ROBERTS, Justice.

For the first time we must decide whether the result of a breath test performed by use of an Alcohol Level Evaluation Roadside Tester (ALERT) device is admissible at trial on a charge of operating a motor vehicle while under the influence of intoxicating liquor (OUI), 29 M.R.S.A. § 1312-B (Pamph. 1988). Robert Ifill appeals his conviction entered on a jury verdict in the Superior Court (Somerset County, Chandler, J.). He challenges the submission, over his objection, of testimony concerning the result of an ALERT test administered by the arresting officer, the admission of an exhibit to the jury and the sufficiency of the evidence. Because we agree with Ifill's first challenge, we vacate the judgment.

I.

According to his testimony at trial, State Police Officer Robert Rowe was stopped at a gas station near the Newport exit of I-95 on October 4, 1986. A motorist pulled in and told the officer that he had been harassed by a rental truck travelling close behind him at speeds up to 90 m.p.h. Rowe relayed this information, as well as the license plate number of the truck, by radio to State Police Officer Robert Williams who was running radar in Pittsfield.

A few minutes later, Williams stopped the same truck, which was being driven by defendant Ifill. Williams smelled alcohol from inside the cab but wasn't sure the odor was coming from Ifill's breath. Williams told Ifill to follow him back to his cruiser, and once both were inside, found that the odor was in fact coming from Ifill's breath. He noticed that Ifill's face was slightly red, that his eyes were bloodshot and glassy, that his speech was thick, and that he misspoke several times. Ifill said he had two beers two hours earlier. asked to recite the alphabet, Ifill stopped midway through it and repeated it incorrectly twice.

Williams testified that he told Ifill he believed Ifill was under the influence but offered to give him an ALERT test. He further testified that the ALERT test is an orange box that a person blows into and that registers "pass" or "fail." Although the test does not register a precise number, a "fail" reading indicates a blood alcohol content above .10% (the statutory level of excessive blood alcohol in 1986, reduced to .08% by P.L. 1987, ch. 791 § 18). Williams was not sure how the machine works but knew that it is calibrated by correlation with an intoxilyzer and that his particular unit had been calibrated the year before. Ifill voluntarily took the test and registered a "fail." Williams then told Ifill he was legally obligated to take a blood or breath test and read him the implied consent form. Ifill refused any further test.

A motor vehicle operator must submit to a blood or breath test to determine blood-alcohol level upon probable cause for OUI. 29 M.R.S.A. § 1312 (1978 Pamph. 1988). Any "self-contained breath-alcohol testing apparatuses" shall be approved by the Department of Human Services. The ALERT instrument is not approved. Nevertheless, Ifill's pretrial motion to exclude testimony about his performance on the ALERT test was denied by the court on the day of trial. The court concluded that the ALERT test is admissible, like the heel-to-toe test, finger-to-nose test or alphabet test, as a field sobriety test. Because there was no official blood or breath test, and because Ifill's refusal to take one was offered against him, the court emphasized in jury instructions the difference between the ALERT test and the test required by Section 1312.

Following a guilty verdict, Ifill moved for a new trial on the ground that an inadmissible statement of Ifill's driving record had been submitted to the jury as part of the Uniform Traffic Ticket and Complaint (UTT). Despite the fact both parties and the court had agreed that the record portion of the exhibit would be excised, the record was submitted to the jury. The court denied the motion, nevertheless, on the ground that the error was harmless.

II.

We have previously considered the ALERT test in an appeal of an administrative suspension, but there the test was not directly challenged. See Mancini v. Secretary of State, 540 A.2d 117 (Me. 1988) (evidence including ALERT and breathalyzer tests sufficient to support finding of excessive blood alcohol by preponderance). In the case before us, the State makes no attempt to defend the accuracy of the ALERT test. Rather, the State contends that its results are admissible to establish probable cause and "to prove the state of mind of the defendant at the time that he refused the [implied consent] test." We reject the first argument because the existence of probable cause is not a proper issue for consideration by the jury. We reject the second argument because the evidence reveals nothing except the defendant's erroneous belief in the validity of the test result. Cf. State v. Trafton, 425 A.2d 1320, 1322 (Me. 1981) (because polygraph test is unreliable, evidence of refusal not probative of consciousness of guilt).

