Gerald Van Griffin was convicted of driving under the influence of alcohol in Lake Mead National Recreation Area, a petty offense under 36 C.F.R. § 4.6 (1986). He appeals. We affirm his conviction.
BACKGROUND
On March 23, 1986 Gerald Van Griffin was observed by Park Ranger Daniel Ol-trogge driving in reverse for approximately one-half mile with the lid of his trunk up, obstructing his rear vision. Oltrogge stopped Griffin at 3:50 p.m. at a checkpoint set up to stop all vehicles leaving this side of the park. Oltrogge detected a “moderate odor” of some alcoholic beverage on Griffin’s breath. According to Oltrogge, Griffin’s speech also seemed to be slurred and his eyes were bloodshot. The ranger asked him whether he had been drinking and he replied that he had had seven or eight beers. The ranger then asked him to take a series of field sobriety tests and Griffin agreed.
In the first of these tests, Griffin was asked to recite the alphabet. He reached m but then skipped to t and continued to z. In the second test he was asked to count backward from 100. He reached 84 then went back to 89 and then back to 91. In *636the third test he was asked to count his fingers. He failed to do so accurately. The fourth test was for nystagmus or jerkiness of eye movement. The ranger held his pen in front of his eyes at a distance of about 12-16 inches and moved it back and forth, asking Griffin not to move his head but to follow the pen with his eyes. The ranger determined that jerkiness in the eyes appeared at an angle between the pen and the eye of about 35 to 40 degrees. Following his instructions, the ranger deducted the degrees from the number 50 to arrive at a rough estimate of Griffin’s blood alcohol content; he concluded it would be between .15% and .10%. The ranger then arrested Griffin.
Griffin was taken to a ranger station where, with his consent, he underwent an analysis of his breath. The breathalizer took two samples of his breathing and registered a blood alcohol content of .12% and .13%. Griffin asked for additional tests of his blood and urine and the requests were rejected.
PROCEEDINGS
Griffin waived his right to a jury trial and was tried on May 28, 1986 before a United States Magistrate. During the trial counsel for Griffin observed that the magistrate had on the bench before him the ranger’s citation of Griffin, to which was attached the ranger’s report of the incident. This report contained Ranger 01-trogge’s observations of Griffin and the results of the sobriety tests administered in the park. Counsel objected to the court having this document. Magistrate Sattler declared, “We’ll state for the record, the court has not read, or reviewed, or examined that report or its contents.” Counsel then moved for the magistrate to disqualify himself. The magistrate responded, “The realities of this proceeding are, I have not looked at, read, reviewed, or examined that document.” Counsel pressed his motion for recusal, and the magistrate denied it.
The principal witness against Griffin was Ranger Oltrogge, who testified to the observations set out above. He was crossex-amined by counsel for Griffin, who asked him if he knew of a publication of the National Highways Traffic Safety Administration, a division of the Department of Transportation, which addressed the proper procedures for testing nystagmus. Ranger Oltrogge said he had no knowledge of the manual and had not based his test upon it. He was asked if he knew that it was critical to have the correct angle from the eye to the top of the pen and he replied that he would not say it was critical, but then added that if it was off by a few inches it “could skew that particular maneuver.” He also said he was aware that everyone had natural nystagmus, and when asked if he was aware that the test was estimated to be only 77% accurate, he replied, “Oh yes sir. We were told it is not a cure all.” He was then asked whether he agreed that, if he had not observed any white of the eye at the extremes to the right and the left of the respective eyes, he should not have “scored” the person being examined. He replied, “Well, I would not quarrel with them, I just ... I was instructed to do it in a different manner. I would have to research it a little bit.” Id.
At this point, defense counsel offered as an admission the government manual he had referred to. It is entitled “Improved Sobriety Testing” and is a 13-page pamphlet setting out the correct procedure to be followed in a variety of sobriety tests including one for nystagmus. The court excluded the publication on the grounds that it was hearsay. Counsel objected and then continued with his cross-examination.
Griffin was acquitted of careless driving, a petty offense under 36 C.F.R. § 4.14 (1986), and convicted of driving under the influence, a petty offense under 36 C.F.R. § 4.6 (1986). He received a suspended sentence of one year, was fined $225 and was placed on supervised probation for one year on condition that he complete forty hours of community service, attend a counseling program, submit to periodic drug and alcohol testing, and stay away from the Lake Mead Recreation Area until November 29, 1986. Griffin appealed.
ANALYSIS
Bias of the Magistrate. 28 U.S.C. § 455 provides in pertinent part:
(a) Any justice, judge, or magistrate of the United States shall disqualify himself *637in any proceeding in which his impartiality might reasonably be questioned, (b) He shall also disqualify himself in the following circumstances:
(1) Where he has a personal bias or prejudice concerning a party, or personal knowledge of disputed evidentia-ry facts concerning the proceeding;
The purpose of the provision is to promote public confidence in the integrity of the judicial process. To achieve this purpose the judge must be disqualified if a reasonable person would believe that the judge was not impartial. Liljeberg v. Health Services Acquisition Corp., — U.S. -, 108 S.Ct. 2194, 2202, 100 L.Ed.2d 855 (1988).
