters of breath.

D. Due Process

The question, then, is whether the margin of error must be applied in Haynes’ favor. As a matter of statutory interpretation, we answer this question in the affirmative. It is well established that

[a] driver’s license is an important property interest, and the driver has a constitutional right to a meaningful hearing before the state can suspend his license. As in a criminal prosecution for driving while intoxicated, the breath test is of central importance in the administrative license revocation proceeding. The ability of the defendant to evaluate these tests is critical to his ability to present his ease.

Barcott, 741 P.2d at 228 (quoting Champion v. Dep’t of Public Safety, 721 P.2d 131, 133 (Alaska 1986)). In Barcott, we held that the hearing officer violated the appellant’s right to due process7 when he failed to consider the inherent inaccuracy of the Intoximeter 3000 breath test in determining that a test reading of .10 satisfied the statutory requirements to revoke the appellant’s license. Thus, we reversed the decision and remanded the case to the Department for further proceedings consistent with the opinion.8 Barcott, 741 P.2d at 230.

Given the .01 margin of error inherent to the Intoximeter 3000, a test reading of .106 could equate to a .096 actual test result. If the .01 margin of error is not applied in Haynes’ favor, the deprivation of an important property interest could result where the actual breath test result was below .10 grams. Absent express legislative intent to the contrary, we hold that failure to apply the inherent margin of error of a particular testing device in favor of the person subject to license revocation violates due process of law as guaranteed by the Alaska Constitution.9 Alaska Const, art. I, § 7.

In summary, we hold that a chemical breath test reading or result which may be reduced below the level of .10 grams per 210 liters of the person’s breath, by applying the margin of error inherent in the particular test used, cannot serve as the basis for a license revocation under AS 28.15.165(c). Extrinsic evidence of intoxication does not mitigate the inherent error; the error remains, and must be applied to the test reading. In light of the .01 acknowledged margin of error in the breath test administered to Danny Haynes in the case at bar, the Department was not authorized to revoke Haynes’ driver’s license under AS 28.15.-165(c). The decision of the hearing officer is *757REVERSED and the case is REMANDED for further proceedings consistent with this opinion.

MATTHEWS, J., and RABINOWITZ, C.J., dissent.

joined by RABINOWITZ, Chief Justice, dissenting.

I.

Under AS 28.15.165(a) and (e), the Department of Public Safety is required to revoke the driver’s license of a person driving a motor vehicle where a chemical test administered to the person “produces a result described in AS 28.35.030(a)(2) ...” namely, “0.10 grams or more of alcohol per 210 liters of the person’s breath....” Today’s opinion construes this statutory system to mean that the department must revoke the driver’s license of a driver who is administered a chemical test where the test produces a result of .11 grams or more of alcohol. This result is required, according to the majority opinion, because “[a]s a matter of statutory construction” (Slip Op. 7, 8) there is no evidence that the legislature considered the .01 margin for error inherent in the testing device.1

‘Where a statute’s meaning appears clear and unambiguous ... the party asserting a different meaning bears a correspondingly heavy burden of demonstrating contrary legislative intent.” University of Alaska v. Geistauts, 666 P.2d 424, 428 n. 5 (Aaska 1983) (quoted in Lagos v. City and Borough of Sitka, 823 P.2d 641, 643 (Aaska 1991)). This means that the plainer the language of the statute, the more convincing the evidence of contrary legislative intent must be. State v. Alex, 646 P.2d 203, 208 n. 4 (Aaska 1982). Here the statute is plain in that it precisely defines the minimum test result which mandates license revocation. Since, as the majority opinion points out, there is no legislative history indicating that the legislature intended to use a different minimum level, we are required to construe the statute to mean what it says. I do not believe that it is right to say that a statute does not mean what it appears to mean because there is no legislative history indicating that the apparent plain meaning of the statute is the actual meaning. That is what today’s opinion does.

II.

While I think the above correctly identifies the logical flaw in the majority opinion, I should add that it also over-simplifies the problems presented by this case. There are two related problems which we should face. The first is that this court has never defined the elements of the license revocation offense. They should be defined. The second is that our prior decision in Barcott v. State, Dep’t of Public Safety, 741 P.2d 226 (Aaska 1987), is poorly reasoned and based on an erroneous assumption regarding the elements of the license revocation offense. It should be overruled. The following discussion addresses these problems.

A. Elements of the License Revocation Offense

The critical debate here is whether a failing test result alone is all that is required to *758revoke a driver’s license or whether a failing test result and an illegal blood or breath alcohol level are needed.2 To fully understand this question our statutes which define the criminal offense of driving while intoxicated as well as the license revocation offense must be examined.