The Supreme Court of North Dakota has described in detail the imprecise nature of the ALERT test:

The ethanol sensor used in the A.L.E.R.T. instrument is manufactured by Taguchi Gas-Electric Transducer (TGS). Quoting from its brochure, the manufacturer . . . states:

The characteristics of the TGS are not at present sufficiently reproducible to make it suitable for use in laboratory or analytical instruments where high accuracy is required

. . . .

Quoted in Erwin, 2 Defense of Drunk Driving Cases § 24.11 (3d ed. 1987) (emphasis added [by court]). . . . [T]he A.L.E.R.T. unit's response is not specific for ethanol and could possibly emit a "Fail" signal due to the presence of some other chemical substance. While a quick and convenient tool for establishing probable cause for an arrest, the A.L.E.R.T. screening test can be unreliable. Not only can the test register an inaccurate percentage of alcohol present in the breath, it may also be inaccurate as to the presence or absence of any alcohol at all. See State v. Deshaw, 404 N.W.2d 156, 158 (Iowa 1987). Therefore, because of the questionable accuracy of the A.L.E.R.T. testing instrument, it lacks the indicia of scientific reliability necessary to provide an evidentiary basis for establishing that an individual was driving in violation of [North Dakota's OUI statute]. . . . See Erwin, 2 Defense of Drunk Driving Cases § 24.11(1) for a discussion of studies done in evaluating the accuracy of the A.L.E.R.T. breath testing unit.

State v. Schimmel, 409 N.W.2d 335, 337 n. 1 (N.D. 1987). The Schimmel court determined that it was error to admit testimony concerning the result of the ALERT test for consideration by the jury.

The North Dakota court concluded, however, that since the State presented evidence that the defendant had taken two blood alcohol tests with the results above .10%, and because the defendant was charged with excessive blood alcohol (EBA), the prejudicial effect of testimony concerning the ALERT test was negligible. The admission of that testimony was thus harmless error. Id. In contrast, the Supreme Court of Iowa has determined that use of a preliminary breath test at trial on an operating under the influence charge (OUI) is reversible error. State v. Deshaw, 404 N.W.2d 156, 157 (Iowa 1987). While not referring specifically to the ALERT test, the Iowa court determined that the "preliminary screening test" could register an inaccurate percentage of alcohol present in the breath, or inaccurately report the presence or absence of any alcohol at all. Id. at 158.

III.

In the case at bar, the trial court determined that the result of the ALERT test should be admitted like any other field sobriety test. We disagree. Tests of physical dexterity, coordination, speech or memory provide observations from which the jurors may draw their own conclusions about a defendant's sobriety. The ALERT test, on the other hand, is meaningless to a jury unless the pass-fail result is perceived as scientifically reliable. It is precisely that false aura of reliability that renders unavailing a trial court instruction that the jury need not accept the test result. Despite the existence of other evidence sufficient to establish alcohol-related impairment, we are unable to conclude to a high probability that the error in admitting evidence of the results of the ALERT test did not affect the jury's verdict. See State v. Reeves, 499 A.2d 130, 137 (Me. 1985). Accordingly, Ifill's conviction must be set aside.

IV.

Ordinarily we would find it unnecessary to discuss the error in submitting the unredacted State's exhibit to the jury because it is unlikely to recur. We discuss the question only to reject the State's suggestion that the error is not preserved. Ifill made clear his objection and the State agreed to excise the inadmissible portion. The record reflects no opportunity thereafter for defense counsel to examine the exhibit. The responsibility for its redaction remained with the prosecution.

We reject Ifill's challenge to the sufficiency of the evidence. Accordingly, we remand for further proceedings.

The entry is:

Judgment vacated.

Remanded for further proceedings consistent with the opinion herein.

All concurring.