Absent a basis for believing that Magistrate Sattler had any reason to lie, a reasonable person would not doubt his denial that he had in any way examined the incident report that was on the bench as he conducted the trial of Griffin. It was not good practice for such an ex parte communication from the police to be in his possession. It was not good practice because a reasonable person might suppose that he would read what he received. This impression is dissipated by his emphatic declaration that he did not read the communication, so that in this case, on these facts, the mere receipt without reading did not create an appearance of bias in violation of 28 U.S.C. § 455(a).
Counsel for Griffin, however, contends that an appearance of bias was created by the magistrate’s continued retention of the report in that he might have been tempted to look at it and compare it with the testimony of Oltrogge and Griffin.
If it was not good practice to receive the communication, it was equally not good practice to retain it. We have no assurance from the magistrate that he did not subsequently look at the communication. Certainly he might have been tempted to do so. The rule requiring disqualification where there is an appearance of partiality is a rule designed to remove the possibility of temptation. Tumey v. Ohio, 273 U.S. 510, 532, 47 S.Ct. 437, 444, 71 L.Ed. 749 (1927). It is not an imputation against the honesty of the magistrate to say that a reasonable person could doubt his impartiality when he kept with him at the trial this ex parte communication. If a jury had received such an ex parte communication that had a reasonable possibility of affecting the verdict, its verdict would be tainted. United States v. Vasquez, 597 F.2d 192, 193 (9th Cir.1979). Here the magistrate was the trier of fact and he did not dispossess himself of a communication which he himself implicitly acknowledged he should not have had. A reasonable person could doubt his impartiality.
The general rule has been that when a judges sits in violation of an express statutory standard, the disqualified judge’s rulings are, on appeal, to be vacated. William Cramp & Sons Skip & Engine Bldg. Co. v. International Curtiss Marine Turbine Co., 228 U.S. 645, 651, 33 S.Ct. 722, 724, 57 L.Ed. 1003 (1913). However, it has recently been indicated that if a judge violates § 455(a) “there is surely room for harmless error.” Liljeberg, 108 S.Ct. at 2203. While the court in Liljeberg specifically refers to error “committed by busy judges who inadvertently overlook a disqualifying circumstance,” and such is not the case here, the Court also said, “[t]here need not be a draconian remedy for every violation of § 455(a).” Id.
The instant case is one where a draconian remedy would be inappropriate. The magistrate is shown to have been guilty of a bad practice but not of actual bias. We have no reason to believe that he read the report. We do not believe he did. But he created the appearance of bias. Nothing in the document before him added to the testimony of Ranger Oltrogge. The incident report and Oltrogge’s testimony are virtually the same. It is a reasonable inference that Oltrogge refreshed his memory by the incident report when he testified. No new facts are in the report that could have influenced the magistrate.
On behalf of Griffin it may be suggested that the very similarity of the report and Oltrogge’s testimony was helpful to the government by enhancing the credibility of Oltrogge’s testimony. Abstractly this possibility can be conceived. Concretely it is extremely difficult to believe that any rational trier of fact would have been im*638pressed with the congruence between an officer's report and the testimony he offered after refreshing his memory by looking at the report. Beyond a reasonable doubt, Griffin would have been convicted whether or not the magistrate had looked at the report.
Admissibility of the Department of Transportation Manual. The basis on which counsel for Griffin sought to introduce the manual was to impeach Griffin but Ranger Oltrogge testified that he had not relied upon or even ever heard of the manual. The manual therefore was not a challenge to the ranger’s testimony and therefore not proper impeachment. McCormick on Evidence § 34 at 72-73 (3rd ed.1984).
The manual, however, could have been introduced by the defendant as part of his defense in order to show the measures that are necessary to be taken in order to have a reliable test for nystagmus. We do not say that every publication of every branch of government of the United States can be treated as a party admission by the United States under Fed.R.Evid. 801(d)(2)(D). In this case the government department charged with the development of rules for highway safety was the relevant and competent section of the government; its pamphlet on sobriety testing was an admissible party admission.
Although the manual should have been admitted, the exclusion was harmless. The nystagmus test was only one of several bases on which Ranger Oltrogge formed his opinion that Griffin was drunk. Oltrogge himself acknowledged that the test was far from being 100% reliable even when perfectly performed. The other indicia of Griffin’s state were more than sufficient to carry the conviction that he was under the influence of alcohol.
The Requested Blood and Urine Tests. Griffin makes no claim that the government acted other than in good faith and in accord with normal practice. Arizona v. Youngblood, — U.S. -, 109 S.Ct. 333, 337, 102 L.Ed.2d 281 (1988). He does not claim that the government knew the test would exculpate him but decided not to perform them. He makes no claim that he requested his own doctor and was prevented from having him perform the tests. Under established law he had no right to require the state to perform these particular tests. Youngblood, 109 S.Ct. at 338 (“the police do not have a constitutional duty to perform any particular tests”). The evidence against Griffin was based on the convergence of four tests administered in the field, the breath analysis and Ranger Oltrogge’s observations of his appearance and conduct. It was overwhelming evidence of his being under the influence while driving in the national park.
AFFIRMED.