The statutory provisions which govern this case are AS 28.15.165, 28.15.166, and 28.35.-030(a). Under AS 28.15.165(a) if a chemical test given to a driver “produces a result described in AS 28.35.030(a)(2)” a notice must be given to the driver that the Department of Public Safety intends to revoke the driver’s license to operate a motor vehicle. (Emphasis added.) License revocation follows unless the driver makes a request for review under AS 28.15.166 within seven days after receipt of the notice. When a request for review is made, the driver is entitled to a hearing before a hearing officer designated by the Commissioner of Public Safety. AS 28.15.166(f). The issues to be determined at the hearing are expressly limited as follows:

(g) The hearing under this section shall be limited to the issues of whether the arresting officer has reasonable grounds to believe that the person was driving a motor vehicle while intoxicated and whether ... (2) the chemical test ... 'produced a result described in AS 28.35.030(a)(2).

AS 28.15.166(g) (emphasis added). If both of these issues are determined in the affirmative by a preponderance of the evidence the license is revoked. AS 28.15.166(j). If one or both of the issues are determined in the negative the license is not revoked. Id. Alaska Statute 28.35.030(a)(2) describes the “result” referred to in AS 28.15.165(a) and AS 28.15.166(g)(2). It provides:

(a) A person commits the crime of driving while intoxicated if the person operates or drives a motor vehicle or operates an aircraft or a watercraft

[[Image here]]

(2) when, as determined by a chemical test taken within four hours after the alleged offense was committed, there is 0.10 percent or more by weight of alcohol in the person’s blood or 100 milligrams or more of alcohol per 100 milliliters of blood, or when there is 0.10 grams or more of alcohol per 210 liters of the person’s breath[.]

AS 28.35.030(a)(2).

The primary function of AS 28.35.030(a)(2) is to express the elements of one of three ways in which a person may commit the crime of driving while intoxicated.3 Although the elements of the subsection (a)(2) DWI offense have never been defined by this court or by the court of appeals, the court of appeals has interpreted language in a municipal ordinance similar to AS 28.35.030(a)(2). See Erickson v. Municipality of Anchorage, 662 P.2d 963, 967 (Alaska App. 1983). The court construed the ordinance to require the driver’s actual breath alcohol level to be above legal levels at the time of operation of a motor vehicle. Id. at 967. According to the plurality of the court, an elevated test result was intended to be presumptive proof of an actual illegal alcohol level. Id. Judge Singleton wrote a concurring opinion in which he expressed the view that a bad test result alone — so long as the driver had not consumed alcohol between the time that he drove and was tested — was the essence of the offense. Id. at 970 (Singleton, J., concurring). Judge Singleton stated: “The jury need not determine the precise blood alcohol level that existed at any given time while the defendant was operating his vehicle.” Id. at 968. This difference of opinion is reflected in the case law of other jurisdictions.

Some courts read their DWI statute, as Judge Singleton did, to create an offense of registering a blood/breath alcohol test reading in excess of the statutory limit. See *759Nugent v. Iowa Dep’t of Transp., 390 N.W.2d 125, 128 (Iowa 1986); Schildgen v. Commissioner of Pub. Safety, 363 N.W.2d 800, 801 (Minn.App. 1985); State v. Lentini, 240 N.J.Super. 330, 573 A.2d 464, 466-67 (N.J.Super.App.Div. 1990). If a properly administered test registers a result at or above the statutory level the offense is automatic.4 Under this view, any margin of error or inherent inaccuracy in the testing technology can be seen as tolerated by the legislature which prescribed the statutory requirements of the offense. See State v. Rucker, 297 A.2d 400, 403 (Del.Super. 1972); Slagle v. State, 570 S.W.2d 916, 919 (Tex. 1978). Thus, evidence concerning the testing equipment’s margin of error is irrelevant because such evidence tends to prove the individual’s actual alcohol content and does not shed light upon the proper functioning or use of the testing equipment.

Under a second approach, courts read their DWI statute, as the plurality of the court of appeals did in Erickson, to require an actual blood or breath alcohol content at or above the statutory level. See e.g., People v. Campos, 138 Cal.App.3d Supp. 1, 188 Cal. Rptr. 366, 368 (Cal.Super. 1982); State v. Prestier, 455 N.E.2d 24, 27 (Ohio Mun. 1982). Under this view, the results of a chemical sobriety test are treated as a means of proving actual alcohol content. Thus, the inherent margin of error for any testing equipment is relevant to the issue of the accuracy of the test equipment’s measurement of actual alcohol content. Courts in these jurisdictions admit and permit the utilization of non-test evidence of intoxication. E.g., State v. Babcock, 227 Neb. 649, 419 N.W.2d 527 (1988); State v. Gates, 7 Haw.App. 440, 777 P.2d 717, 720-21 (1989); State v. Brockway, 2 Ohio App.3d 227, 441 N.E.2d 602 (1981); State v. Keller, 36 Wash.App. 110, 672 P.2d 412, 413-14 (1983).

While AS 28.35.030(a)(2) is ambiguous with respect to the necessary elements of the crime of DWI, does the ambiguity carry through to the conduct necessary to give rise to an administrative license revocation under AS 28.15.165 and .166? Today’s majority opinion states that the criminal DWI offense requires an actual alcohol level at or above statutory limits, and implies that an illegal actual alcohol level is also needed for license revocation. I disagree with the latter conclusion. It seems to me that the most straightforward reading of our statutes is that all that is required for license revocation is a failing test result based on a properly administered test. Alaska Statute 28.15.165(a) requires only “a chemical test” properly administered which “produces a result” at or above statutory limits. Use of the term “result” in .165(a) rather than language which suggests the need for illegal levels of alcohol present in the person’s blood or breath, as in AS 28.35.030(a)(2), points to this conclusion. Moreover, the fact that by the express terms of AS 28.15.166(g) the issue at the license revocation hearing is limited to whether the chemical test “produced a result” at illegal levels also suggests that the test result itself is the critical element.

As noted above, in jurisdictions where an actual level of alcohol is critical, evidence of intoxication independent of the test result is admissible. Today’s opinion concludes that no evidence other than the test result may be admitted to prove actual alcohol levels. This is an accurate reading of the limitation imposed by AS 28.15.165(g)(2) but that limitation only makes sense if the test result rather than actual alcohol level is the critical fact for determination. One must ask why a legislature would enact a statutory scheme in which actual levels of alcohol are critical and then preclude the state from employing non-test evidence which tends to prove that such levels exceeded legal limits in particular cases? The answer is that this would be a very odd thing for a legislature to do. No rationale for a system weighted so artificially and heavily in favor of the drinking driver can readily be hypothesized. Thus, the legis*760lature probably intended that the critical element for license revocation was merely a failing test result.

B. Barcott v. State, Dep’t of Public Safety

As in this case, the driver in Barcott v. State, Dep’t of Public Safety, 741 P.2d 226 (Alaska 1987), was tested by an Intoximeter 3000. The test indicated a .10 alcohol level and the driver’s license was suspended. Id. at 227. Barcott claimed that the administrative hearing officer denied him due process of law by refusing to consider the device’s .01 inherent margin for error. Id. We agreed and remanded for further proceedings. Id.

The Barcott opinion is unclear as to whether a failing test result produced by a properly administered test was sufficient for license revocation or whether a failing test and an illegal alcohol breath or blood level are required. There is much in the opinion which suggests that the court assumed that both a failing test result and an illegal level of alcohol were essential. Throughout the opinion the controlling issue is framed in terms of the hearing officer’s refusal to consider evidence of the margin for error, not that the margin for error was necessarily dispositive. See id.

The rehearing history of Barcott as reflected in the public records of this court, but not in the published opinion, makes clear that our decision was based on the assumption that actual alcohol content was relevant. When the original Barcott opinion was issued the final sentence read: “The decision of the hearing officer is reversed.” The State filed a petition for rehearing, contending that the reversal without a remand instruction amounted to a finding by this court that revocation of Barcott’s license was not permissible under any circumstances and, therefore, Barcott’s license had to be restored without the State having an opportunity to present its case to a hearing officer. The court granted the State’s petition and added the current remand language: “and the case is remanded to the department for further proceedings consistent with this opinion.” Id. at 230. The only sensible explanation for this action is that the Barcott court assumed that a driver’s actual alcohol level was an element of the license revocation offense, therefore the State should be allowed on remand to present evidence of the driver’s actual alcohol level in addition to the test score.

Given my conclusion that a failing test result alone is the critical element in license revocation cases and that a driver’s actual alcohol level is irrelevant, Barcott’s holding that it is a violation of due process not to consider a testing device’s inherent margin for error is plainly wrong. Drunken driving is a social problem of considerable magnitude. The legislature can respond to this problem by making it an offense to drive when a test of the driver’s blood or breath ' yields a given test reading as long as there is a reasonable relationship between the' level established by the legislature and driver impairment. In setting a level, a certain testing margin of error is tolerable within reasonable limits. Clearly the legislature could prohibit driving with an actual alcohol content of .09.5 Thus, there is no reason to condemn as fundamentally unfair, and therefore unconstitutional, a license revocation procedure in which a driver is sanctioned for a test reading of between .10 and .109, given a .01 margin for error.

If, on the other hand, actual alcohol level is the critical element, I agree that it would be unfair not to permit consideration of the testing device’s margin for error. However, it is one thing to allow consideration of a device’s margin for error along with other evidence bearing on the issue of intoxication.6 *761This is what we did in Barcott. It is another and much more questionable thing to conclude that whenever a test result is such that considering the margin for error an innocent alcohol level is possible, there can be no license revocation because no other evidence of intoxication can be received. This conclusion is strange not only because there is no apparent purpose for the limitation, but also because no other jurisdiction in which actual alcohol level is critical employs such a limitation.

III.

In conclusion, I would affirm the decision of the superior court which affirmed the revocation of Haynes’ driver’s license. The critical requirement for license revocation is a properly given test which produces a failing score. Barcott, which assumes otherwise, should be overruled. The legislature has established the point at and above which a test score is to be regarded as failing. That point is .10. The majority opinion errs in changing that point to .11 where an Intoxim-eter 3000 is used. Under our view of the plain meaning rule of statutory construction, the apparent plain meaning of a statute cannot be changed except where there exists compelling legislative history that a different meaning was intended. By relying on the absence of legislative history to alter the plain meaning of the current statutory system, today’s opinion stands the plain meaning rule on its head.